What are good benchmarks for GPU now?

[imaheadcase]

I see that 3d mark scores are all tainted so not worth using to compare anymore, is Heaven 4.0 it?

 

[ElFenix]

what games do you play?

 

[railven]

I see that 3d mark scores are all tainted so not worth using to compare anymore, is Heaven 4.0 it?

If you own an AMD card, any AMD Evolved. If you own NV card, any Gameworks.

If you want something more neutral - good luck with that. I've come to learn based on opinions here - everyone is in someone's pocket.

You can do just like me and have a card from both chip makers

 

[Ken g6]

TechPowerUp? Or do you mean a benchmark you can run yourself?

 

[Headfoot]

Agreed with the TPU chart. the best way to compare GPUs is average performance across the widest possible variety of games. Usually, 12ish modern and mostly modern titles will do the job.

Look for frame-time analysis. It's an underappreciated but very important metric.

 

[IEC]

Whatever games you want to play. Shouldn't you base your decision on what matters to you?

In some cases that's going to "skew" nVidia. In some cases AMD. And that's fine. You shouldn't listen to some random FUD artist with an agenda, you should be basing your purchases on objective data about performance in games that matter to you.

The TPU chart of relative performance is a good gauge of the overall picture of performance.

 

[sirmo]

3D Mark because it's easy to compare to other online scores.. as that seems to be the most used one. Caveats do apply as this is only a DX11 benchmark (they are adding a DX12 one I hear).. and as you said it's a bit tainted.. but it's still an ok reference point when you take the caveats in consideration.

 

[imaheadcase]

what games do you play?

All types of games, casual to FPS to MMO. Playing through Dying Light right now, black desert MMO.

 

[imaheadcase]

3D Mark because it's easy to compare to other online scores.. as that seems to be the most used one. Caveats do apply as this is only a DX11 benchmark (they are adding a DX12 one I hear).. and as you said it's a bit tainted.. but it's still an ok reference point when you take the caveats in consideration.

That is the thing though, you can't use it to compare anymore. All the results are tainted.

You can look at the exact same systems, same hardware/os/game settings and one can be 10k score and some joker has 40k score. Just because he using LN2 to overclocked video card and CPU for the test.

The only thing 3dmark does is just verify a legit driver, they don't take in account modifications to hardware.