Sinatech: ATI/AMD Aquisition Agreement Reached: ATI facing a big shake-up

Status
Not open for further replies.
Instead of fabbing out chips from TSMC, Chartered, etc., ATI will have an access to much more advanced manufacturing technology. AMD wouldn't have capacity? I doubt the high-end graphics chips sales are anywhere near the volume of CPU sales. AMD can just share a fraction of fab space with ATI's high-end chips and could still get by with it. Especially once Fab36 runs in full force.

AMD will have capacity once their new 300mm FAB comes online. Right now afaik AMD is supply constrained and with Dell coming into the picture, it wont get better.

This means, when NV'd be designing chips w/ 90nm or 80nm, ATI would have 65nm for their designs. When TSMC finally catched up with AMD/Intel's current manufacturing (say, 65nm), ATI would be one step ahead with AMD (say, 45nm). Mid-to-low level GPUs can still be outsourced from any of those fabbing houses.

process technology seems to play a lesser role in the GPU world than in the CPU world. I think this is primarily due to the design of the transistors in the GPU. afaik they use bulk design transistors and don't do a lot of customization. Plus the product cycles are a % shorter than the CPU world. AMD and Intel debut a new arch and it lasts 4-5 years with specific goals for each process stepping. ATI and Nvidia debut a new arch with performance goals for the next 6-12 months. Either they hit it or they dont and design the next one to top the competititon. If AMD pours millions into customizing a GPU and tweak the hell out of it for a specific goal. Chances are by the time they did so, Nvidia would release a new GPU with twice the resources and smoke it.


This would put a tremendous pressure on NV and would eventually force NV to build its own fabs. (which would take lots of money and time) In the meantime, this would give ATI a great chance to gain its marketshare.

I dont think so, ATI and Nvidia compete pretty decent in the integrated markets against Intel, who has much more fab capacity than AMD. Nvidia will continue to utilize 3rd party fab partners, and may even get better deals due to a lack of demand from ATI.
 
Maintank said:
process technology seems to play a lesser role in the GPU world than in the CPU world. I think this is primarily due to the design of the transistors in the GPU. afaik they use bulk design transistors and don't do a lot of customization. Plus the product cycles are a % shorter than the CPU world. AMD and Intel debut a new arch and it lasts 4-5 years with specific goals for each process stepping. ATI and Nvidia debut a new arch with performance goals for the next 6-12 months. Either they hit it or they dont and design the next one to top the competititon.

Because they don't have their own fabs? They're relying on 3rd party fabbing houses for their chips. A quick hypothetical example: Current X1900XT manufactured w/ 65nm process - Half the die size, half the power consumption, and 900MHz core speed. Vastly simplified example, but you get the idea.

If AMD pours millions into customizing a GPU and tweak the hell out of it for a specific goal. Chances are by the time they did so, Nvidia would release a new GPU with twice the resources and smoke it.

AMD doesn't pour millions into customizing a GPU. ATI will. Had TSMC offered 65nm process right now, ATI will customize their GPUs for that process, right? Sure there will be trim down here and there but basically it's 2 companies under 1 corporation. Also, to people who say ATI will waste resources on chipset division instead of GPU development - that's what ATI is CURRENTLY doing. They're desparately trying to get in the chipset market right now. With direct access to CPU "marchitecture", they will save tons of resources to develop competitive chipsets.

Am I missing something? It seems like people are more or less paranoid and thinking backwards. Of course this deal won't happen for many reasons, but if it did, it'd benefit both companies big time.
 
Last edited by a moderator:
Because they don't have their own fabs? They're relying on 3rd party fabbing houses for their chips. A quick hypothetical example: Current X1900XT manufactured w/ 65nm process - Half the die size, half the power consumption, and 900MHz core speed. Vastly simplified example, but you get the idea.

My opinion is the product cycle times plays a bigger part in it than a 3rd party FAB. AMD has utilized 3rd party fabs in the past and werent forced into not using their own customizations.

I dont think they would see half the power consumption based on a die shrink without some serious customizations. But that would or could take several months and a few test runs. By the time they do such a thing Nvidia's next core is out and the whole thing is mute.

AMD doesn't pour millions into customizing a GPU. ATI will.

I thought this was a thread about AMD acquiring ATI? I would think if AMD owns ATI, when ATI spends money to customize the transistors and layout so is AMD?

Had TSMC offered 65nm process right now, ATI will customize their GPUs for that process, right? Sure there will be trim down here and there but basically it's 2 companies under 1 corporation. Also, to people who say ATI will waste resources on chipset division instead of GPU development - that's what ATI is CURRENTLY doing. They're desparately trying to get in the chipset market right now. With direct access to CPU "marchitecture", they will save tons of resources to develop competitive chipsets.

Dont know, I didnt make that argument. I have questioned why AMD would buy ATI for an integrated chipset when I am sure they could put one together for less than 4 billion and still have it out within the next 12-18 months.
 
Then let me ask you. If you think this AMD-ATI merger would be basically AMD absorbing ATI and consiquently mid-high end GPU market would be dominated by NV in the future, why would this deal upset NV? (Well at least analysists think so)

Because of chipset market? I don't think so. NV would happily move on with Intel chipset, and would not lose out on AMD chipset market, either.
 
lopri said:
Then let me ask you. If you think this AMD-ATI merger would be basically AMD absorbing ATI and consiquently mid-high end GPU market would be dominated by NV in the future, why would this deal upset NV? (Well at least analysists think so)

Because of chipset market? I don't think so. NV would happily move on with Intel chipset, and would not lose out on AMD chipset market, either.

Its quite a bit more complex then just AMD abosbing ATi, and ATi then will have thier own fabs. First, are thier chip designs ready for AMD fabs? Going from TSMC to UMC ATi would have to layout thier entire chip again. Then you have to concider GPU's are quite a bit more complex then CPU's.

Then you have the dynamics of such a merger, will it really hurt nV or help ATi or vice versa? I really don't think there will be much change in the short term, because the business model of ATi really won't change much. What will hurt them, will be ATi's loss or rather barred by AMD to make Intel's chipsets, and this will likely happen. nV will have intel chips sets all to themselves but not in the short term, it will be more of a mid term gain.

Short term it might help ATi because of AMD's capital. But to get these chips into AMD's fabs will probably be quite a bit work, AMD needs production space which they are sorely lacking right now.

Cost will end up very similiar, the only cost that will be subsiidised will cost of labor to get the chips going in the manufacturing process of the chips, and comparative to other costs this will be very miniscule since AMD will still have to pay thier own employees, and the cost difference is TSMC's or outsourcing fab's profits which if AMD is making in German or where ever thier fabs are thier cost per employee will be quite a bit higher then Tiawan, and other "semi" industrial nations. Raw materials will cost the same. AMD going to 65 nm, with a less complex chip with less transistors per chip, compaired to ATi's top end GPU's, at the moment even if TSMC or any other ousourcing fab had 65 nm working it will take some time to get a chip ready to manufacture on those processes. I'm sure even at 80nm both compaines will have some issues with yeilds in the beginning, possible need 2 or 3 attempts before they get it right. This would be at least a 1 or 2 generation goal not a short term goal going to .65 at AMD's fabs.

As mentioned earlier CPU's life cylces are much longer then GPU's so changing fab processes would grant some benefits but greater benefits are actually design of the GPU and its components. AMD did just fine competing with Intel with larger chips and not as agressive on the fab processes. nV did well also but to a lesser extent.

Now what are the longer term goals of AMD with such and aquisition. Benefits as mentioned earilier would be a CPU/GPU hybrid, which I really don't will sit well against a discreet option and chipsets. Also is AMD's goal to make GPU's right now? Which markets are they looking at. One would think, AMD's goal is to secure the CPU side of things before they move on to other aspects. Chipset side, well yeah AMD can now get thier own chipset but if you look at Intel why did they go to ATi for thier latest chipset even though Intel has the chipset market to itself? Must be a reason. I'm guessing Intel understood if people are going to buy AMD well which chipsets are they getting? Not ones made by AMD since they don't make thier own. Over all Intel chipsets aren't the greatest, so they went with ATi.

The main long term goal from AMD's stand point is diversification hand helds and get a inhouse chipset design team. Not GPU's it might be something they will look into but this will take second seat to stop Intel in the CPU front.
 
Sweet jesus, 11 pages.

And they really believe it too (if the 11 pages of what it is going to mean isn't enough to convince you of that):

Recently X-bit labs has learnt from industrial sources that Advanced Micro Devices is going to acquire ATI Technologies,

Altho they then hedge a bit with "potential".
 
It took Intel nearly three years to recapture performance lead with its Pentium III “Northwood” processor in 2002
Proof reading motherf*##@#. Do you speak it?
 
popcorn.gif
 
Sinistar said:
someone's doing a good job of keeping this rumor alive.

Well, xbit did at least admit that possibility:

“Each time I’ve tried to trace back the source of the rumors and when I have been successful in doing so it's gone back to someone with a financial interest in spreading them,” said Mercury Research principal analyst Dean McCarron.

But the final conclusion still tastes as if Xbit believes it is for real:

While our sources generally tend to be correct, chances that no acquisition is going to take place, or the companies find different ways to collaborate, or just a part of ATI will go to AMD, while the rest will remain independent, still exist.
 
Sinistar said:
someone's doing a good job of keeping this rumor alive.
but why?!
Rumour that AMD will buy ATi - AMD shares fall
In the end the truth that this is not going to happen willbe obvious... and AMD shares will fall (imho) again...
 
Xbit can be wrong like anybody can be wrong. But the amount of effort they put into it, and the tone, strongly suggests to me that whoever fed them their secret info is someone they have some history/trust with.
 
AMD does not have enough cash to make a hostile takeover of ATI. The only possibility is a strategic alliance.
 
phenix said:
AMD does not have enough cash to make a hostile takeover of ATI. The only possibility is a strategic alliance.

I think it would lead to the same outcome in the end.

If Intel thinks ATI is delivering a "preferable treatment customer" to AMD, they might think twice on renewing previous 3rd party chipset contracts for intel branded boards, aswell as Crossfire support on Intel chipsets.


I believe any change in the current balance of power on both GPU and CPU businesses would be bad for end users.
 
Razor1 said:
As mentioned earlier CPU's life cylces are much longer then GPU's so changing fab processes would grant some benefits but greater benefits are actually design of the GPU and its components. AMD did just fine competing with Intel with larger chips and not as agressive on the fab processes. nV did well also but to a lesser extent.

I strongly doubt this is the case. You know, all the AMD fans are waiting for at this moment is the coming of K8L which is 65nm version of dual/quad-core K8. The reason why AMD doesn't have 65nm is not because they don't need it (while the outcome has been true to this until Conroe) but because AMD is not up there yet, be it financially or technically.
 
lopri said:
I strongly doubt this is the case. You know, all the AMD fans are waiting for at this moment is the coming of K8L which is 65nm version of dual/quad-core K8. The reason why AMD doesn't have 65nm is not because they don't need it (while the outcome has been true to this until Conroe) but because AMD is not up there yet, be it financially or technically.

"K8L" is NOT the quad-core 65nm version of the current K8.
"Brisbane" is a quad-core 65nm alright... but based on the current architecture.


"K8L" is a true evolution of the current designs, and it's long overdue ("K9" sounded bad, so they changed it :D).
 
INKster said:
"K8L" is NOT the quad-core 65nm version of the current K8.
"Brisbane" is a quad-core 65nm alright... but based on the current architecture.


"K8L" is a true evolution of the current designs, and it's long overdue ("K9" sounded bad, so they changed it :D).

I stand corrected. Thanks.
 
lopri said:
Then let me ask you. If you think this AMD-ATI merger would be basically AMD absorbing ATI and consiquently mid-high end GPU market would be dominated by NV in the future, why would this deal upset NV? (Well at least analysists think so)

Because of chipset market? I don't think so. NV would happily move on with Intel chipset, and would not lose out on AMD chipset market, either.

I would imagine AMD competing against Nvidia in the chipset market would upset Nvidia.
I also believe AMD becoming a direct competitor in the GPU business would taint their relationship with Nvidia.

AMD will become a direct competitor where they were a partner before. Nvidia is on Intel yes, I am curious what the royalties are however. I remember a few years ago Intel didnt want compeititon on their platform and priced Nvidia right out of the market. However they are back and are either taking it on the chin to increase revenues and marketshare or Intel was gracious. Either way I do believe AMD buying ATI wont please Nvidia, but instead upset them, and rightfully so.

btw I have said if AMD does buy ATI, I think it will be more for the higher margin GPU business than the integrate solution. They see mid to high 30 margins on ATIs GPUs and it is an infusion of cash flow.
 
Status
Not open for further replies.
Back
Top