Titan X vs. GTX 780 SLI: Surround benchmark

[Amon Amarth]

Running at a resolution of 5760x1200 (or 1080), how do two 780's in SLI do compared to a single Titan X in surround? Any benchmarks?

[Delphium]

Found a 780 SLI review, which also has Titan scores in its graph...

Source:- http://www.guru3d.com/articles_pages/ge ... ew,12.html

780GTX in SLI has more performance than a single Titan, the Titan X has a boost over the Titan, but i dont think its enough to beat 780 SLI.

The Titan has the advantage of additional VRAM which may help with antialising and some high level effect at the larger multi mon resolutions although the 780SLI still tends to beat a single Titan.


Source:- http://www.guru3d.com/articles_pages/ge ... ew,22.html

[HaYDeN]

Those benches appear to be referring to the original Titan (article is also old, 05/23/2013 02:58 PM), which even a 780Ti beat or grinded right up on... Titan X is a very different animal... like +50-60% - probably only ~10-1% behind the SLI'd 780's ... gg nvidia naming scheme, it is confusing.

Considering the benchmarks I've seen the Titan X seems to handle 4K reasonably well, sometimes having to compromise on highest settings (eg, 40-50fps target) - it seems like it'd almost be overkill for 5760x1080, good for 60fps+ but not quite good enough for 2560x1440x3

5760x1080 being 6.2mp and 4K being 8.8mp, eg a 30-40% disparity, 2560x1440 being 11mp.

If I could justify the cost, I'd love one, suspect it would be the perfect sweet spot for 5760x1080 without having to deal with the attrocity that is SLI and Crossfire (day one compatibility, profiles and requiring fullscreen exclusivity which some engines don't support)

[scavvenjahh]

I just sold one of my Tis to get an X, so I hope you're right ^^

Titan X vs. 780 Ti benchmarks are hard to come by because drivers and test protocols change in a year and a half... but basically, 780 Tis sit somewhere between 970s and 980s and the scaling is pretty much the same in 1080p and 4K, so Surround shouldn't be different. +15% punch than one X sounds reasonable.

[Amon Amarth]

Thanks for the input guys. savvenjahh, when you get a Titan X, could post your performance on this thread? If is, can you do so on nVidia Surround? (although I'm not certain that surround is possible with a single Titan X without a DisplayPort adapter, since the card comes with one DVI, one HDMI, and the rest displayport).

I'm basically looking to the future, so if and when I can get the funds, I may actually splurge for Titan X SLI at some point. But I'd like to know EXACTLY how a single Titan X does in surround, that way I can buy one replace my 780's then get the other card later.

[scavvenjahh]

Oh it is Surround-ready, just like a single 780 is. No adapter required, I'll be using simple DP/DVI cables. Don't expect in-depth benchmarks (or even graphs) but I'll see what I can do.

If you're looking for a sorta-future-proof dual-card setup though, chances are the upcoming 980Ti and R390X will be more suited to the task. The X's 12 GB of VRAM and horrendous price make little sense beyond their being "premium" features.

[ugzz]

even my 670s sli beat out the OG titan, which even given it's age, i'm quite proud of .
Your big win with switching to the Titan X would be no more SLI woes, less wattage, less heat etc. As far as performance, your going to be in about the same place.

[scavvenjahh]

@ugzz: dude, you own the cream of the crop from three years ago. Of course it rocks ! Not to mention the original Titan turned out to be... a castrated 780 Ti.

OK, here's the kind of framerates you can expect from a stock Titan X at 6060x1200 paired with an i7 @4.2GHz in Win7 x64. I've used max-ish settings to stress the card, but I reckon you can easily gain 5-10 fps everywhere with minimal optimisation.

Format: (min, max) avg

Batman Arkham City
MSAA 4X, DX11 Extreme, PhysX Off: (64, 95) 74
MSAA 4X, DX11 Extreme, PhysX Normal: (54, 82) 65

Metro 2033 Redux
Very High, SSAA Off, AF 16X: (16, 151) 44
Very High, SSAA Off, AF 16X + Advanced PhysX: (12, 156) 39

The Talos Principle Demo
MSAA 4X, Max 3D Rendering MPix Unlimited: (24, 579(!)) 48

Tomb Raider
Ultimate + Shadows Ultra, Hair Quality TressFX: (43, 64) 53
Ultimate + Shadows Ultra, Hair Quality Normal: (55, 80) 66

Total War: Attila
Quality, AA Off, AF 16x, DOF Off, SSAO Off : (28, 46) 38

Middle Earth: Shadow Of Mordor
Ultra + High Texture Quality: (29, 48) 41 - VRAM: 2.9 GB
Ultra + Ultra Textures: (28, 49) 40 - VRAM: 5.2 GB (visual difference: nil)

Some random facts:
- a single 780 Ti OC is 23% behind the X on these tests. Impressive really.
- a Radeon 295X2 beats the X by 20% (Batman AC's PhysX makes it a bit unfair). TressFX hits it just as much as the GeForces.
- 12 GB VRAM on a 16 GB system is not a problem (as long as the framebuffer is not full, anyway). For future-proofing, Nvidia recommends 24-48 GB system RAM (and an OS that supports as much)...
- Surround is generally much less VRAM-hungry than 4K and downsampling. Typical VRAM usage in these tests was 1.5 to 3 GB. Most VRAM used: SoM Ultra/Ultra/200% (12120x2400), 7.9 GB @11 fps !

There's a lot to discuss about this card but time is a bit short atm. I'll answer any questions as best I can.

[Amon Amarth]

Where you running SoM with/without AA? Was it the built in benchmark, or did you play the game itself to record your results?

Thank you for posting your results. Now I have something to compare my performance too.

One last request: Is it possible to benchmark the very first level of Crysis 3 with no AA, everything maxed out on surround? I'd like to see how that also compares to my setup. A great comparison for certain.

[scavvenjahh]

SoM: built-in benchmark, "FXAA and Camera + Object blur". With AA off, avg framerate is 47 (with either High or Ultra textures).
Crysis 3: I can do that, not before next week though.

I played some Dying Light (AO off), 40-90 fps, VRAM usage around 5.9 GB. Even at the highest framerate this game never feels as smooth as it should...

[scavvenjahh]

Crysis 3, first level (Very High, No AA, 6060x1200) - VRAM usage 2.2 GB - Avg: 35.525 fps Min: 24 Max: 60 (graph)