Interview: Daryl Ostrander - AMD R&D vice-president

[Cooper]

Recently PCWatch interviewed Daryl Ostrander - AMD R&D vice-president. As he stated company will introduce new chips production tech using Germanium. This will increase transistors speed up to 15% comparing to current SOI technology.



Also die shrinking shedule was on topic :



45nm build CPUs will have implemented SiGe from the start.

Source

[Bar81]

So, the reports are accurate. AMD 65nm launches in 2007. It'll be interesting to see if they use the K8 to tweak the process before hitting with K8L

[Entity_Of_One]

I wasn't aware of how soon the 45nm builds were coming out. In fact, I didn't even know they were really on the schedule yet. That's quite impressive to hear.

[mesyn191]

Not really, you've just awash in nonsense from all the people who're saying AMD is gonna cave now and they keep falling farther and farther behind, etc.

That chart sums it up nicely, while they've always been behind Intel on process development, the delays have been consistent and not getting farther and farther apart.

[danmoody]

im still amd to the core

[zakelwe]

Why are they building 65nm cores from mid 2006 and then only product launching 6 months later ?

Regards

Andy

[Cooper]

Why are they building 65nm cores from mid 2006 and then only product launching 6 months later ?

Regards

Andy

Apparently this 6 months is term from first prototypes till serial production

[Dimitriman]

I heard a lot of people touting the new SOI process will give a 42% increase in transistor performance over current generation but it seems its not true. What they mean is a 42% increase in performance over no SOI at all. The difference with the current generation will be at best half of that.

[keiths]

From the benching here on xtremesystems, they'll need 3.7 GHz just to match 2.6 Conroe.

[oshua]

From the benching here on xtremesystems, they'll need 3.7 GHz just to match 2.6 Conroe.

That's true with the current K8 arch. with the K8L the clockspeed shouldn't need to be as high the K8.

[lapdog]

Not really, you've just awash in nonsense from all the people who're saying AMD is gonna cave now and they keep falling farther and farther behind, etc.

Sure AMD has something...

AMD is mad if it trusts in the tameness of a wolf, a horse's health, a boy's love or a whore's oath. The firm [Intel. Ed.] said in the Q&A that it had an answer to AMD's next generation too. This was the unkindest cut of all, but what's gone and what's past help should be past grief.

Intel is a moving target.

[Cooper]

Especcialy with Israel team onboard. Those guys put Itanium to it`s knees

[mesyn191]

Intel is a moving target.

Did I say it wasn't?

Both of them are constantly working on something new and tend to leapfrog eachother in developments. The end result is that depending on how much you can spend and when you're looking to buy either one of them may be a better buy over the other.

However as Intel has recently shown us (and AMD to in the not to distant past (ie. Thoroughbred/Barton vs Northwood P4 days...)) it takes a long time to come up with something new and in the mean time all you can do is sit and take the punishment.

So the question isn't if Intel is working on something new to take on K8L, of _course_ they are!!! The question is when they'll have it available AFAIK they haven't released any info. on any K8L-killers that are coming anytime soon (though to be fair all we've had on K8L info.-wise is increased FP units (which does not necessarily equate to increased FP performance) and some vague data on process improvements...).

[WeStSiDePLaYa]

Not really, you've just awash in nonsense from all the people who're saying AMD is gonna cave now and they keep falling farther and farther behind, etc.

That chart sums it up nicely, while they've always been behind Intel on process development, the delays have been consistent and not getting farther and farther apart.

really? i must have forgotten about intels on die mem controller.

[mesyn191]

really? i must have forgotten about intels on die mem controller.

I see absolutely no connection with your comment and what I said...

[mursaat]

You said AMD was always behind Intel in process development, and WSP pointed to the Intel lack of on-die mem controller.

But maybe you was referring only to process size?

[mesyn191]

You said AMD was always behind Intel in process development, and WSP pointed to the Intel lack of on-die mem controller.

But maybe you was referring only to process size?

Process development has nothing to do with the on die memory controller, that would be an architectural design feature that you could build on any modern CMOS process...

[mursaat]

OK, understood, I think WSP meant AMD was ahead in some technologies

[mesyn191]

Even then it still doesn't make any sense though, but whatever...

[lapdog]

The product launch for 45mn is not listed, And where does K10 fit in? I not sure that K8l is going to be a big change. AMD may have a good 2H08, maybe not.

Intel 45nm launch is sooner than the chart said too.

[zir_blazer]

The first photo is interesing. If that reflects Frecuency and leakage scalability, expect that Processors leakage increase as Frecuency increases would be fairly lineal and that most of them will overclock around the same.

[gundamit]

Germanium Am I the only surprised that no Germans are used in its production?

[nn_step]

Germanium Am I the only surprised that no Germans are used in its production?

It is just you.. what do you think the first Transisters were made out of man
Yep you guessed it.. they did it right the first time...

[Cybercat]

Personally I couldn't care less about 65nm K8 CPUs. I'm only concerned with finding out what K8L has in store.

[VulgarHandle]

i'd love to see a 65nm K8, would give me a reason to by the DFI CFX3200