The NEXT LAST R600 Rumours & Speculation Thread

No it does not. But I think you know exactly what I mean (in regards to Amsterdam being cancelled). Sorry if my articulation wasn't up to your standards....

 

I bet Walt thinks 3DRealms Duke Nukem Forever is the greatest business strategy ever devised.

 

Heh I get the sense that geo knows the real deal and he's getting really frustruated with all us headless chickens pecking around in the dark. He seems to have a low tolerance for the whiners and bitchers as of late

 

Maybe far fetched idea, but could it be as simple as "they've seen the oem card already in million pics, we can still keep the retail form secret if we use them on demo machines"

 

Mt apologies if this was mentioned, tough to read the whole thread at work...

For reference, how many flops would a top end AMD X2 6000+ and two x1950XTXs generate?

 

I cant tell you my real feelings on this or I'll get in trouble.

Lets just say the sooner Intel gets into graphics the better, because ATI is not interested in competing, to the point of losing on purpose, imo.

I know that sounds stupid, but it's also very obvious. R600 was ready, and they chose to delay it for no reason at all, essentially. What other explanation is there?

I'm sick of this company. They might as well let Nvidia buy them already, it's what they've wanted all along.

It's going to suck having only one graphics IHV, but oh well. ATI has proved what they want, failure.

Your questions Trinibwoy are obvious ones. Why the hell did they do this. A bug, hardware problems? Those would make sense but they're clearly not the case. So it boggles the mind that they would delay, when they are already so late, it truly does.

There only seem to be two flimsy possible reasons. To launch with Barcelona, or too launch a full lineup at once. Both reasons are emintly stupid and ridiculous, and a true sign of a anti-competitive company. It's not even imaginable that Nvidia would do this, for example.

 

You're forgetting the most probable reason, the possibility of a hard vs. soft launch, and the time needed to create the chips for a hard launch exceeding the previous launch window. It may also allow them to do these others things at the same time by doing so; family launch, K10, better drivers, physics API/CTM apps beyond F@h, cheaper ram prices for cheaper production and perhaps cheaper prices...Who knows what else could happen in that time frame compared to launching in March.

I don't think it's a bad idea when MANY of their high-profile launches over the last several years have been soft, and they BADLY need to reverse the PR on their launches. Nvidia has done it, with the exception of the 7800GTX512, and it's done wonders for them. I think AMD is trying to do the same for ATi.

Cause (creating supply for hard launch) = effect (possible family launch...etc), rather than effect (late launch) = cause (waiting for Barcelona).

 

And if they do that, what will happen over the next several months until 2008 ?
No launches ? No press ?

Will they leave G80 refresh to Nvidia's marketing alone ?
Will they leave Core 2's 45nm refresh to Intel ?

Too many answers are yet to be found in this line of thinking.


IMHO, the reasons for the delay were simple: high speed GDDR4 price and availability.

 

Think about this from a purely business standpoint from AMD. CPU sales account for about $1.35 billion per quarter. ATI sales are about $500 million per quarter total. Say about 60-70% of that is pure GPU sales so let's call it $300 million. Therefore, the CPU business drives 4x the amount of revenue that the GPU business does and brings in much fatter margins.

AMD need to do several things right now to regain traction against Intel:

1. Demonstrate they have the technology lead again.
2. Drive sales of their chipsets, CPUs and GPUs concurrently with a "total platform" approach.
3. Improve corporate gross margins at all costs to improve cash flow.
4. Regain that "top of mind awareness" that comes from having the technology lead and the hot product(s).

AMD needs to regain traction in servers, high end desktop and mobile to keep their company liquid. If Barcelona isn't it, AMD is screwed...totally screwed. So what do they do? They use EVERY tool they have to show Barcelona in the BEST possible light to get the BUZZ going again about AMD products. Launching R600 by itself would only prove that ATI produced a great chip...not what AMD has done to counter Intel.

AMD needs Barcelona to be another Opteron or they are in serious financial trouble...they know it and are betting the farm on Barcelona and using R600 to help the buzz. They MUST stimulate server and high end desktop again or their cashflow will remain negative with their product mix skewed to the low end SKUs.

By showing a Barcelona/R600 launch you get MAJOR buzz. Likely the fatsest CPU/GPU combination on the planet. So instead of getting headlines that say "R600 fastest video card" you get "AMD powered computer sets new records". People can say in one breath...AMD has the fastest hardware on the planet. CPU, GPU and maybe chipset as well. AMD wants to sell CPUs first, followed by GPUs followed by chipsets to improve their gross margins.

In regards to OEMS, having multiple products available in quantity will help them with design wins and help them move product and maybe grab the coveted high end spots like XPS at Dell. Right now they are painted into the low margin low cost corner and it is crushing their margins.

So I can see why AMD is delaying R600 IF and ONLY if Barcelona desktop chips are not far behind. Hard launch R600, hard launch Barcelona servers and paper launch Barcelona desktop parts all on the same day. Otherwise what is the point and there must be a yield/technical issue or supplier availability issue that was delaying the finished product.

What I CANNOT explain is why it appears this was a last second decision? Maybe Q1 unit sales were in the toilet, Q2 was looking the same and they simply HAD to do this or risk major carnage. Or maybe someone had an epiphany???

EDIT* Typos

 

That Barcelona/R600 demo was no coincidence and if you read Henri's comments the past week or so I think they can see where they are going.

 

509Gflops?



What looks like retail R600's using a 4x4 chipset/Agena FX CPUs they used to demo the teraflop:




I'm still betting on 512GFLOPs, and them being clocked at 800mhz, but a few flops are needed for the display etc and are unavailable to brook (F@H)
I guess it could also be a dual opteron board that supports crossfire...Maybe. I like the first option better.

http://hardocp.com/news.html?news=MjQ0MTcsLCxobmV3cywsLDE=

 

Those look just as long as the OEM-version with the same fan placement.

Scary powerfull PSU aswell.

 

No internal/external CF bridge?

Oh and by the way, drool.

*edit* Thank god I have dual PSU's and a Stacker. I was getting worried for a second there

 

How smooth was ATI's integration into AMD?

Because those are OEM cards.

As for the PSU, dual-cpu, dual-gpu system and its obvious.

 

Did i read the article correctly ?
300W power consumption for a single GPU ??? 600W for a Crossfire Setup ??????
Jesus Christ ! (and i'm not even a religious man )


P.S.:
AMD's cable management skills suck.

 

hmm anti christ? room will sure get hot enough!

Where the hell does 509 flops come from, what config lol

 

If Crossfire is 100% efficient in doubling R600, then yes. But that just doesn't happen. So I'm wagering it's actually much higher than 509 Gflops for an individual card.

 

How could they possibly screw up thermal managment so bad that with about the same amount of transistros as nVidia they can achieve double of nVidias power consumption? This is false rumour, or ATi needs to get better engineers...

 

Hey, it was a reasonable guess before.

It seems to me that test is only testing the R600's (surely I could be wrong)...So 1018/2 = 509GFLOPS/S. If nothing else, it seems like they're all MADD...and darn close to the theoretical 100% if I'm right.

Very good point, and you very well could be right. Does anyone know off-hand the difference in usable Flops in F@H with one x19xxxx vs 2x in crossfire?