Do we know why Bulldozer was delayed?

[mikeymikec]

And when I say "know", I really mean it, rather than a lot of speculation.

 

[ElFenix]

no.

 

[Martimus]

I thought it was obvious that it was issues with getting a high enough clock speed on the 32nm process, since the yield for it has been less than stellar to date. I expect that the clock speed with ramp up substantially in the next year as the process matures.

 

[Kyanzes]

I heard that it took AMD's regional sales managers a very long time to be able to say "The best CPU on the market." convincingly. Hence the delay.

 

[StrangerGuy]

Given how huge BD is at 32nm, if it was done in 45nm it would only have 6 cores tops with same or worse clocks.

That will be a disaster when placed next to a Thuban.

 

[Maximilian]

No idea. It would be interesting to know.

 

[Idontcare]

Do we know why Bulldozer was delayed?
And when I say "know", I really mean it, rather than a lot of speculation.

We know exactly why it was delayed. Answer: $

Be it to lose less of it or to make more of it, sometimes the difference is moot.

Now if you mean to frame your question as "what did AMD choose to not spend more money on, such that the opportunity cost of that decision resulted in the delayed release of Zambezi?" then that's a discussion that I doubt even the engineers in the fab or on the design team can honestly answer because those decisions were made way way way up the chain of command.

But at the end of the day, it was because of money. Same as always.

 

[sangyup81]

I'm not really sure why they are pursuing 32nm for CPU and 28nm for GPU at the same time. Why not just skip 32nm all together especially with this news that CPUs from AMD will come out in 28nm next year

 

[IntelUser2000]

I'm not really sure why they are pursuing 32nm for CPU and 28nm for GPU at the same time. Why not just skip 32nm all together especially with this news that CPUs from AMD will come out in 28nm next year

You mean for fully integrated designs like Trinity? Because the 32nm part is for CPU and the transistor is optimized for high performance. The 28nm part is GPU and the transistor is optimized for density. If you put the CPU on the 28nm it'll probably turn out slower than even Llano is.

 

[grkM3]

It was delayed because sandys came out and AMD didnt think they were going to be such powerfull little monsters.

the ceo stepped down a week after intel leaked out its sandy bridge performance numbers.

They were prolly running around crapping them sleves when they first saw the numbers from intels next gen chip.Then they had to raise clocks as much as possible to make them compete.

 

[sangyup81]

You mean for fully integrated designs like Trinity? Because the 32nm part is for CPU and the transistor is optimized for high performance. The 28nm part is GPU and the transistor is optimized for density. If you put the CPU on the 28nm it'll probably turn out slower than even Llano is.

AMD is planning 28nm CPUs for late 2012 or 2013 rather than try for 22nm like Intel is about to do with Ivy. 32nm gave us BD and Llano. Should AMD have bothered?

 

[bryanW1995]

@grkm3: That doesn't make sense. BD is 7% slower clock/clock than x6 is, and the clocks are lower than anticipated as well. It was delayed so much not because it couldn't beat 2600k, but because it couldn't beat i5 870.

 

[Idontcare]

AMD is planning 28nm CPUs for late 2012 or 2013 rather than try for 22nm like Intel is about to do with Ivy. 32nm gave us BD and Llano. Should AMD have bothered?

Not showing up to the game has a guaranteed outcome 100% of the time.

By "bothering" to show up they at least had a non-zero chance of avoiding the otherwise guaranteed outcome.

Business is about risk, and there none greater than that of simply not bothering to try.

 

[IntelUser2000]

AMD is planning 28nm CPUs for late 2012 or 2013 rather than try for 22nm like Intel is about to do with Ivy. 32nm gave us BD and Llano. Should AMD have bothered?

I think that's true for the Bobcat-successors and server, but we haven't seen the "after Piledriver" non-APU desktop and laptop roadmap yet. And Bobcat's successor is due summer of 2012 with 28nm, so that's little earlier, and the predecessor is 40nm so that doesn't matter anyway.

Though if they do 28nm, that would imply a more significant strategy change than implied.

 

[hooflung]

I'm not really sure why they are pursuing 32nm for CPU and 28nm for GPU at the same time. Why not just skip 32nm all together especially with this news that CPUs from AMD will come out in 28nm next year

Their 28nm is a for bulk process. Bulk doesn't have fancy insulators like HKMG or SoI. It is for things such as Graphics chips, Memory, microcontrollers and ARM Cpu's.

 

[IntelUser2000]

Not showing up to the game has a guaranteed outcome 100% of the time.

By "bothering" to show up they at least had a non-zero chance of avoiding the otherwise guaranteed outcome.

Business is about risk, and there none greater than that of simply not bothering to try.

My understanding was that he meant "why not go 22nm" rather than going to 28nm.

But I guess that's another way of answering it. 🙂