Eesh. Looks like RUclips might still be processing 1080p. It was there for a moment! One thing I realized for Scooby Dooby's comment: GTA V specifically has issues with i5 CPUs that hit the 187FPS limiter in the game, which is something we've previously discussed. In this very peculiar scenario, you actually would greatly benefit from a framerate limit.
I would love to see the industry adopt a V-sync standard that aims to reduce latency by only allowing frames to render in the last 8ms of a 16ms frame time. Or even a 4ms target, or better yet allow the player to customize the frame time allotment. CS:GO type players would probably not bother because they'll take frames at any level of completion, but for people wanting to run on a 60hz monitor with low latency and without any tearing (tearing being much more detrimental at 60hz), it would be ideal to be able to just set your frame yourself. The implications on power consumption could be significant as well. Theoretically if you give a 30hz system the same frame time as a 120hz system, it would feel much better for the player and it would consume 1/4 as much power. Nintendo could really use something like that, but if we got the option on PC then it would immediately be an option for Laptops, and more importantly the freshly re-emerging UMPC market with things like the GPD Win.
Steve is correct when it comes to the video editing software. I use Adobe Premiere Pro as well and the time it would take me to learn something else is worth more than a new computer would cost. I could easily spend 100+ hours learning a new program to the level I know Premiere Pro today, even if my time is only worth $10/hr, that is $1,000 worth of time.
It's unfortunate that Blender is rather difficult to learn. What happened though to Adobe's promise last year that they would focus on hw performance and optimisations instead of new features? They said the move was based on customer feedback.
Sadly true, and it makes it all the more difficult for those with older versions who don't like radical changes like dropping RT3D and switching to C4D, especially when they've built up a system optimised for the former; doubly so when Adobe updated its libs for Premiere to support Maxwell CUDA V2, but didn't do that for AE. And then there was the switch to CC and the subscription model. This all hurt a lot of solo pro users.
Tech Deals eh.. if you are a good editor I would think that you could adapt to another video editor fairly easy. Your still doing the same things. It's just a matter of learning keyboard shortcuts and navigating the program. But it will still have a timeline and preview window.
Often, sim racers like to FPS cap at a multiple of the monitor refresh rate. Usually 2x but I run 3x on Iracing which is very lightweight. 180fps on (3) 60mhz monitors. It's more consistent, as Op states, and helps avoid tearing without the input lag caused by V-sync. Switching from 120 cap to 180 cap in Iracing I could feel/see the difference. The information being fed to you is that little bit more up to date, while your input lag is minimized. Sim racers and first person shooters crave frames like an addict fiending for crack. But, capping is still good because consistency > max frames.
overwatch is such a great example, even though the display based option is kinda broken (usually caps fps about 10% higher than refresh rate), but the 30 (if youre on a toaster) or the specific number one is great. I run the game pegged at 144 fps, never goes up or down, smoothest 6.9ms I get to date
This was explained by a Futuremark engineer: If the GPU is fully loaded, everything's generally fine in terms of frametimes. However, if your CPU gets fully loaded by the game engine, frametimes can go haywire because the GPU driver isn't getting enough CPU time to do it's work. In such a case, it's useful to limit the FPS so that the CPU can have some breathing room and the GPU driver isn't starved for CPU time.
Marcelo Tezza That answer is rather simplistic and doesn't actually answer the question of what happens when capping framerate. It talks about performance behavior in general, but you learn nothing about what goes on under the hood of the components. They are addressing the subject in a different way. Our answer looks at it from a canonical pipeline perspective and gives more detail on the backend of the render pipe.
It's fairly common in demanding AAA games for at least one thread of a CPU to be under load at least some of the time. Action games, unless they are a benchmark, are highly dynamic. You can go from low CPU loads to high CPU loads very quickly, depending on what the player is doing A framerate limit or vertical sync can help smooth out that experience. I've played some games (not action games, but demanding adventure games like Detroit) at 30 frames per second just to get a smooth video-like quality.
I love it. It's really great, even on screen when they show graphs for benchmarks they'll have a bar that "counts down" on the very edge of the left and right sides of the screen. At first glance it looks like just graphics/window dressing but it's actually the same thing only on the borders of the screen. Helps so much in being able to find a piece of info again that your looking for for those that are learning this stuff and we need to go back and replay a certain point. I can absolutely see this feature becoming a new norm.
Figured I'd just give a shout and say thanks Steve for getting an answer to my question :) . Don't know how often you actually get a thank you, but figured you deserve one. And thanks for taking my question by the way. Didn't think it'd get picked since it was pretty late into the week when I asked it. I figured it just came down to cost when I asked, but you mentioned a few things like in-store retailers that I didn't think about, so thanks again for the answer and insight. :)
I came here specifically for the second question regarding frame rate limiters and if they just accept the first x amount of fps or if they actually spread the frames throughout each second, and I have to say I couldn't be any happier with your answer. I plan to watch the rest of the video when I have time simply because of how thorough and clear you were with that response - I feel I can learn so much from these videos. Subscribed immediately. Wish I came across this channel sooner!
DJHeroMasta there is always input lag when capping your frame rate because of back pressure. no matter what cap method you use. otherwise freesync and gsync wouldn't exist. you only get stable frame pacing, but it won't fix input delay
Marcus Mathiassen got a link to that? From what I seen when it comes to comparing in game cap and RTSS cap is RTSS is worse for input lag. DJHeroMasta you get tearing if you not using vsync or g/freesync. You not seeing it don't mean it's not there.
Got no link, but from my own experience RTSS seems to cache frames which increases input lag and burn cycles which increases cpu usage to get that perfect frametime. I don't know for sure though.
Question for ask GN: I have recently gotten very into video editing, but I feel like my PC (6700k, GTX 1070, 32gb ram) is being severely bottlenecked when editing 4k footage due to my lack of an SSD. I am currently reluctant to purchase an SSD for editing, for the fact that it would be constantly reading/writing at 100+ MB/s, and I would hate to have the SSD die after just a few months due to me reaching the maximum TBW of the drive. Is this a legitimate concern for people like yourself, who are constantly editing high bitrate video, or should I not be concerned about this at all. And if this is a legitimate concern, would a better solution be to edit off of a pair of hard drives in raid, in order to increase the longevity of the solution? Thanks for reading my question, and keep up the amazing work :)
jonny j Thank you for your response, I'll definitely check out that article. I still don't know if uninformed necessarily equals 'fucking stupid' though :)
+Harrison Glenn First off, never base a decision on feelings. ;) Use system monitoring tools to check what is happening during your workflow (Afterburner, GPU-Z, NV tools, Process Explorer for I/O, etc.) Are you sure it's not a CPU bottleneck? Also, for apps like Premiere, AE and Vegas it's essential to use an SSD as a cache even if the main working store is not an SSD (though it really should be, especially for 4K), while for obscure technical reasons Vegas should use one for its main working area aswell anyway. If you're working with 4K though, then yes it's very likely I/O is holding you back, but check t be sure. As for reaching TBW, don't worry about it. SSDs can usually handle a lot more than their official rating, just look at how much better OCZ's Arc 100s did when they were hammered until failure, all samples went waaay beyond their official limits: www.kitguru.net/components/ssd-drives/zardon/ocz-challenge-kitguru-to-kill-5x-arc-100-ssd-drives/ though I would recommend fitting an M.2 NVMe SSD instead of a SATA in order to get much higher speed, and this will avoid the potential pitfall of some SATA SSDs which have rather low steady state performance (ie. when cache buffers get filled up, etc.) Older SATA SSDs were actually better for this; newer models are kinda meh with the adoption of DRAMless designs and TLC. Kinda ridiculous that my old Vector holds up so well to newer models, but there you go, atm the market is in a push to margins, not quality. One thing though, lesser models in any product range do have lower TBW ratings, so if you can do go for something better anyway, and of course the higher capacities have higher ratings so 512GB is the minimum I'd use for 4K. I recommend the SM951, SM961, 950 Pro or 960 Pro. Note if you lack a spare M.2 PCIe slot, just fit the SSD in a PCIe addin card (I use an Akasa AK-PCCM2P-01, good space for cooling), and a simple heatsink can prevent throttling. Even using a PCIe 2.0 slot, I get 2GB/sec with an SM951 (P67 ASUS M4E, 5GHz i7 2700K)), while a 960 Pro on a 3.0 slot gives 3GB/sec (X79 ASUS R4E, stock i7 4820K), so one doesn't have to be using a newer chipset to get good performance from NVMe SSDs. If you're going to use RAID at all, that's probably best focused on RAID1 for rust spinners to hold all your final data, and don't use lesser consumer drives, get something decent. Personally though I just duplicate data manually, eg. my main photo/video archive is on an SM961 512GB NVMe (the ASUS M4E config mentioned above), and then it's regularly copied/updated to a hot-swap 850 EVO 500GB SATA. 8) Ian.
mapesdhs Thanks for your detailed response. I have been monitoring usage in task manager while editing, and in most cases, after effects lags when disk usage is at 100%, with CPU rarely going over 50%. Even when rendering, this is the case, which boggles my mind. In one case I was attempting to render a particle effect that I created with multiple plugins, and my CPU was under 10 % for all but the first 5 mins of rendering. I'll definitely look more into nvme ssds before I make my final decision. Thanks again for your insight
An nvme ssd (or two if you wanna do 2 in raid on a pcie card since z170/270 m.2 slots are all chipset based) would be your friend, as mapesdhs said, ssd's go well over their write ratings. Since this sounds like a semi professional system, a raid 5 array consisting of 4 2tb drives would make for some nice storage for your data (and ssd backup). If your ssd is backed up (or its a dedicated scratch disk and your os is on a sata drive somewhere) then you dont need to worry about endurance as if the thing starts to fail (which means you will lose capacity slowly) then you will have more than enough time to replace it, and by then the drive will be outdated and m.2 drives on pcie gen 4 will be mainstream. just get yourself a 960 pro of your desired size and call it a day, as the quality of that nand and memory controller is insane, not to mention the speed
+Harrison Glenn, Ah ha, you're using AE! Heh, that app is a right kettle of fish when it comes to performance. First thing to remember is that various plugins for AE were never updated to be multithreaded.Thus, although Task Manager might show an overall low usage, one particular thread may be holding up the overall render process. Use Process Explorer to monitor individual threads to see what they're doing. Friend of mine discovered this when setting up some AE tests for me, and speaking of which if you want to bench your setup then try our CUDA AE benchmark (still writing up a page for it): www.sgidepot.co.uk/misc/cuda.101.zip The default test is just one frame, probably take 10 or 15 mins on your system. Note that rendering the entire animation is for stability testing purposes only, not benching (ie. checking render rig reliability for overnight crunches). The single frame test output looks like this (change the suffix to tga if you prefer the raw Targa): www.sgidepot.co.uk/misc/cuda.101_Frame96.jpg Anyway, here are the most important things about AE (and note you can find much longer and more detailed versions of all this in my posts on Creativecow): - AE is a RAM gobbler. You already have 32GB which is ok, but double that if you can. Also (I'm assuming here your C-drive is already an SSD), it's a good idea to fit a separate SATA SSD to hold the Windows paging file. A 128GB is suitable (something with high IOPS, like an 840 Pro or Vertex4/Vector); the partition should be 1.5X the fitted RAM, so for a system with 32GB RAM the dedicated partition should be 48GB (or to be exact if one is fussy, and I am, 48GB + 6MB, ie. 49158MB, because there's a tiny overhead involved. :D) I normally define this partition to be drive V and call it VMEM. You can setup a Registry entry to prevent Windows moaning about the partition always being full, and another to prevent it being displayed in the Computer devices pane. Not only does this free up a lot of space on the C-drive, it will extend the life of the C-drive aswell. The unused space on the extra SSD can be defined as just a misc scratch area for whatever purpose (I normally call it drive S, named, "SCRATCH"), eg. temporary copies of files, dumping ground for test images, etc. Or just leave the unused space alone. When you have the extra SSD fitted and the logical partition set 49158MB, right-click on Computer, select "Advanced system settings", click "Settings" in the Performance area, select "Advanced", click "Change", untick the Automatic box, then manually define the paging file to be on the V partition and set it to 49152MB. Click the various OKs to confirm and let the system reboot. Once done, you should see the space used on the C-drive has gone down by quite a lot. Speaking of space saving, if you do not use the system Hibernate (sleep) function then you can turn off the hiber file by bringing up a Command Prompt with Admin rights (find the menu option in Accessories from the Start menu, right-click on Command Prompt and select Run As Administrator) and entering this: powercfg -h off Exit the Command Prompt, refresh the devices in the Computer pane and again you'll see the space used on C go down. The above two measures saves a lot of space, especially for a system with 64GB+ RAM (without these measures, using a 256GB SSD for the C-drive does not leave much space for installed apps). - A separate SSD for the different AE cache structures is essential. Doesn't have to be a PCIe, anything will help. If budget is a problem then look for something 2nd-hand, eg. a used Samsung 840 Pro 256GB would be ideal. In other words, whether or not you decide to use an SSD of any kind for the main working store (ie. where the current project data is located, and where rendered output is going to), having an SSD for the AE caches is a no-brainer. Adobe themselves state this on their FAQ page. Btw though, if you have the spare PCIe slots, picking up a used SM951 doesn't cost that much, last one I bought was only 80 UKP for a 256GB (and of all things on receipt it turned out to be new, not used). If you hunt for something used a la SATA, other good models would be the OCZ Vertex4/Vector, Corsair Neutron GTX, Samsung 840/850 Pro and SanDisk Extreme Pro. - Write caching on your disk may be affecting performance, and btw it does of course make sense to be using a different device to the C-drive. Beware though, changing write caching settings can alter whether data may be lost during a power failure. If you switch to an SSD though then there's generally no need to mess with this setting. - RAM speed helps for AE rendering. This is probably not yet an issue for your system as atm you most likely have an I/O bottleneck, but assuming that gets solved in time, do think about maximising the RAM speed if you can. On an X79 system with a 4.8GHz 3930K, my friend observed a 10% render speed reduction when dropping the RAM from 2133MHz to 1866MHz (the system I built for him has 3x GTX 580 3GB, 4.7GHz 3930K, 64GB RAM, etc.) So, first off, get an SSD asap for the AE cache setup, whether it's a SATA or not; I keep hunting for used 840 Pros, but anything good will do (ie. aim for models with high random IOPS ratings). Then later when you can, swtich to an SSD for the main store. Also, have a separate SSD for the Windows paging file. What I do when building such systems for people is have an SSD for the current data being worked on, and then when the day's work is done the results can be copied over to a rust spinner (I've been using 2TB Seagate ES.3 Enterprise disks for a while, getting them new for good prices, bagged another last week for 75 UKP). I normally fit two rust spinners so final data can also be copied as a backup (one can of course use RAID1, but I prefer to do it manually). Lastly, in addition to regular cloning of the C-drive, I fit a lower capacity HDD (usually 1TB atm) to hold general backups, including image dumps of the C-drive, backups of the user folder, etc. If money is tight, here's an example of what is possible re exploiting the used market (a project I'm working on atm): www.sgidepot.co.uk/misc/charitypc1.html One other thing, reading James' reply, note that RAIDing NVMe SSDs won't really gain you very much, and it will hinder random I/O performance a tad. Overall, not worth doing IMO. RAIDing rust spinners can be beneficial though, but I usually stick to RAID0 or RAID10 for a working space (the latter gives both a speedup and some redundancy). RAID5 is a bit more complicated. Mind you, when I was still meddling with rust spinner RAIDs I was using U320 SCSI and later SAS, so the sequential performance was better anyway because of the HBA's abilities, especially a battery-backed HP PCIe card I had way back which had a 1GB DRAM buffer, so it could do 2GB/sec a lot of the time. I don't use such HBAs anymore though, PCIe NVMe makes much more sense. Btw, that's why mbds like the P9X79-E WS are so good, ooooodles of PCIe slots. :D Good luck!! Ian. PS. This is my CUDA research box (ASUS P9X79 WS, 4.7GHz 3930K, 64GB @ 2133, 4x GTX 580 3GB 900MHz, various SSDs): www.sgidepot.co.uk/misc/3930K_quad580_13.jpg For editing though I have an ASUS R5E, 6850K, 950 Pro, etc.
Some comments on case colours: * I totally agree with your point about more colour variants result in higher production costs. * There's a simple work-around for mismatch in colours between plastic and metal parts: Don't try to do them the same colour in the first place! There are many computer cases doing this already. * As a user I've had no problems buying a plain case and can of paint to get the case into my desired colour. Just spray the top and side panels any colour you fancy! * When I was looking for a case to use for my daughter's first computer I initially wanted to go as cheap as possible, but ended up spending twice as much just to get a case that is pink! (It was a better option in other aspects as well, but the colour was what made me look at it in the first place.)
I love this guy. He explains everything so in depth but also simple enough for me to understand. I wish there was a way I could help him expand his viewer base. I have already told everyone I know to watch.
Hi, got another fps related question. Can someone explain why top CPU's, like the i7 7700k or the r7 1800x, draw different fps? I mean- no cpu is at it's limits or am I wrong? So when no cpu core is at a 100% - why did they draw more or less fps? (Without gpu bottlenecking in 1080p for example) Thanks in advance
i always cap my framrate if im getting 100's. it makes the gpu and cpu run at a lower load generating less heat and sound. generaly i cap t around 80-90 or i just put Vsync on.
Back in the 286 through 486 days, our software shop found it worthwhile to have one bleeding edge machine to submit full compiles remotely as a batch job. We'd reinvented the mainframe! Coding, module compiling, & syntax checking would be done locally and package compiling & linking would be done on the 'fast' machine.
I think something relevant to the big discussion on frame rates in this video is that Steve touches on what is knows as frame rate independence - this is the concept of performing actions and animations based on the time between frame switches. Before this was introduced animations etc in games were done entirely on a frame to frame basis, problem being that eventually hardware differences in computers allowed some systems to run the game faster than their slower counterparts.
Question for Ask GN: Something that has been grinding my gears a lot recently is a lot of large tech based channels have moved to using clickbait exclusively for video names and thumbnails. One that quickly pops to mind the (no longer named this way) video named "Ryzen R3 and R5 TESTED" by LinusTechTips with a more fitting name of "AMD Ryzen Emulated On R7". Do you guys plan to switch in the future to get more views, and thus more ad revenue, or stay with the current format and selection of videos?
i'm relatively new to this channel but i just noticed the time markers in the top left that count down for each subject. That is fucking amazing and a feature i never knew i wanted.
Great section on case color! For those that want to change the whole color of the case, there is always paint. It comes in many forms from bottles to spray cans, if the end user wants creativity, then have some. Try not to rely on someone else to do all the work for you. If you have no skills and no ambition, then pay someone else. If you have no money, then ask someone if you can borrow some duct tape and use it on yourself. (be creative)
Regarding colour on cases. I am a cabinet maker and white is one of the hardest colours to do, when It comes to painted kitchen cupboards. Any different light that hits the fronts makes it look like a totally different colour. We often have to take the fronts off and show them in the same light to convince the customer.
I have a question for Ask GN. I'm a fairly new viewer, so I'm not sure if this been covered on here before or not, but I wanted to ask why it seems that the latest trends in PC components have to be lights in everything. On top of that, it is exceedingly hard to find a tower that is closed on all sides so that you can hide all of the lights that have to be on everything. I recently built a PC primarily for gaming, but I have some other workloads as well. I have it in my bedroom, because I share a house with roommates. I realized that everything that I wanted to by for a gaming rig had RGB LEDs on it. Then having to get a case that was closed on all sides, so that I didn't have all kinds of light bleeding into my room, while I was trying to sleep. I realize I might be in a minority in this topic, but I'm curious on your input on this, and figured it would be a good topic for ask GN.
There is one physical reason that capped FPS might get you a higher average (and more importantly, more consistent frame timing). Your minimum FPS, or highest frame latency, occurs infrequently; typically there will be some specific scene at a particular angle that requires a lot more rendering work. If your frame rate is capped such that there is margin, your GPU will have time to cool off more in the usual case, which means there's a larger buffer (in the form of thermal mass) before it needs to throttle when it does need the extra power. Therefore you might get better performance when you do hit the rough patches. On the other hand, engine capping without display sync will yield some missed or skipped frames at a beat frequency. Typical fixed sync monitors are frequently slightly off (for instance, video is standardized at 60/1.001, and I've seen monitors that ask for nearly 62Hz). As for vsync, engines use digitally locked loops allowing the engine cap to be driven by the vsync cap, in particular seen in VR setups that estimate the free time to do some rendering as late as possible. In that case it's not quite as simple as just backpressuring the engine and the frame rate doesn't necessarily drive the latency up.
I think the main reason for improvement of fps lows is the fact that the simulation thread isnt working the whole time (as explained in the video). It should then ideally yield the processor (core) to other threats who can do their work - like the tasks the OS needs to do. If the games threads where using all of the CPU time, the os will yield them from time to time to do its own work. That will most likely result in "fps drops" as some game thread doesnt have his needed 16ms of cpu time
maybe something for ask GN : why does the windowed/windowed boarderless removes tearing? Maybe this is stupid but I have no tearing when I play games in the windowed mode and I have no idea why :-)
Are PCIE lanes with wattage limits stamped on them (for example optiplex small form factor intel motherboards) limited by the motherboards capacity to deliver power over the claimed maximum delivery, or by the ability of the PSU to deliver the stated power draw?
What's the Wattage Limit on the Motherboard PCIe Lanes? is it Less than 75 Watt? Then a good PSU with 8 pins PCIe or more will do more work to feed Wattage to the GPU
if your gpu has a boost feature and your maximum fps is close to your monitor's framerate without a frame limiter, enabling a frame limiter is likely to make your gpu boost higher as it will not use full load in most games and applications 24/7 (as long as it can hit your framerate) allowing your card to cool down and boost higher when it needs to to hit your frame limit.
Awesome episode Steve with all those details about FPS. I wonder if you could get into more details about the whole real-time pipeline and operation of the GPU together with the CPU and the game engine during gameplay for frame generation and displaying. Things moving and changing at their own pace in the game world simulation, while sort of snapshots at given times considered for frame generation, and some time later on that snapshot frame is finished and ready for the display, when the game is actually somewhere else in the "future" of the simulated world already, with respect to that frame, that is. Maybe there are some on-line videos explaining this in graphical detail already, but I have not seen them. So if you could point us to any of that, or actually make an instructive video yourself given your own knowledge and expert contacts about it, explaining all of that to some level of detail, that would be truly fascinating. What triggers what when, when is a frame ready, vs. when is a frame discarded because it wasn't finished in sufficient time, who decides what, and what parameters decide what, things like that. It would help everyone better understand where game engine optimizations apply, vs. where the CPU bottlenecks vs. GPU bottlenecks apply, whether memory size or bandwidth related, or whether cpu simulation vs. gpu parallel calculation intensity related. There is so much stuff happening so quickly for a single frame to appear on the monitor during gameplay in a gaming PC, it would be great to sort of slow down time a bit and better understand what and how it all happens simultaneously for each frame.
Not sure if in your wheelhouse , but what's your opinion or could you test the statements that in a water cooling loop the order of how the coolant flow doesn't matter (ie first the cpu then radiator then gpu). While I understand the argument that the liquid in the loop would (for example in idle conditions) reach thermal equilibrium, the fact of the matter is that computers load and thermal generation is not deterministic and I think you could to some extent (how much this is I have no idea) optimize the order of your loop instead of not caring at all. Any plans to test something like this? or maybe some insight on this?
Can thermal pads be stacked on each other? I have thick pads that barely anyone makes and hard to find. Would stacking two (or more) thinner ones result in a significantly degraded thermal transfer? Or is it "fine".
Question for upcoming episode: Thanks for the great content. Seeing as how you guys started to get into streaming benchmarks (which are very interesting, to say the least), I was wondering if you're willing to go more in-depth on the subject. Couple of interesting points : seeing as how you mostly recommend dual-box setup for those who don't want to lose any performance in gaming, it'd be interesting to see a comparison between using a capture card for said process vs using RTMP over LAN (with gaming machine encoding with NVENC for minimal performance loss at very high bitrate, and then streaming rig using x264 to actually output the stream). Also very curious how upcoming high-core counts CPUs (threadripper) will be able to handle that type of workload, since cost of two rigs can be pretty equal to cost of one threadripper rig. Last but not least - There's very little info on the web in general regarding what type of difference encoding preset makes, how much of and improvement something like "medium" vs "faster" is, and, on that subject, how different games can affect what preset you can select (I'd imagine streaming watchdogs2 that use all the threads is different to streaming something that utilizes only 4 or 6 threads). And the last question on the subject is whether it might make sense instead of changing encoding priority, try and change core-thread affinity of obs process vs game process, so that different CPU cores take care of different load. I imagine that might result in better experience overall. I know this is a long topic, thanks for reading this.
Either way a capture card will benefit you. Capturing from the buffers always hurt performance. It's best when possible to capture from one of the ports that's meant for output to a monitor. If you have enough CPU resources in the gaming rig, I'd try the capture card there before putting it in an entirely different rig.
That's down to the vagueries of market supply/demand for crypto currencies (well basically Ethereum atm), and the degree to which vendors try to stop prices crashing too much when the hordes of unwanted cards suddenly get dumped on the used market. It may already be on the verge of turning around, Ethereum has almost halved in value in the last two weeks, so who knows.
If crypto currencies cash prices will drop a lot. However all these cards where used to mine 24/7 so their lifespan might be quit low. Some might fail any moment
Hi Steve, this question is in response to recent video pitting the R7 1700 against the i7 7700K. I was wondering (maybe you can test this too) will disabling SMT make any difference in gaming and streaming performance of the 1700? I know it sounds counterintuitive but I have read reports of improved gaming performance of the 1700 with SMT disabled. Thanks.
One question... What is the purpose of parallax textures and how the hell do they work? I've been hearing a lot about it, but there's not a lot of info i could find....
Question for next Ask GN: I have an i3 6100 overclocked to 4.6GHz, will a Ryzen 5 1600 with a reasonable overclock be a noticeable single thread regression?
As for the capping FPS thingy... If I get screen tearing that capping doesn't seem to fix and I therefore have to enable Vsync to get rid of it - can I still enable FPS cap as well, to battle frame skips caused by Vsync?
About mining; I recently seen a video about AMD's Instinct cards without any video-out connectors, nor a fan as they are intended to be used in specialized rack cooling systems. It makes me wonder; if they can do that, why hadn't they modified their Instinct cards to use consumer watercooling components so the miners can have their cut-rate Vega's and leave the gamer-oriented cards alone?
Back in my old rFactor racing days someone said that NON Vsync would make their GFX card run hotter. On the flip side, driver input (steering, throttle and such) was calculated faster than 60 fps (if running faster than 60 fps). Input was updated at the frame rate. On the system I was using I had terrible screen tear with Vsync off.
My question is: Why does decreasing the size of a transistor on a dye (16nm-14nm) lower power consumption? Is it because since the size is small it takes less energy to "flick the switch"?
I know that 2 out of the 4 dyes on the Threadripper are active and the other 2 are inactive and used to balance/support the IHS.... But why not just remove the 2 inactive dyes, rotate the other 2 90* and put them side by side? I'm sure there's a reason, I just haven't heard one yet. Is it that the active dyes by themselves can't support the coolers? Side question... Will CPU coolers going forward for Threadripper be covering the whole IHS? I would think more contact is better... is there diminishing returns in heat conductivity and transfer depending on the amount of surface in contact with the cooler? Thanks!
quick question... How do parallax textures and mapping work? how can you make something look so high in polygon count without heavily taxing the graphics card?
I've noticed that you don't often use anything lower than "Ultra" when benchmarking, unless it's a relatively weak GPU. Is there any particular reason why you wouldn't include "Medium" settings as well aside from the extra work involved? I'm curious about this because there are many games don't have dramatic improvements in the visuals above "Medium" or "High", so many people, myself included, will prefer to sacrifice Ultra settings in favor of smoothing out minimum framerates and frametime spikes.
Is there currently a way, or possibly a future way to pool multiple systems to split a project between the pool of two, three, or four systems, and merge the pieces after each system rendered its part? Thank you for breaking down frame capping. Is fast sync only worth it after you're consistently over 2x the refresh rate or more like I've seen online, or would 80-100 fps be an improvement fast synced on a 60hz screen. It seems to me that frame pacing would suffer, but what does everyone think or have you experienced? Fantastic video!
@ASKGN I just watched your video on how heatpipes and air coolers work and I'm wondering how do the heat pipes work when the cooler is mounted sideways, or in the case of a video card, the heatpipes are horizontal. How does the convection work in that orientation?
Capping FPS has always done wonders for me in areas outside of the game performance - heat, power and in the case of some video cards (R9 Nano) massively decreased Coil Whine. When Catalyst came out with the feature - it was the happiest day of my choil whining life.
I can understand the money issue with case color. But an option I think would be to be able to get a primer color so you can paint it yourself or send it to a shop. Even caselabs is lacking these days on color choices. I personally like gunmetal.
Question for ask GN: I have recently updated my setup to a ryzen 7 1700 and a gigabyte gaming k3. I had a watercooling setup and the CPU was overclocked to 4ghz with 1.4v. All the temps were fine but in HWinfo64 i found out the VRM to be extremely high during stress test (120C+) and really high in gaming (90C+), especially in CPU intensive games like BF1. I had to roll back the overclock and put an air cooler. But temps are still bad for the VRM, even at stock clock, unless i put the wraith cooler. Is it a design flaw or all the 4 phase AM4 cards are the same ?
Nice work about the fps cap, but there is a huge missing point. Cooperation ! 90% of the time, a game that doesn't cap it's FPS, get lag not from the game itself but because it get interrupted by other programs (including OS), it was always the case on PC (that's why you need a bigger pc hardware than console hardware to get the same experience) but is now getting more and more true on this generation of console. A game is like a comic stand up, if you do not include time for people to laugh or clap, you might get interrupted by the audience (OS or other programs) at any time, and for longer for a random time, and could not speak as long as the audience decide to keep interrupting you, Strangely, if you write the most efficient program, using all core of you cpu, you can get interrupt many times for 100 or 700 ms, because at one point the OS want to get time for him or other program and when he has the hand he is not giving it up easily, On a single threaded program it's a problem, but on an multithreaded program when you have thousand of tasks with dependencies, if one of the core is given back for a small time to another process you have a cascade of delay, because other task on other thread could wait for the result of the late task to continue, and all work stop. So going the faster you can is not the best solution, you could get a higher max frame rate, but a lower min frame rate or average because of the huge interruptions. And this could also come from your hardware, you talked about it a little by speaking about power usage. Power usage = heat. If you make a CPU or GPU works 100% it will heat up, and then throttle it's frequency to cool down , leading to longer frame time, so possibly another reason from lags, making the overall experience less smooth. And in the worth case scenario it could lead to crashes or reboot, So you have to cooperate with other software and hardware as a game dev, and even like that you could be interrupted leading to lags, that was the idea for windows 10 game mode, allowing an application to get a more exclusive use of your hardware, but apparently they failed. PS, i'm a pro game dev, with multiple million sellers and 20+ years of experience and i still scratch my head to find the best compromise between the game and the OS.
I bought the Phanteks 400s even when your review nearly made me choose another case. I still pulled the trigger and got it. Then my son made my whole computer desk fall. The front panel which covers the fans protected my whole system by absorbing the impact. no damage to anything else then the front panel and of course the acrylic panel which got scrached. Now case build quality is my number one concern. Never thought my pc could drop 4 feet but stuff happens. GN4LIFE
does ambient air pressure i.e altitude affect cooling performance of heat sinks/radiators? I live at 6500 feet. so I wanted to know if altitude effects the temps of a heat sink in any significant way.
Small input on the 2nd question... If you limit the amount of frames the GPU is producing, not only will it lower the power usage like you mentioned, but it will also lower GPU temps and make throttling far less likely.
I have an i5 7500 with a B250-F MB. When would it be a "good" idea to consider upgrading to a K CPU and Z MB? I don't stream, render, and multitask (not heavily), so the criteria would only be for gaming at 144hz (if possible). Thanks!
Ever considered doing a comparison of thermal pads? The Lightning and EVGA 1080 Ti models you did tear-downs of have memory thermal sensors they seem like ideal candidates. Honestly, I cannot find any in depth comparisons of high-end thermal pads with conductivity ranging from 6 to 17 w/m-k.
Irfan Ahmed It's all about price to performance ratio. Copper will cool better, but aluminum is cheaper. Whichever makes more sense for you would be the better option, I suppose.
@Gamers Nexus (2nd attempt) Why is USB 3.1 Type C still so uncommon on (modern) computer motherboards? Most motherboards of $120+ still tend to have only 0-1 type C port. I thought Type C was supposed to be the new standard that is going to replace stuff like type A & Micro B. I see alot more produts (like enclosures) that support Type C, but not a lot of motherboards (and no cases with front panels) that support type C. Thanks!
no i understand that to a degree, but if the consumer products like externals, phones, and yada yada are switching over to it, why is the MB manufactures not stepping up to the plate and including more than 0-1?
Question for Ask GN: Steve, I am curious to know if an Overclocked I7 7700k, paired with some fast memory could potentially bottle neck a pair of 1080tis at 1440p. It would be interesting to know if the higher frequency of the 7700k could compete against the higher PCI-e lane count of an x series chipset when using a dual GPU configuration. Thanks!
As far as coloured cases go, i remember the BitFenix Prodigy. I had the orange one and it was probably the best ITX case I've ever used, athough it was quite big.
Why, oh why did AMD decide to use a 20 degree offset on the 1700X and 1800X cpus? They could have just asked people to set an aggressive fan curve instead. I see the confusion STILL lingers all these months later.
Its not just about render performance. The reason I and a lot of people stopped used Premiere, is its stability it crashes for dumb reasons way too often. I swapped over to DaVinci (some of my industry friends swallowed the bullet and bought Avid for their home use as thats what we were tough in collage) and have never been happier. and its super easy to adjust to.
Ya if it doesn't give you any issues why change? why I stressed stability and barely said anything about performance as thats is just a hardware fix. probably could have worded it better though...
Question: is there a difference playing a game at 60fps on a 60Hz monitor vs 60fps on a 144Hz monitor. Assuming both are displaying 60fps the 144Hz monitor should still be smoother. I guess my point is that do you really need over 100fps to truly take advantage of a high refresh rate monitor like everyone seems to be saying?
I have a question; since my Corsair H100i GTX CPU cooler died recently, should I NOT buy hybrid GPU cards for the same reason? I have the MSI 980Ti Seahawk. And I'm thinking of getting rid of it and buying a 1070 or 1080ti because im worried my card might fail just like my cpu cooler. I game at 1080p and the 980Ti does every game maxed out minimum 60fps from my experience.
I have a question for the next ask GN if it's not too late: Do AIOs Perform aswell as a custom liquid cooling loop, and if they don't why don't they? As far as I can see the only difference between an AIO and a Custom Loop would be the tubing and having the CPU Block seperate from the Pump (and the reservoir if there is one)
Now that you have done a few streams, Do you find live streaming "easier" then your regular content? Does it take a load off of your production pipeline?
Is there something like "cpu only rendered frames"? I know ot sound ridicolous. I ask since some games I play look like they'd scale with cpus indefinatelly. I have 9800 GTX+, and while on some test rigs in f.ex. cs:go ppl get avg 70 on low(on budget cpu), with my 8320 fx I get to the 100fps+ territory.
Just did a new build with a 7700k, gtx 1080 ftw2, MSI z270 m7, 32 gb Corsair vengeance lpx. Premiere is not running as smooth as I had expected. I get significant lag when trying to open the Lumetri color wheels. Just switching from Waveform to the Histogram takes probably 1-2 seconds. It's significantly slower than my 3+ year old laptop. Any idea what is going on??
On the software temp monitoring question. I have the 1600x and HWmonitor along with Ryzen Master both show different temps. Usually around 80C (stock speed) in HWmonitor after just browsing the internet and watching hippoporn uh i mean youtube. Idle temps are around 50c. I have swapped out CPU water blocks, (EK EVO and Swiftech) and motherboards. No difference. I even RMA'd the 1600x with amazon. Still no change. My 1700(@4ghz) and 1500x(4.2ghz) seem fine with both hitting highs around 65c under full load. As i have swapped them out as well. My videocard (GTX 1070) is in the same water loop and it;s max temp is the same with all three CPUs, about 40c tops. Using my Fluke K-type thermocouple and non-contact fluke temp gun along with just feeling with my fingers everything feels just slightly warm or close to room temps. I am going to try the HW64 you mentioned and see how it goes.
I recently bought an oculus Rift and they recommend having 1 Sensor on per USB 3.0 Controller. I was wondering if each USB 3.0 Port on the Rear I/O of a Motherboard would have a separate controller, or if some of them were shared. Looking to do a 4 sensor Room Scale Setup. I have 4 Rear USB 3.0 Ports, and 1 USB 3.1. Trying to find out if i need an add in card or if i would be ok with using the Mobo ports. The Motherboard i am using is the Asus Maximus IX Formula.
Is it possible for RX Vega to use the Infinity Fabric to get better Video Card to CPU coms ( they use IF for CPU to CPU coms in Epyc)? If so, could this be a reason that we have not really seen any testing results from AMD on Ryzen?
Hi GN /Steve, I have two questions. 1, Why after all this time and iteration on multi GPU systems do we still need software drivers for SLI and Crossfire? Why is it not possible to dispense with game developer effort to program for mgpu when we could have a hardware solution within the cards or even the chipsets themselves presenting two card as one for programming purposes. Even if we needed a chip on the card! My point being, it is crazy that some games scale and other don't, this burden should be taken away from the game makers and be done in hardware. Why have we not achieved 70-90% scaling on everything? What happened to AFR? 2, I forgot, had something to do with VESA not borking Gsync by making fast sync better.
You could in theory gain FPS by capping FPS due to the Chip staying cooler so when it gets to more intensive area the CPU or GPU could be able to get full turboboost. This only applies if you are thermal throttling and in rare circumstances. Thought I would just mention this hehe. I actually overclock my turboboost but not stock clocks on my Chip as I am only really bottlenecked hard by CPU in single threaded unoptimized games.
"Adobe is not that well optimized for..." might be a bit of an understatement. ;) Not to mention how often it crashes. The sad thing is that Adobe doesn't seem to care, how difficult can it be for that company to hire some good programmers who can fix all those issues? I don't say that it is easy for those programmers but there certainly are enough programmers available who could fix this if Adobe would let them.
I am thinking of buying dell 7567 I5 , 8 gb ram, 256 gb ssd m.2 1050ti for 900$ is there a better laptop you can suggest for gaming in the same price bracket?
question for you Steve! since 60fps are low, tell us about monitor overclocking, how its done with modern gpus, is it dangerous? why no one does this in the RUclips community (jay, ltt, paul etc)
Hello. What happens if you pair 2800 mhz RAM with a processor that supports slower memory, like the Pentium G4560 which can support upto 2400 mhz RAM? I know you can't overclock the processor, but I am pairing it with a Z270 motherboard I had lying around.
Does Ryzen ram speed above 2400mhz @1440p gaming really matter for the $50-$100 price premium? Building a system soon and would like to know if higher speed 32gb (2x16Gb) kits shipped from the states are actually worth it...
Another good video GN, thanks. Will say though. HWmonitor for Ryzen is still broken. I still have my temps reported in the 80's when I know they are in the 60's. It's a bug on all the X models
Not a bug. That is because AMD still does their TCTL offset silliness, which is only present on X models. Some software accounts for this better than others. HWINFO seems a bit easier to work with.
Hi, I was wondering if you could address what might be the reasons that a Ryzen CPU with an identical core count to an Intel CPU cant overclock as high? Thanks
Eesh. Looks like RUclips might still be processing 1080p. It was there for a moment!
One thing I realized for Scooby Dooby's comment: GTA V specifically has issues with i5 CPUs that hit the 187FPS limiter in the game, which is something we've previously discussed. In this very peculiar scenario, you actually would greatly benefit from a framerate limit.
Still looking to get the info. Would rather have the info than a better bitrate, haha!
Hey steve, what happenned to the loop you built on stream, any leaks?
No leaks. The loop is perfect. Looking forward to testing it once we're back from this next trip! Currently busy with other things.
Steve, Volta or Vega?
I would love to see the industry adopt a V-sync standard that aims to reduce latency by only allowing frames to render in the last 8ms of a 16ms frame time.
Or even a 4ms target, or better yet allow the player to customize the frame time allotment.
CS:GO type players would probably not bother because they'll take frames at any level of completion, but for people wanting to run on a 60hz monitor with low latency and without any tearing (tearing being much more detrimental at 60hz), it would be ideal to be able to just set your frame yourself.
The implications on power consumption could be significant as well.
Theoretically if you give a 30hz system the same frame time as a 120hz system, it would feel much better for the player and it would consume 1/4 as much power.
Nintendo could really use something like that, but if we got the option on PC then it would immediately be an option for Laptops, and more importantly the freshly re-emerging UMPC market with things like the GPD Win.
"You only have two resources, time and money. Only one of those you can get more of." This quote is actually amazing.
AMD fans seem to have more time than others.
PS. The wait for Vega is killing me. I need to complete my AMD ecosystem.
Life lessons by Steve Burke, only on GN.
but it is invalidated by the formula that time=money
What the fuck are you doing reading my username? That conversion only goes one way.
one can always philosophise
i for one take the "in some but not all aspects" approach
Steve is correct when it comes to the video editing software. I use Adobe Premiere Pro as well and the time it would take me to learn something else is worth more than a new computer would cost. I could easily spend 100+ hours learning a new program to the level I know Premiere Pro today, even if my time is only worth $10/hr, that is $1,000 worth of time.
Tech Deals Great seeing you here bud.
It's unfortunate that Blender is rather difficult to learn. What happened though to Adobe's promise last year that they would focus on hw performance and optimisations instead of new features? They said the move was based on customer feedback.
Sadly true, and it makes it all the more difficult for those with older versions who don't like radical changes like dropping RT3D and switching to C4D, especially when they've built up a system optimised for the former; doubly so when Adobe updated its libs for Premiere to support Maxwell CUDA V2, but didn't do that for AE. And then there was the switch to CC and the subscription model. This all hurt a lot of solo pro users.
He does exist!
Tech Deals eh.. if you are a good editor I would think that you could adapt to another video editor fairly easy. Your still doing the same things. It's just a matter of learning keyboard shortcuts and navigating the program. But it will still have a timeline and preview window.
For some games I don't mind capping my FPS, as it allows me to get a more consistent experience, at the cost of a few frames.
Budget-Builds Official, I cap em all!
Budget-Builds Official I agree. I'd rather have the solid frames. I only cap 30 for games like Mafia III that have too much jumps from 60 to 38.
Often, sim racers like to FPS cap at a multiple of the monitor refresh rate. Usually 2x but I run 3x on Iracing which is very lightweight. 180fps on (3) 60mhz monitors. It's more consistent, as Op states, and helps avoid tearing without the input lag caused by V-sync.
Switching from 120 cap to 180 cap in Iracing I could feel/see the difference. The information being fed to you is that little bit more up to date, while your input lag is minimized. Sim racers and first person shooters crave frames like an addict fiending for crack. But, capping is still good because consistency > max frames.
overwatch is such a great example, even though the display based option is kinda broken (usually caps fps about 10% higher than refresh rate), but the 30 (if youre on a toaster) or the specific number one is great. I run the game pegged at 144 fps, never goes up or down, smoothest 6.9ms I get to date
Got it set to 153 fps. My CPU limits me around 150 anyways.
This was explained by a Futuremark engineer: If the GPU is fully loaded, everything's generally fine in terms of frametimes. However, if your CPU gets fully loaded by the game engine, frametimes can go haywire because the GPU driver isn't getting enough CPU time to do it's work. In such a case, it's useful to limit the FPS so that the CPU can have some breathing room and the GPU driver isn't starved for CPU time.
Diwwah Better answer than the video one.
Marcelo Tezza That answer is rather simplistic and doesn't actually answer the question of what happens when capping framerate. It talks about performance behavior in general, but you learn nothing about what goes on under the hood of the components. They are addressing the subject in a different way. Our answer looks at it from a canonical pipeline perspective and gives more detail on the backend of the render pipe.
It's fairly common in demanding AAA games for at least one thread of a CPU to be under load at least some of the time. Action games, unless they are a benchmark, are highly dynamic. You can go from low CPU loads to high CPU loads very quickly, depending on what the player is doing A framerate limit or vertical sync can help smooth out that experience. I've played some games (not action games, but demanding adventure games like Detroit) at 30 frames per second just to get a smooth video-like quality.
that progress bar on subjects is genius. whoever thought of that needs to get a medal.
I love it. It's really great, even on screen when they show graphs for benchmarks they'll have a bar that "counts down" on the very edge of the left and right sides of the screen.
At first glance it looks like just graphics/window dressing but it's actually the same thing only on the borders of the screen.
Helps so much in being able to find a piece of info again that your looking for for those that are learning this stuff and we need to go back and replay a certain point.
I can absolutely see this feature becoming a new norm.
Figured I'd just give a shout and say thanks Steve for getting an answer to my question :) . Don't know how often you actually get a thank you, but figured you deserve one. And thanks for taking my question by the way. Didn't think it'd get picked since it was pretty late into the week when I asked it. I figured it just came down to cost when I asked, but you mentioned a few things like in-store retailers that I didn't think about, so thanks again for the answer and insight. :)
I came here specifically for the second question regarding frame rate limiters and if they just accept the first x amount of fps or if they actually spread the frames throughout each second, and I have to say I couldn't be any happier with your answer. I plan to watch the rest of the video when I have time simply because of how thorough and clear you were with that response - I feel I can learn so much from these videos. Subscribed immediately. Wish I came across this channel sooner!
I always cap my games FPS via Riva Tuner Statistics Tuner. Lower temps, no screen tearing, no input delay, and a lower GPU usage.
DJHeroMasta there is always input lag when capping your frame rate because of back pressure. no matter what cap method you use. otherwise freesync and gsync wouldn't exist. you only get stable frame pacing, but it won't fix input delay
Robin Clarijs, No. Freesync and G-Sync exist for more than just that.
RTSS gives the best frame pacing, actually better than most in-game options. Increases CPU usage though.
Marcus Mathiassen got a link to that? From what I seen when it comes to comparing in game cap and RTSS cap is RTSS is worse for input lag.
DJHeroMasta you get tearing if you not using vsync or g/freesync. You not seeing it don't mean it's not there.
Got no link, but from my own experience RTSS seems to cache frames which increases input lag and burn cycles which increases cpu usage to get that perfect frametime. I don't know for sure though.
Question for ask GN:
I have recently gotten very into video editing, but I feel like my PC (6700k, GTX 1070, 32gb ram) is being severely bottlenecked when editing 4k footage due to my lack of an SSD. I am currently reluctant to purchase an SSD for editing, for the fact that it would be constantly reading/writing at 100+ MB/s, and I would hate to have the SSD die after just a few months due to me reaching the maximum TBW of the drive.
Is this a legitimate concern for people like yourself, who are constantly editing high bitrate video, or should I not be concerned about this at all. And if this is a legitimate concern, would a better solution be to edit off of a pair of hard drives in raid, in order to increase the longevity of the solution?
Thanks for reading my question, and keep up the amazing work :)
jonny j Thank you for your response, I'll definitely check out that article.
I still don't know if uninformed necessarily equals 'fucking stupid' though :)
+Harrison Glenn First off, never base a decision on feelings. ;) Use system monitoring tools to check what is happening during your workflow (Afterburner, GPU-Z, NV tools, Process Explorer for I/O, etc.) Are you sure it's not a CPU bottleneck? Also, for apps like Premiere, AE and Vegas it's essential to use an SSD as a cache even if the main working store is not an SSD (though it really should be, especially for 4K), while for obscure technical reasons Vegas should use one for its main working area aswell anyway. If you're working with 4K though, then yes it's very likely I/O is holding you back, but check t be sure.
As for reaching TBW, don't worry about it. SSDs can usually handle a lot more than their official rating, just look at how much better OCZ's Arc 100s did when they were hammered until failure, all samples went waaay beyond their official limits:
www.kitguru.net/components/ssd-drives/zardon/ocz-challenge-kitguru-to-kill-5x-arc-100-ssd-drives/
though I would recommend fitting an M.2 NVMe SSD instead of a SATA in order to get much higher speed, and this will avoid the potential pitfall of some SATA SSDs which have rather low steady state performance (ie. when cache buffers get filled up, etc.) Older SATA SSDs were actually better for this; newer models are kinda meh with the adoption of DRAMless designs and TLC. Kinda ridiculous that my old Vector holds up so well to newer models, but there you go, atm the market is in a push to margins, not quality.
One thing though, lesser models in any product range do have lower TBW ratings, so if you can do go for something better anyway, and of course the higher capacities have higher ratings so 512GB is the minimum I'd use for 4K. I recommend the SM951, SM961, 950 Pro or 960 Pro. Note if you lack a spare M.2 PCIe slot, just fit the SSD in a PCIe addin card (I use an Akasa AK-PCCM2P-01, good space for cooling), and a simple heatsink can prevent throttling. Even using a PCIe 2.0 slot, I get 2GB/sec with an SM951 (P67 ASUS M4E, 5GHz i7 2700K)), while a 960 Pro on a 3.0 slot gives 3GB/sec (X79 ASUS R4E, stock i7 4820K), so one doesn't have to be using a newer chipset to get good performance from NVMe SSDs.
If you're going to use RAID at all, that's probably best focused on RAID1 for rust spinners to hold all your final data, and don't use lesser consumer drives, get something decent. Personally though I just duplicate data manually, eg. my main photo/video archive is on an SM961 512GB NVMe (the ASUS M4E config mentioned above), and then it's regularly copied/updated to a hot-swap 850 EVO 500GB SATA. 8)
Ian.
mapesdhs Thanks for your detailed response. I have been monitoring usage in task manager while editing, and in most cases, after effects lags when disk usage is at 100%, with CPU rarely going over 50%. Even when rendering, this is the case, which boggles my mind. In one case I was attempting to render a particle effect that I created with multiple plugins, and my CPU was under 10 % for all but the first 5 mins of rendering.
I'll definitely look more into nvme ssds before I make my final decision.
Thanks again for your insight
An nvme ssd (or two if you wanna do 2 in raid on a pcie card since z170/270 m.2 slots are all chipset based) would be your friend, as mapesdhs said, ssd's go well over their write ratings. Since this sounds like a semi professional system, a raid 5 array consisting of 4 2tb drives would make for some nice storage for your data (and ssd backup). If your ssd is backed up (or its a dedicated scratch disk and your os is on a sata drive somewhere) then you dont need to worry about endurance as if the thing starts to fail (which means you will lose capacity slowly) then you will have more than enough time to replace it, and by then the drive will be outdated and m.2 drives on pcie gen 4 will be mainstream. just get yourself a 960 pro of your desired size and call it a day, as the quality of that nand and memory controller is insane, not to mention the speed
+Harrison Glenn, Ah ha, you're using AE! Heh, that app is a right kettle of fish when it comes to performance.
First thing to remember is that various plugins for AE were never updated to be multithreaded.Thus, although Task Manager might show an overall low usage, one particular thread may be holding up the overall render process. Use Process Explorer to monitor individual threads to see what they're doing. Friend of mine discovered this when setting up some AE tests for me, and speaking of which if you want to bench your setup then try our CUDA AE benchmark (still writing up a page for it):
www.sgidepot.co.uk/misc/cuda.101.zip
The default test is just one frame, probably take 10 or 15 mins on your system. Note that rendering the entire animation is for stability testing purposes only, not benching (ie. checking render rig reliability for overnight crunches). The single frame test output looks like this (change the suffix to tga if you prefer the raw Targa):
www.sgidepot.co.uk/misc/cuda.101_Frame96.jpg
Anyway, here are the most important things about AE (and note you can find much longer and more detailed versions of all this in my posts on Creativecow):
- AE is a RAM gobbler. You already have 32GB which is ok, but double that if you can. Also (I'm assuming here your C-drive is already an SSD), it's a good idea to fit a separate SATA SSD to hold the Windows paging file. A 128GB is suitable (something with high IOPS, like an 840 Pro or Vertex4/Vector); the partition should be 1.5X the fitted RAM, so for a system with 32GB RAM the dedicated partition should be 48GB (or to be exact if one is fussy, and I am, 48GB + 6MB, ie. 49158MB, because there's a tiny overhead involved. :D) I normally define this partition to be drive V and call it VMEM. You can setup a Registry entry to prevent Windows moaning about the partition always being full, and another to prevent it being displayed in the Computer devices pane. Not only does this free up a lot of space on the C-drive, it will extend the life of the C-drive aswell. The unused space on the extra SSD can be defined as just a misc scratch area for whatever purpose (I normally call it drive S, named, "SCRATCH"), eg. temporary copies of files, dumping ground for test images, etc. Or just leave the unused space alone.
When you have the extra SSD fitted and the logical partition set 49158MB, right-click on Computer, select "Advanced system settings", click "Settings" in the Performance area, select "Advanced", click "Change", untick the Automatic box, then manually define the paging file to be on the V partition and set it to 49152MB. Click the various OKs to confirm and let the system reboot. Once done, you should see the space used on the C-drive has gone down by quite a lot. Speaking of space saving, if you do not use the system Hibernate (sleep) function then you can turn off the hiber file by bringing up a Command Prompt with Admin rights (find the menu option in Accessories from the Start menu, right-click on Command Prompt and select Run As Administrator) and entering this:
powercfg -h off
Exit the Command Prompt, refresh the devices in the Computer pane and again you'll see the space used on C go down. The above two measures saves a lot of space, especially for a system with 64GB+ RAM (without these measures, using a 256GB SSD for the C-drive does not leave much space for installed apps).
- A separate SSD for the different AE cache structures is essential. Doesn't have to be a PCIe, anything will help. If budget is a problem then look for something 2nd-hand, eg. a used Samsung 840 Pro 256GB would be ideal. In other words, whether or not you decide to use an SSD of any kind for the main working store (ie. where the current project data is located, and where rendered output is going to), having an SSD for the AE caches is a no-brainer. Adobe themselves state this on their FAQ page. Btw though, if you have the spare PCIe slots, picking up a used SM951 doesn't cost that much, last one I bought was only 80 UKP for a 256GB (and of all things on receipt it turned out to be new, not used). If you hunt for something used a la SATA, other good models would be the OCZ Vertex4/Vector, Corsair Neutron GTX, Samsung 840/850 Pro and SanDisk Extreme Pro.
- Write caching on your disk may be affecting performance, and btw it does of course make sense to be using a different device to the C-drive. Beware though, changing write caching settings can alter whether data may be lost during a power failure. If you switch to an SSD though then there's generally no need to mess with this setting.
- RAM speed helps for AE rendering. This is probably not yet an issue for your system as atm you most likely have an I/O bottleneck, but assuming that gets solved in time, do think about maximising the RAM speed if you can. On an X79 system with a 4.8GHz 3930K, my friend observed a 10% render speed reduction when dropping the RAM from 2133MHz to 1866MHz (the system I built for him has 3x GTX 580 3GB, 4.7GHz 3930K, 64GB RAM, etc.)
So, first off, get an SSD asap for the AE cache setup, whether it's a SATA or not; I keep hunting for used 840 Pros, but anything good will do (ie. aim for models with high random IOPS ratings). Then later when you can, swtich to an SSD for the main store. Also, have a separate SSD for the Windows paging file.
What I do when building such systems for people is have an SSD for the current data being worked on, and then when the day's work is done the results can be copied over to a rust spinner (I've been using 2TB Seagate ES.3 Enterprise disks for a while, getting them new for good prices, bagged another last week for 75 UKP). I normally fit two rust spinners so final data can also be copied as a backup (one can of course use RAID1, but I prefer to do it manually). Lastly, in addition to regular cloning of the C-drive, I fit a lower capacity HDD (usually 1TB atm) to hold general backups, including image dumps of the C-drive, backups of the user folder, etc.
If money is tight, here's an example of what is possible re exploiting the used market (a project I'm working on atm):
www.sgidepot.co.uk/misc/charitypc1.html
One other thing, reading James' reply, note that RAIDing NVMe SSDs won't really gain you very much, and it will hinder random I/O performance a tad. Overall, not worth doing IMO. RAIDing rust spinners can be beneficial though, but I usually stick to RAID0 or RAID10 for a working space (the latter gives both a speedup and some redundancy). RAID5 is a bit more complicated. Mind you, when I was still meddling with rust spinner RAIDs I was using U320 SCSI and later SAS, so the sequential performance was better anyway because of the HBA's abilities, especially a battery-backed HP PCIe card I had way back which had a 1GB DRAM buffer, so it could do 2GB/sec a lot of the time. I don't use such HBAs anymore though, PCIe NVMe makes much more sense. Btw, that's why mbds like the P9X79-E WS are so good, ooooodles of PCIe slots. :D
Good luck!!
Ian.
PS. This is my CUDA research box (ASUS P9X79 WS, 4.7GHz 3930K, 64GB @ 2133, 4x GTX 580 3GB 900MHz, various SSDs):
www.sgidepot.co.uk/misc/3930K_quad580_13.jpg
For editing though I have an ASUS R5E, 6850K, 950 Pro, etc.
Some comments on case colours:
* I totally agree with your point about more colour variants result in higher production costs.
* There's a simple work-around for mismatch in colours between plastic and metal parts: Don't try to do them the same colour in the first place! There are many computer cases doing this already.
* As a user I've had no problems buying a plain case and can of paint to get the case into my desired colour. Just spray the top and side panels any colour you fancy!
* When I was looking for a case to use for my daughter's first computer I initially wanted to go as cheap as possible, but ended up spending twice as much just to get a case that is pink! (It was a better option in other aspects as well, but the colour was what made me look at it in the first place.)
I love this guy. He explains everything so in depth but also simple enough for me to understand. I wish there was a way I could help him expand his viewer base. I have already told everyone I know to watch.
Hi, got another fps related question. Can someone explain why top CPU's, like the i7 7700k or the r7 1800x, draw different fps? I mean- no cpu is at it's limits or am I wrong? So when no cpu core is at a 100% - why did they draw more or less fps? (Without gpu bottlenecking in 1080p for example) Thanks in advance
Christopher Zimmermann Awesome question, I've wondered about the same
+1
Frequency is king in gaming
coram yes I know, but if the cores aren't at a 100% load- how can frequencies matter in that case?
50% of 5ghz is better than 50% of 4ghz.
i always cap my framrate if im getting 100's. it makes the gpu and cpu run at a lower load generating less heat and sound. generaly i cap t around 80-90 or i just put Vsync on.
Same a ryzen 5900x is a nuclear reactor so I do cap fps if it doesn't matter a whole lot.
@@tilburg8683 I have a 5600x now and cap games at 72 but leave pvp shooters and racing sims at 144 fps or 120.
7-years ago. Love the channel. Always wondered what happened when I cap FPS at the game level.
Back in the 286 through 486 days, our software shop found it worthwhile to have one bleeding edge machine to submit full compiles remotely as a batch job. We'd reinvented the mainframe! Coding, module compiling, & syntax checking would be done locally and package compiling & linking would be done on the 'fast' machine.
I think something relevant to the big discussion on frame rates in this video is that Steve touches on what is knows as frame rate independence - this is the concept of performing actions and animations based on the time between frame switches. Before this was introduced animations etc in games were done entirely on a frame to frame basis, problem being that eventually hardware differences in computers allowed some systems to run the game faster than their slower counterparts.
Gamers Nexus:
-Really good lighting
-Unassuming backgrounds
-Great camera quality/skills
-Rare video topics
-Professional, consistent assessments
-My subscription.
Question for Ask GN:
Something that has been grinding my gears a lot recently is a lot of large tech based channels have moved to using clickbait exclusively for video names and thumbnails. One that quickly pops to mind the (no longer named this way) video named "Ryzen R3 and R5 TESTED" by LinusTechTips with a more fitting name of "AMD Ryzen Emulated On R7". Do you guys plan to switch in the future to get more views, and thus more ad revenue, or stay with the current format and selection of videos?
i'm relatively new to this channel but i just noticed the time markers in the top left that count down for each subject. That is fucking amazing and a feature i never knew i wanted.
dude, the production quality of your videos is amazing.
Great section on case color! For those that want to change the whole color of the case, there is always paint. It comes in many forms from bottles to spray cans, if the end user wants creativity, then have some. Try not to rely on someone else to do all the work for you. If you have no skills and no ambition, then pay someone else. If you have no money, then ask someone if you can borrow some duct tape and use it on yourself. (be creative)
Regarding colour on cases. I am a cabinet maker and white is one of the hardest colours to do, when It comes to painted kitchen cupboards.
Any different light that hits the fronts makes it look like a totally different colour. We often have to take the fronts off and show them in the same light to convince the customer.
Really Liking the content and the Q&A series of videos. Thanks for making them.
Really like that you show a rundown with times. Nice job.
I have a question for Ask GN.
I'm a fairly new viewer, so I'm not sure if this been covered on here before or not, but I wanted to ask why it seems that the latest trends in PC components have to be lights in everything. On top of that, it is exceedingly hard to find a tower that is closed on all sides so that you can hide all of the lights that have to be on everything.
I recently built a PC primarily for gaming, but I have some other workloads as well. I have it in my bedroom, because I share a house with roommates. I realized that everything that I wanted to by for a gaming rig had RGB LEDs on it. Then having to get a case that was closed on all sides, so that I didn't have all kinds of light bleeding into my room, while I was trying to sleep.
I realize I might be in a minority in this topic, but I'm curious on your input on this, and figured it would be a good topic for ask GN.
There is one physical reason that capped FPS might get you a higher average (and more importantly, more consistent frame timing). Your minimum FPS, or highest frame latency, occurs infrequently; typically there will be some specific scene at a particular angle that requires a lot more rendering work. If your frame rate is capped such that there is margin, your GPU will have time to cool off more in the usual case, which means there's a larger buffer (in the form of thermal mass) before it needs to throttle when it does need the extra power. Therefore you might get better performance when you do hit the rough patches.
On the other hand, engine capping without display sync will yield some missed or skipped frames at a beat frequency. Typical fixed sync monitors are frequently slightly off (for instance, video is standardized at 60/1.001, and I've seen monitors that ask for nearly 62Hz). As for vsync, engines use digitally locked loops allowing the engine cap to be driven by the vsync cap, in particular seen in VR setups that estimate the free time to do some rendering as late as possible. In that case it's not quite as simple as just backpressuring the engine and the frame rate doesn't necessarily drive the latency up.
I think the main reason for improvement of fps lows is the fact that the simulation thread isnt working the whole time (as explained in the video). It should then ideally yield the processor (core) to other threats who can do their work - like the tasks the OS needs to do. If the games threads where using all of the CPU time, the os will yield them from time to time to do its own work. That will most likely result in "fps drops" as some game thread doesnt have his needed 16ms of cpu time
maybe something for ask GN : why does the windowed/windowed boarderless removes tearing? Maybe this is stupid but I have no tearing when I play games in the windowed mode and I have no idea why :-)
Are PCIE lanes with wattage limits stamped on them (for example optiplex small form factor intel motherboards) limited by the motherboards capacity to deliver power over the claimed maximum delivery, or by the ability of the PSU to deliver the stated power draw?
What's the Wattage Limit on the Motherboard PCIe Lanes? is it Less than 75 Watt?
Then a good PSU with 8 pins PCIe or more will do more work to feed Wattage to the GPU
if your gpu has a boost feature and your maximum fps is close to your monitor's framerate without a frame limiter, enabling a frame limiter is likely to make your gpu boost higher as it will not use full load in most games and applications 24/7 (as long as it can hit your framerate) allowing your card to cool down and boost higher when it needs to to hit your frame limit.
Awesome episode Steve with all those details about FPS. I wonder if you could get into more details about the whole real-time pipeline and operation of the GPU together with the CPU and the game engine during gameplay for frame generation and displaying. Things moving and changing at their own pace in the game world simulation, while sort of snapshots at given times considered for frame generation, and some time later on that snapshot frame is finished and ready for the display, when the game is actually somewhere else in the "future" of the simulated world already, with respect to that frame, that is. Maybe there are some on-line videos explaining this in graphical detail already, but I have not seen them. So if you could point us to any of that, or actually make an instructive video yourself given your own knowledge and expert contacts about it, explaining all of that to some level of detail, that would be truly fascinating. What triggers what when, when is a frame ready, vs. when is a frame discarded because it wasn't finished in sufficient time, who decides what, and what parameters decide what, things like that. It would help everyone better understand where game engine optimizations apply, vs. where the CPU bottlenecks vs. GPU bottlenecks apply, whether memory size or bandwidth related, or whether cpu simulation vs. gpu parallel calculation intensity related. There is so much stuff happening so quickly for a single frame to appear on the monitor during gameplay in a gaming PC, it would be great to sort of slow down time a bit and better understand what and how it all happens simultaneously for each frame.
Not sure if in your wheelhouse , but what's your opinion or could you test the statements that in a water cooling loop the order of how the coolant flow doesn't matter (ie first the cpu then radiator then gpu). While I understand the argument that the liquid in the loop would (for example in idle conditions) reach thermal equilibrium, the fact of the matter is that computers load and thermal generation is not deterministic and I think you could to some extent (how much this is I have no idea) optimize the order of your loop instead of not caring at all. Any plans to test something like this? or maybe some insight on this?
Steve casually giving out economics lessons in a tech QA. Love it.
I think this is the first video I considered slowing it down for ease of following it and not having to pause all the time to think. :D
Can thermal pads be stacked on each other? I have thick pads that barely anyone makes and hard to find. Would stacking two (or more) thinner ones result in a significantly degraded thermal transfer? Or is it "fine".
Love this channel. Just good honest information. And no click bait
Question for upcoming episode: Thanks for the great content. Seeing as how you guys started to get into streaming benchmarks (which are very interesting, to say the least), I was wondering if you're willing to go more in-depth on the subject. Couple of interesting points : seeing as how you mostly recommend dual-box setup for those who don't want to lose any performance in gaming, it'd be interesting to see a comparison between using a capture card for said process vs using RTMP over LAN (with gaming machine encoding with NVENC for minimal performance loss at very high bitrate, and then streaming rig using x264 to actually output the stream). Also very curious how upcoming high-core counts CPUs (threadripper) will be able to handle that type of workload, since cost of two rigs can be pretty equal to cost of one threadripper rig. Last but not least - There's very little info on the web in general regarding what type of difference encoding preset makes, how much of and improvement something like "medium" vs "faster" is, and, on that subject, how different games can affect what preset you can select (I'd imagine streaming watchdogs2 that use all the threads is different to streaming something that utilizes only 4 or 6 threads). And the last question on the subject is whether it might make sense instead of changing encoding priority, try and change core-thread affinity of obs process vs game process, so that different CPU cores take care of different load. I imagine that might result in better experience overall.
I know this is a long topic, thanks for reading this.
Either way a capture card will benefit you. Capturing from the buffers always hurt performance. It's best when possible to capture from one of the ports that's meant for output to a monitor. If you have enough CPU resources in the gaming rig, I'd try the capture card there before putting it in an entirely different rig.
Question. When do you think GPU prices will return to "normal"? I found that now even the GTX 1050s shot up 20-30 dollars more
That's down to the vagueries of market supply/demand for crypto currencies (well basically Ethereum atm), and the degree to which vendors try to stop prices crashing too much when the hordes of unwanted cards suddenly get dumped on the used market. It may already be on the verge of turning around, Ethereum has almost halved in value in the last two weeks, so who knows.
If crypto currencies cash prices will drop a lot. However all these cards where used to mine 24/7 so their lifespan might be quit low. Some might fail any moment
Hi Steve, this question is in response to recent video pitting the R7 1700 against the i7 7700K.
I was wondering (maybe you can test this too) will disabling SMT make any difference in gaming and streaming performance of the 1700? I know it sounds counterintuitive but I have read reports of improved gaming performance of the 1700 with SMT disabled. Thanks.
One question...
What is the purpose of parallax textures and how the hell do they work?
I've been hearing a lot about it, but there's not a lot of info i could find....
Hey GN, why does flash perform better when warm? more resistance>less current during erase cycles?
Thanks a lot for such detailed responses. Appreciate your work.
Question for next Ask GN: I have an i3 6100 overclocked to 4.6GHz, will a Ryzen 5 1600 with a reasonable overclock be a noticeable single thread regression?
char whick Probably not too bad if you can get it to 4 ghz
As for the capping FPS thingy...
If I get screen tearing that capping doesn't seem to fix and I therefore have to enable Vsync to get rid of it - can I still enable FPS cap as well, to battle frame skips caused by Vsync?
About mining; I recently seen a video about AMD's Instinct cards without any video-out connectors, nor a fan as they are intended to be used in specialized rack cooling systems. It makes me wonder; if they can do that, why hadn't they modified their Instinct cards to use consumer watercooling components so the miners can have their cut-rate Vega's and leave the gamer-oriented cards alone?
Back in my old rFactor racing days someone said that NON Vsync would make their GFX card run hotter.
On the flip side, driver input (steering, throttle and such) was calculated faster than 60 fps (if running faster than 60 fps).
Input was updated at the frame rate.
On the system I was using I had terrible screen tear with Vsync off.
My question is: Why does decreasing the size of a transistor on a dye (16nm-14nm) lower power consumption? Is it because since the size is small it takes less energy to "flick the switch"?
The last two time stamps in the description (20:33 and 21:23) are swapped.
Looking forward to the Noctua video.
I know that 2 out of the 4 dyes on the Threadripper are active and the other 2 are inactive and used to balance/support the IHS.... But why not just remove the 2 inactive dyes, rotate the other 2 90* and put them side by side? I'm sure there's a reason, I just haven't heard one yet. Is it that the active dyes by themselves can't support the coolers?
Side question... Will CPU coolers going forward for Threadripper be covering the whole IHS? I would think more contact is better... is there diminishing returns in heat conductivity and transfer depending on the amount of surface in contact with the cooler? Thanks!
quick question...
How do parallax textures and mapping work? how can you make something look so high in polygon count without heavily taxing the graphics card?
where is the cat today? by the way i like your intro
Jay Lee Yes, I was sadly disappointed by the lack of the single most important staff member.
I've noticed that you don't often use anything lower than "Ultra" when benchmarking, unless it's a relatively weak GPU. Is there any particular reason why you wouldn't include "Medium" settings as well aside from the extra work involved?
I'm curious about this because there are many games don't have dramatic improvements in the visuals above "Medium" or "High", so many people, myself included, will prefer to sacrifice Ultra settings in favor of smoothing out minimum framerates and frametime spikes.
Is there currently a way, or possibly a future way to pool multiple systems to split a project between the pool of two, three, or four systems, and merge the pieces after each system rendered its part? Thank you for breaking down frame capping. Is fast sync only worth it after you're consistently over 2x the refresh rate or more like I've seen online, or would 80-100 fps be an improvement fast synced on a 60hz screen. It seems to me that frame pacing would suffer, but what does everyone think or have you experienced? Fantastic video!
@ASKGN I just watched your video on how heatpipes and air coolers work and I'm wondering how do the heat pipes work when the cooler is mounted sideways, or in the case of a video card, the heatpipes are horizontal. How does the convection work in that orientation?
Capping FPS has always done wonders for me in areas outside of the game performance - heat, power and in the case of some video cards (R9 Nano) massively decreased Coil Whine. When Catalyst came out with the feature - it was the happiest day of my choil whining life.
Same I use it for the 5900x which is a nuclear reactor(kinda all of ryzen 5000 is fx is nothing compared to it).
hi guys great video, just one question. What is Zotac like as a graphics card manufacture? here in Australia they seem to be cheap than other brands.
I can understand the money issue with case color. But an option I think would be to be able to get a primer color so you can paint it yourself or send it to a shop. Even caselabs is lacking these days on color choices. I personally like gunmetal.
Question for ask GN:
I have recently updated my setup to a ryzen 7 1700 and a gigabyte gaming k3. I had a watercooling setup and the CPU was overclocked to 4ghz with 1.4v. All the temps were fine but in HWinfo64 i found out the VRM to be extremely high during stress test (120C+) and really high in gaming (90C+), especially in CPU intensive games like BF1. I had to roll back the overclock and put an air cooler. But temps are still bad for the VRM, even at stock clock, unless i put the wraith cooler. Is it a design flaw or all the 4 phase AM4 cards are the same ?
Nice work about the fps cap, but there is a huge missing point. Cooperation ! 90% of the time, a game that doesn't cap it's FPS, get lag not from the game itself but because it get interrupted by other programs (including OS), it was always the case on PC (that's why you need a bigger pc hardware than console hardware to get the same experience) but is now getting more and more true on this generation of console. A game is like a comic stand up, if you do not include time for people to laugh or clap, you might get interrupted by the audience (OS or other programs) at any time, and for longer for a random time, and could not speak as long as the audience decide to keep interrupting you, Strangely, if you write the most efficient program, using all core of you cpu, you can get interrupt many times for 100 or 700 ms, because at one point the OS want to get time for him or other program and when he has the hand he is not giving it up easily, On a single threaded program it's a problem, but on an multithreaded program when you have thousand of tasks with dependencies, if one of the core is given back for a small time to another process you have a cascade of delay, because other task on other thread could wait for the result of the late task to continue, and all work stop. So going the faster you can is not the best solution, you could get a higher max frame rate, but a lower min frame rate or average because of the huge interruptions. And this could also come from your hardware, you talked about it a little by speaking about power usage. Power usage = heat. If you make a CPU or GPU works 100% it will heat up, and then throttle it's frequency to cool down , leading to longer frame time, so possibly another reason from lags, making the overall experience less smooth. And in the worth case scenario it could lead to crashes or reboot, So you have to cooperate with other software and hardware as a game dev, and even like that you could be interrupted leading to lags, that was the idea for windows 10 game mode, allowing an application to get a more exclusive use of your hardware, but apparently they failed. PS, i'm a pro game dev, with multiple million sellers and 20+ years of experience and i still scratch my head to find the best compromise between the game and the OS.
I bought the Phanteks 400s even when your review nearly made me choose another case. I still pulled the trigger and got it. Then my son made my whole computer desk fall. The front panel which covers the fans protected my whole system by absorbing the impact. no damage to anything else then the front panel and of course the acrylic panel which got scrached. Now case build quality is my number one concern. Never thought my pc could drop 4 feet but stuff happens. GN4LIFE
does ambient air pressure i.e altitude affect cooling performance of heat sinks/radiators? I live at 6500 feet. so I wanted to know if altitude effects the temps of a heat sink in any significant way.
Is your bench cooler an EK Fluid Gaming A240 or something different from EK?
Small input on the 2nd question... If you limit the amount of frames the GPU is producing, not only will it lower the power usage like you mentioned, but it will also lower GPU temps and make throttling far less likely.
I have an i5 7500 with a B250-F MB. When would it be a "good" idea to consider upgrading to a K CPU and Z MB? I don't stream, render, and multitask (not heavily), so the criteria would only be for gaming at 144hz (if possible). Thanks!
Ever considered doing a comparison of thermal pads? The Lightning and EVGA 1080 Ti models you did tear-downs of have memory thermal sensors they seem like ideal candidates. Honestly, I cannot find any in depth comparisons of high-end thermal pads with conductivity ranging from 6 to 17 w/m-k.
For the next Ask GN: Would you personally recommend a copper based liquid cooling custom loop over an aluminium one like the new EK 'gaming' kits?
Irfan Ahmed It's all about price to performance ratio. Copper will cool better, but aluminum is cheaper. Whichever makes more sense for you would be the better option, I suppose.
@Gamers Nexus
(2nd attempt)
Why is USB 3.1 Type C still so uncommon on (modern) computer motherboards?
Most motherboards of $120+ still tend to have only 0-1 type C port. I thought Type C was supposed to be the new standard that is going to replace stuff like type A & Micro B. I see alot more produts (like enclosures) that support Type C, but not a lot of motherboards (and no cases with front panels) that support type C.
Thanks!
Orvis25
Likely because the average user doesn't use them at all, or very little. For now.
no i understand that to a degree, but if the consumer products like externals, phones, and yada yada are switching over to it, why is the MB manufactures not stepping up to the plate and including more than 0-1?
Orvis25
I compare it to cases only now starting to have USB 3.1 ports.
I guess, but it would still be nice if GN could get some inside input into why from his contacts.
Orvis25
Indeed it would be. I'd be interested in it for sure.
Question for Ask GN:
Steve, I am curious to know if an Overclocked I7 7700k, paired with some fast memory could potentially bottle neck a pair of 1080tis at 1440p. It would be interesting to know if the higher frequency of the 7700k could compete against the higher PCI-e lane count of an x series chipset when using a dual GPU configuration. Thanks!
As far as coloured cases go, i remember the BitFenix Prodigy. I had the orange one and it was probably the best ITX case I've ever used, athough it was quite big.
I read the fps question as... If I lock the max, is the gpu stressed less enough that it can run better, allowing for higher lows?
For ASKGN
can we customise an AIO or CLC are there any aio in market which can be customised?
Why, oh why did AMD decide to use a 20 degree offset on the 1700X and 1800X cpus? They could have just asked people to set an aggressive fan curve instead.
I see the confusion STILL lingers all these months later.
Love this sort of content. Steve and Co. you guys are the best.
Love that thumbnail
I'm still cracking up every time I think about it.
Jack Sparrow's love of rum is a perfect analogy for the gamer's love of frames.
Its not just about render performance. The reason I and a lot of people stopped used Premiere, is its stability it crashes for dumb reasons way too often. I swapped over to DaVinci (some of my industry friends swallowed the bullet and bought Avid for their home use as thats what we were tough in collage) and have never been happier. and its super easy to adjust to.
Glad to hear it works for you!
Premiere works for us, we like it, and we like the videos coming out of it. Until that changes, we will stick with it.
Ya if it doesn't give you any issues why change? why I stressed stability and barely said anything about performance as thats is just a hardware fix. probably could have worded it better though...
These timestamp progress bars are great.
Question: is there a difference playing a game at 60fps on a 60Hz monitor vs 60fps on a 144Hz monitor. Assuming both are displaying 60fps the 144Hz monitor should still be smoother. I guess my point is that do you really need over 100fps to truly take advantage of a high refresh rate monitor like everyone seems to be saying?
I have a question; since my Corsair H100i GTX CPU cooler died recently, should I NOT buy hybrid GPU cards for the same reason? I have the MSI 980Ti Seahawk. And I'm thinking of getting rid of it and buying a 1070 or 1080ti because im worried my card might fail just like my cpu cooler. I game at 1080p and the 980Ti does every game maxed out minimum 60fps from my experience.
I have a question for the next ask GN if it's not too late:
Do AIOs Perform aswell as a custom liquid cooling loop, and if they don't why don't they? As far as I can see the only difference between an AIO and a Custom Loop would be the tubing and having the CPU Block seperate from the Pump (and the reservoir if there is one)
Now that you have done a few streams, Do you find live streaming "easier" then your regular content? Does it take a load off of your production pipeline?
With the latest bios update on Asus Crosshair VI fixed the temp reading on all programs. So it's giving out the actual temps now.
Is there something like "cpu only rendered frames"? I know ot sound ridicolous. I ask since some games I play look like they'd scale with cpus indefinatelly. I have 9800 GTX+, and while on some test rigs in f.ex. cs:go ppl get avg 70 on low(on budget cpu), with my 8320 fx I get to the 100fps+ territory.
Just did a new build with a 7700k, gtx 1080 ftw2, MSI z270 m7, 32 gb Corsair vengeance lpx. Premiere is not running as smooth as I had expected. I get significant lag when trying to open the Lumetri color wheels. Just switching from Waveform to the Histogram takes probably 1-2 seconds. It's significantly slower than my 3+ year old laptop. Any idea what is going on??
On the software temp monitoring question. I have the 1600x and HWmonitor along with Ryzen Master both show different temps.
Usually around 80C (stock speed) in HWmonitor after just browsing the internet and watching hippoporn uh i mean youtube.
Idle temps are around 50c. I have swapped out CPU water blocks, (EK EVO and Swiftech) and motherboards. No difference.
I even RMA'd the 1600x with amazon. Still no change.
My 1700(@4ghz) and 1500x(4.2ghz) seem fine with both hitting highs around 65c under full load. As i have swapped them out as well.
My videocard (GTX 1070) is in the same water loop and it;s max temp is the same with all three CPUs, about 40c tops.
Using my Fluke K-type thermocouple and non-contact fluke temp gun along with just feeling with my fingers
everything feels just slightly warm or close to room temps.
I am going to try the HW64 you mentioned and see how it goes.
I recently bought an oculus Rift and they recommend having 1 Sensor on per USB 3.0 Controller. I was wondering if each USB 3.0 Port on the Rear I/O of a Motherboard would have a separate controller, or if some of them were shared. Looking to do a 4 sensor Room Scale Setup. I have 4 Rear USB 3.0 Ports, and 1 USB 3.1. Trying to find out if i need an add in card or if i would be ok with using the Mobo ports. The Motherboard i am using is the Asus Maximus IX Formula.
Vsync doesn't necessarily introduce stutter, it really depends on the implementation. I always enable it in everything but multiplayer games.
Is it possible for RX Vega to use the Infinity Fabric to get better Video Card to CPU coms ( they use IF for CPU to CPU coms in Epyc)? If so, could this be a reason that we have not really seen any testing results from AMD on Ryzen?
Hi GN /Steve,
I have two questions.
1, Why after all this time and iteration on multi GPU systems do we still need software drivers for SLI and Crossfire? Why is it not possible to dispense with game developer effort to program for mgpu when we could have a hardware solution within the cards or even the chipsets themselves presenting two card as one for programming purposes. Even if we needed a chip on the card! My point being, it is crazy that some games scale and other don't, this burden should be taken away from the game makers and be done in hardware. Why have we not achieved 70-90% scaling on everything? What happened to AFR?
2, I forgot, had something to do with VESA not borking Gsync by making fast sync better.
You could in theory gain FPS by capping FPS due to the Chip staying cooler so when it gets to more intensive area the CPU or GPU could be able to get full turboboost. This only applies if you are thermal throttling and in rare circumstances. Thought I would just mention this hehe.
I actually overclock my turboboost but not stock clocks on my Chip as I am only really bottlenecked hard by CPU in single threaded unoptimized games.
"Adobe is not that well optimized for..." might be a bit of an understatement. ;)
Not to mention how often it crashes. The sad thing is that Adobe doesn't seem to care, how difficult can it be for that company to hire some good programmers who can fix all those issues? I don't say that it is easy for those programmers but there certainly are enough programmers available who could fix this if Adobe would let them.
What are the hardware differences between high efficiency rated(Platinum ) and lower efficiency rated(Bronze) PSUs?
What time do you usually upload? Its 3am here in Brazil
I am thinking of buying dell 7567
I5 , 8 gb ram, 256 gb ssd m.2 1050ti for 900$ is there a better laptop you can suggest for gaming in the same price bracket?
question for you Steve! since 60fps are low, tell us about monitor overclocking, how its done with modern gpus, is it dangerous? why no one does this in the RUclips community (jay, ltt, paul etc)
Look at those sexy Noctua's... those cream and puke brown fans are not only efficient but stylish too! RGB is yesterday... puke brown is in
i actually really like the way noctua fans look. I gives it a simplistic, indrustrial vibe
Hello. What happens if you pair 2800 mhz RAM with a processor that supports slower memory, like the Pentium G4560 which can support upto 2400 mhz RAM? I know you can't overclock the processor, but I am pairing it with a Z270 motherboard I had lying around.
Does Ryzen ram speed above 2400mhz @1440p gaming really matter for the $50-$100 price premium? Building a system soon and would like to know if higher speed 32gb (2x16Gb) kits shipped from the states are actually worth it...
Another good video GN, thanks. Will say though. HWmonitor for Ryzen is still broken. I still have my temps reported in the 80's when I know they are in the 60's. It's a bug on all the X models
Not a bug. That is because AMD still does their TCTL offset silliness, which is only present on X models. Some software accounts for this better than others. HWINFO seems a bit easier to work with.
Hi, I was wondering if you could address what might be the reasons that a Ryzen CPU with an identical core count to an Intel CPU cant overclock as high? Thanks