New GameTech GPU benchmark. Share your results!

[miguel1900]

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.


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 !

Oh... currently it's not possible, I'm sorry. At least not in this early stage.

Couldn't you share mobile data with that PC, just during the startup logo, to pass the internet date check? I would love to see your results over here.

But what did you mean about the msi installer in your first message and about the Process Laso in your second message? Maybe you know something I ignore.

Thanks! (And apologizes too!)

 

[Klemc]

Oh... currently it's not possible, I'm sorry. At least not in this early stage.

Couldn't you share mobile data with that PC, just during the startup logo, to pass the internet date check? I would love to see your results over here.

But what did you mean about the msi installer in your first message and about the Process Laso in your second message? Maybe you know something I ignore.

Thanks! (And apologizes too!)

MSI is the other Gravity*** bench, i don't like installers.

Process Lasso could mess the bench, i asked if better to disable it, if you don't know i will disable it.

No, connexion won't be possible. I"ll be patient, could it be OK when 1.0 is out (0.96 actually) ?

//

OKay, found where he posted after i posted in the driver thread, if ever there is returns :

Info Zone - gEngines, Ray Tracing, AI, DLSS, DLAA, DLDSR, TSR, FSR, XeSS, DirectSR and MODS etc.

I'djust turn off RT and take the giant performance boost. Seems the RT is pretty underwhelming. But you can always flick between em and decide for...

forums.guru3d.com

 

[miguel1900]

MSI is the other Gravity*** bench, i don't like installers.

Process Lasso could mess the bench, i asked if better to disable it, if you don't know i will disable it.

No, connexion won't be possible. I"ll be patient, could it be OK when 1.0 is out (0.96 actually) ?

//

OKay, found where he posted after i posted in the driver thread, if ever there is returns :

Info Zone - gEngines, Ray Tracing, AI, DLSS, DLAA, DLDSR, TSR, FSR, XeSS, DirectSR and MODS etc.

I'djust turn off RT and take the giant performance boost. Seems the RT is pretty underwhelming. But you can always flick between em and decide for...

forums.guru3d.com

I also don't like installers . They generate garbage into the PC

Thank you Klemc, I think Process Lasso is not related with my app nor with Unreal.

Maybe yes, but still not sure. I wanted to publish my bench in this early stage before going for holidays in a couple of days, so some future things are still to be clarified. I'm also speaking with Epic, I want to publish this into their store and in Steam too, etc. But some burocracy still to be solved, and it's slow.

Thank you again

 

[Exostenza]

Cool stuff!

 

[miguel1900]

Cool stuff!

Thank you @Exostenza ! Nice to see you here

Minor update to show Used RAM by the process, instead of by the whole system.

Created too an itch.io project page: https://gametechbench.itch.io/gametech-benchmark

I can't edit my first post, do you know why or could anyone quote a moderator to reeplace the Drive link with the itch.io one, please?

Thank you very much!

 

[neatfeatguy]

Thank you @Exostenza ! Nice to see you here

Minor update to show Used RAM by the process, instead of by the whole system.

Created too an itch.io project page: https://gametechbench.itch.io/gametech-benchmark

I can't edit my first post, do you know why or could anyone quote a moderator to reeplace the Drive link with the itch.io one, please?

Thank you very much!

You can send a report on your original post asking for the mods to unlock it for you.

 

[miguel1900]

You can send a report on your original post asking for the mods to unlock it for you.

Oh thank you @neatfeatguy . Done!

 

[A Computer Guy]

Who knew benchmarking abs was so GPU intensive! Wasn't expecting that. I have to go hit the gym now then then try your benchmark later today.

 

[Fr0stX]

Hi Miguel,

I'm one of the people who questionned the always connected on Guru3d.

I get it, you want to have a valid timestamp for the bench results, and you get it from the internet. This is not ideal as, from a certain point of view, and as I pointed out, you are just a single develloper (not meant to be disrespectfull) posting an .exe on google drive, which then upload data freely from systems. These days with everything going on, I admit, I'm a bit paranoid, and tend to block thru my firewall any .exe I' not 100% sure is safe.

TBH, my intent is not to benchmark my system, but rather to stress test my undervolt and I see no added value to the internet connection requirement for that scenario.

So here is my suggestion: How about creating a stability test mode, with infinite runs, that could run offline?

The benchmark mode would still be an online only mode.

Many people still use heaven bench as their overclock/undervolt stress test. This would make your soft a valid, more modern, replacement for it. Plus, it would help promoting your soft, as it would serve more than one purpose.

For the record, pure ray tracing was my go to stress test, until I blocked the exe in my firewall (and then you pulled the plug on the server side...).

 

[3x0]

v0.97 seems to be bugged with Raster setting, the scenes are very dark and the performance is way higher as a result. 7800XT and Adrenalin 24.4.1

 

[miguel1900]

v0.97 seems to be bugged with Raster setting, the scenes are very dark and the performance is way higher as a result. 7800XT and Adrenalin 24.4.1

Whops, you are right! I will look into this as soon as possible. I just changed a couple of very certain cvars which shouldn't have affected this. Thank you for the report!
EDIT: it should be solved now.

Hi Miguel,

I'm one of the people who questionned the always connected on Guru3d.

I get it, you want to have a valid timestamp for the bench results, and you get it from the internet. This is not ideal as, from a certain point of view, and as I pointed out, you are just a single develloper (not meant to be disrespectfull) posting an .exe on google drive, which then upload data freely from systems. These days with everything going on, I admit, I'm a bit paranoid, and tend to block thru my firewall any .exe I' not 100% sure is safe.

TBH, my intent is not to benchmark my system, but rather to stress test my undervolt and I see no added value to the internet connection requirement for that scenario.

So here is my suggestion: How about creating a stability test mode, with infinite runs, that could run offline?

The benchmark mode would still be an online only mode.

Many people still use heaven bench as their overclock/undervolt stress test. This would make your soft a valid, more modern, replacement for it. Plus, it would help promoting your soft, as it would serve more than one purpose.

For the record, pure ray tracing was my go to stress test, until I blocked the exe in my firewall (and then you pulled the plug on the server side...).

Thank you for sharing your feedback about this concern @Fr0stX ! I will consider it, but I need to ask you more questions to understand it better, please:

The number of loops was thought to be used as a stress test too. Anyway, for that very certain mode, do it need to just run infinitely, without any info, or would it need any data, as average FPS, etc? (Being a mode outside of my "internet desire", I would want to include the less options, if not needed, to not be used this mode as a replacement for the benchmark itself).

Would the "internet concerned" people confortable with this, if the software has one offline mode and one online mode? How would they (you) be confortable without knowing if I'm really making the internet check only when starting the certain mode, or maybe when the whole game starts? Is there anything you use to check it and be sure? (Appart of just trying to execute it without internet connection. Well, or maybe this is the usual behaviour of this kind of user, I just don't know).

What is the general opinion about internet connection if the game would be distributed through Steam, for example? Would it be accepted? They (Steam) make a check to avoid inmoral developments, like strange internet connections, so it gives some confort to internet-aware users? Or the concern would be simply the same and this people would still be against it?

BTW, oh, I didn't know anyone was using Pure RT benchmark as reference. So glad! I have thought about restoring it many times, just as a more simple bench, because it was great for overclocking purposes, as it was very very sensible to it to detect unstabilities.

Thanks again!

 

[oxrufiioxo]

I have no issues with the internet connection being required but it probably should be optional if you want more people to want to use the software.

An offline stress mode and an online benchmark seems like the best of both worlds.

I mainly use 3DMarks Timespy/Firestrike/speedway for my benching stress testing and I'm not even sure you can use them offline lol....

That being said people are always going to be paranoid it's human nature.

 

[QuietBob]

Just tested v0.97 and indeed the resource allocation is lower. Max VRAM usage @ 4K raster is 8.83 GB now, rather than 10.35 GB.

I've noticed two typos still:
- in the overlay the "Operating System" should be abbreviated to OS (not SO)
- in the summary it should read "Operating System" (not Operative)

Thanks for your continued work on the benchmark! It seems like another scene may be in the works?

 

[miguel1900]

You are right @oxrufiioxo ! Thank you.

Thank you @QuietBob ! I didn't notice those. Will update in some weeks, after holydays.

Yeah, a new scene is on the way! Based on one of the Epic examples, but improved, optimized, and with additional special assets (grooming). Probably to be released within the final bench version based on UE5.5 (in september-october, I would say), because of a current Unreal bug. And maybe to be included only in a Pro version sold through Steam (all aspects regarding a paid version are still to be determined, so every detail is still open to your suggestions and thoughts).

Thanks!

 

[jiajo]

Salut,

1080P RT

good job thanks

 

[miguel1900]

Salut,

1080P RT


good job thanks

Thank you!

 

[freeagent]

Looks good

1080p RT off and on:



1440p RT off and on:



2160p RT off and on:



Info:

 

[miguel1900]

Thank you @freeagent !

Didn't you want to test the Path Tracing mode?

And can I ask what kind of screen/monitor do you use?

Interesting results performing above a 3090!

 

[freeagent]

Thank you @freeagent !

Didn't you want to test the Path Tracing mode?

And can I ask what kind of screen/monitor do you use?

Interesting results performing above a 3090!

I can try it.. I am using a TV as a screen, I noticed my resolution was a bit different from everyone else...

 

[yzonker]

Only suggestion I have is to allow control of the horizontal resolution. I get an oddball horizontal res on my LG C1.

 

[agent_x007]

I tried to run multiple loops (3), but I keep getting "internet disconnected" error either during benchmark or when program launches
Is it possible that some firewall/request counter, blocks connection due to too many requests (DDoS prevention) ?
Regardless, I need "internet requirement" to be optionally disabled with corresponding "score invalid due to not being connected to internet" text in whatever place.
I don't care about online validity of the score, when I just want to run some tests on stuff.

Also path trace is good test to draw max. power from card (from what I see).
It's similar to Metro Exodus benchmark at Extreme preset 1080p (no DLSS/FSR).

 

[stahlhart]

So much for that. And I donated, too.

 

[miguel1900]

Hi @freeagent @yzonker @agent_x007 @stahlhart and thank you!

About horizontal resolution, you should not worry about it. It's by design. The vertical resolution will be the important one. The horizontal one, will be rendered maintaining the aspect ratio. The rest will be filled with black bars, without hitting the performance result.

About the connection issues, It's due to a slow response from the server from where I check the realtime global date (http://worldtimeapi.org/). It's solved now, leaving a longer time gap to get a valid server reply and added a second/alternative server check. I don't know how this important server can be so slow these days.

Thanks again!

 

[PierreJG]

I am going to try this on my old system just for kicks. One thing I would love about software like this is a cutdown version to test APU and Older GPUs. Yes, there is 3dmark but it is outdated and then there is userbenchmark which is also inaccurate and is biased in changing their benchmark to benefit Intel/Nvidia when Ryzen came out, they changed their CPU benchmark to base the score based on 80% single core. A Cutdown version would be great but I know all of us are busy and writing software is most prob one of the most time-consuming things on a PC.

 

[stahlhart]

Here are the normie results:

2160

1440

1080

I had some occasional hiccups in the process -- a few times, the bench launched with a black screen and then stopped responding, and one run at 1080p had graphics corruption, so I discarded that result.

These three ran one after another smoothly, though.

Thank you very much for resolving the network issue.