Sign in with
Sign up | Sign in
Your question

ATI could be producing GPU's in Fab 36 by Q4.

Last response: in CPUs
Share
July 28, 2006 4:39:04 PM

http://www.theinquirer.net/default.aspx?article=33320
Now dont take this as anything other than a rumour but it would make sense if the demand for A64's is not as high as they expect. Another article there mentions that AMD gained some more market share at Intel's expense though so the demand is still there.

Anywayz off to work(2nd last day at Hell :D ) so look forward to rampant speculation and good conversation with good friends :wink:

More about : ati producing gpu fab

July 28, 2006 7:28:59 PM

Quote:
http://www.theinquirer.net/default.aspx?article=33320
Now dont take this as anything other than a rumour but it would make sense if the demand for A64's is not as high as they expect. Another article there mentions that AMD gained some more market share at Intel's expense though so the demand is still there.

Anywayz off to work(2nd last day at Hell :D ) so look forward to rampant speculation and good conversation with good friends :wink:



They will probably start doing Fire GL or high end CrossFire but they would be fools to take high margin CPU space for low margin GPUs.

I expect that ASAP they will be makign chipsets in house, but TSMC has no worries. Retooling a FAB for GPUs would be cost-prohibitive right now and wouldn't birng any return on investment.

It maybe that this is a smoke screen to take attention away from server chipsets. AMD needs that more than a mobile chipset, though a mobile chipset would be good since Intel destroyed the desktop market pricing.

I for one almost bought a Turion X2 laptop a few weeks ago.
July 28, 2006 9:33:24 PM

Quote:
They will probably start doing Fire GL or high end CrossFire but they would be fools to take high margin CPU space for low margin GPUs.


What are you talking about? Theres nothing different about Fire GL and "high end crossfire chips" from the normal x1800 and x1900 chips.
Related resources
July 28, 2006 9:48:38 PM

Quote:
http://www.theinquirer.net/default.aspx?article=33320
Now dont take this as anything other than a rumour but it would make sense if the demand for A64's is not as high as they expect. Another article there mentions that AMD gained some more market share at Intel's expense though so the demand is still there.

Anywayz off to work(2nd last day at Hell :D ) so look forward to rampant speculation and good conversation with good friends :wink:


The node transition isn't cut and paste, ATI if they do intend to use AMD's 90nm or 65nm node will have to make adjustments in according to AMD's wafer yields, composition of AMD wafers (doping, SOI, terminal materials so on and so forth) and lithography programs. As well the deal isn't legal yet so this is even more unlikely, unless they have ATI technologists waiting at the door.
July 28, 2006 10:42:32 PM

I can't see ATI using AMD Fabs anytime soon. There's the technical details that Spud mentioned, but it also doesn't make sense for them. ATI is just transitioning nearly their entire product line to 80nm with the RV505, RV535, RV560, and RV570. Using AMD's 90nm Fabs would be a step backwards in that sense. I'm betting most of AMD's SOI advantage would be counteracted by 80nm's smaller size and 80nm is cheaper to manufacture too. That only leaves AMD's 65nm process, but I doubt AMD will want to sacrifice capacity for ATI chips, especially when the process is so new and still ramping. With AMD going quad core, it's not like they have a lot of spare capacity anyways. The Inquirer mentions that Fab 36 is only at 20% capacity, but that's because it never ramped up. They did a small ramp to test all the equipment on 90nm and now their ramping that down as they transition to 65nm. 65nm won't be at 50% until Q1 2007 so the current 20% figure isn't unexpected. It's not like AMD is sitting with 80% capacity on it's hands and is doing nothing with it. It's just that additional 80% potential capacity doesn't exist yet.
July 28, 2006 11:05:08 PM

Quote:
They will probably start doing Fire GL or high end CrossFire but they would be fools to take high margin CPU space for low margin GPUs.


What are you talking about? Theres nothing different about Fire GL and "high end crossfire chips" from the normal x1800 and x1900 chips.

There is a 1300, 1600, 1800, 1900X, a 1900XT, a 1900XTX, and soon a 1950XTX. The higher end cards would be FireGL where the margin sare much higher. It would only help them get used to it.

If stupidity is your job, what's your hobby, monkey boy?
July 28, 2006 11:06:28 PM

Quote:
They will probably start doing Fire GL or high end CrossFire but they would be fools to take high margin CPU space for low margin GPUs.


What are you talking about? Theres nothing different about Fire GL and "high end crossfire chips" from the normal x1800 and x1900 chips.

There is a 1300, 1600, 1800, 1900X, a 1900XT, a 1900XTX, and soon a 1950XTX. The higher end cards would be FireGL where the margin sare much higher. It would only help them get used to it.

If stupidity is your job, what's your hobby, monkey boy?

pO$ER
July 28, 2006 11:45:07 PM

Quote:
WE WERE right about AMD's Dresden and New York AMD fabs and the claim they can and will produce the GPUs. You can read the original part here.


The Inq, so happy that they got something right. Even though they provide no evidence or sources for anything they have said in that article or the one they link to (as if it's some kind of source in itself - some speculation by themselves from a few months before).

I'll believe it when I see it, I think is what I'm trying to circumlocute at this present time in order to get this text on television.
July 29, 2006 12:02:09 AM

Quote:
so look forward to rampant speculation and good conversation with good friends

:) 

Quote:

I for one almost bought a Turion X2 laptop a few weeks ago.


You seem to almost weekly buy or think about buying one thing or another BM. You already claim to have a X2 4400+ so I'm not sure why anything AMD has out now would be of any interest to you.
The rest of what you said doesnt make sense because they will obviously ramp first with less complicated design's if anything.

Quote:

The node transition isn't cut and paste, ATI if they do intend to use AMD's 90nm or 65nm node will have to make adjustments in according to AMD's wafer yields, composition of AMD wafers (doping, SOI, terminal materials so on and so forth) and lithography programs. As well the deal isn't legal yet so this is even more unlikely, unless they have ATI technologists waiting at the door.


One thing to remember that no matter what, GPU's will be clocked much lower so AMD's excellent 90nm process may be enough. 65nm would depend on yield's and about a 100 other factor's which I feel Jack, LTData, and of course yourself, would be more qualified to speculate on. Either way with everything going in house(and I dont see it going sour as they seem to be buddying up like the deal is done and just needs a few signatures) cost's will be reduced and ATI's R&D budget will be internalized by AMD and will be less than when ATI outsourced everything to TMSC.

Quote:
I can't see ATI using AMD Fabs anytime soon. There's the technical details that Spud mentioned, but it also doesn't make sense for them. ATI is just transitioning nearly their entire product line to 80nm with the RV505, RV535, RV560, and RV570. Using AMD's 90nm Fabs would be a step backwards in that sense. I'm betting most of AMD's SOI advantage would be counteracted by 80nm's smaller size and 80nm is cheaper to manufacture too. That only leaves AMD's 65nm process, but I doubt AMD will want to sacrifice capacity for ATI chips, especially when the process is so new and still ramping. With AMD going quad core, it's not like they have a lot of spare capacity anyways. The Inquirer mentions that Fab 36 is only at 20% capacity, but that's because it never ramped up. They did a small ramp to test all the equipment on 90nm and now their ramping that down as they transition to 65nm. 65nm won't be at 50% until Q1 2007 so the current 20% figure isn't unexpected. It's not like AMD is sitting with 80% capacity on it's hands and is doing nothing with it. It's just that additional 80% potential capacity doesn't exist yet.

I completely agree here and as I told BM this will be nothing more than a trial run with low-end, or possibly integrated gfx in chipset's as part of the 65nm ramp process.

Quote:
The Inq, so happy that they got something right. Even though they provide no evidence or sources for anything they have said in that article or the one they link to (as if it's some kind of source in itself - some speculation by themselves from a few months before).

I'll believe it when I see it, I think is what I'm trying to circumlocute at this present time in order to get this text on television.

In my experience 99% of what the inquirer post's has at least some
kind of credibility behind it even if they are the bottom dweller's of the IT news world and that's why I am interested in the in the first place.
July 29, 2006 12:25:40 AM

Quote:
They will probably start doing Fire GL or high end CrossFire but they would be fools to take high margin CPU space for low margin GPUs.


What are you talking about? Theres nothing different about Fire GL and "high end crossfire chips" from the normal x1800 and x1900 chips.

There is a 1300, 1600, 1800, 1900X, a 1900XT, a 1900XTX, and soon a 1950XTX. The higher end cards would be FireGL where the margin sare much higher. It would only help them get used to it.

If stupidity is your job, what's your hobby, monkey boy?

Are you retarded or something? The x1800 is FireGL. Crossfire is mostly done with external chips you moron. The x1950 just uses GDDR4 which is already in the memory controller (the ability to use it, not the GDDR4 itself.)

The x1800 is 288mm2 and the x1900 is 330mm2 (from memory) AMD don't make such large chips.
July 29, 2006 12:46:47 AM

Quote:
You seem to almost weekly buy or think about buying one thing or another BM. You already claim to have a X2 4400+ so I'm not sure why anything AMD has out now would be of any interest to you.
The rest of what you said doesnt make sense because they will obviously ramp first with less complicated design's if anything.



I'm sorry that Optiplex is driving me crazy and I could easily buy a $1200 laptop to take to work with me. I don't remember needing to check with you about my statements so that you won't "get the wrong idea."


R600 IS 65nm designed along with 80nm but if AMD were going to take a chance with Fab36 yields, they may as well go for higher end chips since as "ActionGirl" said there isn't a difference in the chips other than an IGP is worth maybe $50 bucks whereas a FireGL is worth $500. Then there is the issue of PCB versus chip that means an extra expenditure to either fab the card or outsource for the card.

OF course they decied to merge so I don't have an inside track but it's all about ROI. I guess they could substitute single core but the merger won't go through for months, so they can't start fabbing ATi tech. I workd for a comapny that was bought and to a large degree you can't work together closer than the two separate companies.

AMD has no license to Radeon that I know of so they can't do anything. They can work with ATi on designs through the Torrenza initiative but count out the next 6 months.

By the time Fab36 ramps they will need a lot of space for K8L/Brisbane and they won't be 50% before next year. I can see them reserving space in Fab38 or Fab 7 Chartered to do chipsets, but I doubt the new process will be the lowest margin - even though volume costs less.

I have been wrong abotu company operations - ones I don't work at - but AMD has spent a lot of money that will start to come due next year, so they need the hisghest margins and ASP they can get that will come from K8L/Brisbane, though AS I SAID, I don't doubt that they will start to retool a given percentage of FAB38 for GPUs (masks and APM programming) for Fire GL. That way if they don't have great yields - why wouldn' they - but it's better to go fr the high end cards.

Everyone knows that 4x4 is new and needs a specific mobo (2xAM2) with either 3 or 4 PCIe 16 slots. They will limit the bottom to around $350-400 for the chips to keep the ASP for what would be initially low adoption rates.

They can't afford an Intel-like firesale.
July 29, 2006 12:57:50 AM

Actiongirl :lol:  Are you in primary school or something BaronBS?

Quote:
R600 IS 65nm designed along with 80nm


O RLY?
July 29, 2006 1:10:12 AM

Quote:
They will probably start doing Fire GL or high end CrossFire but they would be fools to take high margin CPU space for low margin GPUs.


What are you talking about? Theres nothing different about Fire GL and "high end crossfire chips" from the normal x1800 and x1900 chips.

There is a 1300, 1600, 1800, 1900X, a 1900XT, a 1900XTX, and soon a 1950XTX. The higher end cards would be FireGL where the margin sare much higher. It would only help them get used to it.

If stupidity is your job, what's your hobby, monkey boy?

Are you retarded or something? The x1800 is FireGL. Crossfire is mostly done with external chips you moron. The x1950 just uses GDDR4 which is already in the memory controller (the ability to use it, not the GDDR4 itself.)

The x1800 is 288mm2 and the x1900 is 330mm2 (from memory) AMD don't make such large chips.



Impress me with die sizes?.... I think not. I am visiting ATi as we speak and there is no mention of x1800 being the only x1000 used for FireGL.

Don't you remember saying that bandwidthwas a problem with GPU integration? GDDR4 runs at 3.2GHz. Using 64MB as local with DDR3 - not like this will happen bfore AM3 - will allow for a smooth experience with evrything but maybe FEAR and Oblivion (you only need 30 fps). At least in the first iterations.



They may start with cheap chips as ES but releasing a low margin volume chip when nVidia will make sales would be foolish waste of Fab36 space.

Monkey boy.
Are you going to say that 4x4 demos are faked?
July 29, 2006 1:15:41 AM

Quote:
Impress me with die sizes?.... I think not.


No just stating so facts, something your posts lack.

Quote:
I am visiting ATi as we speak and there is no mention of x1800 being the only x1000 used for FireGL.


:lol:  I so doubt it. Anyways the highest end ones are x1800. They dont use x1900.

Quote:
Don't you remember saying that bandwidthwas a problem with GPU integration? GDDR4 runs at 3.2GHz. Using 64MB as local with DDR3 - not like this will happen bfore AM3 - will allow for a smooth experience with evrything but maybe FEAR and Oblivion (you only need 30 fps). At least in the first iterations.


This sounds stupid, draw a diagram. Also 64mb isnt much memory.

Quote:
Monkey boy.
Are you going to say that 4x4 demos are faked?


Did i ever say that moron? Man I'd hate to be so consistently wrong.
July 29, 2006 1:26:29 AM

Quote:
I'm sorry that Optiplex is driving me crazy and I could easily buy a $1200 laptop to take to work with me. I don't remember needing to check with you about my statements so that you won't "get the wrong idea."


4400+ in an optiplex? Sorry I was under the impression that you used a X2 4400+ which you'd been using for a year. You dont need to check in with me but I was merely pointing out the fact that i've seen you mention that you will pick up the FX60, 5000+, numerous laptops, ect but still complain about the crappy Dell's you work on.

Quote:
R600 IS 65nm designed along with 80nm but if AMD were going to take a chance with Fab36 yields, they may as well go for higher end chips since as "ActionGirl" said there isn't a difference in the chips other than an IGP is worth maybe $50 bucks whereas a FireGL is worth $500. Then there is the issue of PCB versus chip that means an extra expenditure to either fab the card or outsource for the card.


I dont personally believe they will produce R600 there right off the bat which is what the next Fire GL will be based on. Despite the fact that it is higher margin is is also very low volume and would not make sense because as you said. It's all about the ROI. :) 

Quote:
AMD has no license to Radeon that I know of so they can't do anything. They can work with ATi on designs through the Torrenza initiative but count out the next 6 months.


Well not right now but this thread was designed to be based on the not too distant future when we can assume(as long as everything goes through) that AMD will have all of the rights to Radeon and other ATI products.

Quote:
By the time Fab36 ramps they will need a lot of space for K8L/Brisbane and they won't be 50% before next year. I can see them reserving space in Fab38 or Fab 7 Chartered to do chipsets, but I doubt the new process will be the lowest margin - even though volume costs less.


AMD has supplied 20% of the market with a single 200mm 90nm fab for the past 2 years. I think with that + chartered + fab 36 ramping up they will definately have some extra fab capacity to play because with C2D I dont think they will be gaining any more share in the next year. That's just my opinion though.

Quote:
By the time Fab36 ramps they will need a lot of space for K8L/Brisbane and they won't be 50% before next year. I can see them reserving space in Fab38 or Fab 7 Chartered to do chipsets, but I doubt the new process will be the lowest margin - even though volume costs less.

It's really hard to say what they will do but your guess is as good as mine.
Chartered will be using AMD's APM on 65nm so I'm sure they will continue to be good friends next year.

Quote:
Everyone knows that 4x4 is new and needs a specific mobo (2xAM2) with either 3 or 4 PCIe 16 slots. They will limit the bottom to around $350-400 for the chips to keep the ASP for what would be initially low adoption rates.

Not sure where you're going with this but feel free to elaborate further.
July 29, 2006 1:42:43 AM

Quote:
...I am visiting ATi as we speak...
Dont choke on the sweet canadian air while your here :p 











I know I know I know your probably just visiting www.ati.com and I'm being an ass.
July 29, 2006 2:19:35 AM

Quote:
I'm sorry that Optiplex is driving me crazy and I could easily buy a $1200 laptop to take to work with me. I don't remember needing to check with you about my statements so that you won't "get the wrong idea."


4400+ in an optiplex? Sorry I was under the impression that you used a X2 4400+ which you'd been using for a year. You dont need to check in with me but I was merely pointing out the fact that i've seen you mention that you will pick up the FX60, 5000+, numerous laptops, ect but still complain about the crappy Dell's you work on.

Quote:
R600 IS 65nm designed along with 80nm but if AMD were going to take a chance with Fab36 yields, they may as well go for higher end chips since as "ActionGirl" said there isn't a difference in the chips other than an IGP is worth maybe $50 bucks whereas a FireGL is worth $500. Then there is the issue of PCB versus chip that means an extra expenditure to either fab the card or outsource for the card.


I dont personally believe they will produce R600 there right off the bat which is what the next Fire GL will be based on. Despite the fact that it is higher margin is is also very low volume and would not make sense because as you said. It's all about the ROI. :) 

Quote:
AMD has no license to Radeon that I know of so they can't do anything. They can work with ATi on designs through the Torrenza initiative but count out the next 6 months.


Well not right now but this thread was designed to be based on the not too distant future when we can assume(as long as everything goes through) that AMD will have all of the rights to Radeon and other ATI products.

Quote:
By the time Fab36 ramps they will need a lot of space for K8L/Brisbane and they won't be 50% before next year. I can see them reserving space in Fab38 or Fab 7 Chartered to do chipsets, but I doubt the new process will be the lowest margin - even though volume costs less.


AMD has supplied 20% of the market with a single 200mm 90nm fab for the past 2 years. I think with that + chartered + fab 36 ramping up they will definately have some extra fab capacity to play because with C2D I dont think they will be gaining any more share in the next year. That's just my opinion though.

Quote:
By the time Fab36 ramps they will need a lot of space for K8L/Brisbane and they won't be 50% before next year. I can see them reserving space in Fab38 or Fab 7 Chartered to do chipsets, but I doubt the new process will be the lowest margin - even though volume costs less.

It's really hard to say what they will do but your guess is as good as mine.
Chartered will be using AMD's APM on 65nm so I'm sure they will continue to be good friends next year.

Quote:
Everyone knows that 4x4 is new and needs a specific mobo (2xAM2) with either 3 or 4 PCIe 16 slots. They will limit the bottom to around $350-400 for the chips to keep the ASP for what would be initially low adoption rates.

Not sure where you're going with this but feel free to elaborate further.

SInce you're feeling arguma=entative, I have a 4400+ 939 that could uograde to FX60. I have need for a secondary PC so I could sell the 2nd and put the mobo and chip/RAM in a little case and use it. Then I could get the mobo/5000+/RAM.

The Turion was because I go to different assignments and they ahve CRAPTIPLEX. Having a Turion X2 would be a tax writeoff AND a better system.


I didn't even read the rest. I am still at work.
July 29, 2006 2:48:09 AM

Awww I'm at work too bored out of my mind and I want to play....... :lol: 
July 29, 2006 2:14:17 PM

Quote:
Awww I'm at work too bored out of my mind and I want to play....... :lol: 


You spoke against the will of BM, your on his hit list now.
July 29, 2006 9:30:51 PM

According to Henri Richard:

Quote:
Q: Obviously the acquisition does have implications for the fabrication of ATI's products. ATI's products are not based, like AMD's, on SOI technology. So in practice, how will things work from a manufacturing point of view?

A: In practice, we're not going to change anything. ATI is going to continue to rely on a fabless model. We want to nurture the great relationship that exists, here in Taiwan, with TSMC. And you're right, because they're on bulk wafers, and we're on SOI, even if there was a good reason for us to go and think about doing things differently, it's not practical in any short period of time. As I mentioned, one day it will make sense to put a GPU and CPU on the same die. In that case, of course, they would have to be manufactured on the same process.

But right now, AMD's manufacturing strategy is focused on increasing rapidly, as per plan, our CPU capacity, so we don't want to do anything to disrupt that. And then ATI has a very successful model, working very closely with TSMC, and we don't want to disrupt that either. So we'll be both a fab and a fabless company, and we are looking forward to getting the best of both those worlds.


From here.
July 29, 2006 9:39:33 PM

Ya, I'm just putting a few more nails in the coffin of this rumour.
July 30, 2006 2:18:58 AM

Nvidia went from IBM back to TSMC and it didn't really make a difference.
July 30, 2006 3:09:56 AM

It was a fun discussion though even if it was BM bashing for the most part :lol: 

Glad to be part of it though because tommorow is my last day at Hell(where i spend my boring day's posting on here) and then I'm moving back to my hometown(and Spud's) where I will more than likely be spending more time on beer and having fun rather than the internet for the remainder of the summer. My gf's stang got broken into for the 5th time today(they actually almost stole the whole car this time and !@#$'ed over the ignition) in the mid of us moving our things so I cant wait to get out of this big city full of crackhead's, e-tards, kid's on meth, and all the fun they bring with them, and back to where i belong so i can actually make some money.
July 30, 2006 3:15:31 AM

Quote:
Actiongirl :lol:  Are you in primary school or something BaronBS?

R600 IS 65nm designed along with 80nm


O RLY?No, Baron_Matrix doesn't know what he's talking about.
July 30, 2006 3:16:29 AM

Goes without saying.
!