Oh please. Maybe do it on an unexpected subject like GN Steve reviews monitors with you and there is a monitor with monitor Steve visible in the background.
@@Hardwareunboxed "I don't care if the extra price-to-performance would mean you should instead get a dedicated GPU, I DON'T WANT TO!" And as they died they'll be screaming "logic doesn't belong in benchmarks, only the numbers......"
lmao that thumbnail got me; cheers to y'all for taking these comments and turning them into an opportunity to inform and educate and provide interesting content
I would say 3-15 percent is a huge difference so what did we learn? Exactly what everyone said, faster memory increases performance and they left it out of the original review.
@@hill160881 I.. didn't make any statements about the veracity of the comments in the thumbnail, or the validity thereof. Please try not to insert your own subtext in other's comments. I complimented HUB for taking comments that appeared quite antagonistic for the purposes of the thumbnail and using them to create additional content to inform users. And that's it, really.
Yeah, you brain gets weird when you're focusing on the shot through the lens. Or in Steve's case, I'm guessing live view on the back of a camera. Then you see a reflection of yourself and you straight away look at check that it is you for some reason.
I bought an APU and manually overclocked RAM (Crucial ballistics 3200 c16 to 4133 18-22-8-18-44-66 tRRD 4-6 tWTR 4-6 tFAW 16 @ 1,35V). Unfortunately FCLK only goes 2066 on my 4650g so 2066 it is (1:1). Why spend extra for expensive memory when cheap Micron Rev E can max out the CPU and outperform expensive memory?
@@Berserkism I mean given the current economy if you aren't starting and ending the conversation with that in mind you're being disingenuous from the onset...
Your review: "This is a generic setup, different setups will give marginal differences..." RUclips: [Gnnnnrrrrr noises] Your new review: "Marginal differences"
AMD spec for this chip is 3200 so putting faster RAM is is overclocking, so "your millage may vary" comes into play. Paying a lot of money extra, getting the high end RAM for those couple of fps seems like a bad idea to me.
@@aos32 the lowest end cpu one should get with that memory is 5600x, even that may be a stretch..gn showed you can basically equal a 10900k in gaming with 5600x paired with ddr4 3200 cl14 with memory and cpu tuning
Your first video was great already and the follow up is welcome, it shows how much makes sense to spend in RAM, the only point that I think was missed is the purpose of this APU and the 5600G, this are not meant to replace GPUs, they are meant to allow high performing tiny PCs, the kind that you can put under your TV for 720p gaming or 1080p in old titles, to have as a desktop when you don't have much space and gaming is not your main workload.
I think most important thing to keep in mind is to use dual channel, than anything else, there would be some performance difference with high Mhz, but not much after a certain point.
The speed in RAM isn't the frequency itself but the combination with tight timings (and not just CAS : other timings and even subtimings can be fine tuned aswell to make a difference)
Unless you're playing Linpack Benchmark as your favourite game it's not about speed at all. It's all about latency. Games don't require much bandwidth, but they are heavily branched so quite often the CPU comes to the "You wanted to do THAT?" situation where you miss all branch prediction and all caches, and the only thing that can expedite things is the shortest possible delay to fetch from memory. This is why low latency RAM helps measurably with 1% lows. When talking APUs the problem is just doubled. Now you have two things that can miss caches and need to fetch something ASAP. But even with APUs it's only 10% bandwidth (speed) and 90% latency. If you could get a 3200 kit to run CL12 it would run circles around a 4000 CL18 kit.
@@jediii86 I can predict that you are a white supremacist that thinks the "white" (whatever that is) race should rule the world. You believe that you have the right to shoot anyone that you don't like. You own at least three weapons and constantly look for ways to be "prepared". You think all that you don't like in your life or your surroundings is the fault of "Democrats", immigrants and "socialists" though you don't know anything about them or what they are. Trump is your idol. You also want to be at least as racist as him and you envy his endless grift, recknessless for the law and his "happy-go-lucky-know-nothing-approach". You think science is a dirty word and/or socialist, same with education unless it proves your point of view. You work a menial office or blue collar job and you have never been out of the country except for maybe Mexico or Canada. You live in the suburbs or a rural area. All your friends are white. You speak either only english or if you live in the southern part maybe you know some spanish. You think homeschooling is the best. To you "Budweiser" is a real beer. You drive a pickup or similarly sized car. Your interest in the world doesn't spread further than the next baseball/american football game. You *think* you are a christian but never or seldom go to church and you certainly don't live according to the ten commandments. You *LOVE* stereotypes.
@@jediii86 wow... That was perfect. The perfect example of Dunning Kruger effect. I am in my mid 40s. Married with kids with no other relationships ever. A medical professional interested in computers and science. One thing you got right is that I support BLM movement and I am a leftist Marxist who is a member of the Communist party of India and lives in a township and work in rural parts of south India. May be this info can be used to harass me later. Thanks.
Thanks for a follow up to this! This will help APU users in getting an idea of what performance to expect on different memory configs. It's crazy to see how high you got the speeds while maintaining a 1:1 FCLK!
You can take the SKU and search it on the QVL, these generally state whether it's single or dual rank. All 8gb sticks are SR and all 32GB sticks are dual rank. 16gb sticks can be either.
Don't buy ram from off brands and you won't have these issues. Every major brand had this information on there website. Just because it's not on the newegg or Amazon product pages, doesn't mean the information isn't available.
Reminds me of back when ryzen originally came out I specifically sought out one of the few 3200 samsung b-die sticks that was supposed to be overclockable on ryzen. By the time I ordered mine, it was swapped for hynix despite all the professional reviews and customer reviews at the time stating it was samsung.
@@joe_ferreira Doesn't really work like that. Even major brands swap memory IC's all the time. Buy a Corsair kit for example and it could come with any of a dozen different chips inside depending on the week of manufacturer.
Steve, when you review the 5600G, can you included results with the iGPU overclocked? I understand why OC results aren't normally included these days, but other reviewers have shown getting an extra 15%-20% fairly common and not very difficult.
@@Grimmwoldds I think he was talking about gaming focused builds. Obviously if your use case only stresses the CPU, and the iGPU is just a display adapter, it changes the value equation significantly.
There's performance to be gained from premium memory, but I don't know how much performance people are expecting. Thanks for yet another brilliant video thoroughly outlining what to expect at different price points.
So same as the consensus has been for years on memory: buy the most affordable kit you can get at decent speeds (usually 3200 CL16 or 3600 CL18), and if you want more out of your memory use something like DRAM Calculator for Ryzen to manually lower the timings.
dont get 3600 C18 kits, subtimings are usually dogwater and they cost as much as a Patriot Viper Steel B-Die 4400 CL19-19-19-39 kits, meanwhile Corsair 3600 CL18's subtimings are CL18-22-22-22-4x, only 10-15$ difference but the difference in subtimings and quality is really high
@@rdmz135 true... Couldn't get my trident z Neo 32gb 3600mhz cl18 any higher than the xmp value... Even on perfect altered safe settings of the dram calculator. EVERYTHING failed, was dissapointed because my ddr3 ram kits did have a high range of tuning room.
I’m really grateful for the follow-up! Memory scaling was one of my questions, though I’ve watched HUB long enough to know the memory you use is better than average.
As a viewer, I can speak for all of us (we’re good at that) and say that we have no idea what we’re talking about, but everything you do will always be wrong and right simultaneously.
For all the people raging, I'd love them to find a single person actually buying both an APU and the expensive RAM. Reviews should reflect what the average user of that hardware should expect.
the average user doesnt overclock anything the average user doesnt tweak ram settings benchmarks should be from a properly setup system, overclocked cpu, ring/cache with properly tweaked ram as fast as the setup can handle only morons like the standard "benchmarks" that the mainstream do nowadays
You'd do it for building a custom mini-PC. High end APUs + SODIMM RAM are practically made for stuff like TV boxes, custom portables, and anything else where there just isn't space for a bulky GPU. It's a niche, but I'd say in tech reviews, one worth considering given that these applications are just about the only ones where you basically need a good APU.
Memory recently became a big topic of "leaving performance on the table" in recent years since Ryzen's L3 cache performance is impacted. Also, laptop manufacturers are sneaking in 1Rx16 modules with terrible secondary timing Which can be significantly improved by changing to 1Rx8 or 2Rx8 modules with identical speed and timing
Crucial Ballistix 3600CL16 seems like the no-brainer option for Ryzen systems right now. Budget friendly and can OC above their spec depending on silicon lottery.
Not even depending on silicon lottery! Both Micron Rev. E and Micron Rev. B reliably overclock well, and those are the only two dies Crucial uses in its Crucial Ballistix 3600C16 kits.
Great Content!!!! Content and hard work is the key to success. I have following this channel since it had nearly 300k to 400k subs. If Steve and his friends keep up this pace they'll surpass 5 Million subs easily in no time. Until today I don't know single, dual, quad rank memory thanks to your video, I got to learn from it.
Yup. Spend more on RAM than APU that nonsense on most case. But maybe if fhe user a PS giant poster designer maybe but the RAM size whould be more prior than the speed. :)
What I got from this review is that Hardware unboxed still knows what they are doing with hardware and does an excellent job as always while keeping the gamers pocket book in mind.
Crucial Ballistix 3600 MHz kits hit CL14 with 1.45v so would be cracking value for money as the white and red are £71.99 currently in the UK. (local markets will differ)
@@claritoresdiano1021 Not had any issues with the kit my sons uses and I've just picked up the 64GB (2x32GB) kit which also does C14. My 5950x has just been replaced as the IO chip was dying. (dodgy November patch i think)
Thanks for responding to pur comments Steve! This APU is incredibly flexible for people who wants a temporary solution and loves turning. Memory scaling is sure an important topic for APU since the iGPU is always memory bandwidth bounded. That’s also one of the reasons why Renoir and Cézanne only have 8 CUs instead of max of 11 CUs in Raven Ridge and Picasso. BTW, there are also cheap above 4000Mhz DDR4 out there with Samsung B-Die. We can get 2x8GB from Patriot Viper Series at 100 Euro/Dollars and they are good enough. The advantage of APU is that the SOC is in 7nm and could accept 2000-2067Mhz FCLK easily compare to 5000 series. This compensate the lower L3 cache by having lower memory latency. So higher mem frequency, tider mem subtimimgs, high 1:1 FCLK clocks, higher CPU clocks all help with better memory latency. In best case scenario, I assume that APU could be better than CPU. I have a 4750G and tweaked it a lot together with memory. The max potential of it really sits at 3800x level when high L3 size is not needed. Specially, 4750G has a 88W socket limit which the 5700G doesn’t have. You can PBO the 5700G crazy to the 5800x level and still having the efficiency in other scenarios. Jesus I love 7nm APUs.
It's great to see the difference memory makes for APUs, and also how far integrated graphics has come in recent years. Running AAA games at 30 FPS & 1080p without a dedicated GPU seems like witchcraft, and I hope AMD continues to make strides here.
@@frankytanky5076 One day they will be inbuilt into your VR headset & you can watch Klaus Schwab telling you the bugs injected into your stomach are all part of the clotshot game of Matrix.
I found running my (3200 MHz CL-14) G-Skill kit at 3600 MHz CL-16-16-16-40 to work rather well. I still need to tweak the settings a little, but going over 3766 or something like that had a negative effect on the FCLK scaling, effectively making the end result slower
It still would have been cool to see the actual numbers for DD4 3600 16-16-16-36, though, since that has been shown to be a pretty nearly optimal solution by a few other reviewers (when paired with other Zen 3 CPUs), without being ridiculously expensive.
Right? It would only ever make sense ir you are absolutely set on building the smallest posible gaming machine. But I've never understood that either, as a second hand gaming laptop would be faster, smaller and cheaper.
Dual rank as fast and as tight as you can make it. Remember the tertiary and secondary timings as well. This is a monolithic die If should be able to get to 2000 in 1:1 really easily.
@@NotThatGuyJD quad channel for Ryzen. It's also very important you make sure your infinity fabric timing is lined up to your ram Freq for Ryzen eg 3800mhz ram and 1900mhz infinity fabric
@@AllThatJazOfficial Ryzen is dual channel my man. You could get Quad rank but you'd need 4 dual rank sticks as rank is defined per channel and not for the whole system. Yeah I know but these monolithic APUs have much higher IF clock ability and memory controller than the chiplet based CPUs as such can clock higher with good memory sticks than the non APU
@@NotThatGuyJD yeah they do like 2200 1:1:1 with 4400mts mem. I'd like to pick one up literally just for benching. It's a use case...but not a useful one lol
@@fracturedlife1393 yeah they're pretty great for benching, especially the 5300g since it's got much less competition for cpu benching and any of them are great for mem benching
Gave the game away in the second minute. CL14, that stuff is expensive. Good review, well made. I am very interested in seeing how the APU's perform as streaming boxes for steam, xcloud and how frames increase with FSR creeping in. keep up the good work.
@@gamingunboxed5130 Only because it's my weapon of choice, double snipe load out with heavy and boltie. 😂👍 But I am quite partial to the silenced sniper when sniper shootout is on. 😈👍
So, if one were looking to build a "semi-budget" partial gaming/partial video editing PC in the midst of this GPUpocalypse - would you steer them towards a 2nd hand GTX 1060 and the iCore 5 10400 or the 5700G and some decent 3200 ram with the potential to upgrade the GPU later in life, if things ever calm down?
*3200 MT/s, DDR RAMs run at half of its transfer rate because it's Double Data Rate. A 3200MT/s Memory has a memory clock of 1600MHz. See Dr Ian Cutress' RUclips Channel for further explanations about this ruclips.net/video/5fZO77I-6Cg/видео.html
@@erlienfrommars That is correct. The same way what we call Molex and 24 pin ATX connectors aren't. They're the mini-fit jr. or some such. Very few people care. Everyone knows what it means and _that_ is most important. This is what people accepted as the name and it doesn't have to be "correct." Same is with GB and GiB. How often do you hear people going around mentioning gibibytes? Does anyone care? No? Exactly. But fine, I'll call it what the manufacturer labels it.
Including tests like this is extremely helpful as you can find amazing deals on ram kits. I was able to get Viper Patriot 4400MHz 16gb kit for 120CAD(around 100USD) and I run them 3800MHz CL14
Why are some people so biased with tech? It's not like their are winning anything by siding with one of the brands. This channel just shows you test results & comparisons, it's not propaganda. Chill everyone. Thank you for your hard work, Steven. It's really appreciated.
When I look at the difference b/w price and performance it's totally not worth the upgrade. Why upgrade your 3200 mhz ram just to gain a 3-5 fps at the cost of $100? I know some people just want the best but seriously can anyone tell the difference b/w 60fps-65fps or 150fps-160fps.
Probably depends on the person. I most likely wouldn't be able to tell. It took me a long time to adjust to where I could tell the difference between 30 and 60 fps, but I can imagine people with better eyesight than me might be able to tell right away going to 120 fps. I have serious doubts that 240Hz+ monitors are actually meaningful upgrades to anyone though.
@@obake6290 linus already proved it is actually our instincts are lot faster then we think on fps point i played lot of counter strike on 150-160 and when you switch high to low you can notice but if you started with consoles 30 fps then get to 60 or more probably its difficult to know the difference but someone who spend most time playing low res high fps can easily tell the difference
@@SoftnappGAMEPLAY Bro look what I said... 60-65 or 150-160. I didn't compare 65 to 120 or 140 it's obvious u can tell the difference b/w those. What I'm saying it's not worth it to pay $100 more for 3-5 fps since u can't even feel or tell the difference so basically people just need to be smart and save their money on something else.
Seems like a weird choice, to spend so much in RAM with an APU, which usually means people are on a budget... (and buying a 5700G to pair it later with a discrete GPU seems like a bad choice too with those 16MB of cache)
For me personally, it's a temporary stop gap since gpu is so expensive. The recommendation to get I5 10400 with cheap 2nd hand GPU doesn't work around the world since that GPU+CPU combo is still more expensive compared to the APU in my country. It might gimped the performances in the long run but that's the price you paid to game now instead of waiting for God know how long for the GPU price to go down drastically.
Not so weird if people a building an SFF pc which can only hold an APU (Like Asrock Deskmini), so you wanna try to get as much performance as possible from that iGPU
@@STORMFIRE07 This reminds me of all the times I've been asked for recommendations in regards of PC's for people. They'll ask you and give all these limitations and in the end, none of them are actually THAT important. If you're gaming on an Asrock Deskmini, you're either gonna live with whatever performance you get, and be happy to play Fortnite and Sims, or you fork out the money and get a decent _slightly larger_ PC, maybe just bolt it under the desk or something. These limitations are unrealistic and you'll never be happy with the limitated result you'll get out of them. Paying enourmous amount of money for ram is like giving a ready-to-be-scrapped car a paintjob and go on driving it around. It's unsafe, poluting and just a bad time all around. You'll never be able to play Cyberpunk (or whatever) at 120 fps in that thing anyway.
@@stefanejegod8644 But what if people gonna want to improve upon that performance, yes, this use case scenario is really niche but it exists, and some people use it for that, some use it as a tiny server, retro games console, or media player for TVs, because no other CPU in the market pairs a great CPU and GPU in a small package
@@STORMFIRE07 that sounds like a.weird thing to me, To want to play on a deskmini. Most people interested in an Apu for gaming are constrained by budget, not space. I assume there's only a tiny amount of people willing to sacrifice performance for a super small form factor on the same budget for gaming.
Hello Hardware Unboxed. Great video as always! Since you mentioned at around min 3:30 you did testing with more than 2G Frame Buffer Size (8G), did you increase the ingame graphic settings to medium or high until the preset would actually exceed the 2G (max default) VRAM allocation? Would an increased graphics preset (medium/high) lead to decreased frame rate before the VRAM allocation matter, since the integrated gpu has only little ROPs / TMUs?
i’m curious too about what is stated at 3:30 .. forgive my ignorance, but could allocating 8gb ram to gpu still be effective for app like Davinci Resolve (or premiere or after effect i guess) requiring 4 or 8 gb on dgpu?
If you are going to get a dedicated GPU later it could make sense. If you have the budget for it maybe go safe with a 3200CL14 B-die kit. When i tuned my 3600CL15 B-dies to 3800CL15 (1.45V and tightened about every setting) I saw like ~17% improvement in non-GPU bound scenarios with my 3700X. I doubt they could get anywhere near, but sad I didnt compare it to the cheap 3000CL?? D/E-die (samsung) I had before.
Hey Steve, frequency isn't everything, as you have just proven. But Average FPS also isn't everything when it comes to results and by extension the gaming smoothness experience . From our internal testing at my work, we have found that higher frequency+high Lat leads to repeatable stuttering in games vs lower freq+lower lat. As latency for APUs is more critical than pure frequency. Its all about response time for instructions and draw calls etc etc.
Also when u buy a cheaper memory w worse timings, if your mobo can up the voltage you can in most cases get performance of nastly priced kits w cheap ones, i got 1600 ddr3 back in 2015, it ran at 1866 w/1-2 steps better timings than stock, all i did was upped the voltage from 1.35 to 1.5, the memory doesnt even have passive cooler and it never overheated, still working, most kits nowadays have nice passives on em
04:38 CL18 at 2200MHz is not looser than CL14 at 1600MHz. It's very slightly tighter. Only the tRCD and tRP timings are looser. tRAS is also tighter. How loose or tight a timing value is depends on both the cycle count and the frequency.
@@europason2293 Well at that point, just get a 10th gen i5/i3 with a used 1060 or 9xx card, as per recommended. To me, this APU only makes sense if you're building an ultra compact, and/or you have some DDR4 lying around.
@@HodgePodgeProducts But as he showed the review kit performs exactly like the affordable 3600/3800 kits people typically pair with Ryzen. Personally I'm running a 3800 kit at 3200 with timings adjusted accordingly because that was cheaper than a low latency 3200 kit.
As this review (and my own experience) shows, increasing RAM speed alone will not result in significant FPS increases on these APUs. Once you get above 3600MHz the limiting factor is actually the integrated GPU clock, so be sure to try overclocking that if you are running higher RAM frequencies.
This video is probably addressing those APU owners who are more into making a really small APU only gaming PC, and wondering whether faster memory is worth the extra cost, turns out it isn't
Steve... one more good video... now OC the IGPU to 2300mhz and combine it with 4000mhz cl16, and compare it with 3200 cl16 (normal kit) and default gpu speed. I do also use de 3200 cl14 B-Die for all NON monolitic Zen 2+ and Zen 3 cpus. For the Monolitic there are advantages because the memory controllers (there are 2 on the APU´s), are redesigned and allow more 1:1 leverage with faster Ram. That is what helps the IGPU... on the CPU side it also helps but not so much as in the GPU part of the CPU. Best Regards,
Unfortunately we don't know what exactly you tested, except you used a 2x8Gb kit @3600Mhz, which is most likely single rank (but there is a ver 4.24, so it could also be dual rank). Probably the dual rank kits are 2x16Gb, but we don't know that either. Other video's show that does make a difference, however small.
The real question is who the F buys expensive memory? spend the money on a cpu / gpu and buy basic ass ram. ffs (that is if you can only afford an APU)
Since I mentioned *PowerShell* here's one way to check your memory. This is just a single command, i.e. not a Script. Copy/paste that to the PowerShell, by clicking the Right Mouse Button. It's just one *Get-* command for *win32_physicalmemory* so nothing scary there. As default, it shows everything in Bytes. That's why I added those conversions to GB's and rounding, too. Short and simple, but much more lines: *Get-WmiObject* *win32_physicalmemory* After some tweaking: *Get-WmiObject* *win32_physicalmemory* *|* *Select* *Manufacturer,* *PartNumber,* *Banklabel,* *Configuredclockspeed,* *Devicelocator,* *@{Name="Capacity (GB)";Expression={[math]::round($_.capacity/1GB)}};* Manufacturer : G-Skill PartNumber : F4-3600C15-8GTZ Banklabel : BANK 1 Configuredclockspeed : 3600 Devicelocator : DIMM_A2 Capacity (GB) : 8 Manufacturer : G-Skill PartNumber : F4-3600C15-8GTZ Banklabel : BANK 3 Configuredclockspeed : 3600 Devicelocator : DIMM_B2 Capacity (GB) : 8 DIMM_A2 is the furthest slot from CPU. Underneath, A and B refer to two memory channels. So, my two memory sticks are located just like Motherboards manual instructs, i.e. in slots A2 and B2. Thus using both memory channels: ,-------------------------------------------------------------------- / | | ,- BANK 1 ||||||||||||| DIMM_A2 |||||||||||||||| | Ch A | '- BANK 2 DIMM_A1 | | ,- BANK 3 ||||||||||||| DIMM_B2 |||||||||||||||| | Ch B | '- BANK 4 DIMM_B1 | | | ,----------------------, | | | | | CPU | | | | | '---------------------'
Now I wanna see these numbers with tuned subtimings, the higher frequency you run the looser subtimings will run on auto, which means you get even more improvements from tightening them
Doesn't really matter if it boosts performance by 30%, you still have to buy premium high frequency memory and then tune it. But it likely won't make much difference in most games.
Now I've done testing with the 11900k iGPU and seen ~25% boost from memory OC alone, with DJR, your 3200 14-14 kit will easily clock to 4400 and higher
Time to make this thread longer then it needs to be 🙃 but you can also easily make 3200 outperform 4000 (kept on auto subs) with manual tuning. Your 3200 c14 kit literally has the same ICs as the 4000 and 4400 kits, they're just binned slightly better, you can achieve just about the same absolute timings (e.g. 7.5ns tCL) and same subtimings across them
I don´t understand why people actually care about FPS roundups with integrated graphics. iGPUs have the purpose of allowing you to run a PC without a GPU in order to perform BASIC non-gamer activities, such as work or for internet (f.ex. watching You Tube videos). Or to have a backup GPU for basic needs in case your GPU dies. That´s it!
Measuring CPU performance in FPS is like measuring water temperature in eggs boiled per minute. A simple FPS number is all "gamers" want, and good on you for giving them what they want.
![BLACK BAG - Official Trailer [HD] - Only in Theaters March 14](http://i.ytimg.com/vi/Du0Xp8WX_7I/mqdefault.jpg)
Can we get a Hardware Unboxed/Gamers Nexus crossover where the Steves are throwing it back to each other with the Intel "Back to you Steve" LOL
This
would be great for april's fools.
Needs to happen.
Would be so funny 😂
Oh please. Maybe do it on an unexpected subject like GN Steve reviews monitors with you and there is a monitor with monitor Steve visible in the background.
BUT 3600 NUMBER HIGHER THAN 3200 MEAN FASTER BETTER
You make a strong argument, big number is better.
@@Hardwareunboxed "I don't care if the extra price-to-performance would mean you should instead get a dedicated GPU, I DON'T WANT TO!" And as they died they'll be screaming "logic doesn't belong in benchmarks, only the numbers......"
Me after buying 3600Mhz ram: "WHERE FPS?"
Imma gonna buy teh highest CL memory i can find!1!ONE!
@@zenstrata yea, get 5% increase in 10 games then you get a 50% performance increase. BIG BRAIN TIME!
Oh, had been a long time since a responding to comments 😂 Just finished GNs last video so again : THANKS STEVE 😌😄
My classic Steve to Steve morning routine
Back to you, Steve)
lmao that thumbnail got me; cheers to y'all for taking these comments and turning them into an opportunity to inform and educate and provide interesting content
I would say 3-15 percent is a huge difference so what did we learn? Exactly what everyone said, faster memory increases performance and they left it out of the original review.
@@hill160881 I.. didn't make any statements about the veracity of the comments in the thumbnail, or the validity thereof. Please try not to insert your own subtext in other's comments. I complimented HUB for taking comments that appeared quite antagonistic for the purposes of the thumbnail and using them to create additional content to inform users. And that's it, really.
@@hill160881 from 29FPS to a whopping 33FPS for a $220 32GB DDR4 4000 👍🏼
Don't kink shame people for jizzing their pants when they see them 1-3 fps increases
@Steps "You are not the clown, you are the whole circus!"
That look in the reflection I'm dying lmao
When he looked at the camera, the "alert" sound of MGS ringed in my mind.
😂😂
2:25 that face of surprise when he see his reflection XD
Yeah, you brain gets weird when you're focusing on the shot through the lens. Or in Steve's case, I'm guessing live view on the back of a camera.
Then you see a reflection of yourself and you straight away look at check that it is you for some reason.
@@Mach1048 i think he was joking but idk
LOOOL it cracked me up, Steve is legend
If you think through how reflections work, he'd be seeing the reflection of the camera lens not himself, so yes very much on purpose
Commenters logic : Buy an APU and expensive DDR9X to get max fps.
Yeah... That was the problem....
If only we could have GDDR5X VRAm as RAM xd
I bought an APU and manually overclocked RAM (Crucial ballistics 3200 c16 to 4133 18-22-8-18-44-66 tRRD 4-6 tWTR 4-6 tFAW 16 @ 1,35V). Unfortunately FCLK only goes 2066 on my 4650g so 2066 it is (1:1). Why spend extra for expensive memory when cheap Micron Rev E can max out the CPU and outperform expensive memory?
@METhOdhowtoKilLnoObS k then... HBM2 ... that should do it xd
HUB: an increase of 9%
Me: You mean 3 fps.
Strong 33fps gaming, bruh, the power of console for PC. 🤣
Absolutely. Overspending on memory gives you placebo at best.
You can feel the difference between playing at 27 to 33 fps ans playing at 30 to 36 fps.
It's certainly a small improuvement tho
@@MasoMathiou like this is playing a little less shite now, but it’s still pretty shite if I’m honest…
Percentages give a better picture. 3 FPS improvement over 30 is way different to 3 FPS over 90
Randoms: "You don't understand APUs!"
HUB: *laughs in Cost-per-Frame-speak*
No one made a cost to performance argument though. You know what it's called when you misrepresent an argument so you can knock it down? Look it up.
@@Berserkism I just made it for you mate, cheers
@@Berserkism I mean given the current economy if you aren't starting and ending the conversation with that in mind you're being disingenuous from the onset...
Your review: "This is a generic setup, different setups will give marginal differences..."
RUclips: [Gnnnnrrrrr noises]
Your new review: "Marginal differences"
Pretty accurate depiction of the average RUclips audience 😅
people who cried over this are usually the ones buying 3090s off miners/scalpers for 1080p gaming.. they just dont have a clue about whats going on
You’d think they would listen to people who have used many systems. Haha
And this boys is how you get a reviewer to do your testing for you :)
The nicest way of clapping back at the people that can't test on this level but say you're doing it wrong.
I can and do test at this level, hes doing it wrong. Stable 2200mhz infinity fabric clock is impossible
AMD spec for this chip is 3200 so putting faster RAM is is overclocking, so "your millage may vary" comes into play. Paying a lot of money extra, getting the high end RAM for those couple of fps seems like a bad idea to me.
Such high end memory only makes sense coupled with a high end CPU.
I would consider 3200 cl14 high end memory as well.
@@aos32 the lowest end cpu one should get with that memory is 5600x, even that may be a stretch..gn showed you can basically equal a 10900k in gaming with 5600x paired with ddr4 3200 cl14 with memory and cpu tuning
And I'm crying over here 'cause my 5900x will only go up to 3733Mhz 1:1 :( but it is cl 15 tho
5700g is a "waiting" tools to me, bought highest ram speed to get ready system when gpu craziness over that is investment...sorry my english.
Love this kind of "mythbusters" type content. Keep it coming!
Your first video was great already and the follow up is welcome, it shows how much makes sense to spend in RAM, the only point that I think was missed is the purpose of this APU and the 5600G, this are not meant to replace GPUs, they are meant to allow high performing tiny PCs, the kind that you can put under your TV for 720p gaming or 1080p in old titles, to have as a desktop when you don't have much space and gaming is not your main workload.
Thank you for testing this! I was curious, in the end I was slightly let down to say the least
I think most important thing to keep in mind is to use dual channel, than anything else, there would be some performance difference with high Mhz, but not much after a certain point.
After like 3600mhz there is barely an improvement from what I saw, and after 3000-3200mhz the gains get not that large.
The speed in RAM isn't the frequency itself but the combination with tight timings (and not just CAS : other timings and even subtimings can be fine tuned aswell to make a difference)
Unless you're playing Linpack Benchmark as your favourite game it's not about speed at all. It's all about latency. Games don't require much bandwidth, but they are heavily branched so quite often the CPU comes to the "You wanted to do THAT?" situation where you miss all branch prediction and all caches, and the only thing that can expedite things is the shortest possible delay to fetch from memory. This is why low latency RAM helps measurably with 1% lows. When talking APUs the problem is just doubled. Now you have two things that can miss caches and need to fetch something ASAP. But even with APUs it's only 10% bandwidth (speed) and 90% latency. If you could get a 3200 kit to run CL12 it would run circles around a 4000 CL18 kit.
I hope you enabled full spectrum RGB on your test system as everybody knows that can give you up to 10% performance uplift across the board.... ;)
Weird how they stopped painting flames on everything on top of rgb for another 10% extra performance 🤔
Red makes it go faster ... yellow makes it boom bigger
Everyone is a expert. It is called Dunning-Kruger effect.
@@jediii86 I can predict that you are a white supremacist that thinks the "white" (whatever that is) race should rule the world.
You believe that you have the right to shoot anyone that you don't like.
You own at least three weapons and constantly look for ways to be "prepared".
You think all that you don't like in your life or your surroundings is the fault of "Democrats", immigrants and "socialists" though you don't know anything about them or what they are.
Trump is your idol. You also want to be at least as racist as him and you envy his endless grift, recknessless for the law and his "happy-go-lucky-know-nothing-approach".
You think science is a dirty word and/or socialist, same with education unless it proves your point of view.
You work a menial office or blue collar job and you have never been out of the country except for maybe Mexico or Canada.
You live in the suburbs or a rural area.
All your friends are white.
You speak either only english or if you live in the southern part maybe you know some spanish.
You think homeschooling is the best.
To you "Budweiser" is a real beer.
You drive a pickup or similarly sized car.
Your interest in the world doesn't spread further than the next baseball/american football game.
You *think* you are a christian but never or seldom go to church and you certainly don't live according to the ten commandments.
You *LOVE* stereotypes.
@@jediii86This reply is very ironic given what the comment said...
@@jediii86 wow... That was perfect. The perfect example of Dunning Kruger effect. I am in my mid 40s. Married with kids with no other relationships ever. A medical professional interested in computers and science. One thing you got right is that I support BLM movement and I am a leftist Marxist who is a member of the Communist party of India and lives in a township and work in rural parts of south India. May be this info can be used to harass me later. Thanks.
Thanks for a follow up to this! This will help APU users in getting an idea of what performance to expect on different memory configs. It's crazy to see how high you got the speeds while maintaining a 1:1 FCLK!
Something is wrong here
Its impossible to achieve 1:1
With ddr4 4400
@@Adel-sj4ue Yeah, you are wrong, cause it's totally normal for the 5000G APUs.
@@chovekb yes you r right
I searched and found that 5000g differ from the 5000
Thanks
@@chovekb do u have good experience in overclocking???
I have some questions and need an experience one to help??
@@Adel-sj4ue No. You can look for some "hardcore overclocking" channels xD There is one pretty good Actually LOL
The problem is that most RAM brands don’t disclaim if it is Single or Dual Rank or change it after reviews of this RAM
You can take the SKU and search it on the QVL, these generally state whether it's single or dual rank. All 8gb sticks are SR and all 32GB sticks are dual rank. 16gb sticks can be either.
Don't buy ram from off brands and you won't have these issues. Every major brand had this information on there website. Just because it's not on the newegg or Amazon product pages, doesn't mean the information isn't available.
Reminds me of back when ryzen originally came out I specifically sought out one of the few 3200 samsung b-die sticks that was supposed to be overclockable on ryzen. By the time I ordered mine, it was swapped for hynix despite all the professional reviews and customer reviews at the time stating it was samsung.
@@reto. this is not exactly true, older 8GB sticks (~
@@joe_ferreira Doesn't really work like that. Even major brands swap memory IC's all the time. Buy a Corsair kit for example and it could come with any of a dozen different chips inside depending on the week of manufacturer.
Steve, when you review the 5600G, can you included results with the iGPU overclocked? I understand why OC results aren't normally included these days, but other reviewers have shown getting an extra 15%-20% fairly common and not very difficult.
Will do, thanks for the suggestion.
@@Hardwareunboxed Thanks for all your hard work!
@PC_Modder yep, that would be really helpful - since it feels no matter what CPU or GPU you use, the Fusion 360 renderer is horribly slow :)
@Tilapia Aquaponics If you buy a Ryzen APU you don't want a dGPU. I have several systems like that.
@@Grimmwoldds I think he was talking about gaming focused builds. Obviously if your use case only stresses the CPU, and the iGPU is just a display adapter, it changes the value equation significantly.
"oh you think we tested with bad RAM?"
*whips out tray of RAM and Dom Plats*
I like this test video! Searched a lot for such a comparsion. Keep on going sometimes with DDR Tests for grafics!
There's performance to be gained from premium memory, but I don't know how much performance people are expecting. Thanks for yet another brilliant video thoroughly outlining what to expect at different price points.
@@tilapiadave3234 and trolls are trolls no matter what.....
Appreciate the work done to go through this! Thank you
Hardware Unboxed BTS "Y'all wanna whine about RAM?? OK well let's talk about RAM..." Proceeds to whip out massive kit of RAM!!
This video is wrong and full of lies, but this channel relies on your ignorance and it all works out.
@@BlastinRope 100% and its sad
Really appreciate that you reviewed this CPU in this aspect.
So same as the consensus has been for years on memory: buy the most affordable kit you can get at decent speeds (usually 3200 CL16 or 3600 CL18), and if you want more out of your memory use something like DRAM Calculator for Ryzen to manually lower the timings.
I agree. Lower timing is better for stability.
dont get 3600 C18 kits, subtimings are usually dogwater and they cost as much as a Patriot Viper Steel B-Die 4400 CL19-19-19-39 kits, meanwhile Corsair 3600 CL18's subtimings are CL18-22-22-22-4x, only 10-15$ difference but the difference in subtimings and quality is really high
Agreed except dram calculator is unreliable for anything outside of b-die
@@rdmz135 true... Couldn't get my trident z Neo 32gb 3600mhz cl18 any higher than the xmp value... Even on perfect altered safe settings of the dram calculator. EVERYTHING failed, was dissapointed because my ddr3 ram kits did have a high range of tuning room.
3600 CL18 is a trash bin and you shouldn't use DRAM calculator.
I’m really grateful for the follow-up! Memory scaling was one of my questions, though I’ve watched HUB long enough to know the memory you use is better than average.
Get em Steve! :) great Stuff as always
Great follow-up review. Thanks!
As a viewer, I can speak for all of us (we’re good at that) and say that we have no idea what we’re talking about, but everything you do will always be wrong and right simultaneously.
Heisenberg's Interwebz Review Principle.
For all the people raging, I'd love them to find a single person actually buying both an APU and the expensive RAM. Reviews should reflect what the average user of that hardware should expect.
the average user doesnt overclock anything
the average user doesnt tweak ram settings
benchmarks should be from a properly setup system,
overclocked cpu, ring/cache with properly tweaked ram as fast as the setup can handle
only morons like the standard "benchmarks" that the mainstream do nowadays
You'd do it for building a custom mini-PC. High end APUs + SODIMM RAM are practically made for stuff like TV boxes, custom portables, and anything else where there just isn't space for a bulky GPU. It's a niche, but I'd say in tech reviews, one worth considering given that these applications are just about the only ones where you basically need a good APU.
I did bud
At this rate, the CPU could be the cheapest part on/including your motherboard.
Memory recently became a big topic of "leaving performance on the table" in recent years since Ryzen's L3 cache performance is impacted.
Also, laptop manufacturers are sneaking in 1Rx16 modules with terrible secondary timing Which can be significantly improved by changing to 1Rx8 or 2Rx8 modules with identical speed and timing
Crucial Ballistix 3600CL16 seems like the no-brainer option for Ryzen systems right now. Budget friendly and can OC above their spec depending on silicon lottery.
Not even depending on silicon lottery! Both Micron Rev. E and Micron Rev. B reliably overclock well, and those are the only two dies Crucial uses in its Crucial Ballistix 3600C16 kits.
@@daboross2 Well right. But what I am saying is your OC limit with any of those kits is going to vary from CPU-to-CPU when paired with Ryzen.
@@ExoTraveler fair enough! That is true.
Rev. E is solid, I ran my kit at 3800 CL14-19-14-28 @1.55V. Rev. B is even better.
@@-eMpTy- 1.55?! Wow
You know things are getting serious on Hardware Unboxed when the hoodie is on the thumbnail.. bad ass!
Love the reward for watching and not listing like a podcast at 2:25 :D
Great Content!!!! Content and hard work is the key to success. I have following this channel since it had nearly 300k to 400k subs. If Steve and his friends keep up this pace they'll surpass 5 Million subs easily in no time. Until today I don't know single, dual, quad rank memory thanks to your video, I got to learn from it.
Who buys 4000mhz RAM for an APU? Some elitists aren't happy.
Yup. Spend more on RAM than APU that nonsense on most case. But maybe if fhe user a PS giant poster designer maybe but the RAM size whould be more prior than the speed. :)
pretty debatible.
I can get 4400mhz for loading times and such.
I currently has gskill trident z 4000mhz and works wonders.
@@maestrohun b die can be found for what like $30 more than 3200 c16 xmp, wtf are you talking about, 15-20% when tuned for $30 is fucking insane.
cope
What I got from this review is that Hardware unboxed still knows what they are doing with hardware and does an excellent job as always while keeping the gamers pocket book in mind.
Crucial Ballistix 3600 MHz kits hit CL14 with 1.45v so would be cracking value for money as the white and red are £71.99 currently in the UK. (local markets will differ)
Crucial random dies bruh.
you may end up buying a low-mid tier chip.
@@claritoresdiano1021 Not had any issues with the kit my sons uses and I've just picked up the 64GB (2x32GB) kit which also does C14. My 5950x has just been replaced as the IO chip was dying. (dodgy November patch i think)
@@claritoresdiano1021 link to facts
@@fracturedlife1393 can't share link blocked by auto spam
Great follow-up piece, my dude!
Thanks for responding to pur comments Steve! This APU is incredibly flexible for people who wants a temporary solution and loves turning. Memory scaling is sure an important topic for APU since the iGPU is always memory bandwidth bounded. That’s also one of the reasons why Renoir and Cézanne only have 8 CUs instead of max of 11 CUs in Raven Ridge and Picasso.
BTW, there are also cheap above 4000Mhz DDR4 out there with Samsung B-Die. We can get 2x8GB from Patriot Viper Series at 100 Euro/Dollars and they are good enough. The advantage of APU is that the SOC is in 7nm and could accept 2000-2067Mhz FCLK easily compare to 5000 series. This compensate the lower L3 cache by having lower memory latency. So higher mem frequency, tider mem subtimimgs, high 1:1 FCLK clocks, higher CPU clocks all help with better memory latency. In best case scenario, I assume that APU could be better than CPU. I have a 4750G and tweaked it a lot together with memory. The max potential of it really sits at 3800x level when high L3 size is not needed. Specially, 4750G has a 88W socket limit which the 5700G doesn’t have. You can PBO the 5700G crazy to the 5800x level and still having the efficiency in other scenarios. Jesus I love 7nm APUs.
My patreonage once again proved its worthiness! Thanks a lot for the great content!
It's great to see the difference memory makes for APUs, and also how far integrated graphics has come in recent years.
Running AAA games at 30 FPS & 1080p without a dedicated GPU seems like witchcraft, and I hope AMD continues to make strides here.
The next gen am5 ones are gonna be the big jump imo. I'd love one of these as a very portable tiny form factor gaming system.
@@frankytanky5076 One day they will be inbuilt into your VR headset & you can watch Klaus Schwab telling you the bugs injected into your stomach are all part of the clotshot game of Matrix.
Seems more underwhelming than 'witchcraft' when you see the consoles are the same thing, same with laptops.
I found running my (3200 MHz CL-14) G-Skill kit at 3600 MHz CL-16-16-16-40 to work rather well. I still need to tweak the settings a little, but going over 3766 or something like that had a negative effect on the FCLK scaling, effectively making the end result slower
It still would have been cool to see the actual numbers for DD4 3600 16-16-16-36, though, since that has been shown to be a pretty nearly optimal solution by a few other reviewers (when paired with other Zen 3 CPUs), without being ridiculously expensive.
As always in depth analysis 👍🔥
Lesson Learned: Don't buy "Premium" RAMS on APU's.
Right? It would only ever make sense ir you are absolutely set on building the smallest posible gaming machine. But I've never understood that either, as a second hand gaming laptop would be faster, smaller and cheaper.
3200MHz C16 is fast enough to give APU graphics a boost, but not so expensive that you could have bought a discrete GPU for the cost of the RAM
why not? 10-20% increase for max $40 is bonkers lmao?
Getting closer to 1 million. Keep up the good work :)
Short version. You want a good mix of tight timings and high frequency
Dual rank as fast and as tight as you can make it. Remember the tertiary and secondary timings as well. This is a monolithic die If should be able to get to 2000 in 1:1 really easily.
@@NotThatGuyJD quad channel for Ryzen. It's also very important you make sure your infinity fabric timing is lined up to your ram Freq for Ryzen eg 3800mhz ram and 1900mhz infinity fabric
@@AllThatJazOfficial Ryzen is dual channel my man. You could get Quad rank but you'd need 4 dual rank sticks as rank is defined per channel and not for the whole system. Yeah I know but these monolithic APUs have much higher IF clock ability and memory controller than the chiplet based CPUs as such can clock higher with good memory sticks than the non APU
@@NotThatGuyJD yeah they do like 2200 1:1:1 with 4400mts mem.
I'd like to pick one up literally just for benching. It's a use case...but not a useful one lol
@@fracturedlife1393 yeah they're pretty great for benching, especially the 5300g since it's got much less competition for cpu benching and any of them are great for mem benching
Gave the game away in the second minute. CL14, that stuff is expensive. Good review, well made. I am very interested in seeing how the APU's perform as streaming boxes for steam, xcloud and how frames increase with FSR creeping in.
keep up the good work.
How much do they expect from Vega 8 lol?
Now that you've done a whole video to showcase it, YOU CAN LITERALLY SEE IT!
I expect professional reviewers to do something as simple as test a reasonable OC with quality RAM. Well, I don't expect it from HWU...
I enjoyed seeing this. It's good to know how much of a difference things can make.
Steve's about to destroy some RUclips commenters like he does Fortnite players. 🤣👍
Less satisfying than a max pump or snipe though :D
@@Hardwareunboxed 🤣 Gold Heavy Sniper FTW. 😁👍
@@MafiaboysWorld now you're talking :)
@@gamingunboxed5130 Only because it's my weapon of choice, double snipe load out with heavy and boltie. 😂👍 But I am quite partial to the silenced sniper when sniper shootout is on. 😈👍
So, if one were looking to build a "semi-budget" partial gaming/partial video editing PC in the midst of this GPUpocalypse - would you steer them towards a 2nd hand GTX 1060 and the iCore 5 10400 or the 5700G and some decent 3200 ram with the potential to upgrade the GPU later in life, if things ever calm down?
DDR4 3200 --Mhz-- CL14 Samsung B-die is some of the _best_ memory you can get.
Geez...
*3200 MT/s, DDR RAMs run at half of its transfer rate because it's Double Data Rate.
A 3200MT/s Memory has a memory clock of 1600MHz.
See Dr Ian Cutress' RUclips Channel for further explanations about this ruclips.net/video/5fZO77I-6Cg/видео.html
@@erlienfrommars That is correct. The same way what we call Molex and 24 pin ATX connectors aren't. They're the mini-fit jr. or some such.
Very few people care. Everyone knows what it means and _that_ is most important. This is what people accepted as the name and it doesn't have to be "correct."
Same is with GB and GiB. How often do you hear people going around mentioning gibibytes? Does anyone care? No? Exactly.
But fine, I'll call it what the manufacturer labels it.
Including tests like this is extremely helpful as you can find amazing deals on ram kits. I was able to get Viper Patriot 4400MHz 16gb kit for 120CAD(around 100USD) and I run them 3800MHz CL14
based af
those are some of the bang-for-the-bucks RAM ever
That DDR4-4000 DR is really something. 7% to 15% is not shabby. It's a common Intel generation upgrade.
Why are some people so biased with tech? It's not like their are winning anything by siding with one of the brands. This channel just shows you test results & comparisons, it's not propaganda. Chill everyone. Thank you for your hard work, Steven. It's really appreciated.
Imagine buy very expensive ram to maximize the apu rather than buy some cheap dgpu like 1050 and pair it with decent ram for same price
Case closed!. Excellent video!
Steve how much weight have you lost? Looking good dude
no surprise that people making comments don't have the knowledge this channel has on memory specs and usage.
When I look at the difference b/w price and performance it's totally not worth the upgrade. Why upgrade your 3200 mhz ram just to gain a 3-5 fps at the cost of $100? I know some people just want the best but seriously can anyone tell the difference b/w 60fps-65fps or 150fps-160fps.
You're better off buying a $100 used GPU at that point.
You can easily tell the difference between 65 and 120 after 140 its very difficult
Probably depends on the person. I most likely wouldn't be able to tell. It took me a long time to adjust to where I could tell the difference between 30 and 60 fps, but I can imagine people with better eyesight than me might be able to tell right away going to 120 fps.
I have serious doubts that 240Hz+ monitors are actually meaningful upgrades to anyone though.
@@obake6290 linus already proved it is actually our instincts are lot faster then we think on fps point i played lot of counter strike on 150-160 and when you switch high to low you can notice but if you started with consoles 30 fps then get to 60 or more probably its difficult to know the difference but someone who spend most time playing low res high fps can easily tell the difference
@@SoftnappGAMEPLAY Bro look what I said... 60-65 or 150-160. I didn't compare 65 to 120 or 140 it's obvious u can tell the difference b/w those. What I'm saying it's not worth it to pay $100 more for 3-5 fps since u can't even feel or tell the difference so basically people just need to be smart and save their money on something else.
Great comparison👍👍👍😍
Seems like a weird choice, to spend so much in RAM with an APU, which usually means people are on a budget... (and buying a 5700G to pair it later with a discrete GPU seems like a bad choice too with those 16MB of cache)
For me personally, it's a temporary stop gap since gpu is so expensive. The recommendation to get I5 10400 with cheap 2nd hand GPU doesn't work around the world since that GPU+CPU combo is still more expensive compared to the APU in my country. It might gimped the performances in the long run but that's the price you paid to game now instead of waiting for God know how long for the GPU price to go down drastically.
Not so weird if people a building an SFF pc which can only hold an APU (Like Asrock Deskmini), so you wanna try to get as much performance as possible from that iGPU
@@STORMFIRE07 This reminds me of all the times I've been asked for recommendations in regards of PC's for people. They'll ask you and give all these limitations and in the end, none of them are actually THAT important. If you're gaming on an Asrock Deskmini, you're either gonna live with whatever performance you get, and be happy to play Fortnite and Sims, or you fork out the money and get a decent _slightly larger_ PC, maybe just bolt it under the desk or something. These limitations are unrealistic and you'll never be happy with the limitated result you'll get out of them.
Paying enourmous amount of money for ram is like giving a ready-to-be-scrapped car a paintjob and go on driving it around. It's unsafe, poluting and just a bad time all around. You'll never be able to play Cyberpunk (or whatever) at 120 fps in that thing anyway.
@@stefanejegod8644
But what if people gonna want to improve upon that performance, yes, this use case scenario is really niche but it exists, and some people use it for that, some use it as a tiny server, retro games console, or media player for TVs, because no other CPU in the market pairs a great CPU and GPU in a small package
@@STORMFIRE07 that sounds like a.weird thing to me, To want to play on a deskmini. Most people interested in an Apu for gaming are constrained by budget, not space. I assume there's only a tiny amount of people willing to sacrifice performance for a super small form factor on the same budget for gaming.
I'm glad you did this because people were throwing a hell of a fit in the comments
oh c'mon it's still Vega, there's only so much lipstick you can put on it.
Wait(TM) for Rembrandt.
@@vncube1 Wait for Steam Deck
@@vncube1 by the time it comes out there will be cheap gpus available
@@raresmacovei8382 Steam Deck sadly uses Zen 2
@@saricubra2867 no way Zen 2 will bottleneck a few Watts of GPU power
Hello Hardware Unboxed. Great video as always! Since you mentioned at around min 3:30 you did testing with more than 2G Frame Buffer Size (8G), did you increase the ingame graphic settings to medium or high until the preset would actually exceed the 2G (max default) VRAM allocation? Would an increased graphics preset (medium/high) lead to decreased frame rate before the VRAM allocation matter, since the integrated gpu has only little ROPs / TMUs?
i’m curious too about what is stated at 3:30 .. forgive my ignorance, but could allocating 8gb ram to gpu still be effective for app like Davinci Resolve (or premiere or after effect i guess) requiring 4 or 8 gb on dgpu?
I was curious about this as well - especially with games like Doom Eternal and higher texture settings.
2:26 AHHHHHH!
Thanks for doing the hard yards Steve
Also, who uses a 5700G but then pairs it with super expensive, high-end memory just to get the most performance out of it?
If you are going to get a dedicated GPU later it could make sense. If you have the budget for it maybe go safe with a 3200CL14 B-die kit.
When i tuned my 3600CL15 B-dies to 3800CL15 (1.45V and tightened about every setting) I saw like ~17% improvement in non-GPU bound scenarios with my 3700X.
I doubt they could get anywhere near, but sad I didnt compare it to the cheap 3000CL?? D/E-die (samsung) I had before.
nobody lmao, but sane people do buy cheap b die for $30 more which tuned performs (10-20% better) but theyd never show this in a video
Hey Steve, frequency isn't everything, as you have just proven. But Average FPS also isn't everything when it comes to results and by extension the gaming smoothness experience . From our internal testing at my work, we have found that higher frequency+high Lat leads to repeatable stuttering in games vs lower freq+lower lat. As latency for APUs is more critical than pure frequency. Its all about response time for instructions and draw calls etc etc.
That RAM collection is so satisfying to look at
Also when u buy a cheaper memory w worse timings, if your mobo can up the voltage you can in most cases get performance of nastly priced kits w cheap ones, i got 1600 ddr3 back in 2015, it ran at 1866 w/1-2 steps better timings than stock, all i did was upped the voltage from 1.35 to 1.5, the memory doesnt even have passive cooler and it never overheated, still working, most kits nowadays have nice passives on em
3200 CL14 is probably the most optimal kit for Zen 3.
Straight up samsung b die, what more can you ask
also great on my 2700x, I keep beating 3700x benchmarks
3200c14 running super tight 3800c15 is better lol
04:38 CL18 at 2200MHz is not looser than CL14 at 1600MHz. It's very slightly tighter.
Only the tRCD and tRP timings are looser. tRAS is also tighter.
How loose or tight a timing value is depends on both the cycle count and the frequency.
Comments: "You don't understand APU's"
Me: You don't understand common sense. Nobody is going to pair high-end $250 ram with a mid-range APU.
I mean, the 5700G is $360 and you can get a 2x8GB Patriot kit at 4400C19 for $150.
@@europason2293 Well at that point, just get a 10th gen i5/i3 with a used 1060 or 9xx card, as per recommended.
To me, this APU only makes sense if you're building an ultra compact, and/or you have some DDR4 lying around.
@@GENKI_INU for sure.
The review kit they used costs ~$275 so by that logic they shouldnt be using that either.
@@HodgePodgeProducts But as he showed the review kit performs exactly like the affordable 3600/3800 kits people typically pair with Ryzen. Personally I'm running a 3800 kit at 3200 with timings adjusted accordingly because that was cheaper than a low latency 3200 kit.
As this review (and my own experience) shows, increasing RAM speed alone will not result in significant FPS increases on these APUs. Once you get above 3600MHz the limiting factor is actually the integrated GPU clock, so be sure to try overclocking that if you are running higher RAM frequencies.
What happen to APUs being a budget option, lol.
AMD gave up, there was more money to be made selling rx 6700xts at 1299.
@@zen608 you know Amd dont make extra money from inflated retail prices right?
This video is probably addressing those APU owners who are more into making a really small APU only gaming PC, and wondering whether faster memory is worth the extra cost, turns out it isn't
Steve... one more good video... now OC the IGPU to 2300mhz and combine it with 4000mhz cl16, and compare it with 3200 cl16 (normal kit) and default gpu speed.
I do also use de 3200 cl14 B-Die for all NON monolitic Zen 2+ and Zen 3 cpus. For the Monolitic there are advantages because the memory controllers (there are 2 on the APU´s), are redesigned and allow more 1:1 leverage with faster Ram.
That is what helps the IGPU... on the CPU side it also helps but not so much as in the GPU part of the CPU.
Best Regards,
Imagine paying double the price for ram to get 2fps more. LMAO
Imagine sitting at your computer wondering what it would be like if each component was the best of the best
Congratulations almost 800k
2:27 👀
Telling it like it is. Appreciate ya.
02:26 🤣🤣🤣🤣🤣🤣
Unfortunately we don't know what exactly you tested, except you used a 2x8Gb kit @3600Mhz, which is most likely single rank (but there is a ver 4.24, so it could also be dual rank).
Probably the dual rank kits are 2x16Gb, but we don't know that either. Other video's show that does make a difference, however small.
lol buying expensive ram instead of dgpu. I see the logic
It cost less energy at the end.
This kind of content was why I subscribed for.
Next, can you expand this findings for the lower tier or older Ryzen APUs? You know, for science!
The real question is who the F buys expensive memory? spend the money on a cpu / gpu and buy basic ass ram. ffs
(that is if you can only afford an APU)
But but but 6% extra frames for 300 more dollars! Now that's performance
Those who bought 3090 and 5950x)))
Same people who buy expensive high end X570 boards for 3200g or ig now 5600g
Since I mentioned *PowerShell* here's one way to check your memory. This is just a single command, i.e. not a Script. Copy/paste that to the PowerShell, by clicking the Right Mouse Button. It's just one *Get-* command for *win32_physicalmemory* so nothing scary there.
As default, it shows everything in Bytes. That's why I added those conversions to GB's and rounding, too.
Short and simple, but much more lines: *Get-WmiObject* *win32_physicalmemory*
After some tweaking: *Get-WmiObject* *win32_physicalmemory* *|* *Select* *Manufacturer,* *PartNumber,* *Banklabel,* *Configuredclockspeed,* *Devicelocator,*
*@{Name="Capacity (GB)";Expression={[math]::round($_.capacity/1GB)}};*
Manufacturer : G-Skill
PartNumber : F4-3600C15-8GTZ
Banklabel : BANK 1
Configuredclockspeed : 3600
Devicelocator : DIMM_A2
Capacity (GB) : 8
Manufacturer : G-Skill
PartNumber : F4-3600C15-8GTZ
Banklabel : BANK 3
Configuredclockspeed : 3600
Devicelocator : DIMM_B2
Capacity (GB) : 8
DIMM_A2 is the furthest slot from CPU. Underneath, A and B refer to two memory channels. So, my two memory sticks are located just like Motherboards manual instructs, i.e. in slots A2 and B2. Thus using both memory channels:
,--------------------------------------------------------------------
/
|
| ,- BANK 1 ||||||||||||| DIMM_A2 ||||||||||||||||
| Ch A
| '- BANK 2 DIMM_A1
|
| ,- BANK 3 ||||||||||||| DIMM_B2 ||||||||||||||||
| Ch B
| '- BANK 4 DIMM_B1
|
|
| ,----------------------,
| | |
| | CPU |
| | |
| '---------------------'
Haters gonna hate!
HUB
When I saw that thumbnail, I knew this was going to be quite the interesting video!
Now I wanna see these numbers with tuned subtimings, the higher frequency you run the looser subtimings will run on auto, which means you get even more improvements from tightening them
Doesn't really matter if it boosts performance by 30%, you still have to buy premium high frequency memory and then tune it. But it likely won't make much difference in most games.
Now I've done testing with the 11900k iGPU and seen ~25% boost from memory OC alone, with DJR, your 3200 14-14 kit will easily clock to 4400 and higher
Stuff like Crucial ballisitx kits which are very cheap also OC very well since they only use Micron ICs today
Time to make this thread longer then it needs to be 🙃 but you can also easily make 3200 outperform 4000 (kept on auto subs) with manual tuning. Your 3200 c14 kit literally has the same ICs as the 4000 and 4400 kits, they're just binned slightly better, you can achieve just about the same absolute timings (e.g. 7.5ns tCL) and same subtimings across them
nice vid! DDR4 3800 & 4000 CL14 kits are out... any plan on using those in future vids? thanks
I don´t understand why people actually care about FPS roundups with integrated graphics. iGPUs have the purpose of allowing you to run a PC without a GPU in order to perform BASIC non-gamer activities, such as work or for internet (f.ex. watching You Tube videos). Or to have a backup GPU for basic needs in case your GPU dies. That´s it!
Measuring CPU performance in FPS is like measuring water temperature in eggs boiled per minute.
A simple FPS number is all "gamers" want, and good on you for giving them what they want.