AMD Phenom CPUs to see further delays

[polonyc2]

AMD has recently notified its partners that the launch of higher-end quad-core Phenom processors, including the 9700 and 9900, will be postponed to the second quarter of 2008 from the original schedule of early 2008, according to sources at motherboard makers.

However, whether AMD's triple-core Toliman series CPUs will also see delay will be the key decision for the company, since Toliman offers a high price/performance ratio compared with Intel's quad-core CPUs. A delay for Toliman will hurt AMD the most, noted the sources...

AMD Phenom CPUs to see further delays

[Sandog]

Even though AMD is behind Intel whatever AMD does has a big effect on what Intel decides to do. AMD is powerful now. I hope AMD can make a great comeback which I am sure they will.

[polonyc2]

they still havn't found a fix for that errata bug yet??

[Kedas]

If true this means that the 45nm version will be only one quarter behind the real Phenom release.

There is no reason why 45nm would be delayed due to the bug since they get new revisions anyway.

[k0NG0]

Is Tom DeLay on the board?

*ZING*

[XJ.]

That's bizarre, supposedly the B3 stepping fixed the bug - so what happened, it showed up again at higher frequencies??? Something isn't right here...

[Kedas]

I don't think the B3 stepping exist yet, all current CPU's use the bios fix.
(but I'm not sure)

Just to add some info:
The first 45nm (shanghai) bootable sample is expected for January.
and they expect to start production in the first half of 2008.

So we can expect PR news from AMD about 45nm in january.
If they don't then 45nm will have a problem and will be delayed.

[XJ.]

Originally Posted by Kedas

The first 45nm (shanghai) bootable sample is expected for January.
and they expect to start production in the first half of 2008.

There are pretty good odds it will be bootable - the bigger question is will it pass all the QC testing. If it doesn't then Shanghai will need at least one re-spin.

[Gorganzola]

All I know is I wouldn't want to be on that Dev team. I think I saw Acme out back unloading an Guillotine. At least it will be quick and easy.

[Kedas]

Originally Posted by XJ.

There are pretty good odds it will be bootable - the bigger question is will it pass all the QC testing. If it doesn't then Shanghai will need at least one re-spin.

If I'm not mistaken they have time for one respin to stay on schedule.
(mass production start Q2 2008)

[XJ.]

No way are they going to starting mass production in Q208, even with no re-spins.

[k0NG0]

Originally Posted by XJ.

No way are they going to starting mass production in Q208, even with no re-spins.

Agreed. If they manage selling 45nm stuff by early Q4 '08 they've done something really nice for a change.

[Intel_User]

If true this means that the 45nm version will be only one quarter behind the real Phenom release.

There is no reason why 45nm would be delayed due to the bug since they get new revisions anyway.

They won't get 45nm in 2008. The current roadmaps indicate 45nm as H2 2008, but its essentially same as what Intel said with Penryn. Only EE Penryn came in 2007, and that's in November. Some roadmaps say H2 2007 for servers, and Q1 2009 for desktops.

I know it sounds bleak, but there's a difference between going into 45nm production and actually having the part officially introduced to the public. I painfully learned it anticipating for Prescott and such. I have actually tried to see how long it took Intel and AMD to make each process technology advancements. From 0.18u to 65nm, it was consistently 2 years. It was also true with tape-out, it took 9-12 months from tape-out to production. And AMD came with their 65nm in November 2005. Their 45nm should be anywhere from -3 months to +3 months of 2 years from the 65nm introduction.

This is what I anticipate:
Q1 2008-tape-out
Q2 2008-45nm production starts
Q3 2008-45nm final/engineering samples sent to mainboard vendors
Q4 2008-45nm processors available for purchase to public

[Kedas]

Yes this timetable will probably be true if they add new delay's
But it's not like they have to test a new CPU, if 45nm is working they shouldn't have any problem with the CPU.
(except for the cache size there are no big changes, certainly not for the motherboard)
It takes them 3months to do a respin so introducing 45nm 6 months after you have them working does seem possible to me.

And the shift from 65nm to 45nm for AMD isn't 24months but 18months.
(this was planned this way in 2006)

I still think it's possible that Q3 will have the first 45nm CPU's for the public.
(besides that is what their latest roadmap shows)

[k0NG0]

Oh, at 45nm it's basically a new CPU entirely. The logic may be the same functionally, but the layout and stuff is different enough from 65nm to be called a new design. The way the process acts at the different sizes may require significant changes to how you size transistors, buses and whatnot. It's not as easy as "shrink the mask and off you go", far from it.

There is lots of tweaking happening in terms of sizing and positioning before you can ramp to production. A re-spin signifies changing functionality, but there are several mask revisions before a chip starts production.

This day in age, with the validation tools available, it'd be surprising if a first sample didn't boot, be it at a couple of dozen megahertz due to some process weirdness. Getting stuff to run smoothly at high frequencies is where there is need to tweak the layout.