A Gruesome Roadmap . . . "

I read this article the other day and just went and revisited the article and looked at the linked roadmaps myself. And like Ed, it's confirming my suspicions that something is terribly wrong with AMD's 65nm plans. Seeing that the fastest 65nm processors in the roadmaps will only be 2.8 GHz tells me that they must be having problems making the new process go faster, much like the problems that AMD had with the transition from the Palomino procs to the Tbred A series, which weren't much faster than the Pally's. And seeing that the TDP of the new process is actually the same as the mature 90nm process until the 90nm procs get to 2.8 GHz is making me wonder if AMD's processor design is starting to run into the same basic problems with leakage current that Intel ran into with Netburst when it made the 90nm transition. Like Ed, I think the future is looking kind of bleak for AMD for the next 1 1/2 years according to this.

What do you all think?

A Gruesome Roadmap . . . " Linkage

Yea, they "should" have been able to scale 65nm a little better then 90mn.
Hopefully K8L is everything it has been cracked up to be a 15% more

I think there's still a chance...follow me.

First off, I think that roadmap isn't up to date, it shows a couple of oddities such as the FX staying 90nm through the end of next year up to 3.2ghz with the same tdp? Something's not right there...Also, it leaves out the 35w 3800+ which was announced not too long ago and the 3600+ which is looking to become more of a reality as time passes, anyway, here's my thoughts...

(It could be a result of fab constraint. Think about it, AMD has only 1 fab producing 65nm by the end of the year, 1, compared to 2.5 (I'm including chartered) it has going for 90nm. Now, we already know AMD is trying to get the most out of each die by phasing out 1mb chips, so they're going to want to make as many cheaper cpu's as possible. So, what are the big sellers, well, a very low percentage buy the FX line, most fall in the 3800+ (-) 4800+ range. So, instead of making a few FX cpu's at 65nm that may or may not sell, they're concentrating their 65nm production at the lower speeds that will sell to be more profitable. They need their biggest volume movers to be the cheapest to make to maximize profits. There's also the new deal with Dell as well as the other OEMs we have to consider. Heck, for all we know AMD may have promised Dell full Fab36 65nm production to ensure the deal leaving only a partly transfered Charter for everyone else. I also expect to see AMD pushing the server market more aggressively with 65nm cpu's than on the desktop market because they have the most to lose on the server side.

Anyway, the way I see it is that there are just too many variables to consider right now to say anything for sure and we're just going to have to wait to find out come early 07.

Well, an important possiblity was overlooked. We are assuming that the architecture ould remain the same. What if tweaking the architecture would not necessitate monstrous clock frequencies, to stay competitive?

We can only hope. IMHO AMD won't truly be able to compete with Intel until they can beat them down on a clock-for-clock basis. Intel has ALWAYS been one step ahead of AMD in terms of die shrinks, but it has never really hurt AMD before. It clearly is now that Intel has an architecture that not only compete with them, but can beat them...the smaller manufacturing process is just salt in the wounds.

I suspect a lot (if not all?) of AMD's problems are due to it's 65nm process which it paid IBM lots for. Me thinks IBM told porkies about it's process to AMD and even itself.

All we need to know is how bad Cell processor yields really are. That will show if their 65nm process is any good or not.

thingi

Déjà Vu All Over Again . . .
http://www.overclockers.com/tips00993/
I remember stuff too...

I remember faintly that the FX processors (wich are the top end of the MHz spectrum) remained on the mature process (0.13µm) during the switch to 90nm.

This does not only happen with processors, but also with video cards.
Remember the radeon X850? It was a 130 µm part while the low end was upgraded to 110 µm.

As for the wattage, the 90 nm 65 wattage are low voltage parts, but I don't see any 89 watt 65 nm parts. So, there IS a lower power consumption due to the die shrink.

I wouldn't start screaming death and despair yet.

Thingi said:

I suspect a lot (if not all?) of AMD's problems are due to it's 65nm process which it paid IBM lots for. Me thinks IBM told porkies about it's process to AMD and even itself.

All we need to know is how bad Cell processor yields really are. That will show if their 65nm process is any good or not.

thingi

Click to expand...

cell yields are in the toilet .

http://www.theinquirer.net/default.aspx?article=32978

RAMMAN said:

cell yields are in the toilet .

http://www.theinquirer.net/default.aspx?article=32978

Click to expand...

I think current Cells are still 90nm, they not supposed to go 65nm till mid 2007. Seems alot like the AMD roadmap, I reckon SOI is great at 90nm but a nightmare at 65nm.......... That's why the power figures are the same for the 65nm and 90nm processors.

Seems to me AMD have hit a wall like Intel did when presshot has released, only this time it's not the product at fault it's the process.

thingi

problems will always arise when converting to the next shrink size. that is why they use current products to get the feel for it...just the normal shrink process.

usually the higher end chips-speed wise, always return a less yielding wafer, leading to less product...resulting in the "boutique" chips name and price to go with it.

the current arch has it's limits, switching to 65nm won't change too much clock wise, similiar when prescott was shrinked. probably some power consumption benefits. you can only squeeze so much out of it, and looks like it's running out of steam.

obviously will help on cost savings and lead a good path to the next arch.

If they can cut power consumption, they also cut heat wattage. Which means what we have in our rigs now might be the top of the heat ladder, and it'll only go down from here. So with the same cooling, the next gen will clock higher because it's producing less heat watts. And the next gen chips after that will produce even less heat, thus will clock even better on the same cooling.

At least if the current trend of power consumption cuts holds true. And there's also the prognosed rise in HTPC popularity. That's a real push for lower power CPU's since there's less room for cooling in a micro case.

What's ironical about this is the next gen GPU's from ATI and nVidia, they claim that the top end ones will require 1100 watt PSU's. THAT one will go down real well with environment agency people all over the place, when they are pushing active PFC PSU's, lower watt CPU's and RoHs compliance...

We're gonna need scythe ninjas on out graphic cards, while a passive old P4 sink will do for the CPU.

Bad Maniac said:

If they can cut power consumption, they also cut heat wattage. Which means what we have in our rigs now might be the top of the heat ladder, and it'll only go down from here. So with the same cooling, the next gen will clock higher because it's producing less heat watts. And the next gen chips after that will produce even less heat, thus will clock even better on the same cooling.

At least if the current trend of power consumption cuts holds true. And there's also the prognosed rise in HTPC popularity. That's a real push for lower power CPU's since there's less room for cooling in a micro case.

What's ironical about this is the next gen GPU's from ATI and nVidia, they claim that the top end ones will require 1100 watt PSU's. THAT one will go down real well with environment agency people all over the place, when they are pushing active PFC PSU's, lower watt CPU's and RoHs compliance...

We're gonna need scythe ninjas on out graphic cards, while a passive old P4 sink will do for the CPU.

Click to expand...

thus is true, but like i said...90 to 65, you could run into problems intel did, and i'm sure they (amd) will. power/heat consumption can go down, but there are other variables that can prohibit speed increases. im guessing 2-400mhz i dunno.

all in all it's a good thing for all of us.

they'll pull something out of their hat, they always seem to every now and then when things get like this.