Folding GFX card ranking?

[W1zzard]

check one of my recent vga reviews on techpowerup. folding performance is part of our test suite

[PoppaGeek]

check one of my recent vga reviews on techpowerup. folding performance is part of our test suite

That's what I use. Found nothing better.

[W1zzard]

That's what I use. Found nothing better.

define "better". suggestions are always welcome

[MikeB12]

check one of my recent vga reviews on techpowerup. folding performance is part of our test suite.

define "better". suggestions are always welcome

now that you mention it. How do you get around the wu type dilemma when publishing your performance report? if it's not the same wu, then there is no control group in the data published to produce comparable results between cards; or cores if the card is a dual core card on 2 clients. DO you even ppd bench the dual core cards on 2 clients for combined ppd? or is it just whatever wu's were available when the card is tested? if that's the case, then the data is just a snapshot and the cards' data are completely independent of each other and not comparable as a performance bench other than being a individual isolated ppd number for that individual wu and clone in time and scope of the fah project (which changes over time anyway (what wu's are available today, are not available a month from now).

look at all the varying results for just a handful of cards. the wu type variations completely overwhelm the number of models nvidia has.
here's one of our old ppd threads...
http://www.xtremesystems.org/forums/...d.php?t=217372


PG needs a standard benching mechanism like boinc has to overcome this dilemma. until they have it, these performance benches really dont give anyoone comparable data. it's the same concept as someone running a crysis bench and then using that data to determine how it will perform in all games. you just can't make that assumption with any degree of certainty because a bench in crysis will not automatically apply to all games.

[WFO]

now that you mention it. How do you get around the wu type dilemma when publishing your performance report? if it's not the same wu, then there is no control group in the data published to produce comparable results between cards; or cores if the card is a dual core card on 2 clients. DO you even ppd bench the dual core cards on 2 clients for combined ppd? or is it just whatever wu's were available when the card is tested? if that's the case, then the data is just a snapshot and the cards' data are completely independent of each other and not comparable as a performance bench other than being a individual isolated ppd number for that individual wu and clone in time and scope of the fah project (which changes over time anyway (what wu's are available today, are not available a month from now).

look at all the varying results for just a handful of cards. the wu type variations completely overwhelm the number of models nvidia has.
here's one of our old ppd threads...
http://www.xtremesystems.org/forums/...d.php?t=217372


PG needs a standard benching mechanism like boinc has to overcome this dilemma. until they have it, these performance benches really dont give anyoone comparable data. it's the same concept as someone running a crysis bench and then using that data to determine how it will perform in all games. you just can't make that assumption with any degree of certainty because a bench in crysis will not automatically apply to all games.

QFT

Mike is right! Your benchmarks are close Wizard. You should add + or - 2K as a disclaimer. I respect your efforts! Those at F@H are A-holes for not providing accurate benchmarking programs for people to make informed decisions.

For those that haven't seen Wizard's efforts at Techpowerup...