Benchmark Result Discussion

[Djyuran]

Buy a macbook and a high end pc from the price of a gaming laptop as I have done, if mobility matters Browsing and working with mac, and fapping with pc.

I need a mobility for the device where I run VAM.
Should I start looking into the barebone boxes I can install desktop CPU + GPU but still can easily move it around?
Or it is a path to nowhere and high performance desktop system is huge box.

 

[Tomb]

I need a mobility for the device where I run VAM.
Should I start looking into the barebone boxes I can install desktop CPU + GPU but still can easily move it around?
Or it is a path to nowhere and high performance desktop system is huge box.

The ultimate solution to your needs is cloud gaming. Where I live, there is only 4g internet, and can't use it. It is extremly portable, extremly fast, and cheap as hell, but you need blazing fast internet. I'm courious how VAM runnig on that rig: https://shadow.tech/. I think this is the future not, the present...

 

[Sally Whitemane]

The ultimate solution to your needs is cloud gaming. Where I live, there is only 4g internet, and can't use it. It is extremly portable, extremly fast, and cheap as hell, but you need blazing fast internet. I'm courious how VAM runnig on that rig: https://shadow.tech/. I think this is the future not, the present...

I kindly disagree. I think the Cloud is only good to try out VAM so you could avoid spending a ton of money for expansive hardware.
User @lukas.3306 tried shadow.tech here and other Cloud services.
We had a private conversation and there always seemed to be CPU related performance problems. Not sure if he ever found something 'usable'.

Yesterday I already tested the AMD Datacenter GPU Radeon Instinct in Azure https://www.amd.com/en/products/professional-graphics/instinct-mi25 and an AMD EPYC 7V12(Rome) CPU, but this was a complete fail with ~8fps.

Typical Cloud CPU issues:

• potential scheduling CPU core latency issues - non Cloud Threadripper example
• being unable to control the CPU affinity due being in a virtual machine you have no direct control
• Cloud CPUs in general being virtualized and shared costing performance
• datacenter CPUs like Xeon, Epic, Threadripper being high core count, but low single core speed, the opposite of what VAM 1.x needs now

The way I see it, a desktop PC is the only way to get ideal performance in VAM.
Even high end laptops with an external GPU-box suffer from the limited bandwidth / PCI-express lanes to that box.

What could work very well in theory is a Cloud PC where you would have a fully dedicated desktop-class CPU paired with a good GPU.
But who offers that? The price would be high. Obviously CPUs are usually being virtualized to cut costs.

@harms7 Your 12900K broke the record for fastest physics time in baseline 3 of 3.52 with 3.00 milliseconds. Nice!

 

[lukas.3306]

Typical Cloud CPU issues:

• potential scheduling CPU core latency issues - non Cloud Threadripper example
• being unable to control the CPU affinity due being in a virtual machine you have no direct control
• Cloud CPUs in general being virtualized and shared costing performance
• datacenter CPUs like Xeon, Epic, Threadripper being high core count, but low single core speed, the opposite of what VAM 1.x needs now

I guess now I need to invest more time into testing non-HT cloud setups like these mentioned here:

Redirect Notice

I am personally still running the shadow setup, since it's much better than my laptop setup and I did not yet buy a nice local gaming pc.

 

[Tomb]

I guess now I need to invest more time into testing non-HT cloud setups like these mentioned here:

Redirect Notice

I am personally still running the shadow setup, since it's much better than my laptop setup and I did not yet buy a nice local gaming pc.

Please share some benchmark results! It will be interesting to know the VAM performance of the Shadow and other cloud pc-s!

 

[Lytanshade]

What could work very well in theory is a Cloud PC where you would have a fully dedicated desktop-class CPU paired with a good GPU.
But who offers that? The price would be high. Obviously CPUs are usually being virtualized to cut costs.

Wonder if Plutosphere could work with VAM. You'd have to install VAM onto the cloud server every time you want to use it, which would be super annoying. And add to SteamVR as a 'non-steam game'.

Might be interesting as a proof of concept though.

 

[Lytanshade]

Figured I should contribute to the data if i'm commenting in this thread!

fresh install with no overclocks.

Memory DDR4 2666 15-15-15-35 (running at stock. no xmp profile loaded)

 

[Djyuran]

Figured I should contribute to the data if i'm commenting in this thread!

fresh install with no overclocks.

Memory DDR4 2666 15-15-15-35 (running at stock. no xmp profile loaded)

Interesting, 980 Ti + i5 works faster than Ryzen 6900HS + 6700s.
It is really 3440 * 1440 resolution?
I found a bug in the benchmark when it shows wrong resolution.

 

[Lytanshade]

It is really 3440 * 1440 resolution?
I found a bug in the benchmark when it shows wrong resolution.

Yeah 1440p ultrawide

 

[Tomb]

Interesting, 980 Ti + i5 works faster than Ryzen 6900HS + 6700s.
It is really 3440 * 1440 resolution?
I found a bug in the benchmark when it shows wrong resolution.

Just accept notebooks whoever has same peak power as the desktop configs, under continous load they sucks. I have an asus rog strix notebook and a macbook with earth shaking power, they are just so responsive you can't imagine. They load an app, before you hit the button. But when you start continous heavy load, like blender, VAM etc, they just perform like a low end piece of shit...

 

[Djyuran]

Yeah 1440p ultrawide

just curious, could you please put 1080p results?

 

[Mr Explicit 1]

I think i should upgrade from

my 2080

 

[Lytanshade]

just curious, could you please put 1080p results?

I overclocked it to 4.8Ghz all cores. Stress-test stable clock, temps peaked at 74C @ 1.4V on air. I can bring the voltage down a little, but it's end of life for this system and can't be bothered tweaking the details, so just set all clock multiplier to 48 and everything else at auto heh. Still holding up ok for a 6-7 year old CPU and probably a decent cheap VAM box on the 2nd hand market.

And base clock (3.5 - 3.9GHZ)

 

[Sally Whitemane]

@Mr Explicit 1
For correct benchmarks results it is required to turn vertical synchronization aka VSync off. With 'VSync: On' your graphics card waits for your monitors frequency to be ready to preset a frame. Since your monitor is running at 75 Hz the RTX 2080 is limited to a maximum of 75 frames per second. Note how your Baseline 1, ClothSim, Mirror and Baseline 2 tests are all very close to 75 Hz / FPS. They all run into this limit. Correct results would be higher/better.

Using custom settings is fine for personal tests, but makes comparing results here impossible. To get comparable results you'd have to start VAM in the resolution you want to test (VaM (Config).bat in VAM folder) and use the 'Run Official Benchmark' settings.
Except for scenes with a lot physics sim. (clothing, long hair, 3 characters) that hardware should do okay.

Upgrading the GPU would probably result in the biggest FPS gain overall. But then there is the RTX4000-series coming this year. So - buy now - regret later risk. Allegedly the RTX 4090 flagship is being tested right now.

Replacing the CPU using a AM4-socket 5000-series drop-in replacement would improve performance mainly in physics heavy scenes. But then there is ZEN4 CPUs for AM5 not that far away. AM4 is now a dead end-of-life product. While still strong, it cannot beat going Intel 12th gen like the 12700K (Except the 5800X3D maybe). Going Intel now would mean new motherboard + CPU + potential RAM upgrade to DDR5. Very expensive. 13th gen Intel Raptor Lake seems to be coming in 2022 too. My point is - by the end of this year - in only a few month the 'best' hardware will be completely different and it's actually quite rare to have so many big releases in a year. Worth the wait?

 

[Tomb]

@Mr Explicit 1
Upgrading the GPU would probably result in the biggest FPS gain overall. But then there is the RTX4000-series coming this year. So - buy now - regret later risk. Allegedly the RTX 4090 flagship is being tested right now.

Replacing the CPU using a AM4-socket 5000-series drop-in replacement would improve performance mainly in physics heavy scenes. But then there is ZEN4 CPUs for AM5 not that far away. AM4 is now a dead end-of-life product. While still strong, it cannot beat going Intel 12th gen like the 12700K (Except the 5800X3D maybe). Going Intel now would mean new motherboard + CPU + potential RAM upgrade to DDR5. Very expensive. 13th gen Intel Raptor Lake seems to be coming in 2022 too. My point is - by the end of this year - in only a few month the 'best' hardware will be completely different and it's actually quite rare to have so many big releases in a year. Worth the wait?

Maybe they will be realesed at 2022 q4 but who knows the story of the double priced rtx cards and the mystical 12th gen cpus allways disappear before reaching the shops, or the horror priced ddr5 rams won't happen again? Don't feel offended, but I'm a little bit sceptical after the last two years. The big tech companies now learned the lesson, if they say the magic words: covid, crypto, war, supply chain we just pay double price. Just waiting two years for getting a 50% faster pc for an unknown price... And the signs are not so good: 2000-3000 msrp and 600W TGP (in my country it means 300 bucks for a year just to power up daily 4 hours the card) says the rumors for rtx 4090. On the other hand this year comes the 3080 super, for a higher price. If the 4070 has near the same performance as 3080 super, it won't be cheaper when they on the market at the same time, and 4080 won't be cheaper than 4070, the 2000 msrp just a joke for a 4080. That was a brilliant trick, when they said the rtx 30 msrp would be cheaper then the gtx 20 msrp, but the real price was schoking, and they just said the magic words... AAA games developed to run on consoles, and rtx 30 and 12 gen intel can handle it (for a price lower for hardware, higher on sw). Don't expect huge improvements this year in game industry. Vam is the big question, we don't know how much will we benefit upgrading to VAM 2. But when will be it realesed, and when will be it uploaded with contents like VAM 1? Will it benefit that much from the new techs realesed 3 years later the start of developement? Is it worth to skip vam 1 maybe for two years to get unknown fps boost in Vam 2? So many questions and speculations. Just buy the best rig you can afford at the moment

 

[yudodis]

Long time lurker, first time poster. What do you think? My 3080 is severely undervolted to 1710 MHz / 0.750 mv. And my 12700KF has the clocks of the 12900K, although with fewer E cores obviously. RAM at 3600 MHz with 16-18-18-36, CR2, Gear 1.

 

[HolySchmidt]

Finally made one with 12700K stock, 3080Ti stock, 32GB 3200 CL16 // CPU and GPU watercooled

@yudodis
Seems legit

 

[Tomb]

I was curious about the performance impact upgrading to win 11. All day testing...

Real life install, with avg internet security, office ect., and 184 gb vam directory. Picked the most avarage results. First impressions, win 11 super responsive, fast as hell, huge inprovement in user experience, but unstable, two of three restarts just freezes...

Tried to pick up the most avarage results, first win 10, second totally same install win 11.

 

[HiddenSign]

The 5800X3D is in the mail and should arrive tomorrow. I hope that i have time to run some benchs this weekend (with 3080Ti).
I am very curious how it will perform in VAM...

 

[HolySchmidt]

@HiddenSign
You are not alone being curious
I advise creating 2-3 test scenes (1,2,3 persons) where you know exactly how much fps you had before. It would be perfect if scene 1 would be in the gpu limit, scene 2 something in between and scene 3 in the cpu limit.

All tested with a "clean" VaM version please! That would be great

 

[HiddenSign]

I had no real scenes on my VR-Rig, so i ran the benchmark on a fresh download of VAM. Nothing installed but the Benchmark.

System:
AMD Ryzen 3900x PBO on (max boost 4,65Ghz) OR AMD Ryzen 5800X3D stock
Asus TUF gaming B550M-Plus WiFi
32GB (2x16GB) Corsair Vengeance 3200 CL16/18/18/38
GeForce 3080Ti FE @ 114% +140Mhz GPU +1000MHz VRam
Corsair HX 850Watt PSU
LianLi O11 mini Air
Noctua NH-U12A with Thermal Grizzly Hydronaut
Win 10 Pro
VAM on a S-ATA SSD

3900X:
1080p

1440p

4k


5800X3D:
1080p

1440p

4k


Jea, if not GPU limited, its fast. In 1080p its faster than my 5900X with 3090.

 

[HolySchmidt]

So that's exactly a 20% plus in 1080p in relation to a 3900X.
That's a 9% plus to my 12700K stock in 1080p. Even though comparison on this page says there's a 25% single core advantage for the 12700K:
https://cpu.userbenchmark.com/Compare/Intel-Core-i7-12700K-vs-AMD-Ryzen-7-5800X3D/4119vsm1817839
Or this:
https://www.cpu-monkey.com/de/compare_cpu-amd_ryzen_7_5800x3d-vs-intel_core_i7_12700k
Here it is 15%:
https://nanoreview.net/en/cpu-compare/intel-core-i7-12700k-vs-amd-ryzen-7-5800x3d

So what do we have here?
Difference in 1440p and 4K is like zero between your twos, so that's GPU bottlenecked 100%. It would be interesting to know (for me) how the 5800X3D performs vs. the 12700K in a real CPU bottlenecked scene. Cause like I said some pages before: the benchmark is fine but not very orientated to real game scenes where the physics are a huge factor.
You want competition?

 

[Tomb]

So that's exactly a 20% plus in 1080p in relation to a 3900X.
That's a 9% plus to my 12700K stock in 1080p. Even though comparison on this page says there's a 25% single core advantage for the 12700K:
https://cpu.userbenchmark.com/Compare/Intel-Core-i7-12700K-vs-AMD-Ryzen-7-5800X3D/4119vsm1817839
Or this:
https://www.cpu-monkey.com/de/compare_cpu-amd_ryzen_7_5800x3d-vs-intel_core_i7_12700k
Here it is 15%:
https://nanoreview.net/en/cpu-compare/intel-core-i7-12700k-vs-amd-ryzen-7-5800x3d

So what do we have here?
Difference in 1440p and 4K is like zero between your twos, so that's GPU bottlenecked 100%. It would be interesting to know (for me) how the 5800X3D performs vs. the 12700K in a real CPU bottlenecked scene. Cause like I said some pages before: the benchmark is fine but not very orientated to real game scenes where the physics are a huge factor.
You want competition?

12900k cheaper than the 5900x3d, the 12700 just fights in totally different weight class. It is a shame we compare amd high end cpu with intel mid class cpu expecting same performace. Please don't feel offended, just wanted to figure out the consequentis of test results, and just wanted to help with tougths, feel free to tell yours.

 

[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