32nm Nehalem

[VirtualLarry]

Does anyone have links to an advanced roadmap, as to when Intel is going to shrink Nehalem to 32nm? I'm thinking that I may skip first-generation Nehalem rigs (due to teething problems), and just go straight for the 32nm Nehalem rigs, as I don't see how much faster Nehalem 45nm is going to be over a 4Ghz Q9650.

 

[aigomorla]

Originally posted by: VirtualLarry
Does anyone have links to an advanced roadmap, as to when Intel is going to shrink Nehalem to 32nm? I'm thinking that I may skip first-generation Nehalem rigs (due to teething problems), and just go straight for the 32nm Nehalem rigs, as I don't see how much faster Nehalem 45nm is going to be over a 4Ghz Q9650.

Larry you dont understand how much you crack me up.

But im guessing its gonna be around 2-3yrs at most with delays.

 

[PhlashFoto]

If it were to be on schedule, late 2009. More likely into 2010.

 

[VirtualLarry]

Now I've changed my mind (and I should have kept both of these threads in one, I apologize). If the 2.66Ghz Neha is $284 on Q4 2008, and it is equal to a 4Ghz Yorkfield, then perhaps early adoption of Nehalem is in order after all.

 

[Lonyo]

Originally posted by: VirtualLarry
Now I've changed my mind (and I should have kept both of these threads in one, I apologize). If the 2.66Ghz Neha is $284 on Q4 2008, and it is equal to a 4Ghz Yorkfield, then perhaps early adoption of Nehalem is in order after all.

I thought the 2.93GHz was ~= to 3.6GHz and if so you'd need to overclock to more like 3.3GHz or so to match a 4GHz Q9650.
Also, just because the CPU is "only" $284 doesn't mean the DDR3 and motherboard will be hugely affordable.

 

[Tempered81]

isnt this sandy bridge or westmere? Oh i dunno it's so far off

 

[Extelleron]

Originally posted by: jaredpace
isnt this sandy bridge or westmere? Oh i dunno it's so far off

Westmere 32nm = Q4 2009
Sandy Bridge 32nm = Q4 2010

 

[Idontcare]

Originally posted by: Extelleron

Originally posted by: jaredpace
isnt this sandy bridge or westmere? Oh i dunno it's so far off

Westmere 32nm = Q4 2009
Sandy Bridge 32nm = Q4 2010

What's the latest AMD roadmap?

Shanghai 45nm = Q4 2008
(3rd gen Stars derivative) ?? 45nm = Q4 2009
(Bulldozer 1st gen) ?? 32nm = Q4 2010
(Bulldozer 2nd gen) ?? 32nm = Q4 2011

I know Bulldozer is in there somewhere but haven't a clue any more if the timing is relevant.

 

[Extelleron]

Originally posted by: Idontcare

Originally posted by: Extelleron

Originally posted by: jaredpace
isnt this sandy bridge or westmere? Oh i dunno it's so far off

Westmere 32nm = Q4 2009
Sandy Bridge 32nm = Q4 2010

What's the latest AMD roadmap?

Shanghai 45nm = Q4 2008
(3rd gen Stars derivative) ?? 45nm = Q4 2009
(Bulldozer 1st gen) ?? 32nm = Q4 2010
(Bulldozer 2nd gen) ?? 32nm = Q4 2011

I know Bulldozer is in there somewhere but haven't a clue any more if the timing is relevant.

On the desktop front, who knows. Nothing beyond Deneb/Propus & the 800-series chipsets bringing DDR3 support and some new features.

For servers....
Shanghai = Q4 2008
Magny Cours & San Paolo 45nm = Q1 2010
New G34 socket w/ DDR3 support = Q1 2010

AMD hasn't talked about Bulldozer being released since the July 2007 analyst meeting. In response to a shareholder's question in the Q1 conference call Hector or somebody else did say that AMD would be sampling Bulldozer @ 45nm in 2009. But we all know how reliable AMD has been in the past.

I can't imagine you will see Bulldozer until late 2010 at the earliest, probably on a 32nm process. Given the Magny Cours/San Paolo timeframe of Q1 @ 45nm, I think that is pretty optimistic. And by that time, depending on what Bulldozer is, its window of opportunity will be closed. Even if Bulldozer beats Nehalem, it will be competing with Sandy Bridge.

On the server front AMD is going to try to compete by putting more cores together (San Paolo 6-core and Magny Cours 12-core MCM) but that's not going to help them on the desktop where IPC isn't going to be anywhere close to Nehalem. Once LGA1160 CPUs come in Q3 2009 it's all over as far as I am concerned (in terms of AMD competing for even low-end performance).

 

[Idontcare]

Originally posted by: Extelleron
On the desktop front, who knows. Nothing beyond Deneb/Propus & the 800-series chipsets bringing DDR3 support and some new features.

For servers....
Shanghai = Q4 2008
Magny Cours & San Paolo 45nm = Q1 2010
New G34 socket w/ DDR3 support = Q1 2010

AMD hasn't talked about Bulldozer being released since the July 2007 analyst meeting. In response to a shareholder's question in the Q1 conference call Hector or somebody else did say that AMD would be sampling Bulldozer @ 45nm in 2009. But we all know how reliable AMD has been in the past.

I can't imagine you will see Bulldozer until late 2010 at the earliest, probably on a 32nm process. Given the Magny Cours/San Paolo timeframe of Q1 @ 45nm, I think that is pretty optimistic. And by that time, depending on what Bulldozer is, its window of opportunity will be closed. Even if Bulldozer beats Nehalem, it will be competing with Sandy Bridge.

On the server front AMD is going to try to compete by putting more cores together (San Paolo 6-core and Magny Cours 12-core MCM) but that's not going to help them on the desktop where IPC isn't going to be anywhere close to Nehalem. Once LGA1160 CPUs come in Q3 2009 it's all over as far as I am concerned (in terms of AMD competing for even low-end performance).

Thanks for the info. I agree the horizon for AMD is rather dire from our perspective.

I can't imagine what keeps the hope alive internally, probably the beatings I suppose.

 

[KingstonU]

I thought bulldozer was supposed to come out mid-2009, so AMD is stuck with the current architecture until late 2010? Definitely can't wait for that

 

[VirtualLarry]

I just don't think AMD has the resources necessary to compete on a "tick-tock" level with Intel. Sad to say.

 

[Extelleron]

Originally posted by: KingstonU
I thought bulldozer was supposed to come out mid-2009, so AMD is stuck with the current architecture until late 2010? Definitely can't wait for that

That's what the plan was in July 2007, and AMD's roadmap sure looked good back then.

But just about everything has changed since then. Barcelona isn't 40% faster than Clovertown. Phenom didn't launch at close to 3GHz. And most of the roadmap AMD showed at that time has now been completely overhauled (for the worse).

Why exactly, nobody outside of AMD knows. It's possible that AMD targeted Bulldozer for a 2009 launch on 45nm and is simply no longer able to meet that release target, forcing them to push it back to 2010.... at which point they might as well push it back a few months more and release it on a 32nm process.

It's also possible that the roadmap shown in July was complete BS, which looking at AMD's roadmap now compared to then seems credible. San Paolo and Magny Cours are not even hinted in AMD's 2007 roadmap, so either they are being thrown together at the last minute to provide a stopgap product because Bulldozer was delayed or Bulldozer was just shown as targeted for a 2009 release to make the roadmap look better than the truth.

As for AMD being stuck with their current architecture, it certainly looks that way for now. AMD will have to depend on Fusion, graphics (AMD graphics are looking better than ever), and low prices for the forseeable future. And we'll have to see what Hector and the other executives have planned with Asset Lite. Maybe tonight we will hear something about it but Hector has been steadfast in answering absolutely zero questions relating to it in previous conference calls.