• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

AMD Ryzen 7 8700G AI Performance Enhanced by Overclocked DDR5 Memory

T0@st

News Editor
Joined
Mar 7, 2023
Messages
2,077 (3.33/day)
Location
South East, UK
We already know about AMD Ryzen 7 8700G APU's enjoyment of overclocked memory—early reviews demonstrated the graphical benefits granted by fiddling with "iGPU engine clock and the processor's memory frequency." While gamers can enjoy a boosted integrated graphics solution that is comparable in performance 1080p stakes to a discrete Radeon RX 6500 XT GPU, AI enthusiasts are eager to experiment with the "Hawk Point" pat's Radeon 780M IGP and Neural Processing Unit (NPU)—the first generation Ryzen XDNA inference engine can unleash up to 16 AI TOPs. One individual, chi11eddog, posted their findings through social media channels earlier today, coinciding with the official launch of Ryzen 8000G processors. The initial set of results concentrated on the Radeon 780M aspect; NPU-centric data may arrive at a later date.

They performed quick tests on AMD's freshly released Ryzen 7 8700G desktop processor, combined with an MSI B650 Gaming Plus WiFi motherboard and two sticks of 16 GB DDR5-4800 memory. The MSI exclusive "Memory Try It" feature was deployed further up in the tables—this assisted in achieving and gauging several "higher system RAM frequency" settings. Here is chi11eddog's succinct interpretation of benchmark results: "7600 MT/s is 15% faster than 4800 MT/s in UL Procyon AI Inference Benchmark and 4% faster in GIMP with Stable Diffusion." The processor's default memory state is capable of producing 210 Float32 TOPs, according to chi11eddog's inference chart. The 6000 MT/s setting produces a 7% improvement over baseline, while 7200 MT/s drives proceedings to 11%—the flagship APU's Radeon 780M iGPU appears to be quite dependent on bandwidth. Their GIMP w/ Stable Diffusion benchmarks also taxed the integrated RDNA 3 graphics solution—again, it was deemed to be fairly bandwidth hungry.



View at TechPowerUp Main Site | Source
 
Joined
Nov 26, 2021
Messages
1,642 (1.51/day)
Location
Mississauga, Canada
Processor Ryzen 7 5700X
Motherboard ASUS TUF Gaming X570-PRO (WiFi 6)
Cooling Noctua NH-C14S (two fans)
Memory 2x16GB DDR4 3200
Video Card(s) Reference Vega 64
Storage Intel 665p 1TB, WD Black SN850X 2TB, Crucial MX300 1TB SATA, Samsung 830 256 GB SATA
Display(s) Nixeus NX-EDG27, and Samsung S23A700
Case Fractal Design R5
Power Supply Seasonic PRIME TITANIUM 850W
Mouse Logitech
VR HMD Oculus Rift
Software Windows 11 Pro, and Ubuntu 20.04
I wonder why they chose to use DDR5 6000 with relatively pedestrian timings; CL30 kits are available for the same price as CL38 kits. AI workloads are bandwidth hungry so I doubt it would have made a difference, but it would still have been a good data point.

1706728756682.png
 
Joined
Dec 7, 2020
Messages
61 (0.04/day)
Aren't they both GPU, not NPU, tests? Chart titles seem to indicate so; just AI is what is tested on the GPU.
Also, 210 TOPS at Float32 sounds a little bit over the top. Like, about 1 order of magnitude vs AMD slides, and I don't think they were sandbagging here. This looks like a benchmark score, not TOPS indicator.
 
Last edited:
Joined
May 3, 2018
Messages
2,881 (1.20/day)
Aren't they both GPU, not NPU, tests? Chart titles seem to indicate so; just AI is what is tested on the GPU.
Also, 210 TOPS at Float32 sounds a little bit over the top. Like, about 1 order of magnitude vs AMD slides, and I don't think they were sandbagging here. This looks like a benchmark score, not TOPS indicator.
Yes clear as mud. You can run these tests with just cpu, npu or gpu. PC Hardware did just that for Meteor Lake and found the igpu is significantly faster than Intel's much hyped npu. I presume it's the same for AMD as their NPU performance is similar although AMD's is a faster.
 
Joined
May 13, 2008
Messages
762 (0.13/day)
System Name HTPC whhaaaat?
Processor 2600k @ 4500mhz
Motherboard Asus Maximus IV gene-z gen3
Cooling Noctua NH-C14
Memory Gskill Ripjaw 2x4gb
Video Card(s) EVGA 1080 FTW @ 2037/11016
Storage 2x512GB MX100/1x Agility 3 128gb ssds, Seagate 3TB HDD
Display(s) Vizio P 65'' 4k tv
Case Lian Li pc-c50b
Audio Device(s) Denon 3311
Power Supply Corsair 620HX
Sounds about right.

I haven't checked reviews to see where the GPU sits max at stock or overclocked, but it appears to make sense for a tweaker for a light-weight budget build that finds that level of CPU/GPU acceptable.

I definitely think he just about hit the nail on the head with ram speeds. I'd likely shoot for ~7600-8000 decoupled and be perfectly happy with that considering the power of the GPU.

I would probably buy something like $130 7200mhz RAM and then either bump the voltage to ~1.45v (or less if possible) and/or and loosen the timings until whatever GPU clock you end up gives the best performance. While latency matters for CPU performance, I doubt people care THAT much for this product, while it shouldn't really effect the GPU which thrives on raw BW.

I mean, it's not GOOD, but it could work for someone in a pinch, or be a nice toy to play around with.

This is one of the reasons I'll wait for Zen 6. Not just because there will probably be interesting APU options due to SEDs and sufficient cache (+ whatever else they come up with), but that IOD really needs to be updated (not just shrunk to)to what's possible on 4nm so that decent ram can run 1:1 with the fabric. I would imagine 4nm could support 7600, maybe ~7800-8000, closer to the fastest available OC (8400mhz)?
Maybe not. We shall see.

Could be one of those things that ends up making the difference between sticking with 32GB (if fast) versus upgrading to 48GB+ (of more traditional 6000-6400mhz) for some peoples' use-cases.

Still one of the things I anticipate, however much better. AMD has always needed bandwidth, especially for APUs. I hope this will finally be solved shortly (on 3nm APU/GPUs even if through 4nm cache/MC).
 
Last edited:
Joined
Nov 8, 2017
Messages
229 (0.09/day)
Yes clear as mud. You can run these tests with just cpu, npu or gpu. PC Hardware did just that for Meteor Lake and found the igpu is significantly faster than Intel's much hyped npu. I presume it's the same for AMD as their NPU performance is similar although AMD's is a faster.
From the sildes that Intel made for MTL, it seems that it's expected that the OS/apps will use the best hardware for the task. Looks like sometimes the speed gains of the GPU might not always be worthy when you look at the efficiency. And if I understand the graph right, you can run AI stuff of both at the same time. Efficiency has always been the keyword for the NPU
1706770874729.jpeg

View attachment 1706770710476.webp
 

Attachments

  • 1706770568804.webp
    191.6 KB · Views: 22
Joined
May 3, 2018
Messages
2,881 (1.20/day)
From the sildes that Intel made for MTL, it seems that it's expected that the OS/apps will use the best hardware for the task. Looks like sometimes the speed gains of the GPU might not always be worthy when you look at the efficiency. And if I understand the graph right, you can run AI stuff of both at the same time. Efficiency has always been the keyword for the NPU
View attachment 332421
View attachment 332420
Yeah PC Hardware said such a thing, it depends on efficiency. For getting the fastest result possible you'd always run iGPU only, but maybe there are non-time critical AI stuff that would run more efficiently on NPU just take longer. I think when the NPOU gets 3-5x more powerful it'll make more sense. It's mostly for marketing and hype at this stage.
 
Top