Flaws in Video Card Testing?

[cbn]

When shopping for video cards I usually look at the frame rate comparison charts @ X resolution/AA/AF/detail settings.

But how accurate are these tests for making buying decisions?

1. Most (if not all) tests use a quad core processor....but most games don't have 4 threads.

--Could using a quad core CPU during testing mask any benefit from a Nvidia card when the game engine uses Physx? The average gamer with a dual core and might see a larger discrepancy with real world usage if the game engine uses part of the CPU for Physx.

Obviously the above scenario favors ATI cards.

2. Minimum frame rates.

--It seems the most challenging part of the game is what stresses my hardware the most. The more explosions/action.....the more I notice any weakness of my video card. Do average frame rates really give a good all around picture of a Video cards performance envelope (memory bandwidth/VRAM)?

In contrast to #1, I can see the advantage of using a quad core for testing in order to isolate variables.

 

[HannibalX]

I'm only interested in the minimum frame rates. I don't care if the one card has a 20 fps advantage at the top end when they're both within 5 fps on the low end.

 

[cbn]

I'm only interested in the minimum frame rates. I don't care if the one card has a 20 fps advantage at the top end when they're both within 5 fps on the low end.

That is how I feel too.

But if the tester only uses a quad core any real world effect of Physx on a dual core CPU would be missed during testing.

Sometimes I wonder if the Nvidia GPUs may be a better bargain than how they appear to be.....for the average user.

 

[darkswordsman17]

PhysX will only enable if an nVidia card is detected, and I believe is embedded in nVidia's drivers so that you can't just trick it. I could be wrong. Of course, this is actually a bad thing, as while I liked some of the PhysX stuff I saw in videos, quite a bit of it seemed like things that should have been in the game anyways and was coded deliberately to exacerbate the differences.

Well, you're right, average frames don't tell the whole story, but neither does min/max, as it doesn't tell you how long it was at that level. I know some sites post avg and min, and I think there's even some that do charts showing the framerate over the span of their test (so you see the dips and spikes, etc). The important thing is to look at multiple sites and see what the trend is. Generally, while the exact data might not be helpful (using different hardware than you), its usable for comparison (so you can tell that graphics card A is faster or slower than B and C).

 

[cbn]

PhysX will only enable if an nVidia card is detected, and I believe is embedded in nVidia's drivers so that you can't just trick it. I could be wrong. Of course, this is actually a bad thing, as while I liked some of the PhysX stuff I saw in videos, quite a bit of it seemed like things that should have been in the game anyways and was coded deliberately to exacerbate the differences.

I know only Nvidia GPUs can use Physx.....but doesn't the CPU contribute to Physx calculations in the case of someone using a non-Nvidia card? (essentially sapping performance if enough cpu cores are not available)

It seems to me a lot of game engines use Physx now. Even the gamebryo engine (Fallout 3) apparently has it.

 

[darkswordsman17]

I know only Nvidia GPUs can use Physx.....but doesn't the CPU contribute to Physx calculations in the case of someone using a non-Nvidia card? (essentially sapping performance if enough cores are not available)

It seems to me a lot of game engines use Physx now. Even the gamebryo (Fallout 3) apparently has it.

As far as I know, the extra PhysX stuff would only be there if PhysX is enabled, so the only time that any of that extra processing would be needed is if PhysX is running. If its not, then both brands should be running the game roughly the same way.

 

[cbn]

deleted.

 

[GaiaHunter]

When shopping for video cards I usually look at the frame rate comparison charts @ X resolution/AA/AF/detail settings.

But how accurate are these tests for making buying decisions?

1. Most (if not all) tests use a quad core processor....but most games don't have 4 threads.

--Could using a quad core CPU during testing mask any benefit from a Nvidia card when the game engine uses Physx? The average gamer with a dual core and might see a larger discrepancy with real world usage if the game engine uses part of the CPU for Physx.

Obviously the above scenario favors ATI cards.

No.

Only a handful games, most talked being Batman:AA, use GPU accelerated physX effects.

All the other games that use the physX library of effects, use non-GPU accelerated physX, meaning a NVIDIA card has no advantage whatsoever over an ATI card in that scenario.

In conclusion, don't confuse the physX effects library with GPU accelerated physX support - even XBOX 360 that is powered by an ATI chip, run games that use the physX library.

 

[cbn]

No.

Only a handful games, most talked being Batman:AA, use GPU accelerated physX effects.

All the other games that use the physX library of effects, use non-GPU accelerated physX, meaning a NVIDIA card has no advantage whatsoever over an ATI card in that scenario.

In conclusion, don't confuse the physX effects library with GPU accelerated physX support - even XBOX 360 that is powered by an ATI chip, run games that use the physX library.

Thanks. I didn't realize there were different kinds of Physx. (I thought they were all the same)