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

If your aim is to pitch the 780,780Ti, 970, 980, 980 Ti against each other to get a real feel for the comparative performance, shouldn't you test cards at stock as a baseline?

After all, no two cards will overclock to exactly the same degree.

 

My thoughts exactly.

In order to be of objective use we need up to date benchmarks of the cards at stock speeds, ideally all tested on the same platform. Having scatter-gun tests done on a wide variety of hardware configurations, possibly with different drivers, certainly with custom PCBs/BIOSs, isn't going to clarify the picture any more than the multitude of benchmarks already available online.


EDIT:

 

I have got all reference boards. I'll put them on an Asus board with 16gb RAM and 4790K. I'll leave the cards all stock.

I'll give this some time tomorrow!

 

Thanks for the input guys.

I did think about this, however I'm also aware that a lot has changed with the implementation of GPU Boost 2.0, and (to me) it seems counter intuitive to deliberately downclock a card for the sake of benching it. The vast majority of these cards are factory overclocked models so IMO it makes more sense to run the tests "as is" (or with a manual overclock) and report the clock speeds at which they were run. It is for these reasons that I suggest we should state core frequencies with results as it gives meaning to the results. (Case in point: is my 780Ti better than the 980 because it's overclocked, or is it better because the 980 is not overclocked? I don't know the core speed of the 980, so I cannot answer that question!)


Things like custom PCBs and custom BIOSs aren't really variables; the processing hardware is exactly the same, so a 780Ti Classified is still a 780Ti and a 980 G1 Gaming is still a 980. They can overclock, and so can reference design cards.

As I said in the OP I didn't intend for this to be a lab quality comparative analysis; it's just a bit of fun and information for everybody who wants to take part. The tricky bit will be collating the results, but if all the information is there then it shouldn't be too much of a chore.

By all means if somebody wants to do the tests at stock then that's fine, but it will be a lot of work, so kudos in advance to that person!

Edit: just noticed the timely response from Barry Banhammer. He gets it indeed, and he gots all the goodies to get it too!

 

1203/1304 boost

ram @ 3600

 

LR - if I PM you my shared OneDrive/Dropbox folder link, would you stuff that Catzilla benchmark up? Ta!

I'm going to use this as my test bed:

ASUS Z97i Plus
i7-4790K @ stock
16Gb 1600MHz RAM (Crucial Ballistix)
250Gb Samsung EVO
Seasonic 520W PSU

I'll open the choice of OS up to the masses. Happy to oblige with whatever. I'll then fire up each of the cards with the latest nVidia drivers, at stock clocks for each card. They're all 'SC' cards as far as I can remember, so that should make for a fairly 'sensible' pack of results.

I might then throw one of the 780 Classy cards in there for a laugh, and the 980Ti Classy too.

The only cards I have multiple of are the 780 Classy, and I'd have to do the dual and tri-SLI tests on a different board, but I'm not proposing to make that a core part of my testing for now. Does that seems sensible to one and all?

 

Interesting - basically clock-for-clock then. And at 1304 it's quite tame for a 980, which can go north of 1500MHz. Thanks for the clarification!

@Jinksy,

Sure thing - fire me your Dropbox link and I'll try to get Catzilla uploaded tonight.

Are you confident that a 520W PSU will be sufficient for the monster hardware? My rig pulls over 600W at the wall when running 3DMark Firestrike, but that's probably down to the old-ass X58 architecture and the high TDP processor.

I'm happy with your proposed run with the Z97i system. I used Windows 7 for all my tests because everything runs great on it for me and I'll keep using it until I see a reason to upgrade. Feel free to use whichever OS suits you - I don't think there'll be any difference in performance, but Win 8 or 10 would be more "current" I suppose.

 

http://www.3dmark.com/fs/7932780






Palit GTX980 SuperJetStream @1481 / 7600 (bios mod 1281.3mV boost still enabled) (game stable, can go higher, currently does not break past 58C while benching)
i5 3570k @ 4.4Ghz
ASUS Z77Sabertooth
16Gb 1600 RAM
Samsung SSD 840 evo

 

Sweet! So the 980 handily trumps the 780Ti when it's near enough 1500MHz, and by quite some margin in Firestrike too - almost 16K GPU score.

@Jizwizard, your pic isn't showing!

In other news, I have the Catzilla v1.0 Beta installer and I'm a little unsure about sharing it publicly, so if anyone really wants it please drop me a PM and I'll send you a link (I uploaded it myself, so it's safe!).

 

No probs this end Lenny. Strange as I can't see yours in the first post

 

Hmmm... that is strange. Can anybody else not see the images in the OP? The thread kinda depends on it as that's how I'll be sharing the collated results!

 

Morning all - LR - got your PM. I'll get cracking a bit later. Thanks!

 

Awesome

 

i5 6600k + GTX 980ti, both at stock settings:

 

Wow, the 6600k is disappointing at stock speed.

 

do the skylakes oc as well as the haswell e's ?

 

why windowed mode? surely fullscreen is better for heaven

 

Haven't used Heaven before, so I didn't even notice, I just left it on the defaults it installed with. Re-ran with fullscreen:

 

Have you run GPU-Z with the sensors tab open, while running this test to see if there is a reason your min FPS is 19.0? compared to mine and other peoples runs it seems low. PrefCap Reason (GPU-Z) should tell you if the card is throtteling back or limited for a reason.

might be nothing but worth checking as it will effect your score.