NVIDIA GeForce GTX 780 Video Card Review @ [H]

I don't care what motive or competition nvidia has.The card is very costly compared to the last 580/680 cards. The fact that there is an EXTREMELY expensive Titan is causing people to think this card is a "deal" at $650. Its exactly the price conditioning nvidia wanted.

This is the kind of performance jump we all expected to come with from a 680 to a 780. If there was no Titan, would this be so much of a "deal" at $650? Hell, this is priced far enough away from the 680 that they dont even have to lower the price on that, and so on down the line. Its like the low/mid/high end cards have all jumped up a price level. Its bad for everyone except nvidias pockets.

The GTX 780 is actually priced better (you get more performance/increase in cost) relative to its nearest competitor from AMD (7970 GE) than the GTX 580 (6970) and GTX 480 (5870) were.

The 580 was actually the worst deal in comparison to the 6970 and you also got less vram in that case.
 
So, from your apples-to-apples data, on average the 780 is 5 FPS faster than a 7970 for an additional $250 over what you can buy a 7970 non-ghz edition for.

What I'd love to know is how that value proposition changes when you overclock both of those cards to the max!

Check out the Tomb Raider benches. Good old green tinged [H] :eek:
Read from top to bottom and if you don't see whats going on, I have a Titan to sell ya :rolleyes:
 
Agreed.

With the same settings, the 780 is faster than the 7970 by about 5 FPS on average.

Interesting.. same settings
Metro Last Light 780 +10 fps faster than 7970 Ghz
Tomb Raider +1 fps
Crysis 3 +4 fps
FC3 + 6fps

yep, about 5 fps faster.
780 = $650
7970 Ghz = $450

+5 fps is definitely NOT worth $200 to me.
sure, it's great overclocking, but the 7970 overclocks too.

I guess I have a different mindset now that I can make money with Radeons.
Once OpenCL performance improves with Geforce cards by say 10x, then I'd give them a glance. Until then, I will "suffer" with my lowly 7950's.
 
You really think that's the case?

Yep. Look at the Apples-to-apples Tomb Raider:
"In this apples-to-apples graph we are comparing with TressFX enabled at 2560x1600 on all video cards. The GeForce GTX 780 is 22% faster than the GeForce GTX 680. The GTX TITAN is about 11% faster than the GTX 780."

What they DIDN'T say was "GTX 780 is same performance as the 7970 Ghz!"
 

It will be funny if these new drivers release and 7970's all of a sudden provide an equal or better experience than 780's. Makes buying a 7970 in January 2012 for $100 cheaper than a 780 look like a pretty good move!

AMD bests NV without even releasing new hardware, now THAT would be worth a LOL!
 
On page 7 of the GTX 780 review (Overclocking):

continued...
Overclocking
To overclock the GeForce GTX 780 we used a beta version of Precision X provided by NVIDIA.

Perhaps you meant to say EVGA? (Or Rivatuner in that matter...) Because we know that NVIDIA doesn't make anything like that.
 
I am confused, if someone goes titan why... just get a 690 for $1,000

Dual GPUs are problematic. While nVidia does a much better job with them than AMD, they still have trouble. A single GPU just flat out gives you its performance. Dual GPUs vary. Sometimes they double your FPS, sometimes less, sometimes they stutter, sometimes the game just won't work with them and you get only 1 GPU worth of power.

So some people don't want to put up with that, and get a high end single GPU. That's why I have a 680 instead of two 660s which, on paper, are better performing and less cost.

Also for the Titan, it is the kind of thing people who do pro workloads might want. While games don't need 6GB of VRAM, there are other applications that do (like some kinds of 3D previs).

Also some people just have a lot of money, and gaming is the hobby they spend it on, so they'll nab two or three Titans for a really powerful system.
 
Check out the Tomb Raider benches. Good old green tinged [H] :eek:
Read from top to bottom and if you don't see whats going on, I have a Titan to sell ya :rolleyes:

I remember the good old days when people claimed the opposite... and by good old days I mean the 5870 reviews :p

Peoples ability to see bias where none exists never ceases to amaze me.
 
I remember the good old days when people claimed the opposite... and by good old days I mean the 5870 reviews :p

Peoples ability to see bias where none exists never ceases to amaze me.

It is indeed fascinating to see the effects of reader bias on how they perceive neutral statements or objective numbers in reviews.


Yep. Look at the Apples-to-apples Tomb Raider:
"In this apples-to-apples graph we are comparing with TressFX enabled at 2560x1600 on all video cards. The GeForce GTX 780 is 22% faster than the GeForce GTX 680. The GTX TITAN is about 11% faster than the GTX 780."

What they DIDN'T say was "GTX 780 is same performance as the 7970 Ghz!"

Surely you remember the hellstorm of fanboys proclaiming H was completely in AMD's pocket during the 5870 launch, right? Seeing as how you have your forum account from 12 years ago still.... (I lost mine due to the user purge because I didn't login for over a month right around when they were doing it.... :( so I'm a measly 8.5 years now) right?
 
Yep. Look at the Apples-to-apples Tomb Raider:
"In this apples-to-apples graph we are comparing with TressFX enabled at 2560x1600 on all video cards. The GeForce GTX 780 is 22% faster than the GeForce GTX 680. The GTX TITAN is about 11% faster than the GTX 780."

What they DIDN'T say was "GTX 780 is same performance as the 7970 Ghz!"

Most reviews are showing this card to be 20-25% faster on average than the 7970 ghz.
 
Yep. Look at the Apples-to-apples Tomb Raider:
"In this apples-to-apples graph we are comparing with TressFX enabled at 2560x1600 on all video cards. The GeForce GTX 780 is 22% faster than the GeForce GTX 680. The GTX TITAN is about 11% faster than the GTX 780."

What they DIDN'T say was "GTX 780 is same performance as the 7970 Ghz!"

They didn't mention the percentage difference between the 780 and 7970 in ANY of the Apples-to-apples. The only time they mentioned the 7970 in that section was for Metro. Beyond that, the Apples-to-Apples section for TR does not tell the whole story. If you look up the small differences between them allow the 780 to run at higher settings. Even at a lower resolution the 780 is able to push settings higher.
 
Last edited:
It will be funny if these new drivers release and 7970's all of a sudden provide an equal or better experience than 780's. Makes buying a 7970 in January 2012 for $100 cheaper than a 780 look like a pretty good move!

AMD bests NV without even releasing new hardware, now THAT would be worth a LOL!

Well, there are two caveats with that driver.

1. AMD says you have a choice between performance and smoothness. You can set the driver to ignore frametimes or reduce frametime latency at the cost of performance.

2. The driver is still a software solution vs Nvidia hardware solution which is still superior and not driver dependent. AMD will have to wait until the 8000 series to move the fix to hardware, but I'm skeptical of that because they just started working on the fix a few months ago.

No performance gain is coming with the new drivers. If anything, you'll see reduced performance because of reviewers/users choosing smoothness to evaluate AMD cards, prompting another price cut and ridiculous game bundle i.e. BF4.

They didn't mention the percentage difference between the 780 and 7970 in ANY of the Apples-to-apples. The only time they mentioned the 7970 in that section was for Metro. Beyond that, the Apples-to-Apples section for TR does not tell the whole story. If you look up the small differences between them allow the 780 to run at higher settings. Even at a lower resolution the 780 is able to push settings higher.

I wanted to see more games than those that came bundled with the cards.
 
Current Specs : 3770k & GTX 690

I'm currently on the mini-ITX platform and hoping to upgrade to mATX SLI config.

What will be the best upgrade path ? Another 690 or something else?

690 - but I have to qualify that - for highest settings etc another 690 to make quadSLI, when working should give the best performance, however it's more GPUs, more of that random instability. I love my 690 but I have to agree with Sycraft that regardless of the improvement of SLI in recent years, MGPU is STILL problematic, still does random, very aggravating shit, hence why I bought a Titan when I won some money on the punt. Wish I had waited a few weeks but you don't think of these things when alcohol and easy money are involved....If I were you and one of those 'unlimited budget ' guys, I'd ditch the 690 and get a couple of Titans, plus 3 27" 144hz monitors, since you are already familiar with NV MGPU and the 3 144hz monitors, simply because it would be awesome ...Otherwise, if you don't mind or don't have issues with SLI, then another 690 and probably a bigger PSU would be the go...gl!!!

PS - I used the 690 with a Maxi 4 gene gen 3 and they were a good combo - another 690, the right PSU, would make a killer m-ATX build!!
 
Last edited:
Current Specs : 3770k & GTX 690

I'm currently on the mini-ITX platform and hoping to upgrade to mATX SLI config.

What will be the best upgrade path ? Another 690 or something else?

Two 780 cards in my opinion, they also exhaust air out the rear instead of dumping it back into that tiny chasis :).
 
I guess what I got from yesterday's meeting with Kyle at the Tiger Direct store was this:

Galaxy supports the community and gamers more... AND are open to feedback which is HUGE.
Note: They actually hand-carried the cards to the store to put them on the shelf, so they could be bought the day of release... I can't think of ANY other company who did something like that recently... Can we all laugh at the PS4 release now???

nVidia buffers the screens we actually see on the screens in a smoother manner than AMD for the most part... I.e. AMD cards are good, but the DRIVERS are better on the nVidia side for MY purposes which include smoothness on my screen (Note, I use a 24" Trinitron Sony as my eyes are sensitive)

In short, next video card will likely be a Galaxy nVidia card for me, as even the SLI seems to be a bit more fluid than Crossfire... Now, this latter part doesn't even apply to me yet as I really like the 1 video card in a PC thing, but it sounds as if SLI may finally be a way to go for me... Perhaps too cautious, but getting only %50 out of the second card always bothered me.
 
Another quick note... I asked Kyle about monitors, and he suggested I wait for a bit longer, like maybe around Halloween or such... He was very vague, so I am being vague, but yeah, I'm waiting.
 
Another quick note... I asked Kyle about monitors, and he suggested I wait for a bit longer, like maybe around Halloween or such... He was very vague, so I am being vague, but yeah, I'm waiting.

That's interesting.
But it would be really hard to beat the Qnix QX2710 currently selling for about $310 shipped.
2560x1440 Samsung PLS
and it overclocks to 96 Hz and some make it to 120 Hz.
 
Well, there are two caveats with that driver.

1. AMD says you have a choice between performance and smoothness. You can set the driver to ignore frametimes or reduce frametime latency at the cost of performance.

2. The driver is still a software solution vs Nvidia hardware solution which is still superior and not driver dependent. AMD will have to wait until the 8000 series to move the fix to hardware, but I'm skeptical of that because they just started working on the fix a few months ago.

No performance gain is coming with the new drivers. If anything, you'll see reduced performance because of reviewers/users choosing smoothness to evaluate AMD cards, prompting another price cut and ridiculous game bundle i.e. BF4.

1. That's not how the driver works, it simply fixes runt frames. There are no longer any significant amount of runt frames, if you want it performance the solution would be to use the old driver :D

2. Software isn't a "worse" solution, it performs the same, it just is more work for the GPU, possibly hampering performance. But if it works and still is faster, then it doesn't matter if it is hardware or software.

The new drivers have some performance gains, they don't trade FPS for smoothness, they just output a more consistent FPS.
 
Perhaps too cautious, but getting only %50 out of the second card always bothered me.

As WorldExclusive mentioned, SLI scaling has improved significantly in recent years (for 2 cards at least) and seeing anything less than an 80% in most modern games is rare. It is usually closer to 90%.

Since you mention having more sensitive eyesight, I would actually be more concerned with the potential to see micro-stutter. I've been running an SLI setup long enough now that I'm either conditioned to it (and just don't notice) or the latest hardware just doesn't exhibit micro-stuttering anymore. Some folks however still perceive it when first going from a single card to SLI.
 
1. That's not how the driver works, it simply fixes runt frames. There are no longer any significant amount of runt frames, if you want it performance the solution would be to use the old driver :D

2. Software isn't a "worse" solution, it performs the same, it just is more work for the GPU, possibly hampering performance. But if it works and still is faster, then it doesn't matter if it is hardware or software.

The new drivers have some performance gains, they don't trade FPS for smoothness, they just output a more consistent FPS.

I can't bookmark every article I read just in case my claims are disputed but he briefly touched upon it here, but didn't go into detail:

http://www.pcper.com/reviews/Graphics-Cards/Frame-Rating-AMD-Improves-CrossFire-Prototype-Driver

AMD is still planning on releasing this driver in a beta form in the summer but I wouldn't be surprised to see the schedule moved up a bit with some pressure with the Radeon HD 7990 release and better than expected results thus far. AMD continues to promise the ability to enable and disable this feature in the control panel as well as to enable it on a per-game basis, something that NVIDIA hasn't done yet. There are debates on whether or not there are actually input latency benefits to AMD's current method and we are still finding a way to test that at PC Perspective.

If I come across the exact way of how this driver works again, I'll be sure to post it.

As for maintaining or gaining performance with this driver. GPU performance is rated on FPS and AMD are removing runt frames to smooth the framerate, how does that maintain or gaining performance?

The original debate was CF stuttering and suspect FPS numbers that were higher than actually seen on the monitor.
Which means the GPU is slower than perceived. Continue reading the article to see what I'm saying here. FRAPS FPS vs Observed FPS.
 
Nice Review.... Unless you have a 580/570 or 6970/50 class CPU or Lower ... I would probally sit this one out. Sure some guys wanna have the fastest GPU at all times...Dont let anyone tell you how to spend your bread... But its summer time... all my cash is going to my car and bike right now... not much time to sit in the house and game...the ladies are dressing less out there fellas lol
 
That's interesting.
But it would be really hard to beat the Qnix QX2710 currently selling for about $310 shipped.
2560x1440 Samsung PLS
and it overclocks to 96 Hz and some make it to 120 Hz.

Thats what i have the QNIX QX2710 @ 2560x1440 ... can;t be beat.
 
As for maintaining or gaining performance with this driver. GPU performance is rated on FPS and AMD are removing runt frames to smooth the framerate, how does that maintain or gaining performance?

The original debate was CF stuttering and suspect FPS numbers that were higher than actually seen on the monitor.
Which means the GPU is slower than perceived. Continue reading the article to see what I'm saying here. FRAPS FPS vs Observed FPS.

The prototype driver I've been using doesn't have that feature as far as I know.

Think about it. If you meter the FPS, the runt frames simply become acceptable frames. The problem isn't "the GPU is getting more FPS for less work", but rather "the GPU is displaying less results for the work it did". If you meter the frames, it can still draw the same amount of FPS (give or take maybe 5%) while removing runts. Any performance losses in the driver are because it's currently based off of an old driver revision, not the latest Cat beta.
 
The prototype driver I've been using doesn't have that feature as far as I know.

Think about it. If you meter the FPS, the runt frames simply become acceptable frames. The problem isn't "the GPU is getting more FPS for less work", but rather "the GPU is displaying less results for the work it did". If you meter the frames, it can still draw the same amount of FPS (give or take maybe 5%) while removing runts. Any performance losses in the driver are because it's currently based off of an old driver revision, not the latest Cat beta.

Ok, before I continue to speak on this and venture of into ignorance, I'll hold final judgement until it's official released. It seems like you had hands on experience with this.
I hope you're correct because that would make AMD an option for me in the future. The stuttering was main thing and the frequent/late driver releases was the second thing.
 
Ok, before I continue to speak on this and venture of into ignorance, I'll hold final judgement until it's official released. It seems like you had hands on experience with this.
I hope you're correct because that would make AMD an option for me in the future. The stuttering was main thing and the frequent/late driver releases was the second thing.

I can tell you the frame pacing driver is essentially "done", I haven't noticed any issues with it in games. Just needs further qualification I suppose.
 
The problem isn't "the GPU is getting more FPS for less work", but rather "the GPU is displaying less results for the work it did". If you meter the frames, it can still draw the same amount of FPS (give or take maybe 5%) while removing runts.

Agreed - that's also how I understand the workings of it.
Seems to be the consensus.
 
On page 7 of the GTX 780 review (Overclocking):



Perhaps you meant to say EVGA? (Or Rivatuner in that matter...) Because we know that NVIDIA doesn't make anything like that.

NVIDIA provides the beta versions of the program, they work with EVGA to create the new technologies in it for the gpu launches, it is the program they give press to overclock with on new gpu launches now. So yes, I go this directly from nvidia.
 
NVIDIA provides the beta versions of the program, they work with EVGA to create the new technologies in it for the gpu launches, it is the program they give press to overclock with on new gpu launches now. So yes, I go this directly from nvidia.

Ah thanks, I must've overlooked that. I couldn't find it on their website so I posted that.\

EDIT: Wait a second, I was pretty positive that these programs derived from Rivatuner which was a third party program in the first place. I never knew that NVIDIA themselves provided an official overclocking program, unless they started doing this?

According to Wikipedia, Alexey Nicolaychuk said this: "Currently RivaTuner core is powering MSI Afterburner and EVGA Precision (X) only" and the screenshot shows the "EVGA" logo on the tweaking tool in your review.

Unless NVIDIA obtained a license from the original developer (which I'm not aware of at the moment) I apologize if I overlooked information on this and I'm not trying to criticize you, and maybe it's because I didn't have my coffee this morning..
 
Last edited:
I do believe I have found the vid card for my next build in a few months :)

Still running mostly the same build as my sig so it is definitely time to upgrade, that and I need to get prepped for Mechwarrior Online to get the hell outta beta.

Thanks for the great review [H] :cool:
 
Places like Anandtech, PCPer, TechCrunch and Hardware Canucks all realize that FPS measurements are misleading for showing the true performance of a system.. especially since FPS takes into effect runt frames which give the FPS rating a fake boost.

Frame Time is what those sites have shifted to and only include FPS ratings for people who don't know that Frame Time is the new standard and need to be eased into it, or simply as a secondary metric.

I'm very dissapointed to find you did nothing with Frame Time.

Lol.

Nice card but I'd rather spend the money on two 7950s.
 
Lol.

Nice card but I'd rather spend the money on two 7950s.



Frametimes with FCAT combined with FPS graphs are pretty much the new standard for ensuring accurate data is relayed to readers, and that nothing's being gamed (see two 7950s in crossfire with huge latency, jitter, and effective FPS that isn't anywhere near reported fraps fps). Bojamijams is pretty much correct, and I hope to see frametime info in some future [H] articles combined with their FPS info and subjective thoughts (both of which are well done).
 
"The GeForce GTX 780 will have an MSRP of $649. That is currently ~$450 less expensive than a GeForce GTX TITAN."

1000 - 650 = 350.
 
Fortunately the performance of the GTX 780 is much closer to the Titan than the price.

New Zealand's cheapest prices for top-end cards:
EVGA GTX 780 $1137.35 (introductory price)
Asus GTX 690 $1448.99 (introductory price $2500)
EVGA GTX Titan $1799.00 (introductory price $1799)

I could go for GTX 780 (oc) SLI, get close to the performance of GTX Titan SLI and save $1300. Yet GTX 690 SLI doesn't look too bad a deal these days. My 5 yearly build is due January 2013 so I'm prospecting for goodies already.
 
Back
Top