HD6850 VS HD6950 tested: 20 games on 1600x1200

Oh Im a smartass... ... and do have data.

So, if the frame buffer was not the key difference, what is it then? Memory bandwidth (another thing) difference is 30 Gb/s - too small for a notice?

Perhaps it is the texture filtrate? Because HD6950 has 40 texture units more?

I mean, if you bring up examples, dog, you should say your opinions, not just bring the facts.

If I had a decent thought on why, you know I would voice them (judging from your salty opinion of my posts). Your guesses are solid though. I mean, obviously, the 6950 is a more powerful card so logic would take me in the direction you have already went as far as texture units/fill rates, since it appears memory capacity, especially at lower resolutions appears to have a negligable affect if any at all.

Truth be told, I do not know a lot of how these cards tick. I just knew what you stated was contrary to others that have reviewed the same subject as I have read up on them. You asked for proof, and I provided it.

EDIT: Its also possible that the test was bunk. Are your results repeatable? Did you only test once and record the result or did you take a sample of results and discard the anomolous ones (lowest/highest results). When I/staff here review GPU's we make at least 3 runs to make sure a single results isnt anomolous.

EDIT2: Onus is on you, as the person providing the results, to also provide the answers/support of your results.

EDIT3: Ok last edit p)... Your testing is awesome and trust me I KNOW how much effort goes in to getting results like this. The only item I questioned was the 2GB of ram making the difference at a relatively low resolution.

EDIT4: After some crtical thinking, see if you can clock up the memory speeds on the 6850 to match the 6950. That will give you a pretty decent idea on if its bandwidth limited between the two cards... If it doesnt change much, chances are bandwidth has little to do with it, If the results change dramatically, chances are it is a bandwidth thing. Im a betting man and I believe that AA uses memory bandiwidth more so than textures. I would crank the AA to 16x on both cards to stress the heck out of them and then overclock the memory on the 6850 and see what happens.

EarthDog said:

Are your results repeatable? Did you only test once and record the result or did you take a sample of results and discard the anomolous ones (lowest/highest results).

Click to expand...

I made sure to check several times, that the results would have similar pattern.
For example, when i first i checked Necrovision with it's own in-game benchmark, and then i redid the in-game benchmark again - i saw very different results. But when i did my own custom benchmarks several times, the pattern was similar.

Now, i could overclock the memory on HD6850 - that would be an overclock alright..
Also 8X AA is the maximum AA in all of the games, there is no 16X AA - i only saw that in several games, when nvidia cards were inserted, not ATI.

Try forcing it through the control panel. In Nvidia cards its "enhance the application setting". If not, just try 8x and see if its bandwidth.

Also, since ATI cards cannot run PhysX calcs on them, Im curious about this line:

Guess there is no need for NVIDIA card in your other PCI-E slot when you have HD6950.

Click to expand...

If the cards are not doing the calculations, then why would the card matter? Its all done on the CPU. I wonder about why those results ended up that way...

I dont know, when I see 200-300% differences between cards that are from 20-40% apart depending on the game, that makes me scratch my head a bit at those results. Now, I dont have any theories as to why this happened outside of them being anomolous results. But what I do know is those results just dont seem to be in line with other reviews, as far as % difference goes.

Thanks in advance for at least humoring me in trying to understand your results. Please try not to take offense to my questions.

Nice post,be nice to see the 6870 compared

Interesting disscussion, was looking into the 6850 or the 6950. 1gig card vs a 2gig card - plus 2-6850 vs 1 6950. Not sure if this forum has helped me that much but bigger and badder is always good but is it better for the cost??? Well I will keep reading, and by the way smarky agruing doesn't help anyone but i know how gamers are, been there done that.

Case - 932 HAF Coolmaster
CPU - 990X I7 6core
Cooling - Corsar H80
MB - Asus Sabertooth
Ram - 24 gig's of 1600 Rippsaw
Video card - 2 GTX 460 1gig in SLI
Hard Drive - 120gig SSD sata 3