Graphics Nvidia GPUs Comparison - GTX780/Ti, GTX970, GTX980

I've recently seen a lot of confusion about the actual performance of these graphics cards in relation to one another, and it doesn't help that a lot of the performance figures for the 7-series cards are now considerably outdated because of vastly improved drivers (the most recent being version 364.51 WHQL) and many of the reviews are a couple of years old. I regularly encounter statements like "the 970 is about the same as the 780Ti" or "the 980 is a little faster than the 780" and so on.

With that in mind, I thought that a collaborative comparison thread might help clear things up a little bit, especially for people who are thinking of picking up a 7-series card or upgrading to a 9-series card. I've run some tests using three freely available and popular graphics benchmarks (3DMark11, 3DMark Firestrike, and Heaven 4.0), and I'd like to invite owners of the aforementioned graphics cards to do the same and share the results in this thread.

Naturally it would be preferable to have all GPUs tested in the same system, but I figured it'd at least be an acceptable compromise to have all GPUs tested with the same software and driver versions, and things like CPU and RAM can be taken into account when comparing scores. I've done a lot of benching with a few of Nvidia's more recent driver versions (from 361.43 onwards) and there doesn't seem to be any difference in these synthetic tests.

Here are my results for each of the benchmarks. For the Futuremark results, the most important number is the GPU-specific result. I expect more modern systems to achieve higher GPU scores; I'm not sure to what extent there'll be a difference.

My 780Ti is running at 1306MHz core, 7000MHz RAM, and using driver version 362.00.

 

Great idea - I'm definitely going to get in on this at the weekend - I'll run both with 1 GTX780 and 2 in SLI. Last year when I was looking to upgrade I came to the conclusion (from doing a lot of googling and comparing reviews) that 2 780s in SLI were about the same speed as 1 980ti (as long as the game being used scaled well).

 

I've been thinking about this recently and have a decent selection of cards that I can benchmark if of interest!

Titan
780 (up to 3x SLI)
780 Ti
980
980 Ti
Titan X

I can likely run them all in the same mobo for a decent baseline if you like too?

I don't have a 970 any more though.

 

Interesting. My 2x 680s score 15438 in Firestrike and 2639 in Heaven with 104.7 FPS so a single card of the next generation up isn't too far behind. I think I'm gonna wait for Pascal to drop before doing an upgrade though.

 

Surprised to see this generate so much interest - I did not expect that at all!

@jinksy, whatever suits you - it'd be interesting to see how much better they perform in a IB setup.

And yeah, SLI setups will easily beat the pants off a single card. I'd imagine a 980Ti would be pretty close to 680s/770s in SLI, and I did consider adding a second 770 but I've done the SLI thing a few times before and I always end up finding it a bit of a headache so I'm sticking with one card. I must admit, though, 780Ti Classified SLI is mighty tempting...and I'd need a new PSU!

 

I've still got a Titan Black I can run a few benchmarks on. I don't know what this one clocks at (it's an MSI the other two were Palit) but I will have a go.

I know both of my Palit cards were stable with a EVGA SOC bios on them plus about another 100mhz.

If you want a good comparison though you will need to run stuff like Heaven because I 'only' have a 2ghz 8c 16t Ivy with it.

 

I'll have a think for the weekend. I'm not sure my IB board will do tri SLI but I guess that's not a big problem?

 

Not really - I know tri-SLI has improved but two cards is still the sweet spot as far as scaling goes.

 

I might still have a spreadsheet where I tested the 780Ti's in 2 way SLI compared to 980Ti's in 2 way SLI, I will have a look later and see if I do because I tested some games and benchmark tools.

 

Alright. So I'll set up a test rig at the weekend and will do a benchmark of each of the graphics boards I own at the very least at stock settings.

What would we like it running on? I was planning on running it on either:

R3E and X5650, 24Gb RAM;
P8Z77I and 3770K, 16Gb RAM; and/or
Z97i Plus and 4790K, 16Gb RAM

OS and benchmarks, I'm happy to go with the masses. Hit me.

If anyone wants me to fire up a 970 in the rig, feel free to send one to me to play with

 

I was checking 3DMark results and it looks like there's nowt between the X5650 @ 4.3 and the 4790K at 4.5, except of course for the extra cores. The GPU scores are practically identical in Firestrike, so it looks like there isn't any bottlenecking going on despite the 4790K being a far more modern part.

I guess it really depends on the benchmarks, so it's a case of try it and see.

One benchmark I really like is Catzilla, but I'm really disappointed that they disabled the higher res tests in the free version. I still have the beta build so I can do all tests, and damn it's really hardcore shiz... really heats up the GPU.

If there are any suggestions for a place to share the beta installer I'd be happy to do that, but it's a pretty big file, over half a gig.

 

I went on my 3D Mark results list so here they are.

3D Mark Vantage

780Ti SLI

Extreme: 42558

Performance: 53521

980Ti SLI

Extreme: 53742

Performance: 58769


3D Mark 11

780Ti SLI

Extreme: 9291

Performance: 19260


980Ti SLI

Extreme: 14065

Performance: 24933


3D Mark

780Ti SLI

Firestrike: 15636

Firestrike Extreme: 8268

Firestrike Ultra: 2490


980Ti SLI

Firestrike: 20414

Firestrike Extreme: 13417

Firestrike Ultra: 7584


Those scores may increase if I was to turn of shadowplay, but would need to test again to see if nVidia have fixed the performance loss on these benchmarks when running shadowplay.

I will be able to check later to see if I have the game benchmarks still.

 

If the benchmark is reading your GPU core clock right, then the 780Ti is severely handicapped at under 900MHz core. Most 780Tis should do 1200MHz with relative ease, and that would probably make a considerable difference to the SLI performance as well.

 

Since these cards are the chosen ones as far as VR systems go, is it worth putting in the Steam VR benchmark too?

 

3D Mark never picked up the speed of my 780Ti's properly iirc, it always showed them as much less but those tests were ran with everything at there stock speeds, no overclocking other than what they did out of the box and iirc the 780Ti's were around 1130Mhz and above but it's been a while since I used them.

 

^ Ah I see. I'm pretty sure 3DMark reports my GPU clocks accurately, but their Systeminfo software that they bundle with their benchmarks is up to date and works pretty well now I'd say.

@Jehla, I like the idea of doing the Steam VR test and I'll run it when I'm on my other OS.

Edit: on second thoughts, the Steam VR test is near enough a 5GB download(!) and might not be everybody's cup of tea. Anything from the 780Ti is guaranteed to pass anyways (I passed with flying colours, no GPU overclock), but if people want to try it then I guess it's a free country.

 

OK so I ran the three benchmarks on a bone stock Titan Black with an 8c 16t Ivy Xeon running at 2ghz. Here's what I got. First up 3DM.



http://www.3dmark.com/3dm/11260299?

3DM11.



http://www.3dmark.com/3dm11/11072548

Heaven 4.0



I wasn't going to do this just yet but Need for Speed does not work with Crossfire yet (and I desperately need it at 4k) so I am going to play it on this rig.

What tools are you using to overclock Lenny?

 

http://www.3dmark.com/3dm/11262165

http://www.3dmark.com/3dm11/11073206

3770k @4.3 ,GTX 980 phantom , 16GB of 2133 ram

 

I managed to get 1180mhz stable for 3dmark (both 3dm and '11) however Heaven was crashing at launch. What I did notice was the CPU holding back the Titan but then I expected that, given it's a 60w CPU that hits 32c in Prime 95.





So yeah, a slight improvement. I'm not really well versed on voltage and etc given it's been many months since I even turned this rig on. I gave it +50mv though there was room for more.

 

@Vault-Tec, I'm using a software tool that was designed specifically for EVGA Classified cards and their custom power circuitry. I set core voltage to approx 1.27v and then adjust clock speeds in MSI Afterburner (I much prefer it to EVGA's own Precision Tool software).

I can overclock the core to 1200MHz without any tweaks or modding, however with a custom BIOS and voltage increase I can get closer to 1350MHz.

This is ace. What is the core speed of your 980, as this is crucial for the comparison? The 780Ti is marginally ahead of the 980 in these tests but it's overclocked quite a bit more than out-of-box cards would be.