[3dcenter] GK104 specs

[boxleitnerb]

Companies NEVER pass the savings to the customer if they can price according to the competition. Do you expect Ivy Bridge to be far cheaper than Sandy?

This comparison is flawed. Nvidia and Intel are completely different companies with different products. The market is different as is the situation with the competition. Intel can charge the same price and don't increase performance at all because they have no competition in that segment. Also it is much much harder to increase CPU speed than GPU speed. We have come to expect significantly better perf/$ for quote some time now. This has been the rule, not the exception.

And again:
What does Nvidia gain (other than angry customers) if they sell GK104 cards for 400-500$ for a couple of weeks? Prices will fall anyway, that's just the result of competition. Most people are not stupid - they look at the older cards and judge the new cards on value.
Also, HD7950/7970 are already in place. If you cannot take the performance crown (which I assume GK104 can't except for cherry picked scenarios), how do you make people buy your product?
Lower power? Not significant for most people, especially if the difference is really low (like 20-30W).
Price? There you go.

Gaining marketshare can be more important than to make another extra buck. If the pricing scheme some of you suggest would take hold, you would actually pay more for more performance across generations. GTX760 would cost 600, 860 would cost 700...where does that end?

To make GK104 as overpriced as Tahiti would only make sense if GK110 is really really far out - and there is no proof of that.

 

[SirPauly]

Companies NEVER pass the savings to the customer if they can price according to the competition. Do you expect Ivy Bridge to be far cheaper than Sandy?

Companies usually never raise their MSRP by 50 percent.

 

[Arzachel]

This comparison is flawed. Nvidia and Intel are completely different companies with different products. The market is different as is the situation with the competition. Intel can charge the same price and don't increase performance at all because they have no competition in that segment. Also it is much much harder to increase CPU speed than GPU speed. We have come to expect significantly better perf/$ for quote some time now. This has been the rule, not the exception.

And again:
What does Nvidia gain (other than angry customers) if they sell GK104 cards for 400-500$ for a couple of weeks? Prices will fall anyway, that's just the result of competition. Most people are not stupid - they look at the older cards and judge the new cards on value.
Also, HD7950/7970 are already in place. If you cannot take the performance crown (which I assume GK104 can't except for cherry picked scenarios), how do you make people buy your product?
Lower power? Not significant for most people, especially if the difference is really low (like 20-30W).
Price? There you go.

Gaining marketshare can be more important than to make another extra buck. If the pricing scheme some of you suggest would take hold, you would actually pay more for more performance across generations. GTX760 would cost 600, 860 would cost 700...where does that end?

To make GK104 as overpriced as Tahiti would only make sense if GK110 is really really far out - and there is no proof of that.

Was FX5900 cheaper vs Radeon 9700 not to mention 9800? 7900GTX vs x1900xtx? Make no mistake, if Ati had the lead back then, they would've done the same.

 

[LOL_Wut_Axel]

It could be one or the other. It very well may be a typo or miscommunication from engineering to marketing (or intentional) in the document you are citing. This did just literally happen with Bulldozer.
Something similar happened with the reporting of die size of gf104.

Also Nvidia came to market with the GF104-gtx 460 1gb, which was not released in full form until a reincarnation -GF114-gtx 560TI. Even with a partially disabled die, stock @675 was slower than a HD 5850 and @900mhz o/c was equal to a HD 5870.
This could easily play out again.

No, I'm 99% sure that's the size. They weren't wrong about Cayman, so why would they be wrong about Tahiti? AMD's GPU engineers aren't incompetent morons like their CPU ones.

Shifting focus back to these ongoing releases, it appears that unlike earlier rumors were saying it's not just Tahiti that will be GCN. Pitcairn and Cape Verde will use GCN, too.

The HD 7870 could potentially be going against the GTX 670, but it doesn't seem like there's gonna be an HD 7890. I think that would've been the most interesting card in the whole series for enthusiasts if it was priced at around $300-330 given its base architecture would be better than the 7870's and therefore it'd be a further cut down card based on an Enthusiast GPU (kinda like the GTX 560 Ti 448 Core) that would clock like a beast. It could prove a great adversary to the GTX 670 or even 680 depending on exactly how fast they are, but it looks like it's not gonna happen because AMD could easily clock the 7870 high enough for it to have the same stock performance.

 

[Arzachel]

Companies usually never raise their MSRP by 50 percent.

They don't usually halve it between generations too(GTX260 to GTX460). Companies price products according to the competition and to what the market can bear, not according to a perceived value, that you personally hold.

 

[boxleitnerb]

If they had the lead? They did.

It's difficult to find the prices that far into the past. I also don't know how that would apply to today. ATI has since been integrated into AMD, the management changed as has the philosophy to a certain extent. And the market is different today.

I think you can only charge premium prices if you have premium performance across the board (and not just for 1-2 months but for the usual cycle until the next generation or the refresh comes out). Everything else is overpriced and the market will take care of it.

 

[Ajay]

It is? Heh, look at every review on 7970 and you'll see die size described as 365mm2 or somthing like that. Now you'll see people posting that it is 352mm2. This is well after launch and reviews. 340mm2 rumor means nothing at all.

Perhaps, someday, there will be a procedure to help those who lack the sarcasm gene.

 

[Arzachel]

If they had the lead? They did.

It's difficult to find the prices that far into the past. I also don't know how that would apply to today. ATI has since been integrated into AMD, the management changed as has the philosophy to a certain extent. And the market is different today.

I think you can only charge premium prices if you have premium performance across the board (and not just for 1-2 months but for the usual cycle until the next generation or the refresh comes out). Everything else is overpriced and the market will take care of it.

In both cases Nvidia's cards performed worse and were priced higher than the equalent or better performing Ati cards. While Ati had the performance lead, Nvidia had the marketshare lead and so could command higher prices for inferior products, just like Intel could in the Netburst days. While not so pronounced, it's still mostly true today.

 

[AtenRa]

Because that IS the size, and the reviews got it wrong.

It's AMD that's saying the die size for Tahiti is 352mm^2, and I'm sure their engineers know more about that than the reviewers. None of the reviewers bothered to correct it, either.

I have personally asked on the official Tahiti EMEA Briefing Call on December 14 2011 the Die Size and AMD's answer was 365mm2. That was the size every Review worldwide used.

I dont know why AMD changed that again (Bulldozer ???)

Again, AMD gave the 365mm2 size that Reviewers worldwide used in their articles. There is no Official AMD paper stating anything about Tahiti's Die Size being 352mm2 as far as i know. This is the reason why all reviews still have the die size at 365mm2 which was the official size AMD gave.

No, I'm 99% sure that's the size. They weren't wrong about Cayman, so why would they be wrong about Tahiti? AMD's GPU engineers aren't incompetent morons like their CPU ones.

LOL, you know them all up close and personal??

 

[Crap Daddy]

My take on this: NV will launch the GK104 under the name GTX680, a slightly faster card than the 7970 (helped by higher stock clocks) and with certain weak points due to the fact that the chip was initially designed for the performance segment (weaker above 1200p, lower mem. bandwidth) but after AMD's launch it can fulfill other expectations. Price? Neither 300$ nor 550$

 

[SirPauly]

They don't usually halve it between generations too(GTX260 to GTX460). Companies price products according to the competition and to what the market can bear, not according to a perceived value, that you personally hold.

I'm personally not holding anything just stating AMD raised their MSRP by 50 percent. Defend it all you want -- no big deal to me.

 

[Arzachel]

I'm personally not holding anything just stating AMD raised their MSRP by 50 percent. Defend it all you want -- no big deal to me.

Then how is your post relevant to what I said?

 

[Keysplayr]

Then how is your post relevant to what I said?

Wut? I thought it was?

 

[Arkadrel]

(supposedly) leaked GK104 benchmark numbers:
http://videocardz.com/30564/nvidia-geforce-600-specifications-geforce-gtx680-benchmark-leaks-out


GeForce GTX 680 GK104 (GPU Clock: 1050, Memory: 1450 MHz) (overclocked)
vs
Radeon HD 7970 (stock)

=

Geforce GTX680 about ~5-10% faster, when overclocked vs stock 7970, in leaked results.

Rumored default settings are 950 MHz for core and 1250 for memory, while this card was tested with 1050/1450 settings – suggesting that GTX680 has a great overclocking capabilities.

200 Mhz core overclock + 200 Mhz memory overclock = ~5-10% faster than 7970 stock.

Not bad for a card probably gonna be around 399$.

 

[Arzachel]

Wut? I thought it was?

Now I'm even more confused.

Companies don't pass savings to the consumer, they price according to the market, so it isn't likely that Nvidia will try to undercut AMD
The 7970 is 50% more expensive than the 6970!
Well yeah, because they've priced it according to the market
I'm just stating that, defend it how you want.

Am I missing something?

 

[LOL_Wut_Axel]

Again, AMD gave the 365mm2 size that Reviewers worldwide used in their articles. There is no Official AMD paper stating anything about Tahiti's Die Size being 352mm2 as far as i know. This is the reason why all reviews still have the die size at 365mm2 which was the official size AMD gave.



LOL, you know them all up close and personal??

Alright, genius. Then why isn't 365mm^2 stated anywhere on AMD's site at all and why is 352mm^2 used instead? 365mm^2 is probably a ballpark figure an AMD rep gave before they stated the exact size later, which is 352mm^2.

the AMD Radeon™ HD 6970 Series GPU has been calculated at 2.703 TFLOPs of compute power with a measured die size of 389mm2, while the AMD Radeon™ HD 7970 Series GPU has been calculated at 3.789 TFLOPs of compute power with a die size of 352mm2.

http://www.amd.com/us/press-releases/Pages/amd-launches-worlds-fastest-2011dec22.aspx

Scroll to the bottom.

And I know they're not morons judging by the fact they've had overall better products than NVIDIA for two-three years (HD 4850 and 4870; 5770 and 5850; 6850, 6870, and 6950). About the only very good cards NVIDIA has made in recent years for most gamers are the GTX 260 Core 216, GTX 460 (always good and with a headstart, but now the HD 6850 is cheaper and also consumes less power), and the GTX 560 Ti.

 

[Crap Daddy]

(supposedly) leaked GK104 benchmark numbers:
http://videocardz.com/30564/nvidia-geforce-600-specifications-geforce-gtx680-benchmark-leaks-out


GeForce GTX 680 GK104 (GPU Clock: 1050, Memory: 1450 MHz) (overclocked)
vs
Radeon HD 7970 (stock)

=

Geforce GTX680 about ~5-10% faster, when overclocked vs stock 7970, in leaked results.



200 Mhz core overclock + 200 Mhz memory overclock = ~5-10% faster than 7970 stock.

Not bad for a card probably gonna be around 399$.

This is the source:
http://we.pcinlife.com/thread-1831876-1-1.html

It's floating around the net. I'd say that those are the stock clocks and not overclocked and the reason for this is to beat the 7970 out of the box. If this table is close to being real.

 

[tviceman]

200 Mhz core overclock + 200 Mhz memory overclock = ~5-10% faster than 7970 stock.

Not bad for a card probably gonna be around 399$.

Your math is wrong. If GK104's top end part runs at 950 mhz, 1050mhz is a 100mhz overclock, not 200mhz like you say.

 

[tviceman]

And I know they're not morons judging by the fact they've had overall better products than NVIDIA for two-three years (HD 4850 and 4870; 5770 and 5850; 6850, 6870, and 6950). About the only very good cards NVIDIA has made in recent years for most gamers are the GTX 260 Core 216, GTX 460 (always good and with a headstart, but now the HD 6850 is cheaper and also consumes less power), and the GTX 560 Ti.

It's funny that AMD has had the more efficient GPU's at graphics processing, yet they are the ones now confirming and moving in the direction Nvidia has been going in since the 8800GTX introduction (dual-purpose, GPGPU foundation). I think it's a strong indication that Nvidia's engineers saw that it was more efficient, from a business perspective, to sacrifice efficiency in graphics prowess to create GPU's that serve a purpose for more than just rendering graphics. And now AMD's GCN all but confirms Nvidia's vision is the better way to make money off of GPU's.

Weird how it works out, isn't it?

 

[badb0y]

(supposedly) leaked GK104 benchmark numbers:
http://videocardz.com/30564/nvidia-geforce-600-specifications-geforce-gtx680-benchmark-leaks-out


GeForce GTX 680 GK104 (GPU Clock: 1050, Memory: 1450 MHz) (overclocked)
vs
Radeon HD 7970 (stock)

=

Geforce GTX680 about ~5-10% faster, when overclocked vs stock 7970, in leaked results.



200 Mhz core overclock + 200 Mhz memory overclock = ~5-10% faster than 7970 stock.

Not bad for a card probably gonna be around 399$.

So now the GK104 is a GTX 680? What happened to the big Kepler?

It's funny that AMD has had the more efficient GPU's at graphics processing, yet they are the ones now confirming and moving in the direction Nvidia has been going in since the 8800GTX introduction.

Weird how it works out, isn't it?

I can't beleive AMD didn't go this route before nVidia seeing as how they were touting that the future would be fusion lol.

 

[Arzachel]

It's funny that AMD has had the more efficient GPU's at graphics processing, yet they are the ones now confirming and moving in the direction Nvidia has been going in since the 8800GTX introduction (dual-purpose, GPGPU foundation). I think it's a strong indication that Nvidia's engineers saw that it was more efficient, from a business perspective, to sacrifice efficiency in graphics prowess to create GPU's that serve a purpose for more than just rendering graphics. And now AMD's GCN all but confirms Nvidia's vision is the better way to make money off of GPU's.

Weird how it works out, isn't it?

You know it's kinda ironic to say that when we are discussing an arch from Nvidia that is rumored to move much closer to what's traditionaly expected from AMD (no hotclocking, weaker cores, but more of them), right?

 

[SirPauly]

Now I'm even more confused.

Companies don't pass savings to the consumer, they price according to the market, so it isn't likely that Nvidia will try to undercut AMD
The 7970 is 50% more expensive than the 6970!
Well yeah, because they've priced it according to the market
I'm just stating that, defend it how you want.

Am I missing something?

Usually a substantial and significant node and arch change redefine price/performance when compared to the older arch and node. This may be the first time in GPU history that a MSRP increase was more than its performance increase. Execution sure has its rewards.

 

[tviceman]

You know it's kinda ironic to say that when we are discussing an arch from Nvidia that is rumored to move much closer to what's traditionaly expected from AMD (no hotclocking, weaker cores, but more of them), right?

Yep! I was going to add that to badboy's reply. It's all ironically converging together.

 

[Grooveriding]

You are dead wrong. Think about it for a second (compared to 580):

• significantly smaller die
• cheaper PCB, power circuitry and cooling

This card should be only 50-100$ cheaper than a 580 if it performs similarly??? What value is that for a next-gen card? In this market segment, that has never happened to my knowledge. How much was the 8800GT, how much the GTX260, 460? You insist and insist on such nonsense but you don't bring any logical explanations to the table.

GK104 won't be faster than Tahiti XT. It will be a little bit slower but very much cheaper.

If you still insist, answer these questions:
1. If GK104 goes for 400-500$, what will they charge for GK110? How many people would buy that?
2. What card with a 350mm chip has ever cost this much since we have reached the feasibility of 500mm2 dies? Where is the precedent?
3. When they release cards with a performance that you can get today, like the GTX570 or the HD6970 and charge roughly the same how is that a good value? New products will have to compete with existing products, even if they are EOL. Stock still exists.

I'll answer your question once you provide some facts for the your bolded statement you are predicating all your questions on.

 

[blckgrffn]

Commenting mainly to subscribe to the thread... (can this be done without commenting? If so, how? nvm, I fail at using forums evidently. Found it.)

I really hope that this part is competitive to bring all of the prices down. Personally, I don't care how much money AMD/nvidia makes so long as they can pay their employees and fuel their R&D budget. Shareholders be damned, you leeches

The last time I was excited to buy a GPU was the launch day 7800GT that I paid $345 for from Dell of all places. If this card fits into the same niche - good performance, good power utilization, current generation feature set (DirectX 11.1? I haven't seen this mentioned?) and a fair price, hurrah we all win

And maybe I'll be able to get a decent deal on a card for folding (better than a GTX460), finally