[sherlock]

Source

Test set-up





Read the Source for more game(Metro 2033/Skyrim/AVP/F1) tested.

 

[Dreamxtreme]

And this is exactly why i went with this

 

[CrazyHeaven]

Not really question I was ever asking. 3 way sli sounds like more trouble than it's worth. I do like the results of the dual sli 660 ti's.

 

[GerBem]

In the second diagram the 680 SLI is better, or am I reading it wrong?

 

[hzac]

Quote:

Originally Posted by GerBem

In the second diagram the 680 SLI is better, or am I reading it wrong?

Yeah youre right. There is harsh FPS drops on the 660ti tri sli.

 

[kpriess]

I'm assuming these are Ti+s right.. I can't see 2GBx3 handling that resolution that well..

 

[th3illusiveman]

I thought TH said they would include frametimes in their GPU reviews... guess they threw that out the window.

 

[Rayleyne]

Quote:

Originally Posted by th3illusiveman

I thought TH said they would include frametimes in their GPU reviews... guess they threw that out the window.

In order to acheive 90 FPs you need to have a specific frame time.... just saying.

 

[XiZeL]

yeah looks nice on paper and numbers but just doesn't cut it, when you drop just short of a grand on GPU you want to be playing a those hi res at max settings, the 660ti just don't cut it.

 

[chaogui]

What I find funny is that the 660 Ti scaled better in three-way than they did in two-way SLI.

 

[DiNet]

Quote:

Originally Posted by chaogui

What I find funny is that the 660 Ti scaled better in three-way than they did in two-way SLI.

How did you found that out, if that's not a secret?

Price/performance ratio is much better if going 660 route. But... space and SLI... games that won't use SLI and other software... And I'm curious, why they test this on old games?

 

[bojinglebells]

Quote:

Originally Posted by CrazyHeaven

Not really question I was ever asking. 3 way sli sounds like more trouble than it's worth. I do like the results of the dual sli 660 ti's.

3 way SLI supposed to have less microstutter, but with the Titan so close I'd rather go that route

 

[chaogui]

Quote:

Originally Posted by DiNet

How did you found that out, if that's not a secret?

Price/performance ratio is much better if going 660 route. But... space and SLI... games that won't use SLI and other software... And I'm curious, why they test this on old games?

I was always under the impression adding more cards into SLI/crossfire would always give less return than the previous card added. i.e 100% for the original card, +95% for the second card, + 80% for the third card and so on...

However looking at that Battlefield 3 test in particular, at 5760 x 1080 the second card only gave an extra +31% in FPS relative to a 1 card setup. However, the third card somehow managed to give a +45% boost. At such a high resolution with those AA settings, I would have expected the 660 Ti to be so heavily bottlenecked by VRAM that the third card wouldn't give much improvement at all. I have no idea how it managed to give more of a FPS boost than the second card did, could an OCN pro please explain?

 

[DiNet]

Quote:

Originally Posted by chaogui

I was always under the impression adding more cards into SLI/crossfire would always give less return than the previous card added. i.e 100% for the original card, +95% for the second card, + 80% for the third card and so on...

However looking at that Battlefield 3 test in particular, at 5760 x 1080 the second card only gave an extra +31% in FPS relative to a 1 card setup. However, the third card somehow managed to give a +45% boost. At such a high resolution with those AA settings, I would have expected the 660 Ti to be so heavily bottlenecked by VRAM that the third card wouldn't give much improvement at all. I have no idea how it managed to give more of a FPS boost than the second card did, could an OCN pro please explain?

Not really sure it does work in % rate. Drivers+game optimization. Look at metro benches.
Plus they didn't provide single 660ti bench in this graph. Drivers change, game updates change fps too, so looking at older benches doesn't give precise numbers.

P.S. Your % numbers are way too high anyway

 

[chaogui]

Quote:

Originally Posted by DiNet

Not really sure it does work in % rate. Drivers+game optimization. Look at metro benches.
Plus they didn't provide single 660ti bench in this graph. Drivers change, game updates change fps too, so looking at older benches doesn't give precise numbers.

P.S. Your % numbers are way too high anyway

My bad, my graph reading skills must be extra fail today. I looked at the first few benchmarks and assumed that the last one was a 660 Ti, only see that it is a 680 after you pointed it out. Now I feel silly

 

[sherlock]

I am kinda disappointed Tom's didn't throw in another 2 GB GTX 680 to do SLI with so we can see benches of 680 2GB SLI vs 4 GB SLi at 5760X1080, when it is single card vs single card the difference is less than 1 FPS at that resolution and the FPS graph looks the same. I know there are other site that did the 2GB vs 4GB in SLI but wish we had more data.