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

GameTechBench GPU benchmark is already out!

A prototipe for the new scene is ready! If you are curious, you can test it now and I would be glad to read your feedback, please!

Expect lower performance, but I would say higher quality. When migrated to Unreal 5.5, if they fix some bugs in time, performance will improve a little, but quality too.

7,3GB Download (valid until 25th nov.): https://drive.google.com/file/d/1o3gkdh6_mdqGPIOjB-kQoWYgWuJ4HeEV/view?usp=drive_link

View attachment 372339

View attachment 372340

View attachment 372341

View attachment 372342

What do you think? Does it worth the change?

Thank you very much!
Loose the half-naked dude and yeah, it's great!
 
9700x PBO
RTX 4080 Super

GameTech_Path_trace.png
GameTech_R_2k_57.png
GameTech_R_4k_33.png
GameTech_RT_2k_50.png
GameTech_RT_4k_23.png
 
9700x PBO
RTX 4080 Super

GameTech_4k_R_1866.png
GameTech_4k_RT_1343.png
GameTech_2k_R_4803.png
GameTech_2k_RT_3005.png
GameTech_FHD_4803.png
GameTech_FHD_RT_3994.png
 
Thank you all for your tests!!

You may wish to check out Guru3D, MajorGeeks and our very own TPU to see if they'd be interested in hosting. So far this seems legit.
@W1zzard or other TPU admin, Maybe?
Thank you @lexluthermiester , I will consider that option too. For the moment, I am more focused in the development, but I will look those, if they want.

I don't get the 8GB RAM joke @stahlhart ! Could you elaborate? Not very updated about "english happenings". But I think is something wrogn with your 3070 ti test. Look at my laptop:
RT_2024-11_26-15-26-25.png
 
If you're a regular viewer of HU (which is an Australian channel), you would note that this is a pet peeve of the host with respect to Nvidia midrange cards, to the point that you could almost make a drinking game out of it. It was the first thing that came to mind as I was attempting the benchmark at native resolution on that build and watched the GPU getting its ass kicked.

And, speak of the devil...

Could you elaborate on what seems to be wrong...? That first score is 1440p -- and the second is 1080p, which is higher than the laptop score you posted.
 
1440P, Raster and Raytracing Lumen, 2 minute presets
View attachment 353915View attachment 353916
Almost the same system in Fedora Linux with Proton-GE (first Raster, then Ray)

Bildschirmfoto vom 2024-11-27 19-21-19.png

Bildschirmfoto vom 2024-11-27 19-17-19.png

As you can see, the program doesn't utilize the memory correctly, resulting in performance-loss, in Raytracing heavier than in Raster. I wonder what's that about.
 
If you're a regular viewer of HU (which is an Australian channel), you would note that this is a pet peeve of the host with respect to Nvidia midrange cards, to the point that you could almost make a drinking game out of it. It was the first thing that came to mind as I was attempting the benchmark at native resolution on that build and watched the GPU getting its ass kicked.

And, speak of the devil...

Could you elaborate on what seems to be wrong...? That first score is 1440p -- and the second is 1080p, which is higher than the laptop score you posted.

Thank you for the clarification!

My fault about comparing both benchs. I was thinking in my QHD desktop resolution when looking at my FHD laptop benchmark.

But still I think there is something strange there: looking at your graph it seems to be too much flat after the first shot (15th second), however this can be just a percention (seeming flat) depending on the Y axis values, but anyway there is too much difference between the first shot (seconds 1 to 15) and upcmoing shots, after second 15. In summary, all graphs should look very similar, just scaled to their own vertical axis values, but that certain graph is so disruptive.

Almost the same system in Fedora Linux with Proton-GE (first Raster, then Ray)

View attachment 373498

View attachment 373497

As you can see, the program doesn't utilize the memory correctly, resulting in performance-loss, in Raytracing heavier than in Raster. I wonder what's that about.

Hi @leonavis, what do you mean about not utilizing it correctly? Could you elaborate a little more, please? Of course Ray Tracing will suffer more than Raster.
 
Hi @leonavis, what do you mean about not utilizing it correctly? Could you elaborate a little more, please? Of course Ray Tracing will suffer more than Raster.
I mean that it didn't even use 2GiB of Ram (with 64GiB available) and under 7 GiB of VRAM (of 16GiB).

Although, looking at the 4090 result, that might be normal. Dunno! In games like Witcher or Cyberpunk the Raytracing is by now on par with Windows, having - all in all - a few more FPS in Linux.
 
But still I think there is something strange there: looking at your graph it seems to be too much flat after the first shot (15th second), however this can be just a percention (seeming flat) depending on the Y axis values, but anyway there is too much difference between the first shot (seconds 1 to 15) and upcoming shots, after second 15. In summary, all graphs should look very similar, just scaled to their own vertical axis values, but that certain graph is so disruptive.

My assumption would be that the first run was corrupted by the card running out of VRAM virtually out of the gate (hence the HU joke), but I can try running it again, and add a raster run alongside raytracing.
 
I mean that it didn't even use 2GiB of Ram (with 64GiB available) and under 7 GiB of VRAM (of 16GiB).

Although, looking at the 4090 result, that might be normal. Dunno! In games like Witcher or Cyberpunk the Raytracing is by now on par with Windows, having - all in all - a few more FPS in Linux.
I see now what you mean, thanks!

RAM and VRAM are not related and may show quite different values, as they are independent. RAM usage is quite low in this bench, yeah, and lectures seem correct. It's only the RAM used by the process itself. The same for VRAM.

About the performance drop, did you mean comparing your results with the one you quoted? The benchmark has changed, so results are usually not comparable with previous versions.

My assumption would be that the first run was corrupted by the card running out of VRAM virtually out of the gate (hence the HU joke), but I can try running it again, and add a raster run alongside raytracing.
Thank you @stahlhart , that would be nice.

Anyway, it seem it didn't reach the max VRAM of your card (but maybe added to VRAM consumed by other programs did it).
 
Here's my results with the current version. This benchmark makes my GPU sweat like the guy in the garage :laugh:

Lumen in software (raster):
raster_1080p.jpgraster_1440p.jpgraster_2160p.jpg

Lumen in hardware (RT):
RT_1080p.jpgRT_1440p.jpgRT_2160p.jpg

PT:
PT_1080p.jpg

Thanks for your continued efforts to improve the app! Props for syncing the menu to the refresh rate.

And two small suggestions:
- Is it possible to keep the last selected preset after a benchmark run? Currently it keeps resetting to RT with native screen resolution.
- Perhaps adding some background to the menu would make it easier on the eyes? The solid white is almost blinding.
 
Here's my results with the current version. This benchmark makes my GPU sweat like the guy in the garage :laugh:


Thanks for your continued efforts to improve the app! Props for syncing the menu to the refresh rate.

And two small suggestions:
- Is it possible to keep the last selected preset after a benchmark run? Currently it keeps resetting to RT with native screen resolution.
- Perhaps adding some background to the menu would make it easier on the eyes? The solid white is almost blinding.
Thank you @QuietBob !

Hahaha almost every GPU cries here.

Oh, you noticed the vsync in the menu :respect:

- Sure! I will make it remember the settings and will include a button somewhere to reset them. I already thought it and decided to reset them every time, but it's much better to have the option.
- Do you see a white background?:
Captura de pantalla 2024-11-27 23.35.42.png
 
Yep, that's what it looks like on my end:
View attachment 373556
Ouch!

Someone else with the problem?

I see it as it should be in my two computers. Do you notice some kind of "smoke" moving around when you drag the mouse? (It's an effect I did).

And are you using anything "special" regarding the image in Windows? Like colors, HDR... anything could help to think about.
 
Do you notice some kind of "smoke" moving around when you drag the mouse? (It's an effect I did).
And are you using anything "special" regarding the image in Windows? Like colors, HDR... anything could help to think about.
No cursor trails to be seen. I use standard system colors, no HDR. Maybe it's a driver issue?

And I just noticed the "walk mode". Great addition! :clap:
But it's really easy to miss, because you can only get there from the summary screen. Would you consider adding this feature to the main menu, so it gets more publicity? I also think including an overlay with basic controls would make this mode easier to navigate.
 
Ouch!

Someone else with the problem?

I see it as it should be in my two computers. Do you notice some kind of "smoke" moving around when you drag the mouse? (It's an effect I did).

And are you using anything "special" regarding the image in Windows? Like colors, HDR... anything could help to think about.
Yeah, I too have that. In Linux though.
 
Back
Top