GeForce GTX 680 Up To 40% Faster Than Radeon HD 7970: NVIDIA

[Crap Daddy]

Going on all the rumours so far, here is my take and I'm an avid Columbo fan with some Quincy thrown in.

GK110 (call it Big GK) is still in production. It's an insanely powerful gpu but is a bit difficult to make precisely.
AMD release the 7970 and NV are genuinely surprised, expecting something more powerful.
NV realise they have a GK part (104) that was not destined to be directly competitive with HD7970 but now seems that with a little work, may well be (thus the delay).
Initial pricing was for GK104 at mainstream levels. On seeing it's competitiveness with HD7970, NV decide to either capitalise on HD7970's price point and/or the poorer than expected yield from TSMC forces their hand to up prices (as per wafer cost is higher).
This all leads to the release of GK104 (originally a GK660 labelled part) being upped to the top spot (GK680) to take the crown.

Big GK is still coming and will be a monster. But the recently resurrected rumour of a dual GK104 part makes Big GK (as initial rumours suggested) a Q1 2013 part (possibly the GK780).

Dual Tahiti (7990) will be brought in by AMD to counter either the GK680 part or the GK690 dual part.

That's my purely personal opinion but it's based on sound logic of all the rumours thus far. And like I said, I watch Columbo - logic and reason shall conquer.

Oh... one more thing.

Hi Columbo. My name is Sherlock and I got to the same conclusion some time ago.

These slides are based on an old leak which we don't know how accurate were and anyway 95% of the people laughed when they saw. I mean, c'mon "performance" chip to be better than AMD's high-end? This must be a bad joke?

Well, in the meantime the joke is not so much... a joke. Add to this old chart some driver improvements and tweaks and 20% over the 7970 isn't too far fetched. But that was not my idea, it was Dr. Watson's so don't blame me if it's less than 20%...

 

[Dent1]

So 680 will be 10% faster than 7970 !!!! But all 7970 i know can be OCed so to gain at least this 10 % or more.
Will the 680 also have room to be COed that good ? It has only 2x6 pin power !!!

There isnt any benchmarks, so nobody knows how fast it is. For all we know the 680 might be 10% slower than the 7970.

 

[N3M3515]

No one said 40% faster average. Unless I missed that, which I doubt.

EDIT: Kyle from HordOCP did say 45% faster, but I don't know if it meant "up to" or not. And he said it would launch no earlier than June afaik. He suggested a June-August timeframe afaik, so it may very well be GK100. Now it's probably gone.

Nobody said 40%, but everyone said it would be massively faster, including you until 2 days ago.
14% is not massive.....
Anyway, lets end this pointless talk...........it is obvious gtx680 has been over-hyped everywhere.

 

[happita]

So 680 will be 10% faster than 7970 !!!! But all 7970 i know can be OCed so to gain at least this 10 % or more.
Will the 680 also have room to be COed that good ? It has only 2x6 pin power !!!

This makes me chuckle a little bit. What makes you think...or anyone for that matter think that NVIDIA's cards won't be as OCable or even moreso than AMD's 7k cards? However, it is 2x 6pins, so what do I know? I mean, I love AMD cards, but their pricing at launch really bites the dust especially when there is no competition for them. I for one can't wait soon enough for NVIDIA to come out guns blazing and give AMD a swift kick in the ass to lower not only their prices, but also GTX 5xx cards too.
But then NVIDIA will charge a premium for being the fastest....and so the cycle continues

 

[Benetanegia]

Nobody said 40%, but everyone said it would be massively faster, including you until 2 days.
14% is not massive.....
Anyway, lets end this pointless talk...........it is obvious gtx680 has been over-hyped everywhere.

Yes but you have to take into account that "now" it is 20% smaller too. And we don't know performance anyway. These results are fake? And if they are... actual performance is necessarily worse? It could be faster actually, we don't know, so you are jumping to conclusions. As for rumors there's also rumors about a last minute BIOS update that upgraded performance by 40-50%, so according to this rumor the results that indicated GTX680's performance as similar to HD7970 were just an smoke bomb. I'm not giving these rumors any more credit than the rest, but you just can't go by the "it's only 5% faster" rumors either.

 

[borden5]

the key word is Up To

 

[Vulpesveritas]

Yeah that's what rumors suggest to me too, especially looking at them historically. For us the result is certainly unfortunate.

See I've been reading the situation a different way. The quarter ends the 31st of this month, and Nvidia has the NDA not end until about a week before the quarter.
While I can see it your way, I see it more plausible that Nvidia is trying to hold back AMD's first quarter profits back as much as they can, saying "kepler is unbeatable" and the like. If it were truly unbeatable, why doesn't Nvidia just show us the part and benchmarks and show that it is, getting people whom right now are being told to wait instead told to go get a kepler GPU. Surely they have working silicon.

Look at AMD for example with their APUs, they've proven to be effective to SB/IB in laptops, and are marketing them by their benchmark scores, despite not being released. Intel has released benchmarks of Ivy bridge despite only 10-20% increase over SB.

Yet Nvidia, while touting that they're unbeatable with kepler, have it under lock and key until the end of the quarter.

 

[happita]

Yet Nvidia, while touting that they're unbeatable with kepler, have it under lock and key until the end of the quarter.

Yea but, but, but...look at those pretty red and green colors on that graph!!! The higher numbers mean its faster, surely that counts for something?
Oh wait, darn!!!! It's not faster in Civilization 5 !!

 

[Hayder_Master]

we need your chart W1Z

 

[semantics]

oh marketing slides lol i like how they start at .8 times

 

[20mmrain]

Bs!!!!

 

[eddman]

40% in just one specific benchmark. It seems to be around 10% on average, if that slide is even real that is. If true I might say it's a very good result considering that it was supposed to be a performance kepler, not high-end.

 

[Vulpesveritas]

Yea but, but, but...look at those pretty red and green colors on that graph!!! The higher numbers mean its faster, surely that counts for something?
Oh wait, darn!!!! It's not faster in Civilization 5 !!

Yes, well let's also point out that this graph has been specifically made to exaggerate differences, and that we have no solid benches from reputable sources. The NDA isn't up yet, and we don't know this is gk104. it might be gk100 test silicon or simply completely made up by a propaganda source or simply a Nvidia fanboy messing around.

I would say the latter is quite possible.

But we'll find out when the NDA is over I suppose.

 

[Crap Daddy]

I repeat what others have said. This is not even an Nvidia slide, it's made up from a chart posted at pcinlife by somebody one and a half months ago so it has zero relevance. We should stop talking nonsense and make predictions based on this.

 

[dj-electric]

NVIDIA employee 1: So we got that sweet KG104 core
NVIDIA employee 2: Oh kinda like GTX560TI
NVIDIA employee 1: Yeah, its awesome and will have a nice solid 40-45%-ish gain in performance over it
NVIDIA employee 2: So... when are we gonna launch it?
NVIDIA employee 1: Oh wait... it surpasses the 449$ HD7950 in games. Screw every thing, we're not suckers to sell it for 299$, lets name it GTX680 and put a 500-ish price tag
NVIDIA employee 2: Good call... good call...

 

[20mmrain]

The performance difference is on average 10% going both ways. There are some areas where this card does beat the HD7970 but there are a other areas the GTX 680 looses too. What they failed to mention that it also looses sometimes by 20% or more too.

 

[Nihilus]

What do you guys think is it worth the effort to return my 2 7970's and get two 680's? I play on 5760x1080 resolution most of the time. I am worried that the 2gb wouldn't be enough for that type of resolution.

I think it would be a really stupid wast of time and money. On average, this hand picked Nvidia list is giving the GTX 680 a 15% edge. We know for a fact that the HD 7970's are wonderful overclockers. Not so sure about the GTX 680s yet. You may already have the fast card factoring that in.

Just wait for the reviews...

 

[Benetanegia]

See I've been reading the situation a different way. The quarter ends the 31st of this month, and Nvidia has the NDA not end until about a week before the quarter.
While I can see it your way, I see it more plausible that Nvidia is trying to hold back AMD's first quarter profits back as much as they can, saying "kepler is unbeatable" and the like. If it were truly unbeatable, why doesn't Nvidia just show us the part and benchmarks and show that it is, getting people whom right now are being told to wait instead told to go get a kepler GPU. Surely they have working silicon.

Look at AMD for example with their APUs, they've proven to be effective to SB/IB in laptops, and are marketing them by their benchmark scores, despite not being released. Intel has released benchmarks of Ivy bridge despite only 10-20% increase over SB.

Yet Nvidia, while touting that they're unbeatable with kepler, have it under lock and key until the end of the quarter.

My take on that is that Nvidia is not worried about how much AMD profits this quarter. Set aside the fact that volume is so low that it's not going to be much anyway, lower AMD sales on that segment are not going to make them, Nvidia, sell more Kepler cards. It won't make them sell more GTX500 cards either, wuite the opposite. If GTX680 is better it will sell as many cards as they can make. AMD can sell all the cards in the world for all they care, they're just as supply constrained as they are, more probably because practically the only thing Nvidia has said about Kepler is that they have been stockpiling for a hard launch. Whether it's completely true or not remains to be seen, it's the ony official claim either way.

The thing is that for some reason a lot of people think this is an AMD vs Nvidia thing, and that that is the case for them. It's not. Nvidia cares about their sales and profits and nothing more and will do what is better for their profits. Releasing info about Kepler before launch is not going to make them sell more cards, never really did, never will. People buying this kind of hardware waits for reviews. And on the downside, it can tell AMD what to do next. So why do they need to release anything, when silence is just as good or better marketing campaign? So they remain tight lipped and AMD is in the dark.

 

[specks]

Of course they would say 40%. Few would like to buy a newer card if it does not perform very well

 

[TheoneandonlyMrK]

So they remain tight lipped and AMD is in the dark.

yeh but AMD's no more in the dark then you or I, and have still managed to release many a new GPU for near a year or more with little new in reply from nvidia just rehash after more core, so to me they need to get their asses movein as many a sale HAS gone Amd's way because some people couldnt wait a year

 

[Benetanegia]

yeh but AMD's no more in the dark then you or I, and have still managed to release many a new GPU for near a year or more with little new in reply from nvidia just rehash after more core, so to me they need to get their asses movein as many a sale HAS gone Amd's way because some people couldnt wait a year

It does? Nvidia's discreet GPU market share, which is what GTX680 is all abut, grew 4% (to ~64%) last quarter. Reality and your claim just does not compute. During the Fermi time Nvidia has GAINED market share. I could say that people bought GTX500 because they couldn't wait over a year to buy something as fast from AMD.

 

[Nihilus]

Yikes, 1920x1080 resolution? This graph may be a BEST case scenario for the GTX 680. The Hd7970 3 GB of ram is not to be F###ed with. I have a feeling the gap will be in favor for AMD once resolutions are higher. In dual card setup the advantages will be even bigger.

http://www.hardwareheaven.com/revie...ssfire-graphics-card-review-introduction.html

 

[N3M3515]

but there are a other areas the GTX 680 looses too. What they failed to mention that it also looses sometimes by 20% or more too.

Where did you get that info?

 

[beck24]

It does? Nvidia's discreet GPU market share, which is what GTX680 is all abut, grew 4% (to ~64%) last quarter. Reality and your claim just does not compute. During the Fermi time Nvidia has GAINED market share. I could say that people bought GTX500 because they couldn't wait over a year to buy something as fast from AMD.

680 is only the third ranked card nvidia will release this year. nvidia also has proven to have superior driver teams, as many independent sites have attested, and will undoubtedly get more performance out of this series, sometimes very significant increases. It will be a great year for nvidia!

 

[beck24]

great card

This is a great card for what will be the third fastest in the nvidia line up this year. It is supposed to overclock like crazy so I'm excited to see what it can do as drivers mature and vendors customize.