Benchmark Result Discussion

[Sally Whitemane]

Moved the comparison tables to a new thread as promised.

Meanwhile overclocked 5800X results:

• CPU: Ryzen 7 5800X on air with HYDRA OC profile boosts up to 4950 Mhz in single core profile
• RAM: 64 GB DDR4-3000 CL16-18-18-38 OC to 3333 Mhz
• GPU: 2x RTX 3090 Founders Edition OC custom watercooled, power limited screaming for juice during the whole benchmark, one sucks 380W-400W, no SLI, single card run
• MOBO: Gigabyte X570 UD
• PSU: Corsair AX1600i

That's all I can squeeze out my CPU with an air overclock - not worth it - will return to 100% safe & stable clocks after this.
For more I'd need better RAM, faster infinity fabric, better timings, CPU waterblock, etc.
If that 5800X3D would have had overclock support - damn - then I would regret not waiting for it a lot more.

Spoiler: 1920x1080

Spoiler: 3840x2160

 

[HolySchmidt]

Just wait for the next Ryzen generation
I think they will be kind of a "game changer", hopefully.

@Tomb
12700K is no mid-range CPU, sorry. It's the second best Intel is offering right now, and we are speaking multicore only. Singlecore is nearly the same like 12900K.
And the 5800X3D is no "top notch there's nothing better" CPU. In gaming right now: yes. But that's it. And not even in any games, just saying.

Anyway, I was just curious how good this CPU performs with VaM in a real scenario and not with this benchmark, because we got a new technique here. If this really is a "totally" different weight class, would have to be proven.
The physics times are not that much faster than with the 12700K.

I think I just overclock my 12700K a little, what you can't
I'll post some new results later ...

 

[BooMoon]

just an before x after comparison of my gpu upgrade performance itself improved alot , but lately i'm facing a memory leak problem with vam that eats all my 32gb till crash :\ idk if it was any of windows update because before i never ran out of memory , now with 5-10 mins of vam open even on idle on menu it does this memory eat .

 

[harms7]

Thanks @Sally Whitemane !

Those 5800X3D results are looking really spicy, so I did some more benchmarks with a clean VAM install and a 5.2Ghz CPU OC. I managed to lower the physics time quite a bit but still it was unable to beat the average framerate of the X3D .

Edit: My bad, did not notice that the benchmark in windowed resolution was 1920x1009, so here i fixed it, ran again in fullscreen.

 

[Tomb]

Just wait for the next Ryzen generation
I think they will be kind of a "game changer", hopefully.

@Tomb
12700K is no mid-range CPU, sorry. It's the second best Intel is offering right now, and we are speaking multicore only. Singlecore is nearly the same like 12900K.
And the 5800X3D is no "top notch there's nothing better" CPU. In gaming right now: yes. But that's it. And not even in any games, just saying.

Anyway, I was just curious how good this CPU performs with VaM in a real scenario and not with this benchmark, because we got a new technique here. If this really is a "totally" different weight class, would have to be proven.
The physics times are not that much faster than with the 12700K.

I think I just overclock my 12700K a little, what you can't
I'll post some new results later ...

The 12 gen not K Intel CPU-s can be overcloked with a blck motherboard, and just get one class performce increase:

12300 OC = 12400 noOC
12400 OC = 12600K noOC
12700 OC = 12700K noOC

It is possible to OC my CPU, but is it worth to buy high end mobo to OC a "not so mid class" CPU?

My opinion is Apple M1 CPU changed everything, Intel and AMD just given up the looks like competing with each others, and tries to show their real power against a rival has more potential they have ever met. The real threat is not just the Apple cpu, but the fact now anyone can order (huwei, xiaomi, or just a noname brand) a game changing cpu from a taiwanese vendor. Next years will be interesting, the two years slow rotation in nvidia, amd and intel releases will be fired up to keep the peace!

It is good to see you guys to act so friendly and supportive, sharing thoughts and test results to gain from it!

 

[HiddenSign]

[...]

Interesting. Looking at the FPS, it is on par, but I "only" have a 3080Ti in this PC.
I'm heavy trying to resist the urge to buy more faster stuff to see where this CPU can go. I think faster Ram may step it up a notch further.
I have some watercooling stuff lying around, so maybe I will play a litte with that.

 

[harms7]

Interesting. Looking at the FPS, it is on par, but I "only" have a 3080Ti in this PC.
I'm heavy trying to resist the urge to buy more faster stuff to see where this CPU can go. I think faster Ram may step it up a notch further.
I have some watercooling stuff lying around, so maybe I will play a litte with that.

3080Ti is definitely fast enough, I did a couple of 4k tests and compared to yours mine is about 5% slower. I kept my 3090 undervolted and underclocked due to the fact that my OC'd DDR4 (Samsung B-die) with tight timings requires 1.5v voltage, and if the 3090 with OC is running at max speed, it would produces so much heat and the RAM sitting not too far away from it would seldomly get errors due to high temperatures.

I reckon better DDR4 RAMs would definitely help with 1% lows, but not so much on the X3D since it has a f*** tons of L3 cache already. A high frequency and low latency DDR5 would be the way to go I think for the near future, just look at the average FPS Jayjaywon get with 12900k and DDR5, his 1% lows are not that great because of the current DDR5's higher latency compare to good DDR4s, but once faster DDR5s comes out, this is all about to change.

 

[HiddenSign]

Playing with PBO2 Tuner at the moment to undervolt the CPU. All-Core boost ist now at 4,45GHz stable, at the same time Power is down and temps are down a lot. Now checking for stability and when this works i will rebench and look if it makes a difference in performance in VAM.

@harms7
wow, you get memory errors? how hot is your ram? What case do you use?

 

[harms7]

@harms7
wow, you get memory errors? how hot is your ram? What case do you use?

They sit around 47-49c while gaming with UV UC'd 3090, if the card is left running at stock or OC, the DIMMs temperature would hit about 50-52c. Samsung B-die memories are known to throw out errors after hitting 50c+, before I found that out I was getting some weird glitches in games after long playing sessions.

My case is NZXT Nuka Cola which is just a H700 reskin, worse than average airflow. I took the front panel off though and left the dust filter on after upgrading to alder lake from 9900k because the RAM would hit 55C after just 30 mins of gaming

 

[kokmaker]

Was checking up on this thread and apparently I forgot to give an update after upgrading to a 3080ti. Also OCed cpu a little, to 4.4 ghz since then. Running on DDR4 3000 MHz cl15, same as my last post. Was on a fresh install of VAM, SATA SSD, also same as before. Very happy about the performance increase. But my greedy little self was wondering if I could squeeze out just a bit more performance with a better cpu. Also highly curious about 5800x3D, even though AM4 is on it's way out. Wondering if it would be a good, "affordable" stopgap for a slight performance bump until next-gen Ryzen.

 

[Hedgepig]

Slightly off topic, but my 3060 is now about eight months old. I've posted my performance figures. Recently, I've noticed the bottom of the big circular fan is full of dust. The rest of it is dusty inside too. What is the best way to clean the dust, or should it be left alone? It has a glass door which is easy to open and get inside.

 

[HolySchmidt]

Dust should never be left at technical components as it decreases air flow and increases therefore temperatures.

A brush and a vacuum cleaner would be my advice, no big deal. Just turn the computer off and remove the power plug before doing so.

 

[Lytanshade]

Was checking up on this thread and apparently I forgot to give an update after upgrading to a 3080ti. Also OCed cpu a little, to 4.4 ghz since then. Running on DDR4 3000 MHz cl15, same as my last post. Was on a fresh install of VAM, SATA SSD, also same as before. Very happy about the performance increase. But my greedy little self was wondering if I could squeeze out just a bit more performance with a better cpu. Also highly curious about 5800x3D, even though AM4 is on it's way out. Wondering if it would be a good, "affordable" stopgap for a slight performance bump until next-gen Ryzen.

I'd try pushing that OC some more before looking for a stopgap CPU upgrade. I had a LOT of movement in my Baseline 3 results by OC'ing my 6600K to 4.8Ghz. The 8700Ks OC nicely as well (easy 4.8Ghz+), judging from google results.

 

[Alter3go]

my pc died so i had to build something now (wasnt really my plan, i wanted to wait some more months but whatever)
so someone asked me to do this bench and post results. Here you go

 

[keycode]

my results with the standard full screen mode (1920x1080 just for benchmark purpose). My system has windows 11 installed since september 2021. The kingston RAM is at 3600Mhz CL17 and the motherboard is msi z490a pro. My 3070 is an "old" first generation MSI suprim, a model that was released with the normal ddr6, 8GB . It can be useful to compare my results with the above last generation i7 and the upgraded 3070(ti) and also... for those still skeptic about windows 11 actual performance.

 

[keycode]

(of course... you can wonder why the windows 11 is not identified by my Vam benchmark ... this weird absurdity is happening even with the nvidia control panel since the first day I was upgrading my old system with the online protocol... probably in microsoft team someone was forgetting to correct this ... banal formal bug)

 

[Lytanshade]

It can be useful to compare my results with the above last generation i7 and the upgraded 3070(ti) and also... for those still skeptic about windows 11 actual performance.

Pretty nice bump in the Baseline 3 results with the generational improvements between the 10700 and 12700. That's the test that seems to provide the most useful comparative data overall.

 

[enoeht]

Lenove Legion 5 15ACH6 Laptop. Ryzen 7 5800H, 16gb, RTX 3050Ti 4gb Windows 10 Pro 1909. No overclocking. Most Windows 10 resource hogging background apps are either disabled or removed from my system completely, including telemetry. I also have the power mode set to Ultimate Performance. I posted how to enable that hidden option over on the official Virt-A-Mate discord. The Pros: It may give you a little extra boost. The cons: If you're on a laptop, it will drain your battery faster if not plugged in. Here is the link. https://discord.com/channels/363274293112602636/473573078241247246/939948724917010452
Also, I have only tested it on Windows 10 Pro, so i'm not sure if it will work on other editions or Windows 11.

 

[Djyuran]

I am trying to run a benchmark in Oculus VR connected with Oculus Link cable for my 3080 desktop.
I see 72fps even if I used a provided .bat file. Do I need to restart something after I ran .bat file?
I tried to dive deeper and noticed that it changes highlighted in yellow setting, not a ASW setting under the Oculus Link.
Also I don't have a setting in SteamVR to disable MS.
Anyway, even if I change Mobile ASW, nothing changes!

What am I doing wrong? How to get correct numbers?

 

[Ged366]

Gigabyte X570 UD - RYZEN 3600X - RX 6800XT - 32Go 3600MHz C16 - Win11 64bit

Wondering if the 3600x is not holding me down more than needed
Sad that the 5800X 3D is still so much overpriced while the next gen is already almost knocking to the door.
I cant resign to pick a "simple" 5800x, I have the feeling I would miss something


@1920 x 1080 :

@1680 x 1050 :

 

[Djyuran]

Actually, VR test results are weird.
If I use an Official Benchmark, then I get 72 or 36 fps depending on a scene.
But if I set set custom quality (Ultra), then fps is around 56fps in most scenes except mirror one where it is 36fps.
It is desktop 3080 + i12700

 

[Lytanshade]

This might be an interesting comparison if anyone else is thinking of upgrading to an Alder Lake CPU but waiting 'til Ada Lovelace (or dirt cheap mining GPUS) to upgrade the GPU.

This is the same memory (32G DDR4 - CL15 @2666MHz); same VAM install on the same nvme drive; and same GPU (980ti) as my previous 6600k benchmarks

New z690 mobo (asus tuf d4) and a 12600k OC'd pretty nicely - 5.3/5.2/5.0 (2c/4c/6c) and e-cores @4GHz

I'll likely be upgrading the memory before the GPU, so will be interesting (to me at least) to see what effect doubling the memory speed will have.


Obviously the Physics score comparison is the data point of interest. Frames are severely GPU bottlenecked!
In 'real-world' VAM it's a significant improvement. Especially when building scenes in VR with two girls. Gone from chalkboard-nails in my brain to enjoyable






[HWMonitor screenshot taken during a demanding moment of Baseline3 benchmark]

 

[HolySchmidt]

I'm interested in your memory comparison ?

 

[UsernameAndPassword77]

I declocked my memory from 3600 C14 to a bunch of values. I'll post results if interesting...

But before I do... What does WaitTime mean? I ran the same computer with two processors, and overclocked 12600K and a 12900K and got really strange results (great 12600 physics), but the big difference in the comparison was my 12900K run has a longer WaitTime. So what is this?

Hopefully I'll be able to post three complete configs with one or two parts swapped out here and there.

 

[Sally Whitemane]

Correction January 7 2023:
Should have edited this wrong info/assumption a long time ago.
Turns out WaitTime comes from the CPU main thread actually being idle / sleeping because VAM runs into an artificial limits.
VAM has 300 frames per second limit. This limit is 100% verified.
In theory there could be multiple reasons to wait, but the FPS-limit is actually relevant for understanding the benchmark result numbers.
It clearly shows in the results of users with the latest high end hardware. More 'WaitTime'.
When I posted this the available hardware (12900K & RTX 3090 was King) was not fast enough to reach near 300 FPS in a high resolution.
Now a RTX 4090 is already scratching more and more on that limit. Results from faster future hardware will sooner or later all reach around 300 - then only WaitTime will become higher.

It does not mean my old post below was wrong - it's just not verified and speculation / an assumption.
Physics synchronization between threads could still results in 'WaitTime' besides the FPS-limit.

old post:
WaitTime seems to be the time required by Unity to synchronize between the physics- and main-thread.

If somebody want's to dig deep into VAM's code a good starting point would be MeshVR.PerfMon.
Then check where ReportWaitTime() is used. It's only called by the following two functions.

Disclaimer: Based on assumptions just from 'reading' code - correct me if I'm wrong here. Obviously I cannot verify this without debugging/profiling with the full project files.

The physics-thread reports it's wait time from DAZCharacterRun.FixedUpdate().
This function is called at the fixed physics rate frequency you set in VAM's user preferences:

float elapsedMilliseconds1 = GlobalStopwatch.GetElapsedMilliseconds();
  this.fixedStartTime = elapsedMilliseconds1;
  if (this.useThreading && this.fixedUpdateWaitForThread && DAZPhysicsMesh.globalEnable)
  {
    this.fixedThreadWaitTimeStart = elapsedMilliseconds1;
    while (this.pmJointTargetsUpdatingOnThread)
      Thread.Sleep(0); // <- wait time
    if (this.newJointTargets)
    {
      elapsedMilliseconds1 = GlobalStopwatch.GetElapsedMilliseconds();
      this.MAIN_fixedThreadWaitTime = elapsedMilliseconds1 - this.fixedThreadWaitTimeStart;
      PerfMon.ReportWaitTime(this.MAIN_fixedThreadWaitTime);
    }
}

... and the main-thread reports it's wait time from DAZCharacterRun.Update(). This function is called every frame:

float t = GlobalStopwatch.GetElapsedMilliseconds() - this.updateThreadWaitTimeStart;
PerfMon.ReportWaitTime(t);
this.MAIN_updateThreadWaitTime = t;