[wccftech] rumor: amd hawaii benchmarked in 3dmark11 firestrike AMD attacks titan

[Kenmitch]

On the dual gpu theory. Wouldn't it be possible to design dual die gpu? Nothing saying everything needs to be on one die. Nothing saying a dual die chip is sli/crossfire.

Maybe have most on the master die, xtra shaders rops, etc on slave die interconnected on the same package....Kinda modular design I guess with more breathing room and ability to go crazy with rops, shaders, etc.

Something kinda like this image. I know it's an old intel cpu but gpus look similar.

 

[raghu78]

I give them credit where I feel it is due and I don't think I've said anything in a negative manner here in this thread. And don't intend to. Merely discussing opinions and speculations. I would, however, like a straight answer from Raghu in short order.

seriously are you acting naive.

http://www.anandtech.com/show/6751/amd-reiterates-2013-gpu-plans-sea-islands-beyond

"Finally, AMD also used a bit of their time to talk about their plans for the end of the year. With the 7900 series seemingly set as-is for the rest of the year, AMD has formally announced that they will be introducing a new GPU microarchitecture by the end of 2013"

a few days back an image from AMD website went on the web

http://wccftech.com/amd-volcanic-islands-hawaii-gpu-confirmed-amds-training-website/

I cannot clarify things any further. you can wait for Oct to get the so called facts. :whiste: Hawaii is the successor to Tahiti, the flagship product of Volcanic Islands family. a more optimized GCN architecture aka GCN 2.0.

 

[Keysplayr]

seriously are you acting naive.

http://www.anandtech.com/show/6751/amd-reiterates-2013-gpu-plans-sea-islands-beyond

"Finally, AMD also used a bit of their time to talk about their plans for the end of the year. With the 7900 series seemingly set as-is for the rest of the year, AMD has formally announced that they will be introducing a new GPU microarchitecture by the end of 2013"

a few days back an image from AMD website went on the web

http://wccftech.com/amd-volcanic-islands-hawaii-gpu-confirmed-amds-training-website/

I cannot clarify things any further. you can wait for Oct to get the so called facts. :whiste: Hawaii is the successor to Tahiti, the flagship product of Volcanic Islands family. a more optimized GCN architecture aka GCN 2.0.

This is what you've based your "facts" on. My friend, this is speculation. Please don't confuse fact, with speculation. Nothing in your posts indicate anywhere that AMD is going huge die to compete with GK110. Nothing.
It says new architecture. THAT is the only thing I can absolutely agree with and something that I also expect.

So, if this is how you "clarify" things for us, just don't bother in the future.
Now, shall I check off the "NO" box to my question to you? NO as in you do not know anything about Hawaii like you've indicated?

Got my pencil ready.

 

[GlacierFreeze]

delete

 

[ruhtraeel]

I think $450-$500 is definitely possible for a Titan killer.

Matured 28nm, no yield issues, 7970 dropping to $300

Would make sense IMO if the 7970 was $300-$350, 9970 (Titan performance) $450-$500, 7970 (>Titan and GTX 690 performance) $650-$700

I also believe they won't release a dual chip big die card, sort of like how Nvidia didn't with the Titan.

Hopefully this will be the last GPU I buy in another 4-5 years (I think I'm going all out and getting the best giang triple-slot cooler and everything)

 

[sushiwarrior]

This is what you've based your "facts" on. My friend, this is speculation. Please don't confuse fact, with speculation. Nothing in your posts indicate anywhere that AMD is going huge die to compete with GK110. Nothing.
It says new architecture. THAT is the only thing I can absolutely agree with and something that I also expect.

So, if this is how you "clarify" things for us, just don't bother in the future.
Now, shall I check off the "NO" box to my question to you? NO as in you do not know anything about Hawaii like you've indicated?

Got my pencil ready.

The burden of proof is on you, not anyone else. You claimed dual die, you have no supporting argument/proof, it's up to you to prove to us that it's a dual die, not up to us to prove to you that it's a single large die (which it is, honest to god, I swear upon baby jesus's little 7lb 8 oz sweet little face that it's a single large die).

 

[raghu78]

The burden of proof is on you, not anyone else. You claimed dual die, you have no supporting argument/proof, it's up to you to prove to us that it's a dual die, not up to us to prove to you that it's a single large die (which it is, honest to god, I swear upon baby jesus's little 7lb 8 oz sweet little face that it's a single large die).

keys as usual is just keen on derailing this thread , which is about Hawaii, so that it gets locked. he came up a with ridiculous theory about Hawaii GPU. lets not take his bait. just leave him be. :thumbsup:

 

[Gikaseixas]

His job is to ensure the competition is viewed in a negative light here at Anandtech Forums. Why do you guys keep on biting the same bait is just incredible

 

[tviceman]

With AMD refreshing HD7970 series on 28nm, perhaps 20nm is facing delays for whatever reasons (wafer costs, wafer allocation to other customers, etc.).

However, TSMC says 20nm volume production is on track for Q1 2014. Maybe this will be a very short-lived refresh if AMD/NV can launch 20nm parts by say July-August 2014.

For all we know, AMD could be late to the market with Volcanic Islands with regards to their own time tables. The 6970 was only out for a 12-13 months prior to the 7970, so this new product wouldn't have that much of a different lifespan in comparison. Your time frame for 20nm GPU's (July-August) is a good prediction. If anything, Nvidia might even be first to market *gasp* since they only had to release 1 commercial version of each Kepler chip and were able to reassign more engineers sooner (compared to Fermi).

AMD needs a substantial perf/watt improvement within the same node to release a bigger, faster chip. It's not impossible (see GF100 -> GF110) but no one thought Tahiti was "broken" like GF100 was when it was released. It's hard to fix / improve something that isn't broken and could be the reason for a late release. Anyways, synthetic benchmarks are crap, so no matter what fire strike says, I think this new high end card/chip will on average, match the gtx780. Nvidia will be forced to drop prices, and maybe we'll get a 15 SMX 900mhz 7.0ghz vram GTX785 for $699. Ugh. Is it too much to hope for GM104 being a bit faster than my mythical GTX785?

 

[sushiwarrior]

For all we know, AMD could be late to the market with Volcanic Islands with regards to their own time tables. The 6970 was only out for a 12-13 months prior to the 7970, so this new product wouldn't have that much of a different lifespan in comparison. Your time frame for 20nm GPU's (July-August) is a good prediction. If anything, Nvidia might even be first to market *gasp* since they only had to release 1 commercial version of each Kepler chip and were able to reassign more engineers sooner (compared to Fermi).

Anyways, synthetic benchmarks are crap.

The problem appears to be with TSMC, rather than AMD or Nvidia. I'm sure both would have had 20nm done months ago if TSMC was on schedule with their original 20nm predictions.

 

[tviceman]

The problem appears to be with TSMC, rather than AMD or Nvidia. I'm sure both would have had 20nm done months ago if TSMC was on schedule with their original 20nm predictions.

Certainly the slow down with transitioning to a new node is largely at the hands of TSMC, but the point I was making is that AMD has been designing two different sets of chips for the past few nodes and this type of workload may be catching up with them. At 40nm, they made Cypress/Evergreen/etc. then made Cayman/Barts which were not insignificant redesigns. Now at 28nm, they made Tahiti/Pitcairn/etc. and are preparing to release new chips which, again, are not insignificant in their modifications. They are not taking existing chips and tweaking a few transistors like Nvidia did with Fermi going from the 400 to 500 series, they are creating altogether physically different chips which takes way more R&D than simply tweaking existing GPU's.

Nvidia, on the other hand, released one set of chips with 28nm and while they did release significant refreshes with 40nm Fermi chips, all of Fermi v2 chips were tweaked but otherwise based entirely on their counterparts that were replaced and did not differ much (if any) in transistor count or die size. AMD may be executing smaller, incremental design changes between each new family of chips when compared with Nvidia's, but I'm arguing that it ultimately takes more resources to release 2 sets of different / overhauled architectures totaling 8 chips on the same node vs. 5 off of one architecture.

(I researched and added it up here: ANALYSIS. I found that AMD released 8 physical GPU's at 40nm, while Nvidia released only 5. So far at 28nm, AMD has released 5 so far with obviously more coming.)

 

[imaheadcase]

Its as if you guys don't remember what nvidia does every time AMD does a new release..lol

 

[badb0y]

If it doesn't compete with the Titan it doesn't deserve to exist.

Unless it's like GTX 780 performance for $500....that would be quite tempting.

 

[Kenmitch]

Its as if you guys don't remember what nvidia does every time AMD does a new release..lol

Bring up micro-stutter on crossfire? Pimp the advantages of PhysX or Cuda?

 

[Keysplayr]

The burden of proof is on you, not anyone else. You claimed dual die, you have no supporting argument/proof, it's up to you to prove to us that it's a dual die, not up to us to prove to you that it's a single large die (which it is, honest to god, I swear upon baby jesus's little 7lb 8 oz sweet little face that it's a single large die).

Why is the burden of proving anything on me? I said it could be x2. I also said it could be a single huge die, but I was skeptical. So, what is it I have to prove? That I have an opinion? That I'm speculating? Ok here goes:

I admit I have an opinion.
I admit I have been speculating.

That enough proof for you?

You guys couldn't have a decent passable debate to save your lives. Sheesh.

And Sushiwarrior, ( ^ in bold ) you need to enter that you have affiliations with AMD via direct or third party contact in your sig.
If you're going to swear to know things that only affiliates can know, I need that in your sig. So do others.

 

[Mr Expert]

Bring up micro-stutter on crossfire? Pimp the advantages of PhysX or Cuda?

Physx and definatly Cuda can hardly be considerd gaming features seeing as how they are not in every game in fact they are in so few as to be redundant.

 

[GodisanAtheist]

The 5870 to 6970 transition involved a lot of confusion (on the rumor mill's part) thanks to a delayed 28nm process and general misunderstanding of AMD's internal naming scheme.

AMD has previously demonstrated with that transition that they can work around node delays pretty well and come up with SOMETHING to occupy the market while TSMC gets its house in order.

We might be seeing a similar situation here with Volcanic Islands. VI might have originally been intended for 20nm, but when it was becoming clear that 20nm was slipping further and further away AMD retrained what was supposed to be GCN 2.0 to the GCN 1.1 that we see in bonaire. Nvidia had the easy out for the refresh thanks to their GK110 core (much like the GTX480 to GtX580 transition), AMD on the other hand has its work cut out for it.

They have a clearly defined performance target that is very achievable, I would expect them to match or exceed it (although everyone thought the same with the 6970...)

 

[sushiwarrior]

Why is the burden of proving anything on me? I said it could be x2. I also said it could be a single huge die, but I was skeptical. So, what is it I have to prove? That I have an opinion? That I'm speculating? Ok here goes:

I admit I have an opinion.
I admit I have been speculating.

That enough proof for you?

You guys couldn't have a decent passable debate to save your lives. Sheesh.

And Sushiwarrior, ( ^ in bold ) you need to enter that you have affiliations with AMD via direct or third party contact in your sig.
If you're going to swear to know things that only affiliates can know, I need that in your sig. So do others.

Okay, my new theory is that it is an octo-die configuration with LN2 cooling (only the first 10L are free), uses laser based display cables, and is powered by the tears of a dying unicorn. Now we both have equally valid theories that both are equally meaningful (aka useless)!

What if my roomate was buddies with an employee at AMD, would I need to put "lives with a person who knows someone who works at AMD"? What's it to you! I don't have an agenda, I'm just here to discuss.

 

[Keysplayr]

Okay, my new theory is that it is an octo-die configuration with LN2 cooling (only the first 10L are free), uses laser based display cables, and is powered by the tears of a dying unicorn. Now we both have equally valid theories that both are equally meaningful (aka useless)!

Changed my mind. Not worth direct response.

If you have a connection, tell us.

 

[ruhtraeel]

Changed my mind. Not worth direct response.

If you have a connection, tell us.

Isn't that counter-productive if he does that? Confirmed/true leaked specs = competitor info = advantage

Now I don't know if you yourself have leaked any specs or anything before, but unless there's some reciprocal contract that both companies sign in terms of competitor information disclosure, I'd assume it's in the best interest for the people/groups affiliated not to speak about these things.

 

[Red Hawk]

Why is the burden of proving anything on me? I said it could be x2. I also said it could be a single huge die, but I was skeptical. So, what is it I have to prove? That I have an opinion? That I'm speculating? Ok here goes:

I admit I have an opinion.
I admit I have been speculating.

That enough proof for you?

You guys couldn't have a decent passable debate to save your lives. Sheesh.

And Sushiwarrior, ( ^ in bold ) you need to enter that you have affiliations with AMD via direct or third party contact in your sig.
If you're going to swear to know things that only affiliates can know, I need that in your sig. So do others.

What you're suggesting goes against the norm of new GPU launches. Unless you can give a solid reason that this GPU launch will be any different than past GPU launches (ie large single chip that targets matching or beating the performance of the competitor's flagship) your opinion has about as much credence as something like what sushiwarrior described in his last post.

Anyways, OT, I do hope that Hawaii includes a third, or even a fourth, graphics engine in order to improve geometry performance significantly over Tahiti.

 

[tviceman]

If it doesn't compete with the Titan it doesn't deserve to exist.

Unless it's like GTX 780 performance for $500....that would be quite tempting.

I think it will trade blows with the gtx780 and come in at $500-550. I don't think it will be outright faster overall.

 

[Mr Expert]

I think it will trade blows with the gtx780 and come in at $500-550. I don't think it will be outright faster overall.

The 7970 was outright faster than the 580 so i think this next gen will be the same and historically it is usually the way it goes.

 

[Red Hawk]

The 7970 was outright faster than the 580 so i think this next gen will be the same and historically it is usually the way it goes.

The 7970 was also a manufacturing process leap, and a whole new architecture. Hawaii probably doesn't have either going for it, this will be more along the lines of a refresh than a truly new generation of GPUs.

 

[Keysplayr]

Isn't that counter-productive if he does that? Confirmed/true leaked specs = competitor info = advantage

Now I don't know if you yourself have leaked any specs or anything before, but unless there's some reciprocal contract that both companies sign in terms of competitor information disclosure, I'd assume it's in the best interest for the people/groups affiliated not to speak about these things.

You must be on retainer or something.