MSI OCed GTX280 for $459 on newegg

[deadseasquirrel]

Originally posted by: chizow
As for graphs showing SLI being worst than a single GTX 280 in some cases, I think this is what you're looking for:
GTX 280 SLI @ Guru3D
Typically this occurs in CPU bottlenecked games/resolutions, which happens much more frequently with the GTX 280. I believe TweakTown also has some excellent comparisons of GTX 280 SLI/Tri-SLI differences on a 3GHz and 4GHz machine.

I have no doubt that if 10 games are benchmarked, you'll find 1 of them doing what happens here with Quake Wars... where SLI is outmatched by a single card. But that is, like you said, mainly at lower resolutions where the cost of SLI is absolutely wasted.

But, again, my whole complaint is how people throw out that SLI/CF doesn't improve on MIN fps, which recent benchmarks show to be untrue in most situations (exceptions, as always, apply).

Originally posted by: CP5670
Maybe the newer generation cards are different, I don't know.

Maybe, I could see that. I didn't look up benches for any older SLI/CF configs... I just skimmed through the recent ones for gtx280. And man is it hard to find review sites that give MIN fps and test SLI and compare that to the single card at the *same* settings.

So, yeah, previous SLI configs might have shown no MIN fps increase over single cards, but posters continue to say it now without any benchmarks to prove it. As if it is accepted as fact now, yet is proved completely untrue with the benches I linked to.

 

[dadach]

the way this is going, people should wait for total panic attack then buy nvidia...when it gets around 20-30 dollars more than 4870

 

[Munky]

Originally posted by: CP5670

I hear this kinda thing a lot but nobody ever posts links to benchmarks that show this. It's hard enough finding a review that actually shows MIN framerates, but finding one that shows MIN *as well as* CF and SLI compared to their single-card counterparts is even harder.

This can be seen with the 9800GX2 in Xbit's review, where they also have an 8800GTS in the benchmarks. It may be a problem with the 9800GX2 specifically, but n7 and lopri have also noted the same problem with some other setups (3870 X2 and 8800GT SLI, IIRC). Maybe the newer generation cards are different, I don't know.

In that Xbitlabs review there are plenty of examples to make the argument either way. And in some cases Xfire increases min fps in the same game where SLI decreases it and vise-versa. You can't blame that on the system, as they're rendering the same thing, so obviously it boils down to driver-specific behavior and implementation differences between SLI and Xfire.

 

[Cookie Monster]

At the end of the day, it shows that multi GPU setups e.g CF/SLi are all driver dependent since consistency across games is the biggest drawback. That being said, i also finished reading driverheaven.net review and i can tell you that the GTX280 does provide higher min fps (especially at 25x16 with AA/AF) compared to cards in its price range. When the HD4870 was compared with GTX260, the latter had the edge at higher res AA/AF (25x16 most of the time), and provided higher min fps through alot of titles.

It was also evident that there were quite a few cases where the HD4870/GTX260 had similiar avg fps but the former had a higher max fps while the latter had the higher min fps.

 

[evolucion8]

Originally posted by: Cookie Monster
At the end of the day, it shows that multi GPU setups e.g CF/SLi are all driver dependent since consistency across games is the biggest drawback. That being said, i also finished reading driverheaven.net review and i can tell you that the GTX280 does provide higher min fps (especially at 25x16 with AA/AF) compared to cards in its price range. When the HD4870 was compared with GTX260, the latter had the edge at higher res AA/AF (25x16 most of the time), and provided higher min fps through alot of titles.

It was also evident that there were quite a few cases where the HD4870/GTX260 had similiar avg fps but the former had a higher max fps while the latter had the higher min fps.

Jum, the GTX 260 only leads in 6 of 10 games in such resolution in minimum frame rates, but the averages which also matters are in favor of the HD 4870, also the HD 4870 at those resolutions is getting frame buffer limited, 512MB is not enough for that. GRID, Oblivion and Assassin Creed performances on the HD 4870 were astonishing, but at 25x16 diminish considerably (But still faster overall except in Oblivion which the performance dips well below compared to 19x1X because it runs out of VRAM), but Lost Planet runs like crap on the HD 4870, I never understood why that game runs slow on ATi hardware.