Thursday, January 31st 2013

First NVIDIA GeForce Titan 780 Performance Numbers Revealed

The rumor mill is spinning to galeforce (or should we say GeForce) winds. Its newest sack of flour points at what could be the first performance figure of NVIDIA's upcoming GeForce Titan 780 flagship single-GPU graphics card. Circulating among various Chinese tech publications is this 3DMark 11 Xtreme Preset screenshot from the PCinLife community, in which a lucky bloke claimed access to a GeForce Titan 780 engineering sample, and a driver to get it to work. In the scribbled out 3DMark 11 Xtreme Preset score screenshot (below), the source claims the fabled graphics card can singlehandedly score X7107 points. For reference, a GeForce GTX 690 usually scores in the region of X6000 points, and a GTX 680 around X3300. If true, NVIDIA has something truly remarkable up its sleeves, maybe the second coming of 8800 GTX. From older reports, we know that the GeForce Titan is expected to ship sooner than most people think, some time in February.

Source: PCOnline.com.cn
Add your own comment

112 Comments on First NVIDIA GeForce Titan 780 Performance Numbers Revealed

#1
Prima.Vera
by: xvi
The price is ridiculous. I miss the good ol' days where $350 bought you the top of the line card.
I bought my 5870 with EXACTLY the same money. ;)
Posted on Reply
#2
james888
by: Aceman.au
Depending on performance numbers, I'll be getting either a 790 or 8990. (or whatever the dual GPU card will be called.) My 7970's have been disappointing. Low FPS in WoW @ maxxed out 1080p


I would not call what w1zz's review shows bad fps. Sounds like you have a possible crossfire problem, or maybe something else. Why did you get crossfire anyways? Isn't 100 fps enough, if you at least got that?
Posted on Reply
#3
Calin Banc
by: RejZoR
It's somewhat very unlikely that performance will be more than 2x of its predecessor. Usually 2x is the limit at which they hold onto. Though they rarely do it. For AMD i only remember it happening with HD4800 -> HD5800 transition where HD5800 was twice as fast.
As far as I remember, the difference was huge if you'd compare the 4850 512MB model with the 5870 1GB one. A lot of games went beyond 512MB mark and that took a hit. Although 6970 was close to 5870, Crysis 3 seems to show a 57% gap between them - http://forums.overclockers.co.uk/showpost.php?p=23672202&postcount=112
BF 3 and and Max Payne 3 in the TPW "never settle bundle" review, show an almost double performance when 6970 goes head to head with the 7970 ghz.ed. - http://www.techpowerup.com/reviews/AMD/Catalyst_12.11_Performance/5.html . At the end of the day, it depends on the benchmark you put them through and how fast the software makes good use of the new hardware available.
Posted on Reply
#4
tacosRcool
That number is awesome! When the GTX 7XX series comes out, I'll stick to my GTX 670 and will buy another one just cuz everybody is rushing to grab the newer models
Posted on Reply
#6
Aquinus
Resident Wat-man
by: Aceman.au
I get 50 at best, average around 30-40
Are your sure that it's your video card slowing down? WoW has a nasty CPU bottleneck problem. It's not like it can take full advantage of your 6-core SB-E chip. What does your GPU usage look like and does reducing shadow quality make your FPS skyrocket?
Posted on Reply
#7
Aceman.au
by: Aquinus
Are your sure that it's your video card slowing down? WoW has a nasty CPU bottleneck problem. It's not like it can take full advantage of your 6-core SB-E chip. What does your GPU usage look like and does reducing shadow quality make your FPS skyrocket?
I stood still in front of the Stormwind AH. Its fairly populated. About 55 FPS, turning the shadow quality from Ultra to Low bumped it up to 77 FPS.

GPU-Z reports my core clock @ 1000mhz and memory clock @ 1375mhz (these 7970s arent the GHZ editions, but the overclocked editions)

I cant access CCC to see activity. I click the exe and my mouse flickers with a loading circle and nothing...
Posted on Reply
#8
Crap Daddy
I'm starting to think it's not so impossible to be faster than a GTX690 which is basically 2 fully enabled GK104 SLIed running at 915+MHz. The 690 is around 1.6x a GTX680 (correct me if I'm wrong). GK110 has almost double the TMUs 240 vs. 128, almost double CUDA cores 2880 vs. 1536 and memory interface of 384bit giving 288 MB/s bandwidth vs. 256/192MB/s compared to GK104. Let's not forget it has 7 billion transistors, again almost double the GK104 (not that it matters much but it's really a monster chip). If they will clock it at 1GHz and don't care about power consumption since it's uber enthusiast category, I think it's not such a SF presumption that it will top the GTX690.
Posted on Reply
#9
THE_EGG
Hmmm those numbers seem a little too good to be true. But even if it only achieves 6000 or over, Nvidia can still take my money. One of the few top end video cards I'd be interested in getting on release as usually I hold off for the initial price drop but this seems too good to pass up.
Posted on Reply
#11
Hayder_Master
i see it's nothing approaching, it's remind me in nvidia gtx200 series and ATI 4000 series times, Nvidia gtx600 series and AMD 7000 series are doing the job and booth of them is really Satisfied for performance and sales, maybe in Q3 2013 or Q4 wii see something.
Posted on Reply
#13
Aquinus
Resident Wat-man
by: Aceman.au
I cant access CCC to see activity. I click the exe and my mouse flickers with a loading circle and nothing...
Why are you using CCC to view usage? It doesn't even give you vram usage and doesn't graph it or anything so it's basically useless for that purpose. :confused: Use the GPU-Z OSD server or use Afterburner's OSD so you can see it right in the WoW viewport. I suspect you're bottlenecking on your CPU (yes, I know that its blasphemous to say an Intel 6-core is bottlenecking.)

You know, this might sound odd but try disabling hyperthreading and try it again. If your framerate improves it's very likely to be a CPU bottleneck on one or two cores. Could you try and setup turbo so it boosts the first couple cores a bit higher than 4Ghz? Maybe that might help if it is a bottleneck with the most used thread in WoW. Most games won't be able to utilize the full ability of all 6 cores, so no use overclocking all 6 cores unless you need it imho.

Seeing what your GPU usage looks like could also say if its a CPU bottleneck or not. If its anything less than 85%, you know what's going on.
Posted on Reply
#14
Aceman.au
by: Aquinus
Why are you using CCC to view usage? It doesn't even give you vram usage and doesn't graph it or anything so it's basically useless for that purpose. :confused: Use the GPU-Z OSD server or use Afterburner's OSD so you can see it right in the WoW viewport. I suspect you're bottlenecking on your CPU (yes, I know that its blasphemous to say an Intel 6-core is bottlenecking.)

You know, this might sound odd but try disabling hyperthreading and try it again. If your framerate improves it's very likely to be a CPU bottleneck on one or two cores. Could you try and setup turbo so it boosts the first couple cores a bit higher than 4Ghz? Maybe that might help if it is a bottleneck with the most used thread in WoW. Most games won't be able to utilize the full ability of all 6 cores, so no use overclocking all 6 cores unless you need it imho.

Seeing what your GPU usage looks like could also say if its a CPU bottleneck or not. If its anything less than 85%, you know what's going on.
Wow what, the GPU usage is at 78 percent on GPU2 and 24 percent on GPU1
Posted on Reply
#15
Finners
3dmark is fake just an OC'd 690. From Devast8nDiscoDave over on aria forums

Posted on Reply
#18
Gradius2
It was already confirmed to be *FAKE*.

People should investigate a bit before posting such obvious (fake) news. :rolleyes:

Every GPU bench pre-official-release specially for China is meant to be fake. :cool:
Posted on Reply
#19
ThunderStorm
by: repman244
And? I still don't see how the 780 can be faster than the 690 when you compare the specs.

K20X

Shading Units: 2688
TMUs: 224
ROPs: 48
SMX: 14
Pixel Rate: 41.0 GPixel/s
Texture Rate: 164.0 GTexel/s
Floating-point performance: 3,935.23 GFLOPS

GTX 690:
Shading Units: 1536 x2
TMUs: 128 x2
ROPs: 32 x2
SMX: 8 x2
Pixel Rate: 29.3 GPixel/s x2
Texture Rate: 117.1 GTexel/s x2
Floating-point performance: 2,810.88 GFLOPS x2
Exactly. The specs speak for themself. There is no way the 780 can beat 690, that's just way unreal.

With pure mathematical calculation: 2688/1532 = 1.75 ergo a theoretical 75% performance increase over the 680 with no power leak and 100% effective input power ( sorry I forget the exact phrase for this particular expression, I'm just 19 you know. )
If you insert the exhaust power into the theoretical equation, so let's say 20%, that 75% would now ve 55%.

And now the price comparison: 900$ / 1.55 vs 469$ ( the cheapest one i can find on newegg )/ 1. I would say NV is applying the same strategy as the one ASUS are now using with the ARES II, so anyone who is bitching about it now can stay silenced. Thus, Videocardz recently posted an article about the delay of next generation to Q4 2013. Therefore, I would say buying either ARES II or 780 would not be stupid.

by: Gradius2
It was already confirmed to be *FAKE*.

People should investigate a bit before posting such obvious (fake) news. :rolleyes:

Every GPU bench pre-official-release specially for China is meant to be fake. :cool:
Not just every GPU bench pre-official-release, everything related to China is FAKE......, merchandise/people/culture... EVERYTHING.
Posted on Reply
#20
erocker
by: syeef
so you mean around 90 posts over nothing?
Lol, yes. But really, it is entertaining to see people get carried away with obviously bad/fake information. :)
Posted on Reply
#21
Fluffmeister
by: Finners
3dmark is fake just an OC'd 690. From Devast8nDiscoDave over on aria forums

http://i.imgur.com/oGPVPHY.jpg
Whilst I appreciate the score is most likely bull, why does this magic technic apparently reveal the cards name, yet not even a smidgen of the time and date it was run?

The original pic my be fake [score], but the fake to prove it's a fake is also a fake. :P
Posted on Reply
#22
Gradius2
Well, that was good for the LOL (post #95 to the end). :)
Posted on Reply
#23
Crap Daddy
Information is that PCPer and TechReport have the card for testing. Fake or not I'm sure we'll find out soon what's in the bag.
Posted on Reply
#24
Fluffmeister
by: Crap Daddy
Information is that PCPer and TechReport have the card for testing. Fake or not I'm sure we'll find out soon what's in the bag.
Indeed, and regardless it will be interesting to see what the fastest GPU on the planet can do.
Posted on Reply
#25
jmcslob
by: erocker
Lol, yes. But really, it is entertaining to see people get carried away with obviously bad/fake information. :)
I think TPU should have an official turd pic for instances just like this...Perhaps one of somebody holding a fresh squeezed one in their hand with a toilet in the background...just saying
Posted on Reply
Add your own comment