NVIDIA GeForce GTX 780 SLI Video Card Review @ [H]

FrgMstr

FrgMstr

Just Plain Mean
Staff member
Joined
May 18, 1997
Messages
55,634
NVIDIA GeForce GTX 780 SLI Video Card Review - Take two GeForce GTX 780 video cards, enable SLI, add GeForce GTX TITAN SLI, and 3-way Radeon HD 7970 GHz Edition CrossFire, sprinkle with a sampling of today's most demanding games, and add a dash of NV Surround and Eyefinity 3-display testing. We find out out how tasty GeForce GTX 780 SLI is; the real world gaming results may surprise you.
 
Last edited by a moderator:
At high resolutions, in a 3-display setup, we are able to run every game at its "maxed out" graphics settings. That is, we are able to run with the highest quality, highest tessellation, and highest graphics settings. We do not have to skimp or sacrifice any quality settings with GTX 780 SLI at 5760x1200.
Click to expand...

This is not true. You turned down settings to "high" in Crysis 3, and you had to disable PhysX in Metro:Last Light.
 
BigMack70 said:
This is not true. You turned down settings to "high" in Crysis 3, and you had to disable PhysX in Metro:Last Light.
Click to expand...

Uhh, Only to compare MSAA 4x testing - about the most memory/horsepower intensive form of post-processing available in modern gaming.

With normal AA methods, at normal levels, very high was used.

As for Physx, well, I suppose that's true.
 
But Kyle, Titan's have 6gig Vram buffer over the GTX780 3gig Vram buffer. We all know 6 is faster than 3 and more future proof. LOL j/k

Good to see the 780sli doing so well. Think I may wait to see what ATI have in store for the next release and decide if I'm going to drop some loot on the 780's or 9970.... :cool:
 
I have the distinct pleasure of having two gaming computers currently, that are nearly identical, save for one has Crossfired 7970s and the other has SLi'd 780s.

Currently in all my games, I fully agree with the H opinion and findings......despite at times identical or even faster FPS on the HD 7970 pair, the is just something odd about the pacing and visuals to the Crossfire setup.

The 780s are smoother across the board, now I OC my 780s to 1160 Core in game so my actual frames are slightly quicker than the article. My 7970s run at 1125 Core.

I'm anxious for the new drivers to trickle up to EyeFinity.:D
 
Very interesting comparisons. I would have liked to seen a 3-way with the 770 also as they are currently running at $399 and are pretty comparable to the 7970s.
 
Just some constructive feedback on these video card reviews: I enjoy the testing methodology of telling the readers what is subjectively "playable" and then backing those claims up with fps data. Even if my view of playable is different (eg. Crysis 3 multiplayer is not playable to me at an average of 46.9 fps), it's good to hear a subjective viewpoint and descriptions about perceived smoothness and playability. However, a very small percentage of people play games with multiple monitors and limiting testing to this single extreme resolution of 5760x1200 leaves a lot to be desired. I'd love to also see whats "playable" at more common resolutions of 1920x1080 (or 1920x1200) and 2560x1600. I can try to derive what my experience might be at other resolutions with your data but usually end up referencing reviews at other websites for this information.
 
Arcygenical said:
Uhh, Only to compare MSAA 4x testing - about the most memory/horsepower intensive form of post-processing available in modern gaming.

With normal AA methods, at normal levels, very high was used.

As for Physx, well, I suppose that's true.
Click to expand...

No, the Crysis 3 page definitely states that the system spec was set to "High" and not "Very High". When they did test "Very High" with AA turned down to 1xSMAA, they commented that it was "not playable, because of performance".

It is simply not true that they found playable performance with 780 SLI on highest settings in Crysis 3, unless they incorrectly entered what settings they tested on.
 
AuxNuke said:
Just some constructive feedback on these video card reviews: I enjoy the testing methodology of telling the readers what is subjectively "playable" and then backing those claims up with fps data. Even if my view of playable is different (eg. Crysis 3 multiplayer is not playable to me at an average of 46.9 fps), it's good to hear a subjective viewpoint and descriptions about perceived smoothness and playability. However, a very small percentage of people play games with multiple monitors and limiting testing to this single extreme resolution of 5760x1200 leaves a lot to be desired. I'd love to also see whats "playable" at more common resolutions of 1920x1080 (or 1920x1200) and 2560x1600. I can try to derive what my experience might be at other resolutions with your data but usually end up referencing reviews at other websites for this information.
Click to expand...

Yes I feel the same way about this. Test with more common resolutions such as 2560x1440 or 2560x1600. Testing at 5760x1200 is nice and all but the data is irrelevant for a good portion of the viewers I would imagine. I think testing 780 SLi at 1920x1080 is not necessary though, its going to max out everything.
 
pandora's box said:
I think testing 780 SLi at 1920x1080 is not necessary though, its going to max out everything.
Click to expand...
You'd be surprised. Using GTX 780's in SLI; Crysis 3 at 1920x1200 with maximum AA and very high settings averages in the 60's and dips into the low 30 fps! Not fully playable in multiplayer, IMHO.
 
I don't find triple-wide resolution testing to be irrelevant at all. Sure, most people play 1080/1200, but a SLI 780/Titan article about 1080p would be boring... "Well, at this scene we dropped from 60 FPS to 58."

I've always found it a pretty easy task to interpolate performance with lower resolutions from the [H] results at higher ones.
 
While I appreciate all the disclaimers in the article about frame pacing not working at eyefinity resolutions, I have to wonder why you'd bother to run the tests at this point in time then and make conclusions like 780 SLI is superior. Wouldn't it make more sense to wait until next month when frame pacing isn't an issue and provide a more balanced look at what the configurations can do? (maybe you're planning a follow up)

I think it was same to assume that everyone knew that this test wouldn't favor AMD in any way as soon as it was announced that frame pacing doesn't work at these resolutions yet.

Anyway, good article otherwise, interesting data. I have to keep telling myself "I do not need 3x1080p, I do not need 3x1080p," Hehe :)
 
Can't belive 2 780's beat 3 7970's by that much.. Hmm might have to re think my build. Was looking at 2-3 7950's since there only 220 now.
 
twzTechman said:
Very interesting comparisons. I would have liked to seen a 3-way with the 770 also as they are currently running at $399 and are pretty comparable to the 7970s.
Click to expand...

This. Heck, 770's tend to OC very well, and can get within 5% of the 780. I really don't understand why anyone would go for the 780 over the 770, especially you can SLI them for the same price as a single 780.
 
gamerk2 said:
This. Heck, 770's tend to OC very well, and can get within 5% of the 780. I really don't understand why anyone would go for the 780 over the 770, especially you can SLI them for the same price as a single 780.
Click to expand...

Where did you get "within 5%" from? You mean an overclocked 770 vs a stock 780? Maybe...but you can overclock the 780 also.

I upgraded from 1250 core sli 770s to 1200 core sli 780 and the gains are approx 30-40%. The extra ram is also nice since I run surround.
 
I would have like to see a cost for cost comparison, as in 3 way 780 SLI vs 2 way Titan SLI. That would be a good article.
 
gamerk2 said:
This. Heck, 770's tend to OC very well, and can get within 5% of the 780. I really don't understand why anyone would go for the 780 over the 770, especially you can SLI them for the same price as a single 780.
Click to expand...
Because some of us don't want to deal with SLI... that, and there's no guarantee that you'll get a GTX 770 that's a golden overclocker (let alone two of them)
 
AuxNuke said:
Just some constructive feedback on these video card reviews: I enjoy the testing methodology of telling the readers what is subjectively "playable" and then backing those claims up with fps data. Even if my view of playable is different (eg. Crysis 3 multiplayer is not playable to me at an average of 46.9 fps), it's good to hear a subjective viewpoint and descriptions about perceived smoothness and playability. However, a very small percentage of people play games with multiple monitors and limiting testing to this single extreme resolution of 5760x1200 leaves a lot to be desired. I'd love to also see whats "playable" at more common resolutions of 1920x1080 (or 1920x1200) and 2560x1600. I can try to derive what my experience might be at other resolutions with your data but usually end up referencing reviews at other websites for this information.
Click to expand...

I think it's pretty safe to say that anything less than Surround or Eyefinity resolutions using the graphics horsepower these cars represent will be top notch.
A single monitor won,t even require full gpu(s) utilization with ALL the features turned up.

If you want to see what happens at 30" resolutions, look at the single card reviews.......SLi will be better.;)
 
I'd like to see this comparison done at 1600p -- I'd like to know how well the 7970s perform with frame pacing enabled in comparison to the 780s and Titans.
 
twzTechman said:
Very interesting comparisons. I would have liked to seen a 3-way with the 770 also as they are currently running at $399 and are pretty comparable to the 7970s.
Click to expand...

All in due time, 770 SLI is the next SLI config to test on the list.
 
[L]imey;1040101218 said:
While I appreciate all the disclaimers in the article about frame pacing not working at eyefinity resolutions, I have to wonder why you'd bother to run the tests at this point in time then and make conclusions like 780 SLI is superior. Wouldn't it make more sense to wait until next month when frame pacing isn't an issue and provide a more balanced look at what the configurations can do? (maybe you're planning a follow up)

I think it was same to assume that everyone knew that this test wouldn't favor AMD in any way as soon as it was announced that frame pacing doesn't work at these resolutions yet.

Anyway, good article otherwise, interesting data. I have to keep telling myself "I do not need 3x1080p, I do not need 3x1080p," Hehe :)
Click to expand...

This review was started before I knew that Frame Pacing was coming. In reality, we are a little late on covering 780 SLI/770 SLI/760 SLI to begin with.

The experience we receive right now, is what everyone else will experience. It is our obligation to let you know what you will experience at any given time. Can't blame us for AMD not supporting Eyefinity yet. It'd be unfair to CrossFire owners to wait.

Of course we'll cover CrossFire in Eyefinity when Frame Pacing is active in Phase 2.
 
BigMack70 said:
This is not true. You turned down settings to "high" in Crysis 3, and you had to disable PhysX in Metro:Last Light.
Click to expand...

If you turn off AA, or down to SMAA Low 1X, Very High becomes playable, but we wanted some form of higher AA.
 
Good to see that the 780's are still within overclocking distance of the Titans at surround resolutions. Not that I really expected any different.
 
AuxNuke said:
Just some constructive feedback on these video card reviews: I enjoy the testing methodology of telling the readers what is subjectively "playable" and then backing those claims up with fps data. Even if my view of playable is different (eg. Crysis 3 multiplayer is not playable to me at an average of 46.9 fps), it's good to hear a subjective viewpoint and descriptions about perceived smoothness and playability. However, a very small percentage of people play games with multiple monitors and limiting testing to this single extreme resolution of 5760x1200 leaves a lot to be desired. I'd love to also see whats "playable" at more common resolutions of 1920x1080 (or 1920x1200) and 2560x1600. I can try to derive what my experience might be at other resolutions with your data but usually end up referencing reviews at other websites for this information.
Click to expand...

Anything single-display resolution isn't demanding enough to show off the potential and performance that high-end SLI can provide. 1080p is useless. You'd have a better argument at 2560x1600 in SOME games, but not many. To really stress the GPUs, and keep everything GPU limited, multi-display must be used. (Until we are all using 4K monitors)

We figure people who have the money to equip $1300 worth of graphics are the ones that are also going to opt for the multi-display experience. We lack the time and resources to cover every resolution. We must stick to one platform generally. Eyefinity/NV Surround is that platform to test out SLI and CrossFire on high-end cards. I think you'd agree 3 GPUs in 3-way 7970 GHz Edition CF might be overkill for a single display.

We base the platform on the cost of the cards and what makes sense. Comparing $1300 worth of graphics on a $150 1080p monitor doesn't make sense. Now when we get to 760 SLI, I may include some 1600p resolutions then, that makes more sense with that level of card. And especially down to 7790 CF and 650 Ti Boost SLI, up to 1600p will be the max.

Performance can be extrapolated.
 
Last edited:
gamerk2 said:
This. Heck, 770's tend to OC very well, and can get within 5% of the 780. I really don't understand why anyone would go for the 780 over the 770, especially you can SLI them for the same price as a single 780.
Click to expand...

Some people don't want the complication of more than 1 video card in their system. 780 SLI is going to be a lot faster than 770 SLI. And of course 770 SLI will be faster than 1 780 single card. We'll get to 770 SLI testing later this month.
 
inglewood78 said:
Where did you get "within 5%" from? You mean an overclocked 770 vs a stock 780? Maybe...but you can overclock the 780 also.

I upgraded from 1250 core sli 770s to 1200 core sli 780 and the gains are approx 30-40%. The extra ram is also nice since I run surround.
Click to expand...

Yup. Gk110 scales a lot better than gk104.
 
Very interesting. I would have liked to see a test performed with Frame Pacing enabled on those 7970s but as the author said, the new driver doesn't support Eyefinity yet.

I'd be interested in a couple updated benchmarks after Phase 2 is out for sure :)
 
Lots of hard work there, thank you. But with monitors well down in price, I'd like to have seen tests with both 3 and 5 monitors. I can now buy 5x 24" 1080p gaming monitors at £120 each for about the same price as the 22" Iiyama I purchased so many years ago. Come to think of it, I purchased a 17" CRT over 20 years ago for about that.
 
Brent_Justice said:
If you turn off AA, or down to SMAA Low 1X, Very High becomes playable, but we wanted some form of higher AA.
Click to expand...

OK, then you should edit the article - either in the conclusion or the Crysis 3 page - to reflect that, because that is not what the article says. In the article, when you tested at very high settings with SMAA Low 1X, you got the following results for 780 SLI:
Min - 27
Max - 48
Avg - 38.1

And you commented "the gameplay experience was smooth, consistent, and fluid. While not playable, because of the performance, it was still a smooth experience with no stuttering"

Not sure how you guys missed this...
 
Need to get an LG 29EA93-P in for your "1080p" tests. ;)

Still, unless we're talking 4k resolution, single-screen isn't going to be a problem for any high-end card that's even remotely recent.
 
How come you never did a 7990 with a 7970. That seems like it would have been a closer match in price, as well as performance. It would still count as TRI-FIRE
 
flooberjobby said:
How come you never did a 7990 with a 7970. That seems like it would have been a closer match in price, as well as performance. It would still count as TRI-FIRE
Click to expand...

No point, and it would be slower than 3 separate 7970 GHz Edition video cards, the 7990 is about 15% slower than 7970 GHz Edition CrossFire. The price works out the same, around $1200, and it wouldn't be as fast as 3 separate cards, so no point, we used the best match up.
 
inglewood78 said:
Where did you get "within 5%" from? You mean an overclocked 770 vs a stock 780? Maybe...but you can overclock the 780 also.

I upgraded from 1250 core sli 770s to 1200 core sli 780 and the gains are approx 30-40%. The extra ram is also nice since I run surround.
Click to expand...

Yep, I went from a GTX 670 @ 1372mhz core/7110mhz mem for 24/7 use (and that's a definite golden card and beat out most 680 oc's) to a GTX 780 @ 1202mhz core/6400mhz mem for 24/7 use and saw gains of, on average, 45-50% or more. Huge, huge upgrade... 770 oc sli to 780 oc sli would be fairly similar ;) since a 770 is basically a 680 with faster memory chips stock.
 
Last edited:
Enjoyed reading this article- this is the last 'wave' before the next-gen cards get here, and we get 4k monitors on the low-end; hopefully there will be more 6GB-8GB cards available then!
 
I guess since no one is talking about the obviously huge white elephant in the room .. I will.

2 x used 7970's for around $300 each off eBay or Craigslist. Total - $600 all day every day.

12.8 drivers and a $600 pair of 7970's Xfired can't be beat.
 
You must log in or register to reply here.
Back
Top