AMD to launch Volcanic Islands GPU (HD8000s) in 2013

[MaddMuppet]

http://wccftech.com/rumoramd-volcani...am-processors/


If true... sweeeeeet

Here's hoping.

[kromosto]

yes if true that is really will be a nice gpu but if it is at 1000usd segment doesn't mean a good thing to us.

is compute unit numbers placed wrong?

[aussie-revhead]

Too late to stop people buying 700 series right? Or is it late enough to be another upgrade after 700 series?

[Olivon]

September 2012

http://www.semiaccurate.com/forums/s...28&postcount=7

I call fake.

[boxleitnerb]

Opteron APU? Certainly not VI with a DDR3 memory interface.

[Rollo]

That's why people buy cards like that in the first place - to go on forums and brag about their overpriced hardware.

That is why some people do, I agree.

What I was replying to though is "No one is buying Titans" and as a guy who's been trying very hard to resist the urge to do so, I can say a. they go in and out of stock daily, everywhere b. I see a lot of people on forums with them.

My personal "I won't support a $1000 GPU" protest is apparently meaningless as they seem to be selling all they can make.

[Rollo]

Whether these cards really launch or not doesnt matter much. This news is in response to Nvidia's 700 series trying to get customers away. All AMD has to do is say right now is that the timetable has been moved up, which will keep a few people from buying a new Nvidia card. Then AMD can either stick to their new timetable or can come back in 2-3 months from now and say "well we have some setbacks so now it is back to 2014 again". people will be dissapointed if that happens for sure, but AMD will have achieved the primary goal they were going for, to keep some people from buying the new Nvidia cards.

"Maybe Q4 13" is a weak attempt to do that.

My guess is most people in the market for a card in Q2 13 will not see this and say "Holy cow! I'd better wait to see what AMD may have up their sleeves in Q4!". Too far off, and still just a rumor.

If I was running the show at AMD, I would have launched a 7970 GHz+ edition with a water block and closed loop HSF for $600-$650 as a stop gap product. No CFx worries, if you can add 10% to the GHz edition performance you've got a product firmly between the 680 and Titan, likely to be not far below Titan LE.

I'd buy that card.

[Sushi Warrior]

First batch of Hawaii is already out of the oven btw.

[Rollo]

First batch of Hawaii is already out of the oven btw.

OK but if that's A1 silicon how often is that sold? Isn't it usually A2 - A3, with A3 even more likely on a big chip?

Or is Hawaii one of the small chips we generally don't care about on forums like this?

[Sushi Warrior]

OK but if that's A1 silicon how often is that sold? Isn't it usually A2 - A3, with A3 even more likely on a big chip?

Or is Hawaii one of the small chips we generally don't care about on forums like this?

Pre-A1 still! Very first batch, not close to retail at all. A1 will be coming too, then A2 and probably A3, so a few months to go still. This is the full HD8000 series chip, top of the line.

[boxleitnerb]

20nm or 28nm?

[Hornet331]

hmm I am skeptical about that, even at 20nm 4k SPs would make quite a huge die.

[Dimitriman]

OK but if that's A1 silicon how often is that sold? Isn't it usually A2 - A3, with A3 even more likely on a big chip?

Or is Hawaii one of the small chips we generally don't care about on forums like this?

Isn't it usually 6 months between tap-out and retail production? This would put the chip on Q4'13...

[Sushi Warrior]

20nm or 28nm?

I didn't get any info on that, never saw any specs. Just that "it exists!".

Isn't it usually 6 months between tap-out and retail production? This would put the chip on Q4'13...

Sounds right to me. Sounded like there were very few hardware bugs though, so it should only need A2 and then be good to go. By the end of the year I would think. Nothing ASAP.

[tajoh111]

hmm I am skeptical about that, even at 20nm 4k SPs would make quite a huge die.

Completely impossible at 28nm with the 64rop and 512 bit memory. Even tough at 20nm. AMD graphics haven't ever been that big and getting a working chip like that fully enabled might seem impossible for them to do at this point. Also AMD has been more in the business of making their shaders larger but with a higher utilization, hences its modest shader increase for the last 2 generations.

Doubling up shaders, might not do them much good if they are not utilized.

When has that website ever been taken as credible? Its takes any remote rumor and turns it into a story that looks true. The only time they have been accurate is when they steal the information from some other website that has some credibility. They usually miss a dozen times before this happens.

[informal]

I doubt this new core will have 4096SPs and 512bit memory interface.

[LordEC911]

They have had silicon in house for over a month now.
There are at least 3 plausible rumors floating around out there right now but with a key source keeping quieat it might be awhile until things get semi-confirmed.

FYI- They are/were targeting a ~60% increase over Tahiti.

[iMacmatician]

There are at least 3 plausible rumors floating around out there right now

Is the 20 nm rumor one of them?

FYI- They are/were targeting a ~60% increase over Tahiti.

That seems like a lot. So something like 48 CUs and a 512-bit bus at higher clocks could get there?.

[Chrono Detector]

If the 512 bit memory bus is true (Which I doubt) I wonder how much it would cost, and probably have have at least 4GB or even 8GB of VRAM.

[Sushi Warrior]

They have had silicon in house for over a month now.
There are at least 3 plausible rumors floating around out there right now but with a key source keeping quieat it might be awhile until things get semi-confirmed.

FYI- They are/were targeting a ~60% increase over Tahiti.

Over a month? Well, either AMD is telling me the wrong thing or your source is exaggerating 2 weeks old is exactly what I was told (not a rumor, absolute proof haha). But not a huge difference.

[PedantOne]

it is absolutely impossible that VI will be launched this year on 20nm, maybe on 28nm

[Dimitriman]

If 20nm they will be competing with Maxwell. Then it better be 4096 shaders and at least 60% faster than Tahiti...

[solofly]

With this news I'm not so sure I'll be getting what I was planning to get which was two 780s. I might just get one more 680 and wait until Maxwell...(benches and price will make that decision for me)

[Dimitriman]

With this news I'm not so sure I'll be getting what I was planning to get which was two 780s. I might just get one more 680 and wait until Maxwell...(benches and price will make that decision for me)

you are plenty fast with 2x 680s. My opinion: wait for PS4 to launch and then we will see some new games in the market that will make good use of the PC and then you will have a good reason to buy into the next gen GPUs.

[zalbard]

FYI- They are/were targeting a ~60% increase over Tahiti.

Then it's not 4096sp.

60% is probably an increase over regular 7970, which is around 50% faster than GHz edition, which is attainable by 3072sp / 512 bit GDDR5 configuration with slight architectural tweaks.