[Sweclockers] Titan X in SLI

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

SimianR

Senior member
Mar 10, 2011
609
0
81
#26
Taking things like feature set and the displays I use into account, I'd still prefer Kepler atm. An upgrade of displays would change that though.
At the 780 Ti's release price of $699.. I would hope so. The fact that the 290 comes anywhere near, but actually surpasses it in crossfire is insane.
 

Granseth

Senior member
May 6, 2009
258
0
71
#27
NOT a single review site has put Titan X Sli under the FCAT microscope
I wonder why?:sneaky:
you find frametimes ( Renderingstider in swedish) in this review
 

omeds

Senior member
Dec 14, 2011
646
0
81
#29
Not very good frame times there, but that's of little consequence to me. I play BF4 at 1080p with a 147fps cap, at which frame times are very good. Good job from AMD all the same though.

At the 780 Ti's release price of $699.. I would hope so. The fact that the 290 comes anywhere near, but actually surpasses it in crossfire is insane.
It can surpass them at stock clocks and high resolutions sure. Both oc'd at lower resolutions is a different story as I tried to explain in my previous post. Kepler is an old architecture for older resolutions, at which they excelled. They're basically Nvidias Tahiti. Hawaii is a more forward thinking and modern architecture, better suited for 4k and the like.

I actually run vanilla 780 SLI btw, at 1250mhz at 1080p144 and 2560x1080p75, at which they hold their own against oc'd Hawaii just fine. They're great for high frame/refresh rates at lower resolutions, where processing speed is the major factor. As resolution increases and other factors become more limiting, like pixel fill rate, Hawaii pulls away convincingly. You're looking at 64 ROPs vs 48, meaning Kepler has no chance at modern 4k or 21:9 (3440x1440) resolutions compared to Hawaii. Plus 3GB is limiting at those resolutions too.

Anyway, that's why I said with my current displays, and taking into account feature set, TDP, support etc, I'd still prefer to have Keplers atm. A monitor upgrade will change that though, and that will be soon, but by then they'll have served their purpose and held up perfectly fine in relation to what the competition could offer for the same config. My next displays will be 1440p144 and 4k and they'll require new GPU's. I'd like to stay with Nv for feature set and eco system, but will go back to AMD (most of my GPU's have been ATi/AMD) if the 390x proves to be superior for 4k.
 
Last edited:

SimianR

Senior member
Mar 10, 2011
609
0
81
#30
There's definitely a case to be made for both architectures/brands. To be honest, I was running crossfire 290's for some time but found it to be overkill for my needs. In the past few years I've found myself playing less demanding titles (MOBA's.. heroes of the storm atm) and on a 60hz monitor it's even more of a waste for me. If I move to a 4K monitor I'll probably need some serious GPU horsepower. My next build is going to be shoving as many high end parts into a Raven RVZ02 as possible :cool:
 

gamervivek

Senior member
Jan 17, 2011
478
0
76
#31
Large Kepler overclocks quite well in comparison though, but sure Hawaii has certainly improved more since launch, much like Tahiti did. Hawaii really shows its strength in higher resolutions too, largely thanks to ROP count and pixel fill rate. Keplers performance degrades much quicker due to a large deficit in this area, despite generally having faster processing speed (as is evident when fill rates are not impacting performance or creating a bottleneck). That is to say Kepler is better suited for lower resolutions, and Hawaii is better suited for higher resolutions.

Taking things like feature set and the displays I use into account, I'd still prefer Kepler atm. An upgrade of displays would change that though.
The problem with that explanation is that Tahiti begins catching up with Kepler with increasing resolution despite the situation being the exact opposite of what it's with Hawaii and Kepler.
 
Aug 10, 2009
11,951
0
106
#33
Aren't those just Fraps frame times?I'm looking for some FCAT action.
I think it's pretty safe to assume if sites aren't using nVidia's own bench tools it's because they would lose that comparison.
 

Eymar

Golden Member
Aug 30, 2001
1,645
0
81
#34
Nvidia really needs to fix SLI frame pacing. G-Sync fixes most frame pacing issues I've seen with 980 SLI (haven't tried with G-Sync off on SLI TiX's), but can't expect all SLI users to buy a G-Sync monitor.
 

omeds

Senior member
Dec 14, 2011
646
0
81
#35
Yeh, there are work arounds but really there's no excuse to not have it on par with the competition after all this time.

The problem with that explanation is that Tahiti begins catching up with Kepler with increasing resolution despite the situation being the exact opposite of what it's with Hawaii and Kepler.
290x has 56% higher pixel fill rate than GTX780. GTX780 has 28% higher pixel fill rate than Tahiti.

Tahiti and GTX780 both have 288GB/s memory bandwidth, 290x has 320GB/s

As resolutions increase these two factors are likely hindering performance much sooner on the 780 and allowing the gap to grow in relation to 290x, while at the same time to close in relation to 280x.

280x isn't that close and 290x isn't that much faster either, 780/Ti just don't come with "uber" mode out of the box lol. Raise TDP and temp limits by the same amount and similar increases are seen. That's why review sites like computerbase.de are smart enough to test 780, Ti, 970, and 980 at two separate settings - default and "max", performing the exact same function as "uber" mode.

In a review they did about a month ago across 15 games at 1600p, 780 is 13% faster than 280x (Tahiti has done well to even get that close in 280x form given 7970's launch state) and 290x is 6% faster than 780. In "uber" mode the 290x is dead even with a default Ti, and 7% slower than the Ti when set to "max".
At 4k the "uber" 290x gains ground and is 7% faster than default Ti as expected, but 1% slower (I'd say even) than the Ti at "max".

A couple side notes too, 780 has conservative clocks with more oc headroom than Hawaii, my 780's have a ~35% oc on the reference coolers for example. Secondly, every modern review I've seen that includes a 280x is with an aftermarket design.

Make of that what you will, but it all contributes to real world user experience.
 
Aug 10, 2009
11,951
0
106
#36
It's interesting that "max" and "Uber" (which Hardware.fr calls their nVidia settings similar to AMD's Uber) are used to stop throttling, but we hear nothing about nVidia throttling at stock settings from the English speaking sites. Double standards FTW.
 

gamervivek

Senior member
Jan 17, 2011
478
0
76
#37
Yeh, there are work arounds but really there's no excuse to not have it on par with the competition after all this time.



290x has 56% higher pixel fill rate than GTX780. GTX780 has 28% higher pixel fill rate than Tahiti.

Tahiti and GTX780 both have 288GB/s memory bandwidth, 290x has 320GB/s

As resolutions increase these two factors are likely hindering performance much sooner on the 780 and allowing the gap to grow in relation to 290x, while at the same time to close in relation to 280x.

280x isn't that close and 290x isn't that much faster either, 780/Ti just don't come with "uber" mode out of the box lol. Raise TDP and temp limits by the same amount and similar increases are seen. That's why review sites like computerbase.de are smart enough to test 780, Ti, 970, and 980 at two separate settings - default and "max", performing the exact same function as "uber" mode.

In a review they did about a month ago across 15 games at 1600p, 780 is 13% faster than 280x (Tahiti has done well to even get that close in 280x form given 7970's launch state) and 290x is 6% faster than 780. In "uber" mode the 290x is dead even with a default Ti, and 7% slower than the Ti when set to "max".
At 4k the "uber" 290x gains ground and is 7% faster than default Ti as expected, but 1% slower (I'd say even) than the Ti at "max".

A couple side notes too, 780 has conservative clocks with more oc headroom than Hawaii, my 780's have a ~35% oc on the reference coolers for example. Secondly, every modern review I've seen that includes a 280x is with an aftermarket design.

Make of that what you will, but it all contributes to real world user experience.
Where are you getting the numbers from, and it seems to be going by the base clock numbers, 780 surely boosts beyond it to make it a level 50% increase in ROPs from Tahiti. Besides if they're still using the uber mode on 290X, they should upgrade to a custom cooler one.
 

omeds

Senior member
Dec 14, 2011
646
0
81
#38
Numberz r from teh webz. 780 boost clock is 902mhz. Max boost can vary quite a lot. I disabled it and just set +350 constant. Reference coolers have their place, especially in MGPU - I actually prefer reference (providing it can do the job).
Tahiti has boost clocks also. In fact it's the first AMD GPU to use it.
 

Granseth

Senior member
May 6, 2009
258
0
71
#39
Aren't those just Fraps frame times?I'm looking for some FCAT action.
I don't know, they don't state how they measure frame times. But they add their subjective observations as well about how they game was perceived, but they are still in swedish :(
 
Thread starter Similar threads Forum Replies Date
M Nvidia 5
I Nvidia 12

Similar threads



ASK THE COMMUNITY

TRENDING THREADS