800 Million Transistors go Missing

[dma0991]

According to Anand's writeup, GloFo's 32nm process takes both the top and bottom spots for 32nm xtor density. Or am I reading this chart incorrectly?

Yes you are. GF's 32nm might take top and bottom spots in terms of transistor density but they are definitely not taking the top spots in terms of overall performance.

Higher transistor densities are generally more desirable to a manufacturer (fewer defects per die, more die per wafer, lower costs), but from the end user's perspective the overall price/performance (and power?) ratio is what ultimately matters.

 

[Kalessian]

I didn't say anything about performance, but people keep blaming GloFo's process and I want someone in the know to comment on how GloFo was able to get llano xtor density so high.

 

[Idontcare]

I didn't say anything about performance, but people keep blaming GloFo's process and I want someone in the know to comment on how GloFo was able to get llano xtor density so high.

see here

 

[Kalessian]

Ah thanks, missed that bit somehow, carry on.

 

[Blue Shift]

OK, so now it seems like issues with GF's 32nm process had a higher impact on BD's performance/watt than was previously thought? I'm not sure whether that's a good or bad thing for AMD in the long run. What use is a decent chip layout if there are no fabs that can compete with Intel?

Edit: What doe you guys think would happen if Intel was vertically trustbusted and forced to spin off its fabs? Not that such a thing is at all likely...

 

[LoneNinja]

OK, so now it seems like issues with GF's 32nm process had a higher impact on BD's performance/watt than was previously thought? I'm not sure whether that's a good or bad thing for AMD in the long run. What use is a decent chip layout if there are no fabs that can compete with Intel?

Edit: What doe you guys think would happen if Intel was vertically trustbusted and forced to spin off its fabs? Not that such a thing is at all likely...

That's like asking what if Intel and AMD negotiated a deal for Intel to manufacture AMDs processors, which honestly sounds more likely to me.

If Intel got rid of its fabs it would probably be another GF all over again. lol

 

[Vesku]

It seems that GF has spent at least twice the amount of $ that AMD could have invested for 32/28nm. If AMD had held onto their fabs I think those who have been chanting "AMD is dead" since Phenom I would have finally had their wish.

 

[exar333]

It seems that GF has spent at least twice the amount of $ that AMD could have invested for 32/28nm. If AMD had held onto their fabs I think those who have been chanting "AMD is dead" since Phenom I would have finally had their wish.

And what do they have to show for it? A poor 32nm process and a scrapped 28nm CPU altogether. What a success?

 

[Tsavo]

Wow. If GlFo keep this up, all they'll be making are hub caps for nanobots.

 

[Vesku]

Yes, and what would AMD have to show for it since they would be trying to implement the same processes with half the money? Stems from poor choices and a lack of production level R&D. Trying to take academic oriented R&D straight into production often leads to these kinds of issues. Seems like AMD and now GF are IBM's trailblazing patsy, perhaps the sackings at GF will see this change.

 

[exar333]

Yes, and what would AMD have to show for it since they would be trying to implement the same processes with half the money? Stems from poor choices and a lack of production level R&D. Trying to take academic oriented R&D straight into production often leads to these kinds of issues. Seems like AMD and now GF are IBM's trailblazing patsy, perhaps the sackings at GF will see this change.

I don't think you can say 'Company A' with XYZ dollars would fail because 'Company B' with XYZ*2 failed.

It is likely? Maybe. But GloFo is not looking out just for AMD, so their expenses are different. Building a commercial fab vs. a private fab is not the same. Nor are the goals.

 

[frozentundra123456]

AMD's design sucks to a point but Global/TSMC are the real reason AMD can't compete.

This may be true, but does it really matter? AMD made the decision to sell of their fabs and now are dependent on someone else. Bottom line is what the performance of the final product is. In a sense it is irrelevant who is at fault.

 

[Vesku]

I don't think you can say 'Company A' with XYZ dollars would fail because 'Company B' with XYZ*2 failed.

It is likely? Maybe. But GloFo is not looking out just for AMD, so their expenses are different. Building a commercial fab vs. a private fab is not the same. Nor are the goals.

GF had almost the same staff in charge of production as AMD had, until recently that is. So it's extremely likely. Now the response to it would be different. AMD couldn't throw money at it like GF has. They would also be focused on just fixing it for their chips where as GF needs to fix it for general use.

 

[Idontcare]

This may be true, but does it really matter? AMD made the decision to sell of their fabs and now are dependent on someone else. Bottom line is what the performance of the final product is. In a sense it is irrelevant who is at fault.

I look at it like this. GloFo's 32nm/28nm situation can only be better than whatever it would have been had AMD held onto the fabs.

GloFo, to their credit, are not going down without trying. They are throwing gobs of money at it.

It just goes to show that doing this stuff is really hard, it is rocket science and there are no shortcuts to be taken.

And likewise with AMD, I can't possibly imagine a superior path for them to have taken than the one they took in which they acquired ATI and then spun off their fabs while maintaining contracts with them in good faith.

It was the cleanest and most tractable method of separating the two businesses.

The only way AMD could have done anything different is if someone had come along and simply gifted them $10B circa 2002. Barring that, they've made the best of a very untenable economic reality.

 

[jones377]

I don't think you can say 'Company A' with XYZ dollars would fail because 'Company B' with XYZ*2 failed.

It is likely? Maybe. But GloFo is not looking out just for AMD, so their expenses are different. Building a commercial fab vs. a private fab is not the same. Nor are the goals.

Running under utilized fabs like AMD was doing before spinning off GloFo is what bled them so much money. Did you miss the fact that AMD is now back in the black for several quarters? If profits go sour again in the future they can downsize their company without worrying about some big expensive fab that they couldn't do much about in the past except watch it bleed.

 

[Vic Vega]

Hilarious situation when they don't even know wtf they are making.

A total joke, everyone involved with the BD development in the AMD CPU division should be fired if this actually occurred (and not just a lame PR spin).

I'm amazed they haven't hired you as CEO yet! :whiste:

 

[antisocialmunky]

I kinda wonder what a 45nm bulldozer without the terrible power bleeding at load could do.

 

[Vesku]

Power 7 is 2 billion transistors on 45nm. Doubt consumers would want to pay the going rate for that though.

 

[soccerballtux]

If somebody had come along with $10B to get rid of hector ruinz then lol that would have been the best for AMD....

 

[jones377]

Power 7 is 2 billion transistors on 45nm. Doubt consumers would want to pay the going rate for that though.

Is that a veiled joke? Power7 is also about 1.2B transistors Regardless, it's an amazing chip really. 8 cores, 4 threads/core and 32MB L3 made of eDRAM running at over 4GHz!

 

[Vesku]

Using AMD Marketing math. ;p But yeah, AMD could have made a x86 knock-off of the Power 7 on 45nm let alone 32nm yet we get the current Bulldozer. Things at AMD and GF are very strange, at this point I think Dirk must have purposely pissed off the board just so he could bail out with some extra cash.

 

[aigomorla]

guys i think the 2Billion transitors they are talking about is the machine which litographs the cpu.

leave it to marketing to leave out what the 2B is actually used.


And yes i HATE AMD MARKETING with a passion... I still blame them for everything.
If AMD came out true.. showed us really what bulldozer could do, and then didnt deny it so we could tier it properly, i dont think their launch would of been as barf'd as it was.

 

[busydude]

So.. according to AMD math 2Billion transistors = 1.2 Billion transistors?

Is there a white paper by AMD to prove this?

 

[exar333]

Running under utilized fabs like AMD was doing before spinning off GloFo is what bled them so much money. Did you miss the fact that AMD is now back in the black for several quarters? If profits go sour again in the future they can downsize their company without worrying about some big expensive fab that they couldn't do much about in the past except watch it bleed.

Not really. It just means they have no actual hard assets any more. They don't have anything 'real' to sell-off if they hit issues again.

 

[BD231]

This may be true, but does it really matter? AMD made the decision to sell of their fabs and now are dependent on someone else. Bottom line is what the performance of the final product is. In a sense it is irrelevant who is at fault.

So irrelevant AMD decided it was pointless to peruse intel due to process limitations?? So irrelevant AMD cpu's have for some time now offered less performance per watt than the competition despite efficient/high performance designs?? So irrelevant AMD has decided to change their entire position in the market? So irrelevant its affected both AMD's GPU and CPU markets in a negative fashion, actually requiring a " skip " in process nodes to compensate for lost time?

Being responsible for AMD's lack of ability to innovate is hardly irrelevant. The smartest thing I've heard AMD say in a long time is bowing out the high end race, because in the end you're only as good as your weakest link and I think AMD has proven it's capeable of making good CPU's.