AMD to launch Volcanic Islands GPU (HD8000s) in 2013

[Olivon]

We will see what happens. There might be something else in the works to compete with Maxwell...

Do you really think AMD got 50-60% better with a 28nm thing ?
Next round is 28nm again, no ?

[LordEC911]

I have not seen a post by anyone on the forums about their 7990, it may be the worst selling high end launch of all time. So I checked at newegg, all out of stock, but only about 20 reviews for all five cards.

Then I went to Amazon, not selling them. So I went to Tiger Direct, one model, no reviews.

I tried Google Shopping, got the one on TigerDirect, and a couple more on Ebay.

The 7990 seems to truly be The Card That Time Forgot.

You'd think AMD would have lots of these on hand as the chips are 17 months old, but no one has them and no one seems to be buying them the few that exist.

Contrast this with NV's last two $1000 chip launches: Lots of vendors had the cards, product going in and out of stock daily, lots of people on the forums posting about their new cards.

AMD could learn something about timing from this. If they put out a quad 7970 card next year as their high end, people will give it the same reception.

Don't know if they were duped by TSMC on 20nm process availability, but two whole years without new high end parts is disastrous for them. About the only way they'll be selling 7970s at any decent price is if they throw in a FX8350.

LoL... Nvidia just admitted that GTX690 had a very poor showing sales-wise.

Do you really think AMD got 50-60% better with a 28nm thing ?
Next round is 28nm again, no ?

Quite possible, yes.
Not necessarily.

[boxleitnerb]

50-60% better on 28nm is impossible imo. At 300W TDP with aggressive binning maybe, but why would they do that? Makes no sense to me.
AMD needs to do more. They increased performance by about 60-65% while increasing power consumption considerably (7970 GE vs 6970). Nvidia increased performance by about 80-85% while power consumption decreased a little bit (Titan vs 580). If that happens again, AMD has a real problem on their hands. They lost all the competitive edge they had with the HD5870 in 2009.

[zalbard]

LoL... Nvidia just admitted that GTX690 had a very poor showing sales-wise.

Nowhere they said GTX 690 sold poorly. Titan just sold much better than they expected.

You seem to be coming with your opinions as facts a lot recently. You should cut the BS. It's getting old.

Nvidia's dual-GPU behemoth, the GeForce GTX 690, has been out in the wild for over a year now, but the equally freakishly expensive GTX Titan has outsold it in less than three months. Now either the GTX 690 sold a pifflingly small amount (quite possibly) or the GTX Titan has been better received than even Nvidia thought.

According to Nvidia, it's the latter.

[Dimitriman]

Nowhere they said GTX 690 sold poorly. Titan just sold much better than they expected...

Either way you look at it, Titan outselling the 690 seems like a clear indication that a big monolithic chip will sell better than a dual chip card in most scenarios. Excluding all other market forces at play...

[raghu78]

AMD has done itself and the consumers no good by skipping a refresh. AMD could get 30% more perf from Bonaire HD 7790 compared to HD 7770 Cape verde at the same 85 W TDP. A similar update to Pitcairn (Hainan) and Tahiti (Curacao) was talked about. but that seems to have been scrapped.

http://videocardz.com/39041/meet-aru...hd-8000-series
http://videocardz.com/34981/amd-rade...fiation-leaked

AMD could have addressed the weaknesses of Tahiti, by adding more front end resources and going for 2560 SP, 4 ACE, 4 tesselators, 4 raster engines, 48 ROP, 384 bit memory bus chip at 420 - 440 sq mm. This chip could have easily improved per sp performance and with the newer power states and voltage binning (1.15 - 1.175v) AMD could have easily got 25 - 30% more performance at 250W TDP.

Maybe AMD were trying to cut costs or were constrained on resources due to PS4 / Xbox Next production and launch this year. That seems to be the only logical explanation. The bleeding CPU division hasn't made things easier for AMD. HD 7970 was the first GPU chip to be launched on 28nm and by this year end would have completed 2 years. AMD needs a very strong reply to resurrect its poor competitive situation and loss of brand value. The lack of competition has helped Nvidia to increase prices in the high end with Titan selling for USD 1000 and post record margins. But a strong reply from AMD is what will bring lower prices at the high end. Thats good for everbody , even people who prefer Nvidia cards.

[boxleitnerb]

AMD has done itself no good by skipping a refresh. AMD could get 30% more perf from Bonaire HD 7790 compared to HD 7770 Cape verde at the same 85 W TDP.

Yes, but TDP != power consumption. If you look at the reviews of the 7790 that measure the power consumption of the card only (not the whole system), you will see that the 7790 doesn't do exceptionally well - the 7770 did bad compared to Pitcairn. With Bonaire, AMD reached this level of efficiency in the lower segment that it had since Q1 2012. It is doubtful that you can transfer this to Tahiti or Pitcairn.
http://www.3dcenter.org/artikel/laun...d-7790-seite-2

The 7850 uses about 20% more power but is also 20% faster.

[SKYMTL]

LoL... Nvidia just admitted that GTX690 had a very poor showing sales-wise.

That is patently false. The GTX 690 exceeded the expectations of retailers, board partners and NVIDIA alike. So much so that they produced a second run months later and the cards you see in stock now are from a third, more limited run.

[raghu78]

Yes, but TDP != power consumption. If you look at the reviews of the 7790 that measure the power consumption of the card only (not the whole system), you will see that the 7790 doesn't do exceptionally well - the 7770 did bad compared to Pitcairn. With Bonaire, AMD reached this level of efficiency in the lower segment that it had since Q1 2012. It is doubtful that you can transfer this to Tahiti or Pitcairn.

http://www.3dcenter.org/artikel/laun...d-7790-seite-2

The 7850 uses about 20% more power but is also 20% faster.

HD 7770 rarely got close to 85w tdp at stock. as long as you can stay within same TDP its fine. definitely a 10w higher power consumption for 30% higher performance is expected even after optimizations. the laws of physics cannot be broken. Hainan was expected to be a 1792 sp GPU at 1.1 Ghz with 256 bit memory bus at 6 Ghz (192 Gb/s). performance would have been around GTX 680 range at around 160 - 170W TDP. We already know that HD 7950 competes with GTX 680 at the same clocks. basically Hainan would be AMD's GK104 equivalent. in fact slightly better perf/watt. The newer power states combined with voltage binning would have easily allowed to achieve these goals. Finally Curacao would have been even more easier as Tahiti had significant bottlenecks. Tahiti rarely was 50% faster than Pitcairn even though it had 60% more shaders. also HD 7970 Ghz was binned at 1.25v which adversely affected power efficiency. 1.175v would be ideal for Curacao. Even with 950 - 1000 Mhz clocks , addressing the front end bottlencks and increasing sp count to 2560 and ROP count to 48 would have enabled a gain of 20 -25% more performance at 250w TDP.

[boxleitnerb]

Performance/clock alone is irrelevant. And all of what you're saying is pure theory and speculation. A lot of "would", "if", "could".
I proved that there is nothing special with Bonaire in terms of efficiency and it is only natural that you have to be careful to take one conclusion and apply it to a different GPU family.

[raghu78]

Performance/clock alone is irrelevant. And all of what you're saying is pure theory and speculation. A lot of "would", "if", "could". I proved that there is nothing special with Bonaire in terms of efficiency and it is only natural that you have to be careful to take one conclusion and apply it to a different GPU family.

what are you saying. Bonaire was more power efficient than Cape Verde. if you can't even acknowledge the most basic of facts there is no point in speaking to you.

http://www.hardocp.com/article/2013/...card_review/10

"The new ASUS Radeon HD 7790 DirectCU II OC idled at the same level as the Radeon HD 7770, which was slightly more than the GeForce GTX 650 Ti. It also ate more power at full-load compared to the Radeon HD 7770. Think of it this way though, it only consumed 5% more power, but is delivering 30% more performance. That is quite a good efficiency. It's on part power wise with the GTX 650 Ti, but in a lot of games it was also faster. "

[boxleitnerb]

what are you saying. Bonaire was more power efficient than Cape Verde. if you can't even acknowledge the most basic of facts there is no point in speaking to you.

http://www.hardocp.com/article/2013/...card_review/10

"The new ASUS Radeon HD 7790 DirectCU II OC idled at the same level as the Radeon HD 7770, which was slightly more than the GeForce GTX 650 Ti. It also ate more power at full-load compared to the Radeon HD 7770. Think of it this way though, [B] it only consumed 5% more power, but is delivering 30% more performance. That is quite a good efficiency. It's on part power wise with the GTX 650 Ti, but in a lot of games it was also faster. [\B] "

Clearly you either haven't read my post or didn't understand it.
Yes, Bonaire is more efficient than Cape Verde.
No, that level of efficiency isn't new - AMDs Pitcairn is equally efficient give or take 2% or so.

I hope it is clear now...AMD didn't make new strides in efficiency, but they improved the weakest part of their GPU family in that regard to get it on par with Pitcairn. Nothing more, nothing less.

Oh and btw:
HardOCP doesn't measure the card only but the whole system, so your calculations with these values are completely off. Refer to the numbers I posted instead.

[Rollo]

That is patently false. The GTX 690 exceeded the expectations of retailers, board partners and NVIDIA alike. So much so that they produced a second run months later and the cards you see in stock now are from a third, more limited run.

You do not understand the fan and/or viral marketing world very well Sky.

If you're an AMD fan or viral marketeer, NVIDIA says,"We are shocked at how many people paid $1000 for 690s, but we're even more shocked a lot more people paid $1000 for Titans! Thank you for giving us record margins all you people with $1000 in your pockets!" but all you see is "690s sold very badly. Titans are doing a little better, but can't really be expected to succeed due to the lower performance.".

Raghu78 got it right- AMD needed a refresh, badly, to compete in the Q2-Q4 2013 market. People are going to get really, really tired of discussing the 7970. They'll get even more tired of buying it.

18 month old tech is 3/4 of the way down the EOL hill in the GPU and CPU world. About the only thing AMD can do to salvage is release the frame metering driver in working order, and fire sale 7970s so people buy another to save money.

[tajoh111]

Nowhere they said GTX 690 sold poorly. Titan just sold much better than they expected.

You seem to be coming with your opinions as facts a lot recently. You should cut the BS. It's getting old.

Lordecc always takes the most negative view on Nvidia. It isn't a surprise.

Gtx 690 had to sell reasonable well for Nvidia to even consider the 1000 dollar price point for the gtx titan. Also, the gtx 690 never had a price drop which typically happen after sales are bad, e.g like the whole 7xxx series after the gtx 6xx series came out(this is to just rub it in lordecc face a bit).

[raghu78]

Clearly you either haven't read my post or didn't understand it.
Yes, Bonaire is more efficient than Cape Verde. No, that level of efficiency isn't new - AMDs Pitcairn is equally efficient give or take 2% or so. I hope it is clear now...AMD didn't make new strides in efficiency, but they improved the weakest part of their GPU family in that regard to get it on par with Pitcairn. Nothing more, nothing less.

Oh and btw:
HardOCP doesn't measure the card only but the whole system, so your calculations with these values are completely off. Refer to the numbers I posted instead.

Pitcairn's efficiency was good. But what AMD desperately needed was a larger Pitcairn to fight GK104. A 1792 sp GPU with 256 bit memory bus at around 280 sq mm die size with similar or better perf/watt than GK104 and similar perf. That would help margins.

http://videocardz.com/34981/amd-rade...fiation-leaked

With the GTX 760 Ti (a faster GTX 670) expected to launch next month at USD 300 , this Hainan part would have been the right competition and good for AMD's margins.

[Dimitriman]

You do not understand the fan and/or viral marketing world very well Sky.

If you're an AMD fan or viral marketeer, NVIDIA says,"We are shocked at how many people paid $1000 for 690s, but we're even more shocked a lot more people paid $1000 for Titans! Thank you for giving us record margins all you people with $1000 in your pockets!" but all you see is "690s sold very badly. Titans are doing a little better, but can't really be expected to succeed due to the lower performance.".

Raghu78 got it right- AMD needed a refresh, badly, to compete in the Q2-Q4 2013 market. People are going to get really, really tired of discussing the 7970. They'll get even more tired of buying it.

18 month old tech is 3/4 of the way down the EOL hill in the GPU and CPU world. About the only thing AMD can do to salvage is release the frame metering driver in working order, and fire sale 7970s so people buy another to save money.

You're looking at things at a very opinionated point of view. Yes Nvidia has the lead right now but it will be at least 12 months before Maxwell will be released as I doubt the 780 series will have just a 6 months life span. If rumors are true and AMD does release a 20nm chip in Q4 you can be sure that it will at the very least beat the 780 series and will likely do so for a good 4-6 months before Maxwell is released. Then you will have the opposite situation as you have today. What would be said then about Nvidia and its delay in releasing 20nm chips?

Everything that happens is not only due to good/bad management decisions, production cycles have a lot to contribute, and while Nvidia is releasing new products now, it won't have much of an answer at the end of the year when its AMD's turn. Don't be so sure Nvidia will dominate the market for so long.

[SKYMTL]

You're looking at things at a very opinionated point of view. Yes Nvidia has the lead right now but it will be at least 12 months before Maxwell will be released as I doubt the 780 series will have just a 6 months life span. If rumors are true and AMD does release a 20nm chip in Q4 you can be sure that it will at the very least beat the 780 series and will likely do so for a good 4-6 months before Maxwell is released. Then you will have the opposite situation as you have today. What would be said then about Nvidia and its delay in releasing 20nm chips?

Everything that happens is not only due to good/bad management decisions, production cycles have a lot to contribute, and while Nvidia is releasing new products now, it won't have much of an answer at the end of the year when its AMD's turn. Don't be so sure Nvidia will dominate the market for so long.

The issue here is that AMD publicly said they won't have anything to replace their high end cards this year. To me, that holds more truth than rumors.

As it stands now, NVIDIA is essentially AT LEAST a half generation ahead of AMD. AMD needed a refresh which is something that should have happened five months ago. Instead, they will suddenly find themselves competing against refreshed NVIDIA cards that will walk all over them, necessitating price cuts.

To make matters worse, if AMD can't achieve at least a 75% performance increase over Tahiti XT with the next generation, NVIDIA won't even need Maxwell. They'll just reduce the price of TITAN, release a fully enabled GK110 and wait until Maxwell is fully baked.

And let's not forget AMD's track record on launches. We have constantly seen them release a card to reviewers only to make it available weeks or sometimes even months later. So Q4 2013? In the world of AMD launches, that means absolutely, positively nothing.

[boxleitnerb]

Wasn't the wording "until the end of the year"? That wouldn't exclude the possibility of seeing something in Q4.

As much as I like Titan (I have two) - it's "only" 30% faster than the 7970 GHz while consuming even more power. If AMD manages to gain 40-50% on the 7970 GHz (which would be little to normal with a new node), they would be comfortably faster than Titan and considerably cheaper to produce and sell.

[raghu78]

The issue here is that AMD publicly said they won't have anything to replace their high end cards this year. To me, that holds more truth than rumors.

As it stands now, NVIDIA is essentially AT LEAST a half generation ahead of AMD. AMD needed a refresh which is something that should have happened five months ago. Instead, they will suddenly find themselves competing against refreshed NVIDIA cards that will walk all over them, necessitating price cuts.

To make matters worse, if AMD can't achieve at least a 75% performance increase over Tahiti XT with the next generation, NVIDIA won't even need Maxwell. They'll just reduce the price of TITAN, release a fully enabled GK110 and wait until Maxwell is fully baked.

And let's not forget AMD's track record on launches. We have constantly seen them release a card to reviewers only to make it available weeks or sometimes even months later. So Q4 2013? In the world of AMD launches, that means absolutely, positively nothing.

Nvidia products based on GK110 are not going to be competing with AMD products. GTX 780 and Titan are in a different class of perf and price. obviously you should know more details on exact pricing but I expect GTX 780 to be around USD 650. GTX 770 based on GK104 will go head on with HD 7970 Ghz. thats a close battle. even when your consider both cards overclocked they are going to trade blows. GTX 760 Ti will have a much easier time against HD 7950 boost. AMD needs a 1 Ghz HD 7950 without any clock throttling to compete against a 1.1 Ghz GTX 760 Ti.

Because AMD skipped a Tahiti refresh Nvidia has no competition in the USD 500+ single GPU market. dual GPU products like HD 7990 should be competitive with the July frame pacing driver. but make no mistake AMD has never had two generations were they were behind significantly. after the HD 2900XT generation their response was aggressive with the HD 4870 / HD 4850. AMD has chosen to prioritize resources in difficult times. But you can count on a strong response with Volcanic Islands. Historically AMD has been first to new memory technologies as they are the driver behind the latest GDDR technology. A GDDR6 based Volcanic Islands product will be able to match the bandwidth of Tahiti / GK110 with a 256 bit memory controller at 9 - 10 Ghz. That would allow more shader power to be fitted into a given die size / TDP.

also a 75% faster product than HD 7970 Ghz is excellent and quite frankly very difficult on a 350 sq mm die at 20nm. Titan is roughly 33% faster than HD 7970 Ghz. a fully enabled Titan with 7.5% more shader power would be 40% faster. At 1.75x a Volcanic Islands chip will be 25% faster than a fully enabled Titan (1.4 x 1.25 = 1.75). such a jump from 28nm to 20nm is outstanding as the jump from 40nm to 28nm included transition to high k metal gate which provided better than historical gains in transisitor efficiency. the expectations for 20nm are quite low. thats also one of the primary reasons for TSMC to go to 16nm FINFET a year after 20nm launch.

http://www.cadence.com/Community/blo...hnologies.aspx

" While TSMC has four "flavors" of its 28nm process, there is one 20nm process, 20SoC. "20nm planar HKMG [high-k metal gate] technology has already passed risk production with a very high yield and we are preparing for a very steep ramp in two GIGAFABsTM," Sun said. Sun noted that 20SoC uses "second generation," gate-last HMKG technology and uses 64nm interconnect. Compared to the 28HPM process, it can offer a 20% speed improvement and 30% power reduction, in addition to a 1.9X density increase"

if AMD hit 1.75x of HD 7970 Ghz with the HD 9970 and stay around 350 sq mm they would have done extremely well. Maxwell GM104 doubling GK104 perf at similar die size / TDP will also be difficult. even 1.75x would be a huge achievement.

[hurleybird]

AMD has already had 2/3 the console market for years now with no appreciable impact on PC gaming.

Out of the three last gen consoles, one was too weak to be targeted in cross platform games, one had an Nvidia GPU, and the one with an ATI GPU had a unique architecture that never showed up on the PC. Of course this never had a benefit on PC gaming. It also meant that cross platform games could not be optimized as heavily for dissimilar consoles.

This generation, both of the consoles that actually matter for cross platform games (Xbox and Playstation) are using the same base architecture, and so is AMD on the PC side. They've given hints that GCN is going to be fairly long lived as well. It's an obviously different situation.

[solofly]

The issue here is that AMD publicly said they won't have anything to replace their high end cards this year. To me, that holds more truth than rumors.

I'm surprised you haven't seen this yet...

http://www.rage3d.com/board/showpost...8&postcount=53

http://www.rage3d.com/board/showpost...9&postcount=55

And I hope they do release something cause I'm tired of Nvidia charging arm and a leg for their stuff...

[highoctane]

This generation, both of the consoles that actually matter for cross platform games (Xbox and Playstation) are using the same base architecture, and so is AMD on the PC side. They've given hints that GCN is going to be fairly long lived as well. It's an obviously different situation.

Windows, OSX & linux are all using the same base hardware, so common hardware does not mean cross platform optimizations will definitely apply universally under different OS's.

It doesn't really matter how long GCN lives, console games are going to be optimized to run on lower end hardware than we already have on the PC platform today. Why in the world would there need to be concern about any high level of optimization going forward for PC gaming hardware from either Nvidia or AMD. They already have enough performance today compared to yet to be released consoles and PC hardware will be on new next gen uarch in the near future.

While technology and hardware performance jumps every generation consoles will still be in the same place they are today, do you remember the level of hardware you where running 3-5 years ago compared to what you have today.

The situation I see is still the same as it was before, different platform is a different platform even if hardware is common.

Even different releases of Windows OS can have an impact on overall performance and compatibility on the same hardware.

[SKYMTL]

I'm surprised you haven't seen this yet...

http://www.rage3d.com/board/showpost...8&postcount=53

http://www.rage3d.com/board/showpost...9&postcount=55

And I hope they do release something cause I'm tired of Nvidia charging arm and a leg for their stuff...

Of course I have but I put about zero credibility behind rumors.

Instead, I refer you to this official AMD slide:



If you want a transcript of their Q&A session with analysts, I can provide that as well. But here is a direct passage:

Brandon (press): Are you saying that the HD 7970 gigahertz edition will be your flagship single gpu card for the rest of 2013?

Devon Nekechuk (AMD): Yes, that is exactly what we are saying. While other cards will be rolled out throughout the year to fill out some segments, the ghz edition will remain in a top tier spot for the remainder of the year.
----------------

I can provide links to plenty of sites from TR to Anandtech to Tom's which received the exact same info.

So either one of AMD's premier product managers and pretty much their entire international product team told us the wrong info (which I highly doubt) or there's some viral marketing going on in an effort to blunt NVIDIA's upcoming launches.

Take your pick but I'm certainly not going to hold off on any purchases based on a rumor that runs contrary to everything AMD has been saying for the past five months.

To me, this smells of the same BS that went down when people and several sites hilariously assumed the "Solar System" parts were AMD's next gen desktop parts. Oh how the rumors flew then.

Again though; I say this in relation to high end SKUs. It's completely possible that AMD will have a lower end 20nm pipe-cleaning product in 2013 followed by a broader rollout in 2014.

[Piledriver]

As it stands now, NVIDIA is essentially AT LEAST a half generation ahead of AMD.

Great spin. So basically Fermi wasn't late it was a half generation ahead?

[solofly]

@SKYMTL

One can only hope...