Intel Corp (INTC)

[chipguy]

Interview with Phil Hester: AMD's CTO

http://www.amdzone.com/amdzone/index.php?option=com_fireboard&Itemid=27&func=view&id=134918&catid=52

When would you guys plan to launch Phenom 9900? Nowadays, all tech sites (or specialized media) have one, but still it's not available in the market.

That particular product will be launched in the second quarter of this year. Still, we are not publishing a particular launch date yet.

Let's talk about frequency scalability. When you all designed Phenom, were you expecting that your top of the line would be clocked less than 2.3GHz?

It's still on the minimum expectations we were all waiting for. Even still, this new architecture works extremely well. When you design a new architecture, the people in charge of the manufacturing will give you reports about how the design works through many different conditions. When you start to ramp up volume, you get reports about how manufacturing and performance are working out (what to expect from them). In Phenom's case, the projected statistical distribution was different from the real one, although, it's something that we can fix in time by adjusting transistor ratios.

What type of frequencies do you expect from Phenom in 65nm?

Let's not talk about specific frequencies. The question that we should make is: how much work does a processor performs in each clock cycle?
A processor that runs at 2.5GHz may perform much faster than one that runs at 5GHz if the first one performs more than double of instructions for each clock cycle. We do expect to have the leadership in performance.

Are you expecting greater performance than Yorkfield at the same frequency

In an apples to apples comparison, yes.
Intel is focusing more on lithography, and we should tell you a detailed story about...what we're trying to make is a good combination between lithography performance and transistor performance.

What intel doesn't publish is that their manufacturing performance is not the same as 5 or 6 months before, because they have a learning curve to master. When they introduced their 45nm transistors, our 65nm processors had better yields because they (intel) were in their first 45nm generation, while we were at the third or fourth 65nm generation. Intel loves to say that they are ahead of the industry when it comes to lithography, that’s not the how way we work. We ask our selves where we are in terms of transistor performance, and to put it this way, we believe that the technological gap is only 4 months.

How would you manage the pricing pression that intel can apply with their 45nm products?

I think you should aske them what type of yields are they having with 45nm. If they answer you that yields are good, then why aren't they releasing all their offerings in this technology? (Note: during the interview, intel announced 16 new 45nm products).

Did you read the article about Phenom’s AM2 motherboard compatibility, published by Tom’s Hardware?

No, I haven't.

Well, basically they state that two out of ten motherboards worked with Phenom

Did they had BIOS updates?

Indeed, the latest ones.

The only thing I’ve heard is that BIOSes are not in time. The electrical and logical side of the motherboards are compatible with Phenom. We've even tested them on our labs and they work flawlessly. The point is time coordination for the BIOS updates and I think you should ask your motherboard manufacturer to release their respective BIOS updates.

Isn’t AMD working with motherboard manufacturers?

We work with them, we have a validation platform, the problem is that we don’t know what type of internal cycles they have for their developments.

After the episode in Lake Tahoe, the weak availability during the first weeks and the lack of BIOS updates, all of this make us think that Phenom had a rushed launch. Don’t you think the same?

It’s all about expectations. In some moments they criticize us because they want to see our latest technological improvements as soon as possible, and in this occasion, they also criticize us because the launch was too soon. We need to find the perfect balance for our future launches.

After AMD signed a deal with Dell, seems that you guys forgot who gave you to eat (talking about the enthusiast community). What can you tell them?

If we look at the third quarter of 2006 and the first quarter of 2007, it took us alot of work to manufacture all the units needed for the big OEMs (DELL, etc.). I think the commentary is valid, but if you look nowadays, things are much more balanced.

Volume manufacturing, ¿what plans do you have?

We will keep working with UMC and TSMC for the ATI side. In the case of CPUs, we'll keep with the current hybrid model: in house processor manufacturing with our fabs, and some support from Chartered. There won't be any changes in the short to mid term since we think this model works very well for us.

Un-frigging-believable!