All AMD R6xx chips are 65 nanometre chips, now

[Greenman]

I think they should delay until they have a 25nm fab, and can put a gig of DDR4 on it. Make the wait worth while.

 

[Cookie Monster]

Originally posted by: Greenman
I think they should delay until they have a 25nm fab, and can put a gig of DDR4 on it. Make the wait worth while.

Yea, and with a launch date of Q1 2009...

:roll:

 

[Matt2]

A reason for this are problems or the bad yield of 80 nm of the process which should not probably have unimportant difficulties with leakage streams.

I still have a hard time believing that yields are better with 65nm than 80nm.

Doesnt TSMC fab Nvidia GPUs as well?

If rumors are true that G81 is 80nm and is done, then I dont see how AMD is having so much trouble with 80nm. One of these rumors is false.

 

[Acanthus]

Originally posted by: Cookie Monster
65nm? what so the INQ thinks you can just jump to 65nm in 5 months?

What people dont understand is that from 90nm to 80 nm is possible, because its just an optical shrink of the GPU architecture. However from 80nm to 65nm is another story. 65nm is a totally different process, and the fact is that you cannot just shrink to 65nm. The architecture itself has to be revised, tweaked and changed to accomodate the new process.

Unless they had two teams working on both 80nm and 65nm from a long time ago, i just dont see this as possible. Scraping the 80nm design is financially illogical to do so. The time period for such move is too short to be actually believable.

It would explain why the core has been respun 15 times. As opposed to the industry standard 1-3.

 

[Acanthus]

Originally posted by: munky
Well, that would be one hell of a surprise if true. However, I don't think either AMD's or TSMC's 65nm process is mature enough to result in big improvements over 90nm or 80nm silicon for now.

TSMC has been making 65nm chips for about 4 months now.

They *CAN* make them, but who knows how reliable the process is or what the yields are.

There is no way in hell AMD is producing the GPUs in house, they just dont have the capacity.

 

[Janooo]

Originally posted by: Matt2

...
If AMD delayed R600 just so RV610 and RV630 could come out on 65nm, I will be extremely upset.

It appears they don't want you upset
If the 65nm is true then some of those AMD exec's comments about the delay make sense now.

 

[Cookie Monster]

I dont think A15 means the 15th revision. We dont even know if A15 actually exists.

This is how i think of it.

A10 = Original
A11 = 1st revision
A12 = 2nd revision
and so on.

 

[ShadowOfMyself]

Originally posted by: Greenman
I think they should delay until they have a 25nm fab, and can put a gig of DDR4 on it. Make the wait worth while.

Actually, it already has a gig of DDR4 on it by the time we get to 25nm we should probably be on 8 gb or something

 

[Acanthus]

Originally posted by: Cookie Monster
I dont think A15 means the 15th revision. We dont even know if A15 actually exists.

This is how i think of it.

A10 = Original
A11 = 1st revision
A12 = 2nd revision
and so on.

A0
A1
A2...

Its the standard for pretty much every company. Intel goes a little weird with theirs though.

 

[Cookie Monster]

Originally posted by: Acanthus

Originally posted by: Cookie Monster
I dont think A15 means the 15th revision. We dont even know if A15 actually exists.

This is how i think of it.

A10 = Original
A11 = 1st revision
A12 = 2nd revision
and so on.

A0
A1
A2...

Its the standard for pretty much every company. Intel goes a little weird with theirs though.

It could be a standard, but doesnt mean you have to follow by them. Although i could be wrong, i mean 15 respins?!?

 

[Acanthus]

Originally posted by: Cookie Monster

Originally posted by: Acanthus

Originally posted by: Cookie Monster
I dont think A15 means the 15th revision. We dont even know if A15 actually exists.

This is how i think of it.

A10 = Original
A11 = 1st revision
A12 = 2nd revision
and so on.

A0
A1
A2...

Its the standard for pretty much every company. Intel goes a little weird with theirs though.

It could be a standard, but doesnt mean you have to follow by them. Although i could be wrong, i mean 15 respins?!?

Thats why i was so alarmed when i saw it in the other thread.

However if they went 90nm... 80... 65... that would certainly add up.

 

[hardwareking]

uh guys i have a question
doesn't power leakage increase when reducing the size of the manufacturing node?Wasn't that the problem with 90nm prescotts?
So this could very well hinder AMD than help it.Please confirm

 

[ViRGE]

Originally posted by: Acanthus

Originally posted by: munky
Well, that would be one hell of a surprise if true. However, I don't think either AMD's or TSMC's 65nm process is mature enough to result in big improvements over 90nm or 80nm silicon for now.

TSMC has been making 65nm chips for about 4 months now.

They *CAN* make them, but who knows how reliable the process is or what the yields are.

There is no way in hell AMD is producing the GPUs in house, they just dont have the capacity.

Frankly it sounds like it has all the makings of the NV30 debacle. TSMC has never been a bleeding-edge powerhouse, I would be absolutely shocked if they could get good yields of such a big chip this early.

 

[Killrose]

I dont see why this whole 65nm rumor seems so unlikely to some of you. It makes perfect sence as to why R600 has been delayed and the release of low to mid-tier R610 and R630 will appear at the same time. For every 65nm R600 GPU there will be dozens or maybe hundreds of R600 GPUs that wont make validation and become low to mid tier product.

I'm sure they would have loved to release it at 80nm too, but now that the industry is closing in on 65nm GPU capability, it makes perfect sence to wait. I bet they could show or preveiw R600 at 80nm which we would never be able to buy, but then the whole web community would scream paper launch! That kind of bad press hasnt done any company any good.

 

[Gstanfor]

Well, if AMD's excuse for lateness is leakage on the 80nm process node, its going to be fascinating to see how the 8900 from nvidia (rumored to use the 80nm process node) turns out... Someones either telling porky pies or has clueless designers...

 

[coldpower27]

Originally posted by: hardwareking
uh guys i have a question
doesn't power leakage increase when reducing the size of the manufacturing node?Wasn't that the problem with 90nm prescotts?
So this could very well hinder AMD than help it.Please confirm

The 90nm process was fine, look at the Pentium M Dothan for confirmation, it's the high clockspeed was what kills it.
With GPU's when they shrink and keep the same performance targets you actually do get cooler and less power hungry.

 

[apoppin]

Originally posted by: Janooo
This is from xbitlabs. It's 10 days old: R600 - 80nm; 610-630 - 65nm

...
Advanced Micro Devices, the company that supplies 25% of the world?s x86 microprocessors that has been making inroads into graphics technologies business since it acquired ATI Technologies in late October, 2006, demonstrated the delayed code-named R600 graphics boards at a conference and said it would introduce the board along with its derivatives made using 65nm process technology.

?We pushed out the launch of the R600 and people thought ?it must be a silicon or software problem, it?s got to be a bug?. In fact, our mainstream chips are in 65nm and are coming out extremely fast. Because of that configuration, we have an interesting opportunity to come to market with a broader range of products,? said David Orton, the former chief exec of ATI who now leads AMD?s graphics product group.

Earlier AMD already notified its partners and media that it planned to delay the release of the code-named R600 graphics chip, but did not provide any approximate timeframe for the commercial launch of graphics boards based on the R600. However, a document presumably sent to AMD?s add-in-card partners, revealed that certain versions of R600 graphics cards would reach the partners only in late March, days before the quarter ends.

While it was generally believed that the reason for the delay was ATI?s intention to develop an R600-based board to compete against Nvidia?s GeForce 8800 320MB graphics card. However, right now it is claimed that the company plans to align the launch of the R600 made using 80nm process technology with the launch of DirectX 10-supporting products made using 65nm fabrication process, which code-names are RV610 and RV630.

?Instead of having them separate, we thought, lets line that up, so we delayed for several weeks,? David Orton is reported to have said according to ExtremeTech web-site.

In a bid to prove the existence of R600, AMD demonstrated at the press event in San Francisco, the so-called ?Teraflop in a Box? system running a standard version of Microsoft Windows XP Professional that harnessed the power of AMD Opteron dual-core processor technology and two next-generation AMD R600 ?stream processors? capable of performing more than 1 trillion floating-point calculations per second using a general ?multiply-add? (MADD) calculation, according to AMD.

still makes no sense

IF we are to believe them ... and Orton still did NOT deny there was a HW bug in r600 ... wth would they WAIT on the flagship just to bring out "midrange" or mainstream GPUs in 65nm?

instead of launching as scheduled, "let's line that up, so we delayed for several weeks?
:Q

i call BS

Orton is full of ..it

 

[Keysplayr]

Look, it's obvious we are being fed a lot of BS. But you know what? It all comes around. Karma works...

 

[TanisHalfElven]

Originally posted by: Cookie Monster
I dont think A15 means the 15th revision. We dont even know if A15 actually exists.

This is how i think of it.

A10 = Original
A11 = 1st revision
A12 = 2nd revision
and so on.

i am no expert but from what i read A15 mean 6th revision. 5 of one thing and 1 of another. (wish i could remember which number refered to what)

 

[Cooler]

My Guess is that some of the 80nm R600 will be sold as OEM DragonHead which explains the size of the huge cooling fan and more power it requires then DragonHead2. DragonHead2 65nm R600 will run much cooler an use less power hence it being 9.5 inch card.

 

[hardwareking]

that could also explain the OEM version consuming 270W and the retail version consuming 240W
But the shrink from 80nm to 65nm only brings a 30W drop in peak draw?

 

[apoppin]

Originally posted by: keysplayr2003
Look, it's obvious we are being fed a lot of BS. But you know what? It all comes around. Karma works...

that's why Standard Oil, Haliburton, RAMBUS and MS are doing so 'poorly'

:roll:

"karma" doesn't always manifest itself in a single generation

 

[Cooler]

Originally posted by: hardwareking
that could also explain the OEM version consuming 270W and the retail version consuming 240W
But the shrink from 80nm to 65nm only brings a 30W drop in peak draw?

They could have increased the clock speed as well for DragonHead2

 

[RyanVM]

I don't find this too hard to believe and here's why. It makes sense to me that a company planning to release a chip roughly around the same time they're preparing to do a die shrink would spend some time doing the designs (80nm and 65nm) concurrently in the event it becomes necessary for whatever reason. I highly doubt the decision was "Hmm, 80nm doesn't work. We need to design a 65nm version quick!" I'm guessing the conversation was more along the lines of "OK, 80nm isn't working well. Let's delay a month and switch to plan B (65nm)."

In other words, I think they planned R600 to be an 80nm chip. I also think they worked on a 65nm design at the same time as a contingency plan. When they saw they weren't going to have an acceptable 80nm chip, they delayed the release to ramp up 65nm production. Finally, many months later they announced the delay under the silly premise of having a coherent product line (as if that's ever stopped them in the past), since there's no way you can switch process technology 6 weeks before release.

Makes sense to me anyway

 

[apoppin]

Originally posted by: RyanVM
I don't find this too hard to believe and here's why. It makes sense to me that a company planning to release a chip roughly around the same time they're preparing to do a die shrink would spend some time doing the designs (80nm and 65nm) concurrently in the event it becomes necessary for whatever reason. I highly doubt the decision was "Hmm, 80nm doesn't work. We need to design a 65nm version quick!" I'm guessing the conversation was more along the lines of "OK, 80nm isn't working well. Let's delay a month and switch to plan B (65nm)."

In other words, I think they planned R600 to be an 80nm chip. I also think they worked on a 65nm design at the same time as a contingency plan. When they saw they weren't going to have an acceptable 80nm chip, they delayed the release to ramp up 65nm production. Finally, many months later they announced the delay under the silly premise of having a coherent product line (as if that's ever stopped them in the past), since there's no way you can switch process technology 6 weeks before release.

Makes sense to me anyway

then AMD's total confusion and screwed-up marketing "plan" worked for you



i think r600 had a serious HW bug just like x1800 series... while they were *looking* for it ... and 14 respins later - found it - the OTHER product they were working on at 65nm [anyway] are ready and they might as well release it all together as THOUGH they had "planned" it

makes much more sense for a proud company like AMD ... ATi had no pride and was more likely to tell the truth about x1800 ... i fear we will never know about what REALLY *happened* to r600

at least it looks to be ready ... a year late

good job
:Q