Automatic Super Resolution (Auto SR)
HTML-код
- Опубликовано: 28 сен 2024
- Sources:
devblogs.micro...
support.micros...
What equipment do I use to make my videos?
Camera: Sony a6100 amzn.to/3wmDtR9
Camera Lens: Sigma 16mm f/1.4 amzn.to/36i0t9t
Camera Capture Card: Elgato CamLink 4K amzn.to/3AEAPcH
PC Capture Card: amzn.to/3jwBjxF
Mic: My actual mic (AT 3035) is out of production but this is a similar mic (AT 2020) amzn.to/3jS6LEB
Portable Mic attached to camera: Rode Video Micro amzn.to/3yrT0R4
Audio Interface: Focusrite Scarlett 2i2 3rd Gen: amzn.to/3wjhlad
Greenscreen: Emart Collapsable amzn.to/3AGjQXx
Lights: Neewar Dimmable USB LED amzn.to/3yw4frD
RGB Strip Backlight on desk: amzn.to/2ZceAwC
Sponsor my channel monthly by clicking the "Join" button:
/ @danielowentech
Donate directly to the channel via PayPal:
www.paypal.com...
Disclaimer: I may earn money on qualifying purchases through affiliate links above.
![[Limbus Company - Teaser] Season5 - Oblivion](http://i.ytimg.com/vi/8anKuEeZLeI/mqdefault.jpg)
They're probably using Borderlands 3 as an example because the art style easily hides poor upscaling and over-smoothing.
If you use one of those old free online image upscalers, you know what I mean.
Yeah lol, I can clearly tell that upscalat has problems when it changes the gender of your charscter lol
@@DragonOfTheMortalKombat my man, thats a different run with a different character... its even a different jacket. the girl has short sleeves.
@@steinkoloss7320 the kid is just dumb to be dumb.
The foliage still gives it away. Native res looks like grass, the upscaled is just a garbled mess of mushy pixels, lol.
@@DragonOfTheMortalKombat tell me you cant read without telling me you cant read
The AI at 4:50 is so advanced it upscaled a male character into a female one.
Microsoft knows what we gamers want 🤷♂️
I couldn’t tell if it was a female or a fsr type shimmering lol
Based Microsoft
It's a diversity upscaler.
The AI baby inc 🤣🤣🤣🤣
It seems like basically just a solution to compensate for the reduced performance, due to emulation, of Windows games on ARM CPUs. It can probably get away with extremely low starting resolutions like 720p because the display size is likely to be 17" or smaller (i.e. a laptop screen). That could also help make the increased latency harder to notice. I don't imagine that this is something that most of us would use because we're heavy gamers and probably not going to buy an ARM-based laptop to game. I see this being useful more for people who get one of the first generations of Copilot+ laptops for work or school use but still might want to be able to do a little casual gaming on the side at playable framerates.
Yep. ARM won't be replacing CISC for higher end gaming machines
The CPU isn't affected all that much by render resolution. That's mostly an issue for the GPU.
@@TreyMotes The CPU is mostly affected by raytracing and high framerates
The foliage is always a dead giveaway of such upscaling.
DLSS does foliage the best?
@@viking9442 is that a question?
If so, I guess it does, out of the upscalers available. Though it depends on the initial input resolution and how the game renders alpha textures.
It's always best to start with the highest base res possible before upscaling. For me 1440p as a base res (upscaled to 4K or up) seems good. Anything lower and the artifacts become too bothersome (smearing, weird mushy pixels/edges, blotching, etc).
Native res with DLAA is currently the gold standard, though.
@@TheUltimateBlooper Thanks for the feedback. Do you think DLSS quality will still be useable at 3440x1440 resolution?
Another one are fences. The jagginess and shimmering.
@@Gl0ckb1teDLSS has non existent shimmering tbh atleast compared to FSR
I think this is very good for older titles with no Upscaling features and sometimes locked resolutions
Microsoft entering the upscaling scene with a subtlety of a rhinoceros, while not really providing anything of worth and doing a disfavor to the players and anyone that is sick of diluted standards and directX...
As someone who plays mostly non competitive games like survival or story games, latency isn't something I usually notice or care about unless it's extreme, like half a second. What I look for in my gaming experience is smooth frame rates and no stuttering. I play at 4k 60fps with vsync on because my TV doesn't have freesync. I do have a 1440p 144hz monitor and I can't really notice the difference in smoothness while actually playing a game. Only when swirling my mouse around on the desktop. I'm sure there is a small improvement when playing something like battlefield or something but they haven't even been good enough to play lately. I'm honestly all for these upscaling techniques. Better quality and more fps for weaker hardware is a good thing. Especially when Nvidia is charging a kidney for a gpu. Love the work you do!
The most interesting comparison here would be AutoSR vs LosslessScaling's upscaling options.
You already have it. GPU drivers offer native support for DirectSR, an API that standardizes Super Resolution (SR) in D3D12 titles. It permits users to choose between upscalers during runtime and functions as an independent solution, eliminating the need for vendor-specific integration. AMD FSR 2.2 includes a built-in version of DirectSR. NVIDIA drivers, starting with version 560.38, have embedded DLSS Super Resolution accessible via DirectSR, supporting NVIDIA RTX 20-Series and later models.
DirectSR and AutoSR are not the same thing.
Sounds like a way to take away control from users to make life of future Linux expats difficult. I have zero issues choosing what upscaling I want to use on game's graphic settings.
I'm not convinced until they put this AutoSR thing into Cyberpunk 2077 or other realistic game to work reasonably.
"Where's my unrecord button."
My favourite part of the video, thanks for leaving it in.
A single frame latency in upscaling is not the same than frame generation, because in frame generation the frame time doesn't decrease, so if the game was running at 30fps and interpolated to 60fps the frame delay will still be 33ms but with upscaling a 30fps with each frame taking 33ms will become much faster when decreasing resolution, so each frame will be for example 16ms and a frame of delay will be much lower than a frame of delay in frame generation.
Plz AMD if ur seeing this release 3.2 or something when the hell its going to release on doomsday
AMD already checked the FSR box and can go back to ignoring the quality of the feature.
@@HappyHubris Amd is not done fsr is far from fine quality lost is huge compared to dlss while moving even xess 1.3 is better may be they're limited cuz of ps5
@@mrshinobigaming8447 I use fsr quality on ghost of tsushima and can't notice any quality loss at all.
@@mrshinobigaming8447 they shouldn't be done,. But the company often focuses on slapping features on the box without making them good.
Thank you for these videos, I only watch you and Zach on this platform for tech news.
As long as there's at least 100% of frame rate increase to an additional delayed frame, or 50% for half a frame of delay, the latency would actually be better than native.
D.O. talks, I listen!
He is my fav
Indeed
he literally just reads from other sites lol
@@najeebshah. just think for a second, how would he know how AutoSR works if not from a Microsoft blog is no like he can go to their office and ask the people that coded all of that, use you brain brother
@@_jn0298 im not disputing that, im just surprised at how people fanboy over that ☠️
No hate but if I need a new cpu which most likely will be very expensive, I don't see why would I need this, would probably have money to also pay for a good graphics card and this would not be needed.
I think this could be good only to the new laptops.
Back when I was in Junior High when Call of Duty 4 onward were out on and being played or referenced often by everyone the Xbox 360 they upscaled a 1024x768 image to 720p with checkerboarding and then typically again to 1080p using the internal hardware upscaler. Results were extremely good, it looked sharp and crisp especially on a nice television or even a terrible one. Fast forward to today were on 4k approaching 8k beyond. Without question this will be a positive thing and end up being a godsend for every title even if it doesn't look quite right in certain scenarios playing a game running this smoothly and this amount of detail or clarity was second to none especially for ease of access and splitscreen, although I'd imagine it will be best used in the OS which is what I'm most excited for. AI is going to shape the world for the better, let's continue this trend. This will help new console revisions immensely to enhance visual quality and keep PC on an equal opportunity playing field.
Microsoft seems to be completely allergic to the GPU AI accelerators we already have on Nvidia and Intel gpus
That's nothing new? I thought MS did not like those tensor core like hardware since the tech originally coming from their competitor haha.
They are coming up with new ways to upscale with AI for a broader user base. Nvidia and Intel already have their own solutions and I wouldn't be surprised if AMD makes an AI based version of FSR too. Microsoft is making them easier to implement with DirectSR, so there isn't much of a need for Microsoft to make yet another version of the same thing. With AutoSR, upscaling will be available for anyone purchasing a new laptop/mobile device with an NPU(which will pretty much be almost everything from here on out it seems). It's great because most people won't have laptops with dedicated graphics, and the mobile market is the market that would benefit the most from upscaling due to a lack of dedicated GPUs in most mobile platforms.
@@geo_beans FSR already runs on everything, even on 8-year old Nvidia graphic cards. Whats the point of making the same thing but with worse latency?
@@Boris-Vasiliev FSR looks bad on low resolutions. Maybe when 3.1 releases, it might look good, but we'll have to wait and see about that. Again with 3 upscaling technologies for GPUs already, there is no reason for Microsoft to make their own. I don't think the added latency will also be that bad on APUs bc everything is on one chip, so latency should be even lower. There's absolutely nothing wrong with more unique options.
@@Boris-VasilievMost people don’t know what FSR is. If they add AutoSR to Windows as a toggle option, a lot of people would just turn it on.
I do like Frame Generation. There's a very slight difference in feel, but you also feel the smoothness as well.
Current NPUs are lacking in memory bandwidth, a metric just as crucial as TOPs for new AI applications.
So they're pretty slow for LLMs which are memory bandwidth bound and I guess it's also part of why running upscaling on the NPU adds so much latency.
I hope that memory bandwidth will become part of the basic specifications when talking about NPUs, not just a single, extremely obscure TOPs number (which doesn't even specify at which precision)
Green light under the tree is presumably a crate with loot, we lose it in lower res.
As far as I understand there is nothing stopping AMD and Nvidia running any "copilot" NPU algo on their modern GPU's.
And it seem that Microsoft will open up the ability for them to do so.
At least Nvidia seem to hint at that.
And there is no technical reason for why GPU's could not run this code.
The NPU's so far are way weaker than most GPU's for similar code, but they are focused on providing "TOPS" in much lower power envelope.
Superb. Thanks for the insight.
Yea if you're PC has an RTX GPU then it should handle this with ease.
@@Spright91 For sure. And just for comparison for anyone who don't know who stumble on this thread.
An RTX 4090, have 660 TOPS.
Just divide that with your current GPU's power and you get a pretty good idea of where it stand.
@@cajampaAI tops 1321?
@@racoonchief Yeah, you are correct that is the stated "spec".
But I have never actually seen anyone be able to get that much out of it in any app or benchmark.
They highest claim I remember I have seen measured is 660 TOPS.
And I think that is closer to what ever here NPU hardware is claiming. Because they are not so power limited for example and are running limited models in size and complexity because they just don't have memory bandwidth or speed to run anything massive.
If you or anyone know any place where there is a stated higher measured value than that. Please correct me with details.
Unfortunately I can't remember where is saw the 660 TOPS claim but it was in some LLM discussion.
It seem like there is some limitations maybe in how apps are using the GPU.
Or maybe I just am misunderstanding something or my basic reasoning is wrong.
So please correct me if someone know for sure I am wrong or way off here.
Thank you for posting this video. I was wondering if you might have an idea as to how many upscaled 720P to 1440P frames per second could be achieved by the 45 TOPs of the Hexagon NPU? Hopefully it's at least 60 upscaled frames per second for smooth gameplay.
LOW-Midrange laptop probably has the best benefit with that 700-900p spec, 1080p/14440p laptops with middle class mobile spec GPUs. I guess the HUD elements will be upscaled with the rendered image as well.
allways learning something awesome thank you so much sensei
The 'one frame extra latency' must be an estimate assuming 30-ish base FPS. Because the NPU's frame rate cannot become faster just because the GPU's frame rate is higher, unless it decreases the quality of the upscale. Also consider that if the NPU took more than one frame to upscale, it would become the bottleneck and would end up decreasing the output frame rate (compared to source, not native) in addition to adding latency.
Another point I should make is that DLSS upscaling is run before the UI rendering and before a lot of the post-process effects so those are done at the output resolution and this also contributes to lower-than-source frame rates with DLSS. It isn't just the render time of the upscaling algorithm itself. So Microsoft's AutoSR is more similar to a driver-based upscaler like Radeon's RSR in that respect.
I am not surprised that Microsoft's AutoSR is this slow, considering that 40 TOPS is very little compared to the average RTX GPU.
I think that this is intended for laptops with an SoC containing a low end GPU that doesn't explicitly do matrix maths and an NPU which specialises in that. Becauses it's all in the SoC, the RAM pool is the same and you don't have as much latency from moving data between the GPU and NPU, meaning all the latency comes from processing the algorithm. This will probably come to similar chips from Intel and AMD. From the Microsoft standpoint, the competition is Apple, as their SoC has all the components to do this, but doesn't have this software feature to run on it. It's a reason for a casual gamer to buy a copilot plus PC instead of switching to a MacBook.
When it comes to an Nvidia GPU, the tensor cores aren't that dissimilar to an NPU, as they also specialise in matrix maths. By being on the same silicon as the rest of the GPU and using the GPU RAM, the latency is minimised, so DLSS makes sense. Shunting data back to an on-CPU-die NPU would add extra latency and isn't needed because DLSS is effectively running on equivalent hardware in a more appropriate location.
NVidia already makes ARM CPUs with custom cores, so they could make an SoC if they wished. They would probably tweak the tensor core section of the GPU and call it their NPU, and you'd probably be able to run either Microsoft's upscaling or DLSS on the same hardware. I expect they'll just wait to see whether Windows on Arm takes off or flops first.
They could use this to upscale streaming video or upscale low resolution video in real time. Maybe even upscale photos or other types of media.
So Auto SR is the Microsoft version of PSR, Sony rumoured to be using a dedicated NPU on the PS5 Pro rumoured to be released this November but it maybe delayed to next year to launch alongside GTAVI as a launch title.
Be interesting with the Xbox Next idea which it's rumoured to go 3rd party next gen which in all fairness it's basically going to be a Mini ITX PC with Xbox OS as Tom Warren already predicted, but the slides from the FTC case did indicate a NPU as part of a hybrid console, most likely with a big push on Xcloud as well so Xbox Next would have to have that NPU processor as well on it's motherboard or within it's specs to meet the Xbox Next standard.
It's all looking very interesting for the next generation of consoles with a big push for AI.
1:53 can't really agree, all of the slopes of the house are extremely aliased after "Auto SR" applied, won't be surprising if it shimmers a lot in dynamics
One of my problem with this technology is, that it needs an NPU which not a lot of people will have at the time (Edit: ok just limiting it to one thing is more questionable, but still interesting to see). Another problem I could imagine is that Linux won't be able to use it. But DirectSR sounds nice
I noticed using "Lossless Scaling" (the image upscaler, not FG) along with DLSS FG in supported games eliminates ghosting and artifacts. Maybe the AI just removes the ghosting or something
Yo you got the best content in the game keep it up
Cool i guess, but if you have an Nvidia card DLSS for Performance gains and DLDSR for quality gains /upscaling past your monitors native resolution are just better options.
Yes you may have a tiny performance penalty, but the quality will be much better and add no latency.
Interested where this will go though.
but if u play at 30 frames and then u use framegen and play at 60 or 70 frames for example latency don't make sense here because the half frame or 1 frame letency is better than playing the game at 30 because 30 fps u become the latency especially in online games so its better to play with framegen
i like using upscaling for games that struggle gettting over 60 fps or just in general as i think most games in 1080p out of the box is kind of blurry. but upscaling from native monitor resolution 1080p, is just a even more blurry mess.. so when activating DSR in nvidia control panel to 1620p at 5% smoothness. you will raise the upscaler to run at 1080p and itll result in a very nice picture again.. and to my experience itll still run at a bit higer/same fps campared to native.. this might also work with amd. as you can add a custom resolution.. doing it this way on nvidia also works great so there is 2 ways to implement it.. i use rtx 3080, so performance might differ from older models.. but in general i actually use this to get a sharper image..
Microsoft: AutoSR only for Arm
meanwhile 95% of people running on x86 architecture....
is this some kind of joke?
How do you think they hype up stuff lol.
Give it a few years and we will see how x86 holds up. Who knows.
@@zushikatetomotoshift1575 Yeah x86 will be history . The only problem is gaming as it will be difficult to emulate
I think MS co-develop this with Qualcomm. For x86 there already several options. FSR, DLSS, XeSS and Loseless.
@@zushikatetomotoshift1575Months even.
The 720p to 900p base makes me think this was developed to help the underpowered Series S. However, the NPU requirement wouldn't allow it to work on the Series S. It would be interesting to know the origins of this feature.
Or possibly for the Xbox handheld they confirmed they were working on today.
If it's a next gen device it needs to be better than series s to play next gen Xbox games, and a series s strength handheld would already be expensive.
Perhaps with this they can make a 1080p handheld that can keep up using their own tech.
1.) DLSS does NOT use an AI based upscaling algorithm. It works oretty much the same as FSR2 upscaling: They calculate some pixels and reuse the other pixels from the previous frame. Because in a game stuff is usually moving they need motion vectors to identify where that pixel was in the previous frame.
Where DLSS uses machine learning is to solve certain edge cases where either motion vectors or the pixel in the last frame. These are all the situations that can potentially cause artifacts and this is also the reason why DLSS and FSR2 are pretty much the same in upscaling itself, but DLSS has less artifacts.
2.) Auto SR is a competitor to AMDs RSR and Nvidias NIS. We will see how image quality looks like, but I think RSR will be the better option because:
a) It does not add latency
b) It also has neglectable overhead, but does not use the NPU so there is full power budget available to CPU/GPU
c) It does not need a high end NPU that is only available on APU that are not the best gaming devices anyway
3.) Inam really wondering: Why do we need DirectSR as a 4th solution? It will obviously not replace the existing 3 technologies
3: It's probably going to be a common API to speak to DLSS, FST, XeSS, and anythinglike them. Implement it once and it'll work for all of them.
we will need devs to port ARM native game, other worse solution is byte code on vm or emulation. and ARM we talk about is focused on low power processor
Them presenting it on game which does not have sharp textures and fine details speaks for itself.
Example is Garbage-IN => bit less Garbage-OUT.
But once it is on High-Quality-IN => results are Garbage-OUT.
Finally someone use NPU for Frame Generation and Image Upscaling
Imagine a PC with CPU from 2024(because you need NPU) and a GPU from before 2019 (when rtx2000 was introduced with first version of DLSS) running windows 11.
I dont see such a configuration, where a 5-year old GPU is paired with modern CPU and has no support for DLSS or FSR so you have to use AutoSR.
If AutoSR introduces latency then I'd much rather use FSR over that, even if it looks worse. The whole point is for the game to feel smoother, not just look smoother. Unless you're already running at high framerate, that doesn't help you much
so I can render 1080p or 720p and use autoSR to upscale to 4KUHD? :o
Looks like next year we will run games 4k DLSS/FSR quality and also hit Auto SR to get 500 FPS without loosing visual quality or adding latency.
I usually play games on my junky office laptop in 25 FPS (I have 75 Hz monitor). The latency is already noticeable. I can't really imagine the 2x extra latency when playing games because with 40 ms latency +- error is already bad enough.
here's hoping to see it on the next xbox (handheld?) and upcoming arm PCs
I remember Maxwell and Vega or Polaris having DSR or VSR but it didn't do well as it impacts performance when you have SSAO and almost felt like physx's choppy gameplay. Feels more like justifying WoA for enthusiast or gamers why x86 is bad for higher market and mind shares
In my experience Windows display features haven’t been great.. I think it’s a frequent problem for people with multiple monitors. I just want my screensaver to turn on consistently and I still haven’t seen them handle that update. Lol
Its like if i wanted latency in my game. I'd just use Frame Generation right ?
1440p is 2560x1440, which is four times larger than 1280x720, hence QuadHD.
Wait. Frame Gen require 2 frame so it is like if you are running game at half the frame/rate in terms of latency, here is telling you you may acquire up to 1 frame of latency so it is less at the beginning and the same or less once is running.
You need a de-esser sir
The Microsoft upscaling seems like something I'd have thought of before hearing of it but surprisingly this is the first time I've considered OS level upscaling specifically. AMDs driver level upscaling is actually really good and a lot of time I would prefer going 1728p upscaled to by AMD to 2160p as opposed to using fsr 2 quality 1440p to 2160p. I can't imagine the new tech will be better than that but certainly the more the merrier.
Yeah FSR 1 didn't look that good but didn't had any ghosthing either, so using RSR to stabilize 1% lows was a great tool, these days i prefer Xess 1.3 for in game upscaling and LSFG to cap my 165hz monitor
Curious on running it on a ps1 emulator running on native resolution 🤔
Requiring copilot makes this a big "NO" for me, it's way too intrusive for me to even consider it, every time microsoft does something like this, it just pushes me to switch away from windows.
Dlss renders internally in a lower resolution also usually 12:49
So it's not even for normal pc users?
AutoSR are the upscaler's equivelant of the AFMF's equivelancy to frame generation?
wonder relatively cheap A card + auto SR will be the GPU solution in the future
So it's only for laptops due to npu
I still don't see a point in frame generation, i still see it as a solution looking for a problem. As a dad to small children i find I'm predominately playing single player games when i get the time, and i still don't have any desire to insert fake frames and increase latency instead of just turning down some settings
So what is the point of this? Why not use fsr which is available on any system?
It is reinventing the wheel. I thought that.
Perhaps they made it for Xbox but I'm sure AMD let them use FSR so again what's the point?
one you'll need to patch games to remove frame boosting technology and run better
I will be buying this sh1t, no doubts
Fan of LEGO games and Horizon too
Where is my unrecord button? There it is
They should enable this for video too, imagine you enabling this for a video youtube at 120fps... or for a video or movie at 120fps... by default
I am enjoying a lot, using Lossless for RUclips videos/VLC movies at double speeds
these is pretty much porman version of nvidia NIS and amd RSR
Are they trying to kill high end GPUs?
Microsoft also has "Auto HDR" and it's pretty awful. This being tied to Copilot automatically makes it a huge fail.
Why are they using a low details cell shaded game Borderlands 3?
Once again a technology that will only work for controller based adventure games and nothing with mouse input. XeSS is our only hope for a one size fits all upscaler I guess.
The difference between the responsiveness of Auto SR and frame generation is that Auto SR will be increasing the actual FPS. Unlike generated frames this will result in a reduction of input latency.
It's an upscaler just like DLSS or FSR. There is no need to compare it to frame generation.
@@jorge69696 except it adds latency the way frame gen does, and upscaling hasn't until now.
@@jorge69696 He's comparing to Frame Generation in reference to the increased latency Auto SR imposes by taking an extra frame to do the upscale after it's been rendered.
It's a comparison Daniel Owen makes in this very video @ 5:28
@@SpecialEllio lol upscaling does allways increase latency becaue its chained in the pipeline there is literaly no way around it....all existing once included..a full frametime is rather long however....typicly you end up with lower overall latency due to higher frame rate....ie if it takes 25% longer per frame to show up on screen but then you are doing 40% more frames thx to reducing their size you still end up with less latency afterall...this is realy no diffrent except beeing on a entirly seperated track and hence much slower....but OP is right fromthe tech its an upscaler and has zero to do with frame gen...
@@proxis9980 that's not adding latency, it has a frame time cost for the calculations, but that's offset by the lower base resolution. But it still has whatever latency it's supposed to for a given framerate.
AMD is working to use NPUs to reduce power usage by RDNA4 gpu by 20% while gaming. Which is pretty cool. And yeah if microsoft wants to impress me, show us super resolution with an ingame video like ghost of tsushima in front of a flower field and lets see the quality and flickering then... guarantee its worse than fsr-dlss-xess.
Npu, has been cracked. Be careful
i wonder if MicroSauce know about this video 😂 0:14
So it's kind of autoHDR but for resolution scaling. Don't think it would be usable in many games, as most games have DLSS, FSR, XeSS... And old games should not need it... if you don't have a potato PC from 2010.
The images look totally different.
Cool but I really don't want to switch to windows 11 for this :(
The real problem to solve is money the "frames" cost too much so players must get lower spec cards. And new tech like ray tracing lower frame rates even more.
Still waiting for some Super Code Optimization technology to fix CPU bottlenecks
So it's pretty decent upscaling but with the frame time cost of frame interpolation. Nope, not worth it.
And I wonder if they did the testing on an SoC where the time to send the data to the NPU, calculate the upscaling, and then send it back to the GPU is lower than it would be for a dedicated card where it would be sent over PCIe. The latency could be worse than "1 frame on average" if that's the case.
It's trash. It does what FSR1 does. AMD users already have that, and it's at a driver level and called RSR. It makes stuff look like a watercolor painting. that's why they picked this game for this test. The game already looks painted. You can't tell how bad it is, because the art style hides it.
DLSS decrease your framerate in exchange of visual quality.
Should have been called Automatic Super Sampling. Missed opportunity...
All the time they changing stuff and we haven't seen a good game. It started to get annoying. Fsr's , dllss , Ai , double frames, ARM based processor to play normal windows games with Neural Processing whatever it is whats going on here. My mind is going to explode.
I feel ya we aren't seen games to truel advantage of alot of the newer technologies
I disagree. I think there are many good games coming out all the time
Yeah I'm a stick to Nvidia/AMD and Intel as long as Co-pilot is required
Neural Processing Unit!? Did they learn nothing from terminator!?
Unrecord 🤔?
at this point i rather have maxed out game at low fps and just interpolate to get 2x the fps
So how will this work in desktop PC's then i wounder? If you use a RTX3080 as an example and you have something like an AMD Ryzen AI370, will the game render from the RTX card then auto SR is done on the iGPU then outputted onto the monitor? Or is it you must use the iGPU for it all?
I don't think that is really feasible. I don't know much about that technology. But you'd practically have to send your frame buffer from the GPUs Vram to the System Ram then the IGPU needs to do upscaling and send it back to the GPU because the monitor cable is plugged in there. And for doing that there would have to be a lot of virtualization involved too. Both of which would add a massive lag to every frame. If i'd have to estimate I'd say it will probably be 50-100 ms of lag which will make every game unplayable. I think you need to use the iGPU for all if you wanna use that.
probably looks best with a game like Borderlands due to the very basic graphic art style
FSR 3.0 and FSR 3.1 goated #amd
It's shameful to skew a test in this manner. Testing a low texture cartoon like that game to share results. Nonsense, how much did they pay you to select such a poor example.
Using still images to compare the framerate?
Come on Microsoft
getting tired of all the tricks from all teams, long for a return to proper optimization and powerful components that aren't a rip off
Agreed, it started to get ridiculous. Why they change damn processors and technologies whats going on, i never seen in my life such a big mess.
Buy a used, top spec, last gen card. You have options.
Just don't buy your game at launch.
i dnt even see this in mine windows
Microsoft has no choice but to make gaming on arm work because their next consoles will be arm based. They literally have no choice
idk man i can tell a big diff
Fo sho
It's part of the overarching goal to provide real life levels of motion fluidity to rendering. Graphically, games are finally at the level where real life can be simulated. But the frame rate just can't seem to catch up, especially for competitive play (500hz+). The gaming industry has moved away from direct X implementations because Microsoft has proven they will arbitrarily end support for previous versions to force consumers into upgrading/spending more money. If they can do this with vulkan it will catch on.