NVIDIA GTX TITAN vs. SLI & Crossfire Comment Thread

[SKYMTL]

NVIDIA's GTX TITAN is the highest priced graphics cards on the market so parallels will naturally be drawn between it an dual card setups. In this article, we look at now it lines up against SLI and Crossfire setups of the GTX 680, GTX 670, HD 7970 GHz Edition and HD 7950 Boost, all of which happen to be less expensive.

Read more here: NVIDIA GTX TITAN vs. SLI & Crossfire

[ipaine]

Very interesting results. About what I expected and all of them beat my once mighty 580SLI setup. I know you didn't have a second Titan to be able to do Titan SLI testing, but I am wondering never the less, do you think that the Titan would experience the same degradation in frame times as the other SLI solutions? Or would it be significantly better than other SLI/crossfire solutions due to it being so good in the first place?

[Bungwirez]

Great Review Skymtl. Frame Time metrics education is very enlightening. thanks, now I can just hold off getting a second 680

[ccm001]

Great review SKYMTL!

I liked the frame time metrics. If AMD could find a fix for their crossfire stutter issues they would easily increase their sales even more! If there was no or very minimal stutter issues people might flock to get a crossfire 7950 setup and save big $$.

P.S. I noticed a slight misprint in the review; the last sentence of the Crysis 3 page states "Against the HD 7970 Boost and GTX 670 SLI it does much better. " - obviously it should say HD 7950 Boost.

[IRQ Conflict]

Sweet! Good review Sky. Would you speculate as to the cause of the extreme frame time lag? Could it be solved with better drivers or is this mostly a hardware issue? I have heard that in some cases CPU's can be blamed for some of the frame time issues, notably hyper-threading enabled CPU's.

I have also seen some people report better (smother) gameplay in some titles with HT enabled. Doesn't seem to be ant rhyme or reason for it. No one seems to be able to tell for sure what the problem is, just that it's there with some titles and not others.

Edit: I remember playing Far Cry 2 a few years ago and with a single 4890 I had to either turn off DX10 and run in DX9 mode and/or cap my frame rate to 30 in order for the game to stop stuttering. Wild fluctuations in FPS iirc. Not sure if that is similar to frame time/micro stuttering or not.

[SKYMTL]

Originally Posted by ipaine

Very interesting results. About what I expected and all of them beat my once mighty 580SLI setup. I know you didn't have a second Titan to be able to do Titan SLI testing, but I am wondering never the less, do you think that the Titan would experience the same degradation in frame times as the other SLI solutions? Or would it be significantly better than other SLI/crossfire solutions due to it being so good in the first place?

I am working my ass off in order to get another TITAN in but I haven't had any luck yet. NVIDIA sent out all their TITANs with the prebuilt systems to reviewers. I asked for the SFF chassis which means I ended up getting only one. Even ASUS and EVGA are pretty much coming up dry these days.

If I had to make a prediction, it would be that two TITANs in SLI would exhibit the same issues as two lesser cards. HOWEVER, the actual visual effect would be minimized since extremely high framerates would be achieved (and thus lower overall frametimes) but there would still be some issues in the games I mentioned in the review.

With a TITAN SLI setup am most interested in the large buffer's effect on those large 70ms++ hesitations SLI exhibited in Hitman and some other titles. It could very well be they are due to the memory buffer being unable to keep up with certain processing requests, which means the TITAN should minimize them due to its massive 6GB of memory.

Originally Posted by IRQ Conflict

Sweet! Good review Sky. Would you speculate as to the cause of the extreme frame time lag? Could it be solved with better drivers or is this mostly a hardware issue? I have heard that in some cases CPU's can be blamed for some of the frame time issues, notably hyper-threading enabled CPU's.

I have also seen some people report better (smother) gameplay in some titles with HT enabled. Doesn't seem to be ant rhyme or reason for it. No one seems to be able to tell for sure what the problem is, just that it's there with some titles and not others.

I have no idea why AMD's solutions stutter so much. I've heard some back-room buzz that there's communication issues between their primary processing stages and the memory controller but those are simply rumors right now.

With that being said, I have indeed encountered higher instances of stuttering on AMD systems but there's just no way to tell whether that was due to the lower starting framerates or something else.

That brings up an interesting test though: on an AMD vs Intel basis, with framerates being equal, which will display higher frame times? Looks like my next article has been decided....

[IRQ Conflict]

Can't wait. Should be interesting if not enlightening.

[beck2448]

Great review. Frame time latencies are the new frontier. I have even read reviews where Crossfire fps numbers were inflated by runt frames, or slivers of pixels, yet still counted in the FRAPS fps number as a normal frame. When these runt frames were removed from the count, Crossfire was no faster than a single card. Your data bears that out, as these big fps numbers don't deliver a better real world experience.

[great_big_abyss]

Originally Posted by SKYMTL

I am working my ass off in order to get another TITAN in but I haven't had any luck yet. NVIDIA sent out all their TITANs with the prebuilt systems to reviewers. I asked for the SFF chassis which means I ended up getting only one. Even ASUS and EVGA are pretty much coming up dry these days.

What about teaming up with another review website, asking that if they send you theirs for a month to do SLI testing, you'll send them yours for a month, so they can do SLI testing.

Originally Posted by SKYMTL

That brings up an interesting test though: on an AMD vs Intel basis, with framerates being equal, which will display higher frame times? Looks like my next article has been decided....

Looks like the current matchup that people are interested in is 8350 vs 3570K. The video from TekSyndicate was very interesting, and Linus just put out a video testing the same thing. Linus's results weren't really in-line with Tek's, but nonetheless, the 8350 did okay. The fault with both those videos though, was lack of detail. It would be nice to see an article with both chips, using a cost relevant card (so, not a Titan, but something that an average user would buy with those chips...a 660Ti, for example) with multiple benchmarks at multiple settings, and lots of glorious numbers to pore over.

[IRQ Conflict]

I would like a link to those "reviews". I seriously doubt it.