GTX 590 reviews

[E.R]

what sticks out to me now is just how broken SLI is in stalker and metro. something about those darn eastern europeans. it is clearly not the 1.5gb of memory because 580 and 6970 are close just like they're supposed to be. but the SLI pairs suck donkey dong compared to the crossfire pairs.

if a reviewer compares their 6990 vs 580 results to their 580 vs 580 sli and 6970 vs 6970 crossfire results, I expect they will find that all the games where 590 sucks compared to 6990 are the games where the 580 sli scaling sucks. meaning some software guys somewhere need to fix their crap.

40,1 vs. 60,1 for Stalker in SLI and 28,74 vs. 52,31 for Metro2033 in SLI dont look that bad but not perfect either.

[bamtan2]

you compare a 580 clocked at 772mhz to a dual 580 but who are clocked at 607mhz .. you can't use the 580 score as comparaison... I don't say you are wrong, but it's really hard to compare with thoses numbers.

40,1 vs. 60,1 for Stalker in SLI and 28,74 vs. 52,31 for Metro2033 in SLI dont look that bad but not perfect either.

Let me help you understand what I mean.

SLI scaling in crysis is 85%. But in metro it is 64%.
CF scaling in crysis is 77%. But in metro it is 80%

By scaling I mean how much extra performance is gained by adding 100% extra (one more) graphics cards. SLI is unusually bad in metro. This explains why reviewers are complaining about GTX 590 performing well in some games, but bad in others like metro. The SLI scaling is bad.

Look at the anandtech numbers yourself. GTX 580 beats 6970 in metro. But when you SLI it, it loses.

crysis numbers: http://images.anandtech.com/graphs/graph4239/36042.png
metro numbers: http://images.anandtech.com/graphs/graph4239/36049.png

CONCLUSION: SLI SCALING = BAD IN METRO AND STALKER

[E.R]

Let me help you understand what I mean.

SLI scaling in crysis is 85%. But in metro it is 64%.
CF scaling in crysis is 77%. But in metro it is 80%

By scaling I mean how much extra performance is gained by adding 100% extra (one more) graphics cards. SLI is unusually bad in metro. This explains why reviewers are complaining about GTX 590 performing well in some games, but bad in others like metro. The SLI scaling is bad.

Look at the anandtech numbers yourself. GTX 580 beats 6970 in metro. But when you SLI it, it loses.

crysis numbers: http://images.anandtech.com/graphs/graph4239/36042.png
metro numbers: http://images.anandtech.com/graphs/graph4239/36049.png

CONCLUSION: SLI SCALING = BAD IN METRO AND STALKER

Correct me if i am wrong but it smells Vramlimit as hardwarecanuck show over 82% and metro2033 uses over 1,4GB in 2560x1600 with light settings and anand uses more AA right? so if they used the 3GB in reviews it would be more interesting as the limit would not be a factor that can impact to mutch even on highest setting as Metor dont uses 3GB in 2560x1600.

[bamtan2]

Correct me if i am wrong but it smells Vramlimit as hardwarecanuck show over 82% and metro2033 uses over 1,4GB in 2560x1600 with light settings and anand uses more AA right? so if they used the 3GB in reviews it would be more interesting as the limit would not be a factor that can impact to mutch even on highest setting as Metor dont uses 3GB in 2560x1600.

the memory has nothing to do with the sli scaling I am talking about. any damage done by low memory is done to both configurations, the single and the sli. the performance gained by sli is lower than the performance gained by crossfire. but only in some games.

[E.R]

the memory has nothing to do with the sli scaling I am talking about. any damage done by low memory is done to both configurations, the single and the sli. the performance gained by sli is lower than the performance gained by crossfire. but only in some games.

It is not lower, just differtent than in that anand review as hardwarecanuck shows other things and in same leauge as CF in same game as anandtech.