GameTechBench GPU benchmark is already out!

[miguel1900]

Play it for free on STEAM , or buy the Pro version!

GameTech is a benchmark based on Unreal 5 that aims to measure performance and verify stability on today's and tomorrow's modern PCs. Utilizing the most cutting-edge and innovative gaming technology, it leverages Lumen, Nanite, Virtual Textures, Virtual Shadowmaps, Metahumans... to present an environment that is as realistic, demanding, and optimized as possible, while prioritizing visual quality.

GameTechBench definitve 1.0 version is already out!

What's Lumen? A new global illumination system implemented in Unreal Engine 5 that, preferably using Ray Tracing, allows for dynamically and realistically lighting entire scenes, generating diffuse and occlusion shadows. This achieves a quality similar to what was previously obtained by baking the lighting, but instantly, albeit at a much higher cost.

What's Nanite? A new mesh rendering system that allows for displaying meshes with millions of polygons without heavily overloading the scene, though it has a high base cost. The meshes only show the necessary polygons on screen based on the pixel surface they occupy, allowing for "infinite" completely smooth and partial transitions, even along a single mesh.

There are several benchmarking modes:

• Without Ray Tracing (Raster): Will use "Lumen Software" if the PC is not compatible with Ray Tracing, or if the user explicitly selects it. Included simply so that older or less powerful cards (within reasonable limits) can still be accommodated.
• Ray Tracing: Will use "Lumen Hardware." The default standard gaming mode.
• Path Tracing: Offline rendering of the highest quality. Will use a fixed image iteratively and a fixed resolution of Full HD so that VRAM size is less of a burden and focuses more on measuring performance itself. Additionally, this standardizes the result for all PCs.

(Very) minimum requirements for Lumen Software:

• Windows OS
• 1920x1080 (at least) display's native resolution
• Internet connection (Date check)
• 8GB RAM
• GPU compatible with SM6 and DirectX12
• GPU with 6GB of VRAM
• GPU equivalent to GTX 1660 (1080p @ 20fps)

Minimum requirements for Lumen Hardware and Path Tracing:

• Windows OS
• 1920x1080 (at least) display's native resolution
• Internet connection (Date check)
• 16GB RAM
• GPU compatible with SM6 and DirectX12
• GPU with 8GB of VRAM
• GPU equivalent to RTX 2060

Recommended requirements for Lumen Hardware and Path Tracing:

• Windows OS
• 1920x1080 (at least) display's native resolution
• Internet connection (Date check)
• 16GB RAM
• GPU compatible with SM6 and DirectX12
• GPU with 10GB of VRAM
• GPU equivalent to RTX 3060

C&C are welcome! I can't wait to see your results!

 

[QuietBob]

Wow, you did all this? Amazing job
Here's native 4K with software RT, will post more results tomorrow:

4K hardware RT and 1080p PT


1440p


1080p


And a few observations, if I may. It would be good to include a caption in the main menu to indicate how to start the test. Now it isn't very clear where you should click to run the benchmark. Also, maybe the main font in the above summary could be a different color than white? The "min FPS" line is barely readable.

 

[3x0]

1440P, Raster and Raytracing Lumen, 2 minute presets

 

[oiawshboiwsahnoidaws]

nice work on the benchmark
downclocked/undervolted 4090

 

[oxrufiioxo]

Nice work, was more impressed by this than the latest 3Dmark test.








This was an interesting test not what I expected but cool.

 

[miguel1900]

Thank you all !!

Thank you @QuietBob ! Except modeling/texturing the individual meshes, I did all the rest, including remeshing the meshes. From lighting to the hardware-info retriever.

In fact I already thought about the same you have suffered today, but maybe I tried to convince myself, however you have confirmed it so I have added some changes based on your feedback! (Link already updated).

Thank you @oxrufiioxo , may I ask what were you expecting? Maybe I can get some feedback from it.

 

[DR4G00N]

"WX 9100" @ 1600/945 220W
The res is slightly off, probably just because of my 720P TV.

1440:


4K + 125% scale (Have to use that 16GB vram for something):

 

[miguel1900]

Well done!

Should I add 8k for those with a 4090?

By the way, I was waiting for a non standard monitor like yours. Did you see the image stretched or deformed? Or maybe just with widther black bars?

I think displays with a ratio less than 16:9 may be render less pixels, so not directly comparable to others. If 16:9 or higher, should be equivalents in pixels (compensated with more or less black spaces, which don't hit the performance); a 16:9 1080p monitor should perform the same as a 21:9 (for example) 1080p monitor in the same PC.

 

[oxrufiioxo]

Thank you @oxrufiioxo , may I ask what were you expecting? Maybe I can get some feedback from it.

Same benchmark just Pathtraced.

What you have is fine though.

 

[miguel1900]

Same benchmark just Pathtraced.

What you have is fine though.

I see, thank you!

Unfortunately if the same sequence would be rendered with path tracing, would take hours. Path Tracing is currently not a real time tech in Unreal. In Cyberpunk, if you were thinking about it for example, it's not (pure) path tracing too, anyway, but a "fake" one; more or less the equivalent to Unreal's Lumen.

In this benchmark, path tracing is intended to be compared (or to replace) other offline rendering GPU benchmarks, more like Vray, Blender, Octane...

Thank you again

 

[Lew Zealand]

Radeon 780m, it really wanted to use more than 8GB VRAM even though I only have 8 allocated.

1080p RT

1440p Raster

 

[AMF]

10 BIT COLOR HDR 10 ON ALL TESTS

 

[3x0]

@miguel1900 Is it possible to monitor GPU power consumption during the test? Would be useful to get performance-per-watt numbers as a statistic directly in the benchmark window.

 

[Deleted member 223428]

In Linux, just for fun. Manjaro, GE-Proton-9-9, Nvidia 550.90.07 Drivers

Good size performance hit over native Windows.

RT:

No RT:

 

[Sunny and 75]

In Cyberpunk, if you were thinking about it for example, it's not (pure) path tracing too, anyway, but a "fake" one; more or less the equivalent to Unreal's Lumen.

Good to know!

 

[neatfeatguy]

75% power limit set on my 3080Ti which is about a 5-10% loss in performance overall.

 

[QuietBob]

In fact I already thought about the same you have suffered today, but maybe I tried to convince myself, however you have confirmed it so I have added some changes based on your feedback! (Link already updated).

Thanks for these, I may have found a bug though. When running the bench for the first time, the selected resolution does not get applied if it's different than the native one.
For example, upon starting the app I select 1440p, but the benchmark begins in my native 4K. Interrupting the test and re-selecting 1440p fixes this. Also, running the benchmark in native res first and then changing it without exiting the app seems to work fine.

Perhaps someone else could confirm it?

 

[Arctucas]

GPU stock:

GPU OC:

 

[miguel1900]

Thank you all!

@3x0 I will look into it, but maybe I already tried it and was not able to achieve, unfortunately, with a generic API (I wouldn't want to use propietary APIs of every vendor). But added to my TODO list, and maybe in a future I will be able to try more complex aproaches, but this will take a while. Thank you.

@QuietBob you are right, thank you. Yesterday I noticed it in the test by @Lew Zealand and I asked him for some feedback. I have (I hope) fixed it some hours ago (already updated to v0.96).

Thank you all too for the feedback provided!



NEW POST:

Updated to v0.96!

Fixed a selectable resolution issue. Reduced VRAM usage. Added 8k resolution option.

Here my results (I finally had time to restar my PC and make a clean test):

RTX 3090 @1440p and 4k:

RTX 3070 Laptop @1080p:

 

[Lew Zealand]

New v0.96 results (using a different GPU than last time, I'll do that one next), Radeon RX 6600 XT 8GB.

Notes: It looks like the RT section isn't engaging the GPU well as power use in the Raster sections is in the typical (reported) 135W range while the RT tests has GPU use at 50-70W. The power use fades during the RT tests from 70-50W, suggesting a bottleneck is slowly developing and I see VRAM use go over 8GB pretty quickly even at 1080p, so it could simply be swapping memory though the test runs smoothly with no obvious frame drops, which is good.

1080p Raster

1440p Raster

1080p RT

1440p RT

 

[miguel1900]

Thank you @Lew Zealand !

I don't know why your card goes above 8GB consumption. I have reduced a bit more the VRAM consumption, if you want to redownload, but I won't be able to reduce it more.

Unfortunately, about the "sleeping" during the RT benchmark (so extrange!) I can't do anything, as it's just Unreal with it enabled. Maybe the GPU gets quite blocked when high VRAM % is used. You could try 1080p combined with 66% TSR upscaling and even with the 1 minute sequence, which consumes slightly less VRAM too (but it won't be comparable with other sequences, as they show different images)

 

[Klemc]

7800x3d 4070ti (pny) here (556.12)

//

I'll do bench, if you can tell me it's a portable app, that won't write stuff in appdata and/or registry, if it does give me the folders dir please so i'll remove them after use ( ihate app writing everywhere).

//
Oh well, it's just a UE app, so it's settings in appdata, just that, i'm on it (oh, it has sound).... OH NO it needs internet, no way for me.
//

BTW:

GeForce 556.12 WHQL driver download & discussion

AW2 with the latest updates. I replaced all three files and the image in the cut scenes became cloudy, plus stutters began to appear in the gameplay....

forums.guru3d.com

+

GeForce 556.12 WHQL driver download & discussion

AW2 with the latest updates. I replaced all three files and the image in the cut scenes became cloudy, plus stutters began to appear in the gameplay....

forums.guru3d.com

The other recommended uses msi installer, no way

 

[miguel1900]

7800x3d 4070ti (pny) here (556.12)

//

I'll do bench, if you can tell me it's a portable app, that won't write stuff in appdata and/or registry, if it does give me the folders dir please so i'll remove them after use ( ihate app writing everywhere).

//
Oh well, it's just a UE app, so it's settings in appdata, just that, i'm on it (oh, it has sound).... OH NO it needs internet, no way for me.
//

BTW:

GeForce 556.12 WHQL driver download & discussion

AW2 with the latest updates. I replaced all three files and the image in the cut scenes became cloudy, plus stutters began to appear in the gameplay....

forums.guru3d.com

+

GeForce 556.12 WHQL driver download & discussion

AW2 with the latest updates. I replaced all three files and the image in the cut scenes became cloudy, plus stutters began to appear in the gameplay....

forums.guru3d.com

The other recommended uses msi installer, no way

Hi @Klemc !

I try to be totally transparent:

The app is portable, yes, so it doesn't write anything over there.

The internet usage is ultra simple: it just allows me to set some settings or to modify it over distance (for example, it could load my "logo.png" texture from my server, so I could easily change it in all distributed benchmarks over there, like I did in my older Pure Ray Tracing benchmark).

However, in this case, the only thing it does is to read the internet date, just to allow me to stablish an expiration date (so you can't change the Windows date to just keep it working).
Why? Well, this is just a rudimentary way to keep my project updated and to be sure that outdated versions of my benchmark will not be used for a long time. That's all.

In addition, come on! I include even my personal full name in the credits, so nothing to hide.

Regards

 

[Klemc]

Hi @Klemc !

I try to be totally transparent:

The app is portable, yes, so it doesn't write anything over there.

The internet usage is ultra simple: it just allows me to set some settings or to modify it over distance (for example, it could load my "logo.png" texture from my server, so I could easily change it in all distributed benchmarks over there).

However, in this case, the only thing it does is to read the internet date, just to allow me to stablish an expiration date (so you can't change the Windows date to just keep it working).
Why? Well, this is just a rudimentary way to keep my project updated and to be sure that outdated versions of my benchmark will not be used for a long time. That's all.

In addition, come on! I include even my personal full name in the credits, so nothing to hide.

Regards

Oh but my PC has no internet connexion, here is on DeskMini i access web, the Desktop PC is totally offline, if there is a way i'll do all benchs possible.

7800x3d
4070ti
1440 27" 240hz OLED DHR...

Thanks for the aswer.

it doesn't write anything over there

Yes, that's true, perhaps his *.ini settings in an appdata like all UE apps but idc.

OOPS, edit:
would i disable Process Laso ?.. if we find a way to have it working offline !

 

[Dahita]

Thanks for the bench!