GameTechBench GPU benchmark is already out!

[yzonker]

Wow! How did you get that outstanding score? It's around 300pts over the second best 4090 user



Nice performance improvement! A pro overclocker was telling me this is a great OC tool as it's quite instability-sensitive too, when the OCs are too high

Chiller + EVC2

 

[miguel1900]

Chiller + EVC2

Really impressive! Would you mind testing in Path Tracing too? Just curious

 

[remekra]

Regarding Path Tracing:

Seems Nvidia really did improve their RT cores. That score is basically on par with 4090 scores. And that's back on stock settings without OC.

 

[miguel1900]

Regarding Path Tracing:

View attachment 383001

Seems Nvidia really did improve their RT cores. That score is basically on par with 4090 scores. And that's back on stock settings without OC.

Yeah! In offline rendering (using Unreal) it's even a little faster than the 4090, which is quite interesting! Curiously, it isn't in render engines like Blender or Vray, which is also a 'plus' for Unreal.

 

[remekra]

Yeah! In offline rendering (using Unreal) it's even a little faster than the 4090, which is quite interesting! Curiously, it isn't in render engines like Blender or Vray, which is also a 'plus' for Unreal.

Must be using some new features that they introduced in RT Cores on Blackwell I think.

 

[remekra]

After lots of stability testing in various software, final OC for now seems to be +400 on core and +2000 on memory. That got me within 1% of average 4090 score so I'm happy

Also not sure if somebody reported it but there seems to be strange issue when first running Path Tracing and then Raster.
Basically if I run PT first and after that Raster (only tried 4K) the benchmark dips to really low fps in two places. First it got me scared that my memory OC is not stable, but then I noticed that in mode tab it says that I'm running PT mode even though it's raster.
So I went back to stock settings and did the same thing and the same thing happened. If I run raster first - no issues.

 

[miguel1900]

Thank you @remekra !

Funny bug . I have reproduced it and it's a typo error during the run. At the end, in the summary screen, it shows the correct mode, but I will fix it for next update.

However, I got the exact same performance when directly running Raster than when doing it after a Path Tracing test. Of course, running any test before another one, could affect the next tests because of temporal files accumulated in GPU memory etc, and it can affect different PCs in different ways, so I don't know if it could mean some kind of weakness or not, but maybe professionals can "read" it as useful information of a PC's health. I can only say it didn't affect at all in my weak 3070 laptop, but I will also try in my 3090 desktop.

Thank you!

 

[miguel1900]

v1.12 is live:

• Improved anonymity protection for reviewers with "secret" GPUs. If you have a secret GPU, you will still see your own results and username, but other users won't see either your username or your score, ensuring complete anonymity.
• A screenshot is now taken automatically at the end of the benchmark. It is saved in your ...\Documents\GameTechBench Results folder.
• Fixed a minor typo that displayed an incorrect benchmark mode name during Raster mode if Path Tracing mode had been previously run.

And a 5070 Ti it's over there!

 

[miguel1900]

Hi all!

A new approach is coming! But any feedback is always welcome (please):

- Game will still have a Free version and a Pro one and current owners will still own the Pro version.
- It will change from base Pro game + limited Demo to Free-to-play limited game + Pro DLC.
- DLC will be just an "unlocker".
- This will allow to have just one installer.
- Free and Pro would receive updates at the same time, and more unified in general.
- Free scores will be also added to the leaderboards. Currently, Free scores are discarded (they are forced to use different leaderboards).
- Could be easier to add future scenes, as independent DLCs.
- Free version would be slightly less limited, detailed below.

Improved Free version would include (feedback specially welcome too):
- A couple of free Pro usage days each month (the 1st and the 15th). This could change randomly but punctually, to promote its use and scores additions to leaderboards.
- A free Pro version usage the very first time you execute the leaderboard.

Cons:
- Haters could review it negatively because they could consider they should own the full pro version for free. (Actually an issue? Even if I actually don't care a lot, as I prefer raising its propularity)


What do you think?

Thank you very much!

 

[miguel1900]

RTX 5070 spotted in!

27% slower than 5070 Ti / 4080. In pair with 3090

First screenshot shows general performances (5090 as 100%):

Second shows 5070 relative (shadowed bar) performance:

Third shows 5070 Ti relative (shadowed bar) performances:

Data is becoming quite accurate, considering the TechPowerUp's all-games averages (but this is only one 'game'! ):

Would also be so nice if you would consider publishing this on Reddit, please!

 

[wNotyarD]

@miguel1900 is the embargo on the 5070 already lifted?

 

[3x0]

@miguel1900 is the embargo on the 5070 already lifted?

No, and it's not him testing it and publishing the results. It's some reviewer who accidentally submitted/leaked the results.

 

[yzonker]

I'm more curious about the 9070 XT. Nvidia is just releasing paper. Although there are a few more 5070 Ti's. Maybe they'll actually make some 5070's to compete with AMD.

 

[miguel1900]

No, and it's not him testing it and publishing the results. It's some reviewer who accidentally submitted/leaked the results.

This is @wNotyarD . I was just informed about. Data is hidden to protect his anonymity until the embargo is over.

I'm more curious about the 9070 XT. Nvidia is just releasing paper. Although there are a few more 5070 Ti's. Maybe they'll actually make some 5070's to compete with AMD.

Me too! If you know somone with a 9070 XT, let him know!

 

[wNotyarD]

This is @wNotyarD . I was just informed about. Data is hidden to protect his anonymity until the embargo is over.


Me too! If you know somone with a 9070 XT, let him know!

Yeah, I noticed the blurried names. I just don't recall if the card results were blurried as well when 5090 and 5080 were first tested with your benchmark.

 

[miguel1900]

Yeah, I noticed the blurried names. I just don't recall if the card results were blurried as well when 5090 and 5080 were first tested with your benchmark.

I see. This time I blurred the whole Top players chart just to center the focus in the other two, but only the "secret" usernames are hidden, with "*****", the same as their scores, to not be recognizables.

All those private names were hidden since the beginning in different ways, until each embargo date. It was a must have to protect the reviewers.

 

[miguel1900]

9070 XT on par with the 5070 Ti.

Trustable data, matching the same % gap published by VidecardZ versus the 7900 GRE and 6900 XT:

 

[Arctucas]

New version Steam only?

 

[miguel1900]

New version Steam only?

Hi @Arctucas !

From final Release v1.00, yes. For the moment I have no the time or resources to extend it over other platforms (not easy), but I hope to do it in the future, as well as a limited but standalone version.

BTW: current leaderboards use the integrated Steam's Leaderboards API. Additionally, and just to clarify (if it were your concern), Steam doesn't affect the performance of the benchmark at all, confirmed.

 

[Arctucas]

Hi @Arctucas !

From final Release v1.00, yes. For the moment I have no the time or resources to extend it over other platforms (not easy), but I hope to do it in the future, as well as a limited but standalone version.

BTW: current leaderboards use the integrated Steam's Leaderboards API. Additionally, and just to clarify (if it were your concern), Steam doesn't affect the performance of the benchmark at all, confirmed.

I just do not use Steam.

Thanks

 

[miguel1900]

I found the leaderboards system so useful so I decided to integrate it into Steam, first. I think it worths it, iy you don't have other impediments. But I also know some users don't like it and I will try to search alternatives, but with quite a bit time (I wish I could do it faster).

Thank you again @Arctucas !

 

[AVATARAT]

9070 XT on par with the 5070 Ti.

Trustable data, matching the same % gap published by VidecardZ versus the 7900 GRE and 6900 XT:

Why does the benchmark detect the 9070 XT with 15GB VRAM?

 

[miguel1900]

Why does the benchmark detect the 9070 XT with 15GB VRAM?

It rounds the actual megabytes, so it might have something between 15.00 and 15.49 GBs, actually. (I will round the numbers to ceil, to match expectations and prevent confusions)

 

[lexluthermiester]

Hi @Arctucas !

From final Release v1.00, yes. For the moment I have no the time or resources to extend it over other platforms (not easy), but I hope to do it in the future, as well as a limited but standalone version.

BTW: current leaderboards use the integrated Steam's Leaderboards API. Additionally, and just to clarify (if it were your concern), Steam doesn't affect the performance of the benchmark at all, confirmed.

Sadly, until you do port over to other distribution methods some of us won't be using this latest version. Some of us don't and won't use Steam. A stand alone download is still preferred.

 

[miguel1900]

Sadly, until you do port over to other distribution methods some of us won't be using this latest version. Some of us don't and won't use Steam. A stand alone download is still preferred

I know, and it is indeed sad! That's why I have you on my TODO list, but there are still many things to do first. On the other hand, that version wouldn't have leaderboards, and some anti-piracy measures would need to be considered. Another option could be making only the limited free version available for distribution, but unfortunately, all of this will take time, as I mentioned. But I won't forget this.