Benchmark Result Discussion

I want to try it, but I don't know how to configure it for my hardware. Can you help mi with that?
I was searching for a tutorial/guide but I think there is none
The guide is on the hub page that you get the software.
 
The guide is on the hub page that you get the software.
IIRC that dude actually has a CPU that isn't listed on the CPU performance upgrade page, so he's probably wondering the specific settings for his CPU. But I could be completely wrong, i just remember wanting to help him one time on this matter and i gave up once i saw that.
 
13600k rtx4070 ram 32gb With performance patch
스크린샷 2024-10-01 133037.png

스크린샷 2024-10-01 133012.png
 
In 1920x1080 im getting 300+ fps in desktop but same scene only gets 30+ fps in VR (also in 1920x1080). tried oculus, steamvr and virtualdesktop but they all get pretty much same vr performance. idk what im doing wrong

Ryzen 9 5900X
RTX 3060
32gb ram
 
Anyone plans to get their hands on the new stuff? I am wondering how vam with the optimizations would react to the new intel architecture. Reviews are mixed, but certain cases shows improvement, may be here too. I am curious if its a step ahead or the new x3d will get just further ahead.
 
For anyone with a 7900, how does VAM run purely in Desktop for you? My current rig is 6700xt, 5600x, 32gb. I'm thinking of grabbing a 7900 gre so curious how it'd run in Desktop.
 
Anyone plans to get their hands on the new stuff? I am wondering how vam with the optimizations would react to the new intel architecture. Reviews are mixed, but certain cases shows improvement, may be here too. I am curious if its a step ahead or the new x3d will get just further ahead.
Yeah, new x3d will be way ahead of the new intel in vam. New intel is pretty much just a refinement of power usage. New X3d is minimal gaming increase with much better all core; but still way ahead of intel in gaming and sims.
 
Yeah, new x3d will be way ahead of the new intel in vam. New intel is pretty much just a refinement of power usage. New X3d is minimal gaming increase with much better all core; but still way ahead of intel in gaming and sims.
Most likely yes. Tough Intel is finicky, they can be altered quite far from default, where amd is like plugnplay.
Intel still has the highest ST score and can be tuned with insane memory frequencies.
I am just sad I never saw a result here with 8k+ ram configs with tuned ring and such.
Plus if you run stuff constantly to enhance vam (like browsers/ai stuff) then you have an additional parasitic loss on an 8 core platform which is not appearing while you benchmark.
 
Most likely yes. Tough Intel is finicky, they can be altered quite far from default, where amd is like plugnplay.
Intel still has the highest ST score and can be tuned with insane memory frequencies.
I am just sad I never saw a result here with 8k+ ram configs with tuned ring and such.
Plus if you run stuff constantly to enhance vam (like browsers/ai stuff) then you have an additional parasitic loss on an 8 core platform which is not appearing while you benchmark.
It's not just about ST score, for VAM it's about memory latency. The unity engine does a lot of physics which is wasn't designed for. The physics simulation does a lot of memory reading/writing in a non-optimized memory layout. Crazy memory speeds or huge caches lower this latency and that's why intel cpus with 8000mhz ram or ryzens with x3d are so good for VAM. Previous gen CPUs might still beat current gen CPUs if they can drive high RAM speeds.
 
13900KS All P core 5.8G
Turn off E-core and Hyper-Threading
With CPU performance patch
Compare the differences between DDR4 and DDR5

Conclusion: DDR5 has better min1%

Hope to see the results of 9800X3D



DDR4 4000 C20

Benchmark-20240819-035630.png



DDR5 8000 C38

Benchmark-20241108-120400.png
 
13900KS All P core 5.8G
Turn off E-core and Hyper-Threading
With CPU performance patch
Compare the differences between DDR4 and DDR5

Conclusion: DDR5 has better min1%

Hope to see the results of 9800X3D



DDR4 4000 C20

View attachment 428869


DDR5 8000 C38

View attachment 428870
Thats what I missed for soo long! Thank you!
Did you raised ring and uncore ratio too?
Which had the lower memory latency? DDR4 has usually, but then it seems vam benefits bandwith too somehow.
Also with cpu patch or processlasso you can set which cores use vam, rest of the OS overhead could go onto the E cluster. I hoped somehow with intel the baseline 3 avg physicstime could break the 1 ms barrier.
Altough I think we should need a different more cpu heavy benchmark or at least add some more script to this scene.
 
Last edited:
Thats what I missed for soo long! Thank you!
Did you raised ring and uncore ratio too?
Which had the lower memory latency? DDR4 has usually, but then it seems vam benefits bandwith too somehow.
Also with cpu patch or processlasso you can set which cores use vam, rest of the OS overhead could go onto the E cluster. I hoped somehow with intel the baseline 3 avg physicstime could break the 1 ms barrier.
Altough I think we should need a different more cpu heavy benchmark or at least add some more script to this scene.
NO raised ring and uncore ratio
DDR5 had the lower memory latency(DDR4 60ns、DDR5 55ns)
I tested switching the E core on and off and there was no big difference in performance. I turned off the E core because it can significantly reduce power consumption and overclock the P core to a higher frequency. A higher P core frequency can significantly improve game performance.
 
Thats what I missed for soo long! Thank you!
Did you raised ring and uncore ratio too?
Which had the lower memory latency? DDR4 has usually, but then it seems vam benefits bandwith too somehow.
Also with cpu patch or processlasso you can set which cores use vam, rest of the OS overhead could go onto the E cluster. I hoped somehow with intel the baseline 3 avg physicstime could break the 1 ms barrier.
Altough I think we should need a different more cpu heavy benchmark or at least add some more script to this scene.
Yeah best i could do on the 7800x3d was 1.09ms. Im sure the 9800x3d probably breaks the 1ms barrier easily.
 
Dam, im really thinking about. Theres plenty just sitting at my local microcenter calling out to me.
7800x3d can be sold pretty well now, and its a simple drop in. I dont like giving purchase advices, but if you are tempted anyway.. :)
Im sure lots of folks are eager to see what that brings too.
 
NO raised ring and uncore ratio
DDR5 had the lower memory latency(DDR4 60ns、DDR5 55ns)
I tested switching the E core on and off and there was no big difference in performance. I turned off the E core because it can significantly reduce power consumption and overclock the P core to a higher frequency. A higher P core frequency can significantly improve game performance.
RPL can handle ring usually up to 5.4 if I remember correctly, that would speed up the internal connections everywhere. There is still potential left in your system.
How was power anyway? Allowing E cores really raised that much to worth disable?

Anyway it seems if intel would release bartlett lake - posssibly with cudimm support - might just be the saving grace.
 
Last edited:
7800x3d can be sold pretty well now, and its a simple drop in. I dont like giving purchase advices, but if you are tempted anyway.. :)
Im sure lots of folks are eager to see what that brings too.
Yeah any smart 7800X3d owner right now : sell your 7800X3D... and buy a 9800X3D... it'll cost you like noting or very little and you'll get +10%-15% performance.
 
Thanks for sharing your insight on what is important for VAM. I have always kinda wondered if x3d was helping in this game or if cores was more important. I also agree that once you introduce AI your system is humbled very quickly lol. It looks like 9800x3d was a W curious to see what improvements we see. I think for me though I'll be waiting the 9950x3d/5090 combo to see what she can do if they are good.

Just sharing another 7800x3d 4080 results. Clean install 24H2 windows. Small improvement over lots of vars w/ cleaned dup morphs (recommend doing it at the least for dup vars). PBO, but no other tweaking. 6000 cl30 ram. These are patched. I was 198FPS avg before patch with a lot of vars, 250FPS with a lot of vars patched, but with some vars removed for dup morphs. (combo reading unfortunately). 251FPS clean patched on 23H2. So cleaning those morphs and patch really helped.
255FPS at 1080p CLEAN + PATCHED 24H2.png



Here is the 1440p results
206FPS at 1440p CLEAN + PATCHED 24H2.png
 
Last edited:
Man, again one of your answer or thread, like in this case, is saving my time (or mi life? time is money xD) so thanks again MacGruber!
 
Back
Top Bottom