[SA] GK110 aka GTX 680 release date: Late Q3 '12

[MrK6]

I was told in the "nvidia will win big" Charlie article thread that Charlie was full of it and can't be trusted. Now that Charlie is saying kepler will be extremely late he's now legit.

Disappointing news. August/September seems way too late. Not really sure it matters anymore though. Heck my overclocked GTX 480 pretty much chews through any game I can throw at it and I just turn down a couple settings and BF3 runs great too. If pc gaming doesn't keep progressing these launches lose more and more importance.

Anyway, that photo of the leaked Sapphire product line showed a 1300MHz "7970" part, which I don't think is impossible if 28nm matures the same as other process nodes. There's a lot of potential left in 28nm, and it seems AMD went extremely conservative just to get something to market. Judging by the varying overclocking results (and assuming those reporting them are decently competent), there's still a bit of maturing/tweaking to go before we'll see a guaranteed 1200MHz+ part. That said, if this maturation is what NVIDIA needs just to get it's flagship off the ground, I'd believe they overestimated the capabilities of a new process (or their engineering team) yet again.

 

[blackened23]

They need to release 7970's at 1300mhz since all of them can do that. Strange though that many can't get above 1050mhz once they reach us.

Just like all GTX 480s couldn't do 950, huh

A refresh is a different chip based on the same architecture, refined with better TDP and higher clock speed. Just FYI. See also: GTX 580

 

[tviceman]

They dont even have to do a "refresh", as process matures the clock head room should go up. The cards aren't stuck on reference designs, this opens up AIBs to make special high OC cards and price it accordingly.

Or heck, they can do a 7970 XT with an extra 200-300 mhz at default settings.

This sounds pretty likely. I could see an XTX with a 1200-1300mhz clock. Thats still a huge performance leap over a stock clocked 7970, 30-40% higher as stated earlier.

Why don't they do it NOW and price it $100-125 higher than the hd7970?

 

[blackened23]

Why don't they do it NOW and price it $100-125 higher than the hd7970?

Why is peanut butter so expensive?

 

[Grooveriding]

Why don't they do it NOW and price it $100-125 higher than the hd7970?

Why ? They have the best cards on the market, the entire high end cornered and no competition in sight. Release their refresh early to show off and lose money ?

We've seen a 1200 core 7970 crush overclocked 580s by 50-80% and leaks of AIBs with plans for 1300core cards. The 7970 has perfect overclocked%= performance% improvement scaling.

They can sit on their hands and keep selling $550 7970s until the fall of 2012 when nvidia gets out a competitive card and refresh then.

 

[Quantos]

Anyway, that photo of the leaked Sapphire product line showed a 1300MHz "7970" part, which I don't think is impossible if 28nm matures the same as other process nodes. There's a lot of potential left in 28nm, and it seems AMD went extremely conservative just to get something to market. Judging by the varying overclocking results (and assuming those reporting them are decently competent), there's still a bit of maturing/tweaking to go before we'll see a guaranteed 1200MHz+ part. That said, if this maturation is what NVIDIA needs just to get it's flagship off the ground, I'd believe they overestimated the capabilities of a new process (or their engineering team) yet again.

I hope that Nvidia is scared of messing up a new architecture again. That'd really look dumb if they did. That said, I don't see them waiting for the process to be fully mature. If the speculation about the Tahiti refresh is correct, that would mean that AMD would not only have sold a card months before Nvidia, they would also come back once the process matured with a card to compete against GK110 (Tahiti).

 

[JAG87]

Why don't they do it NOW and price it $100-125 higher than the hd7970?

Because some people in this forum actually believe that AMD and NV would rather play a cat and mouse game, instead of maximizing their profit as fast as possible.

It's called the logic of the fanboy.

Why ? They have the best cards on the market, the entire high end cornered and no competition in sight. Release their refresh early to show off and lose money ?

We've seen a 1200 core 7970 crush overclocked 580s by 50-80% and leaks of AIBs with plans for 1300core cards. The 7970 has perfect overclocked%= performance% improvement scaling.

They can sit on their hands and keep selling $550 7970s until the fall of 2012 when nvidia gets out a competitive card and refresh then.

Completely illogical from a business perspective.

A 1300 Mhz 7970 for $550 right now would sell 2 times better, pull in money now to invest in R&D, and cherry on the cake, make their competitor look really bad when they release a product late that still can't keep up.

You're argument is like saying that X car company has an engine that can do 100m/g, but they are saving it because they are already leaders in fuel mileage.

If you have it, you release it, and profit from it.

 

[blackened23]

butterscotch

 

[MrK6]

I hope that Nvidia is scared of messing up a new architecture again. That'd really look dumb if they did. That said, I don't see them waiting for the process to be fully mature. If the speculation about the Tahiti refresh is correct, that would mean that AMD would not only have sold a card months before Nvidia, they would also come back once the process matured with a card to compete against GK110 (Tahiti).

Yep. Once again, prudence wins. Right now NVIDIA really has a Catch-22: if they wait long enough to get a proper part to market, they're missing a ton of sales. If they release early to grab sales, the product might be worse than intended and receive a not-so-receptive welcome (a la Fermi). I think they'll mitigate some of this by releasing a mid-range part early and competing with AMD in the higher-yield performance segment, but it may not be enough marketing-wise. You can tell AMD is eating it up right now, just check their main page: http://www.amd.com/us/Pages/AMDHomePage.aspx . I think AMD's marketing team might finally have someone with brains calling the shots, as this is how you get hardware sold.

Completely illogical from a business perspective.

A 1300 Mhz 7970 for $550 right now would sell 2 times better, pull in money now to invest in R&D, and cherry on the cake, make their competitor look really bad when they release a product late that still can't keep up.

No, I don't believe so at all. In fact, that move would be completely illogical from a business perspective. The amount of excess cost they'd have to absorb in R & D and binning chips would be astronomically disproportionate to whatever revenue they'd pull in from such a maneuver, especially in such a discriminating market segment. They chose specs that would handily beat NVIDIA's top offering while maximizing the amount of chips they can push to market.

 

[thilanliyan]

I really hope nV releases something competitive with the 7900 series at a competitive price. I couldn't care less if nV wins by 5% or 50%, as long as it's priced competitively to make AMD reduce prices.

 

[JAG87]

Yep. Once again, prudence wins. Right now NVIDIA really has a Catch-22: if they wait long enough to get a proper part to market, they're missing a ton of sales. If they release early to grab sales, the product might be worse than intended and receive a not-so-receptive welcome (a la Fermi). I think they'll mitigate some of this by releasing a mid-range part early and competing with AMD in the higher-yield performance segment, but it may not be enough marketing-wise. You can tell AMD is eating it up right now, just check their main page: http://www.amd.com/us/Pages/AMDHomePage.aspx . I think AMD's marketing team might finally have someone with brains calling the shots, as this is how you get hardware sold.




No, I don't believe so at all. In fact, that move would be completely illogical from a business perspective. The amount of excess cost they'd have to absorb in R & D and binning chips would be astronomically disproportionate to whatever revenue they'd pull in from such a maneuver, especially in such a discriminating market segment. They chose specs that would handily beat NVIDIA's top offering while maximizing the amount of chips they can push to market.

I disagree. The bulk of the revenue doesn't come from high end GeForce cards, it comes from low end first (OEMs) and mid range second (OEMs and most gamers). The enthusiast crowd that spends $500 on a GPU is tiny with respect to those groups. The much larger group that group that spends that much on a GPU are the workstation and HPC crowd, and NV doesn't have to worry about those customers since ATI has nothing to counter Quadro or Tesla. The Geforce brand (like the Radeon) pulls 90% of it's revenue from low and mid-range, ATI being first to the market with a high end 7970 means very little to them, because the customer base for that product is very small.

Exactly, they chose a spec that would allow them to maximize yield, which means that the 7970 can't do 1300 Mhz to begin with. In fact, there's a bunch of people who can't even get it to 1100. If it could do it in numbers, there is NO REASON for them not to have released it with a higher clock. The idea that they are waiting on their competitor is asinine.

 

[blackened23]

I disagree. The bulk of the revenue doesn't come from high end GeForce cards, it comes from low end first (OEMs) and mid range second (OEMs and most gamers). The enthusiast crowd that spends $500 on a GPU is tiny with respect to those groups. The much larger group that group that spends that much on a GPU are the workstation and HPC crowd, and NV doesn't have to worry about those customers since ATI has nothing to counter Quadro or Tesla. The Geforce brand (like the Radeon) pulls 90% of it's revenue from low and mid-range, ATI being first to the market with a high end 7970 means very little to them.

Exactly, they chose a spec that would allow them to maximize yield, which means that the 7970 can't do 1300 Mhz to begin with. In fact, there's a bunch of people who can't even get it to 1100. If it could do it in numbers, there is NO REASON for them not to have released it with a higher clock. The idea that they are waiting on their competitor is asinine.

If you track market trends you would see that intel is dominating the low end market by a LOT, and a lot of PC makers are removing discrete graphics in favor of IGPs to offset costs of HDD's. Further, AMD actually has far more oem sales than NV does, although NV has a higher discrete share. The ramifications of this remain to be seen, but one can guess that AMD and NV will concentrate more on high end graphics, and then apply those graphic technologies to their other product lines - APUs for AMD and tegra for NV.

So with the low end out of the picture, that leaves the mid range aka 300$ and up market. There is no reason to chase the low end market. It is dead. Gone. Stick a fork in it. RIP. Intel has a 60 something percent share there, and it is increasing every quarter.

Exactly, they chose a spec that would allow them to maximize yield, which means that the 7970 can't do 1300 Mhz to begin with. In fact, there's a bunch of people who can't even get it to 1100. If it could do it in numbers, there is NO REASON for them not to have released it with a higher clock. The idea that they are waiting on their competitor is asinine.

This argument is ridiculous. A refresh entails making a new chip with better TDP and higher clocks, but based on the same architecture. Now using your argument I could say that the Fermi could never do 787 - 850 mhz to begin with . The 1900XTX used a different (refined) chip than was used in the 1900. The GTX 580 has a different (refined Fermi) chip than the 480. Its NOT a hard concept to understand. Although the counter argument holds as well - AMD will have 6 months of competition free sales, so there's little point at this stage of the game. Keep in mind that AMD could probably do a 1300mhz part now, but yields would not be ideal (i'm speculating here) - if they are indeed making a part (and i'm not sure), they are liking waiting for improved yields.

 

[tviceman]

No, I don't believe so at all. In fact, that move would be completely illogical from a business perspective. The amount of excess cost they'd have to absorb in R & D and binning chips would be astronomically disproportionate to whatever revenue they'd pull in from such a maneuver, especially in such a discriminating market segment. They chose specs that would handily beat NVIDIA's top offering while maximizing the amount of chips they can push to market.

The sapphire docs are either showing a physically different Tahiti chip or, more likely, were faked and/or misinformed papers because it showed more shader cores than what Tahiti has.

Either way, if AMD is to release a faster Tahiti part at some point in the near future, it makes no sense to delay it's release. Why not just release it now and charge an ultra premium for it? If they are going to have to pay for binning expenses, it would make more sense to maximize the profit margins as much as possible to recoup losses. One chip would then be in cards selling for $450, $550, and lets just say $650 for hd7970 XT. That makes a hell of a lot more sense than waiting for the competition to release, dropping your prices, and bringing in the hd7970 XT at either the same or lower price as what the current hd7970 is selling for now.

 

[kevinsbane]

Completely illogical from a business perspective.

A 1300 Mhz 7970 for $550 right now would sell 2 times better, pull in money now to invest in R&D, and cherry on the cake, make their competitor look really bad when they release a product late that still can't keep up.

That is, if they could keep up with the additional demand. But they already have a product that makes their competitors look bad* and is selling reasonably if not very well. Will releasing a faster product really increase their marginal profit overall, as opposed to releasing that faster product after a period of time once demand dies down for the initial product?

Basically, if AMD will make $100m revenue with the 79xx series as it is now, but would make $110m revenue with an improved 79xx series, then it may not make any business sense to release the improved version now. Especially if the improved version has lower yields at current process maturity (and thus higher cost), which seems to be a reasonable assumption to make.

*YYMV on my wording. Suffice to say the 7970 is undisputedly faster than the 580, and Nvidia's 28nm parts have been rumoured to only have recently taped out.

 

[AdamK47]

It would appear as if I made the right call in getting the 7970s now rather than wait until the next high end nVidia card is released.

 

[OILFIELDTRASH]

Why ? They have the best cards on the market, the entire high end cornered and no competition in sight. Release their refresh early to show off and lose money ?

We've seen a 1200 core 7970 crush overclocked 580s by 50-80% and leaks of AIBs with plans for 1300core cards. The 7970 has perfect overclocked%= performance% improvement scaling.

They can sit on their hands and keep selling $550 7970s until the fall of 2012 when nvidia gets out a competitive card and refresh then.

This makes perfect sense and would explain why most 7970 owners are getting only 100-150mhz overclocks. Amd is binning all the monster 400+ mhz overclockers for the refresh. When nvidia releases a faster card amd releases a 400mhz OC cards although most only get 100-150mhz overclocks currently.

 

[skipsneeky2]

Frozen bloody tampon popsickle=kepler.

Math is not my biggest major sorry.

 

[sontin]

Because they had smaller dies and the legwork was already done with GF100. You know why both AMD and Nvidia usualy release the biggest chips before smaller ones? Because bugs, that can slip by unnoticed with a small chip, are much more likely to be an issue for a larger one. GK104 isn't that much of a help and even less so, if GK110 has additional compute features that need debugging.

So, the first card based on kepler will have a BigK Chip? :whiste:

 

[Grooveriding]

Completely illogical from a business perspective.

A 1300 Mhz 7970 for $550 right now would sell 2 times better, pull in money now to invest in R&D, and cherry on the cake, make their competitor look really bad when they release a product late that still can't keep up.

You're argument is like saying that X car company has an engine that can do 100m/g, but they are saving it because they are already leaders in fuel mileage.

If you have it, you release it, and profit from it.

I think about the only illogical perspective is the one from the consumer choosing not to have the card available to buy now for $550 instead of the lower clocked version available.

Why would AMD blow their proverbial refresh load when they have no competition ? Are they even in a position to put out cherry 7970s that consistently clock to 1200 and 1300 with voltages they feel are reasonable ? 28nm is still in its infancy at TSMC and a lot of customers are not even getting usable chips out of the process yet.

...Bryant said that there are 10 designs in manufacture from seven companies. "We're now hearing none of them work; no yield anyway,"...

It would be illogical not to wait for their manufacturing to improve, making releasing a faster stock 7970 easier on them. They don't need to out-do themselves - they need to out-do nvidia, this is not Intel against themselves in the high end here. There is real neck and neck competition between AMD and nvidia in the GPU area. I can't see why AMD would squander what they still have left to put under the hood of Tahiti just to out-do their current best in class card on the market.

 

[tviceman]

I think about the only illogical perspective is the one from the consumer choosing not to have the card available to buy now for $550 instead of the lower clocked version available.

Why would AMD blow their proverbial refresh load when they have no competition ? Are they even in a position to put out cherry 7970s that consistently clock to 1200 and 1300 with voltages they feel are reasonable ? 28nm is still in its infancy at TSMC and a lot of customers are not even getting usable chips out of the process yet.



It would be illogical not to wait for their manufacturing to improve, making releasing a faster stock 7970 easier on them. They don't need to out-do themselves - they need to out-do nvidia, this is not Intel against themselves in the high end here. There is real neck and neck competition between AMD and nvidia in the GPU area. I can't see AMD would squander what they still have left to put under the hood of Tahiti just to out-do their current best in class card on the market.

Why not just release a higher priced part now and maximize their profits? $650 hd7970 XT. It doesn't make sense to me that they are going to wait and may end up having to readjust prices to match what Nvidia comes out with. Then their refresh will be the same price or maybe even cheaper than what the current hd7970 sells for now, and they would have lost the money from releasing it before hand and pricing it higher than $550.

 

[Grooveriding]

Why not just release a higher priced part now and maximize their profits? $650 hd7970 XT. It doesn't make sense to me that they are going to wait and may end up having to readjust prices to match what Nvidia comes out with. Then their refresh will be the same price or maybe even cheaper than what the current hd7970 sells for now, and they would have lost the money from releasing it before hand and pricing it higher than $550.

Why doesn't nvidia just release their high end GPU and maximize profits ? $650 GTX 680...

Oh wait.. it's not ready and there is still a long wait ahead.


Why doesn't AMD just release a higher priced refresh of a part they released under a month ago and maximize profits? $650 7970XT

Oh wait... the GTX 680 is nowhere in sight.. it's not ready and there is still a long wait ahead.

 

[GaiaHunter]

A 1300 Mhz 7970 for $550 right now would sell 2 times better, pull in money now to invest in R&D, and cherry on the cake, make their competitor look really bad when they release a product late that still can't keep up.

So the 7970 could go OoS twice as fast?

AMD not selling higher clocked parts is probably due to 2 reasons: not enough chips to make it worth all the hassle (especially if that would mean another sku) and those chips would require too much power.

 

[Olikan]

Why not just release a higher priced part now and maximize their profits? $650 hd7970 XT. It doesn't make sense to me that they are going to wait and may end up having to readjust prices to match what Nvidia comes out with. Then their refresh will be the same price or maybe even cheaper than what the current hd7970 sells for now, and they would have lost the money from releasing it before hand and pricing it higher than $550.

Doing a refresh is just not that simple, it takes a lot of money, time and people.

Right now, amd just want to release the 78XX and 77XX series...
Than they probaly will want to release the workstations and HPC cards.

Other thing, doing a refresh right now, it will make every partner that desiged a custom cooler lose alot of money...
heck, custom coolers take months to show up, tahiti could probably have been released way early, if amd didn't want to hold up and be nice with the partners

 

[tviceman]

Why doesn't nvidia just release their high end GPU and maximize profits ? $650 GTX 680...

Oh wait.. it's not ready and there is still a long wait ahead.


Why doesn't AMD just release a higher priced refresh of a part they released under a month ago and maximize profits? $650 7970XT

Oh wait... the GTX 680 is nowhere in sight.. it's not ready and there is still a long wait ahead.

My question had absolutely nothing to do with Nvidia and had everything to do with AMD having the ability to do it now and maximizing their profits. Way to not answer the question. :thumbsup:

Doing a refresh is just not that simple, it takes a lot of money, time and people

We are not talking about a refresh. We are talking about the currently existing Tahiti chip, clocking it higher, and officially sanctioning it as new model. Everyone says Tahiti can overclock like crazy with very little power draw. Why not capitalize on it, bin faster Tahiti chips, and sell it for $100 more than what the hd7970 sells for.

 

[tviceman]

AMD not selling higher clocked parts is probably due to 2 reasons: not enough chips to make it worth all the hassle (especially if that would mean another sku) and those chips would require too much power.

No one wants to admit reason #3 - AMD has no intention of releasing a higher clocked Tahiti chip.