AMD ditches GloFo, goes with TSMC

Wow. That's a pretty expensive move. Hopefully it ends up being a winning move in the long run and results in better yields.
 
[QUOTE="10-15% increase in performance each year"[/url][/QUOTE]

Great! So in 5 years it will in the i7 performance range.
 
That means I have a piece of history on my hands!!! The only CPUs made by GloFo. :D
 
This is just the low end right? The Brazos and the ultraportable? This would not include the successors to the Llano and Bulldozer? Or am I confused.
 
I don't envy Rory Read. This is like being promoted to pilot but being given the controls in midair once the engines have flamed out. Sure, he's doing the right thing by dumping GloFo, but considering that the fab was AMD (before Ruiz The Terminator got his hands on the company) this does not bode well for the whole "former company."
 
Weeth said:
I don't envy Rory Read. This is like being promoted to pilot but being given the controls in midair once the engines have flamed out. Sure, he's doing the right thing by dumping GloFo, but considering that the fab was AMD (before Ruiz The Terminator got his hands on the company) this does not bode well for the whole "former company."
Click to expand...

An apt analogy. I agree that Rory Read has his work cut out for him, but I also think that he knows what he's doing. For years, AMD simply attempted to out-gun Intel and slowly but surely built itself into a company that could actually do it (circa 2003-2006), with Dirk Meyer and the Alpha CPU design team and the Athlon 64 X2s. However, this strategy always put profitability as a distant second priority -- Gerry Saunders probably always knew that this 'slowly but surely' strategy would doom AMD to being barely profitable enough to survive, but he had a goal, and that goal was to out-engineer Intel. He achieved it just before he left.

Now AMD needs to take it's resources and strengths and become a profitable company. They still have the engineers to build world-class x86 CPUs, GPUs and even the remnant of an ARM team. Really, what AMD needs is to merge with a company like NVidia, or to be bought out by IBM or Apple -- but they could really use a large infusion of cash. If that happened, they could reconstitute themselves and become serious competitors on the CPU side once again. Personally, I like the NVidia/AMD merger scenario, because it allows both companies to merge their major synergies: the best GPUs in the world for both consumer and supercomputing applications, a renewed x86 team, and with NVidia, a leading ARM processor team. Oh well, it may never happen because it would make too much sense for both companies!
 
anubis44 said:
An apt analogy. I agree that Rory Read has his work cut out for him, but I also think that he knows what he's doing. For years, AMD simply attempted to out-gun Intel and slowly but surely built itself into a company that could actually do it (circa 2003-2006), with Dirk Meyer and the Alpha CPU design team and the Athlon 64 X2s. However, this strategy always put profitability as a distant second priority -- Gerry Saunders probably always knew that this 'slowly but surely' strategy would doom AMD to being barely profitable enough to survive, but he had a goal, and that goal was to out-engineer Intel. He achieved it just before he left.

Now AMD needs to take it's resources and strengths and become a profitable company. They still have the engineers to build world-class x86 CPUs, GPUs and even the remnant of an ARM team. Really, what AMD needs is to merge with a company like NVidia, or to be bought out by IBM or Apple -- but they could really use a large infusion of cash. If that happened, they could reconstitute themselves and become serious competitors on the CPU side once again. Personally, I like the NVidia/AMD merger scenario, because it allows both companies to merge their major synergies: the best GPUs in the world for both consumer and supercomputing applications, a renewed x86 team, and with NVidia, a leading ARM processor team. Oh well, it may never happen because it would make too much sense for both companies!
Click to expand...

I do not like a nVidia/AMD merger. It's like a merger between Intel/AMD but for GPUs. They will have a complete monopoly on the GPU market, you wouldn't want that happening would you?
 
anubis44 said:
An apt analogy. I agree that Rory Read has his work cut out for him, but I also think that he knows what he's doing. For years, AMD simply attempted to out-gun Intel and slowly but surely built itself into a company that could actually do it (circa 2003-2006), with Dirk Meyer and the Alpha CPU design team and the Athlon 64 X2s. However, this strategy always put profitability as a distant second priority -- Gerry Saunders probably always knew that this 'slowly but surely' strategy would doom AMD to being barely profitable enough to survive, but he had a goal, and that goal was to out-engineer Intel. He achieved it just before he left.

Now AMD needs to take it's resources and strengths and become a profitable company. They still have the engineers to build world-class x86 CPUs, GPUs and even the remnant of an ARM team. Really, what AMD needs is to merge with a company like NVidia, or to be bought out by IBM or Apple -- but they could really use a large infusion of cash. If that happened, they could reconstitute themselves and become serious competitors on the CPU side once again. Personally, I like the NVidia/AMD merger scenario, because it allows both companies to merge their major synergies: the best GPUs in the world for both consumer and supercomputing applications, a renewed x86 team, and with NVidia, a leading ARM processor team. Oh well, it may never happen because it would make too much sense for both companies!
Click to expand...

I'm not clear on what you're saying as Dirk was after Horrible Hector as CEO. He was only CEO from July '08 to January '11 and by then it was too late to do much of anything. IMHO it was Ruiz who dynamited AMD to smithereens. Dirk just tried to pick up the pieces and Rory... well... all the well meaning leger de main in the world is not going to help a company that has so much fundamentally flawed in its operations. I really do wish them well... but that doesn't mean I'm going to buy their stock.
 
Tsumi said:
I do not like a nVidia/AMD merger. It's like a merger between Intel/AMD but for GPUs. They will have a complete monopoly on the GPU market, you wouldn't want that happening would you?
Click to expand...

Yeah, AMD/nVidia is a horrible idea, same with AMD/Apple, since I think Apple would likely make everything AMD produces proprietary, and force you to buy a shitty Mac to get any of it. I would much rather see them remain independent, but if they simply must merge, go with IBM.
 
IBM would be bad for us, they aren't a consumer brand, they wouldn't think twice about screwing us over and making only business parts, that's who they are.

plus I read an article about IBM years back where a top executive was predicting only them and intel would be producing CPUs in 5 years, that was about 10 years ago. I would hate for AMD to get caught up in a company that arrogant and lacking in market foresight.
 
anubis44 said:
Now AMD needs to take it's resources and strengths and become a profitable company. They still have the engineers to build world-class x86 CPUs, GPUs and even the remnant of an ARM team. Really, what AMD needs is to merge with a company like NVidia, or to be bought out by IBM or Apple -- but they could really use a large infusion of cash. If that happened, they could reconstitute themselves and become serious competitors on the CPU side once again. Personally, I like the NVidia/AMD merger scenario, because it allows both companies to merge their major synergies: the best GPUs in the world for both consumer and supercomputing applications, a renewed x86 team, and with NVidia, a leading ARM processor team. Oh well, it may never happen because it would make too much sense for both companies!
Click to expand...
I don't think the regulators would allow NVIDIA and ATI to become part of the same company and I don't think it's feasible for AMD to split from ATI at this point.
 
Cool Vibrations said:
Thanks for nothing, Hector Ruinz.
Click to expand...

Couldn't agree with you more. No matter what thread you try to follow to get to the root of AMD's ruination it always runs back to Ruinz. Splitting off GloFo ended up disembowling both companies until now there's virtually nothing left. The only reason he did it was so he could gorge on Abu Dhabi megabucks. Why he's still out of jail and enjoying the millions he stripped from AMD is one of the great mysteries of the decade.
 
I am reluctant to say Bulldozer is a design disaster. I wonder if the cache problems are manufacturing related as we don't know for sure what speeds the uncore runs at.
 
Poisoner said:
I am reluctant to say Bulldozer is a design disaster. I wonder if the cache problems are manufacturing related as we don't know for sure what speeds the uncore runs at.
Click to expand...

I think it's pretty safe to say that from a design standpoint bulldozer is impressive.

I think it's also safe to say the architecture, if given time to to mature(software and hardware) will show itself to be pretty darn impressive, practically.

Intel will be doing some of the same things AMD tried to do with bulldozer in the next couple of years.
 
This is very bad

"AMD are now looking to move production of 28nm chips from GlobalFoundaries to TMSC, as well as starting completely from scratch. As ExtremeTech have said, "the implications and financial repercussions could be enormous," there's just no other way to put it. Moving 28nm APUs from GloFo to TMSC means completely scrapping the existing designs and laying out new parts using gate-laste compared to gate-first manufacturing currently in place at GloFo."

http://m.tweaktown.com/news/21677/amd_cancels_28nm_apus_from_globalfoundaries/index.html
 
stealthy123 said:
god knows tsmc can't do worse than glofo.

I bet I regret saying that...
Click to expand...

its TSMC you damn well will regret saying that after the 3rd or 4th delay and 30% yields..



Tsumi said:
I do not like a nVidia/AMD merger. It's like a merger between Intel/AMD but for GPUs. They will have a complete monopoly on the GPU market, you wouldn't want that happening would you?
Click to expand...


nope, the day that happens the desktop market will be dead anyways and it only happens because AMD some how leapfrogged nvidia in the mobile market with their APU's. odds of AMD some how making nvidia's mobile market obsolete are pretty slim right now so i wouldn't really worry about it.
 
Last edited:
So is Intel the only one with reliable fabs anymore? What about IBM? I haven't kept up with fab tech and capacity, but is TSMC really the only other option than GloFo for these sort of chips?
 
Reus said:
So is Intel the only one with reliable fabs anymore? What about IBM? I haven't kept up with fab tech and capacity, but is TSMC really the only other option than GloFo for these sort of chips?
Click to expand...

Actually, there are a lot of other chip manufacturers: Tostitos, Doritos... :)

UMC, SMIC, Samsung, Toshiba, Micron Hynix, Powerchip, Texas Instruments, Renesas, Inotera, Elpida, STMicro, Fujitsu, NEC Electronics, ProMOS, Nanya, Matsushita, NXP, Winbond, IBM Micro, Sharp, Freescale, Infineon, Spansion, Magna Chip, X-Fab... but which ones can actually produce a 28 or 22 nm CPU? Very few.
 
Poisoner said:
I am reluctant to say Bulldozer is a design disaster. I wonder if the cache problems are manufacturing related as we don't know for sure what speeds the uncore runs at.
Click to expand...

They initially designed the chips to run at over 4ghz stock, which would have put them within spitting distance Phenom II IPC -- which was their intent from the beginning. Keep IPC the same and improve the threading. The cache latency issues have existed since the phenoms and still haven't been addressed but rather become progressively worse. But a 2 billion transistor chip running at over 4ghz stock clocks would have been nearly impossible to put in a 120W package even under 32nm. They had to drop them to mid 3 ghz's to keep them from requiring a nuclear power plant to power. Then there's the issue of shitty overclock headroom. I just don't understand how people can pin the blame on GloFo for the processor's sub-par performance. The yields may be low and they don't have enough of them to provide to potential customers (mostly Llano, if we're honest), but for its performance? Really? No matter which way you spin it this was a design disaster and they were hoping GloFo could pull a magic gold-shitting rabbit out of a hat and when AMD realized they couldn't they just blamed GloFo -- again, mostly for the low 32nm yields for Llano/bulldozer and not its performance.

I think they're moving Brazos to TSMC because they believe they'll have less issues with yields, even if it requires investing some extra time for the transition. It's the Fusion tech that's making AMD money, so this is a smart move IMO.
 
Reus said:
Yeah, AMD/nVidia is a horrible idea, same with AMD/Apple, since I think Apple would likely make everything AMD produces proprietary, and force you to buy a shitty Mac to get any of it. I would much rather see them remain independent, but if they simply must merge, go with IBM.
Click to expand...

Right, Apple is like a bizarro google, buying stuff and trying to figure out how to monopolize and control it, where as google will buy a company and flatout give away what they were producing just to quickly as possible gobble up some market share.
 
sirmonkey1985 said:
its TSMC you damn well will regret saying that after the 3rd or 4th delay and 30% yields..

...
Click to expand...

30% is like hitting the jackpot compared with GF's first attempt which produced 0 yield and the 2nd which gave 1-2 working dies per wafer.
 
Weeth said:
Actually, there are a lot of other chip manufacturers: Tostitos, Doritos... :)

UMC, SMIC, Samsung, Toshiba, Micron Hynix, Powerchip, Texas Instruments, Renesas, Inotera, Elpida, STMicro, Fujitsu, NEC Electronics, ProMOS, Nanya, Matsushita, NXP, Winbond, IBM Micro, Sharp, Freescale, Infineon, Spansion, Magna Chip, X-Fab... but which ones can actually produce a 28 or 22 nm CPU? Very few.
Click to expand...

Correct me if I am wrong, but aren't IBM's fabs now part of GlobalFoundries? Especially the one in Malta, NY? Also, wasn't UMC bought by GlobalFoundries?
 
No, you have UMC confused with CSM. I think IBM's fabs and GloFo are kinda just hanging out together for a while too.
 
stealthy123 said:
I think it's pretty safe to say that from a design standpoint bulldozer is impressive.

I think it's also safe to say the architecture, if given time to to mature(software and hardware) will show itself to be pretty darn impressive, practically.

Intel will be doing some of the same things AMD tried to do with bulldozer in the next couple of years.
Click to expand...

I would expect Intel not to want to copy anything at all from bulldozer. I mean the the module design that combines 2 weak cores is just the opposite of Intel's strong over provisioned core splitting into 2 concept with HT. Intel had turbo boost first and also power gating of cores and they certainly would not want to touch AMD's slower higher latency cache.
 
Last edited:
pelo said:
They initially designed the chips to run at over 4ghz stock, which would have put them within spitting distance Phenom II IPC -- which was their intent from the beginning. Keep IPC the same and improve the threading. The cache latency issues have existed since the phenoms and still haven't been addressed but rather become progressively worse. But a 2 billion transistor chip running at over 4ghz stock clocks would have been nearly impossible to put in a 120W package even under 32nm. They had to drop them to mid 3 ghz's to keep them from requiring a nuclear power plant to power. Then there's the issue of shitty overclock headroom. I just don't understand how people can pin the blame on GloFo for the processor's sub-par performance. The yields may be low and they don't have enough of them to provide to potential customers (mostly Llano, if we're honest), but for its performance? Really? No matter which way you spin it this was a design disaster and they were hoping GloFo could pull a magic gold-shitting rabbit out of a hat and when AMD realized they couldn't they just blamed GloFo -- again, mostly for the low 32nm yields for Llano/bulldozer and not its performance.

I think they're moving Brazos to TSMC because they believe they'll have less issues with yields, even if it requires investing some extra time for the transition. It's the Fusion tech that's making AMD money, so this is a smart move IMO.
Click to expand...

Only thing I got out of this post was the fact that you have no idea what IPC is.
 
savantu said:
30% is like hitting the jackpot compared with GF's first attempt which produced 0 yield and the 2nd which gave 1-2 working dies per wafer.
Click to expand...

Source? Not that I do not believe you but I have never heard any numbers and I am interested.
 
drescherjm said:
I would expect Intel not to want to copy anything at all from bulldozer. I mean the the module design that combines 2 weak cores is just the opposite of Intel's strong over provisioned core splitting into 2 concept with HT. Intel had turbo boost first and also power gating of cores and they certainly would not want to touch AMD's slower higher latency cache.
Click to expand...

I am not sure turbo boost is a feature, if we are going that route then AMD offered unlocked processors first, so meh.

it's not two weak cores, it's two real cores with shared resources, and ya, in time intel will adopt the same strategy. It makes a ton of sense, never mind the gloflo fuck ups.
 
Yeah AMD and NVIDIA merger is such a great idea remember what NVIDIA did to 3DFX?? They bought them, then shut them down,they gave no customer support,no driver support for us owners, except to give you $30.00 trade-in for your 3DFX card if you bought their card in place of your 3DFX card!!! I went to 3DFX website and they had that posted in the driver search one day,before sell was made totally public!! Then they used their technology about 3-5 yrs later, that's how far ahead 3DFX was in their technology! This all made me so mad I started buying ATI products instead and I have hated NVIDIA ever since!!! If they got permission to buy AMD what do you think they would do to the Graphics side and CPU side?? I'll bet you History would repeat itself again you wanna bet??
 
You must log in or register to reply here.
Back
Top