Mario Rivas, Executive VP of the Computing Products Group sat down for a very enlightening interview hosted by CRN. The reporter asked tough questions, and Mario gave some surprisingly honest and straightforward answers. As we all know, AMD has had a rough time of it, and the editorials about the perception of their external marketing and product placement have been rearing their heads as of late. In this interview Mario is amazingly forthright about the issues AMD has been having, how they have been addressing them, and what things they could have done better. We should all probably remember, nobody is perfect, especially not a company as large as AMD. What we are starting to see though is that the company is slowly, but surely, responding to the issues that major partners and consumers are having with AMD’s latest products.
There are a few things of great interest throughout the
interview, but the one that should catch everyone’s attention is the upcoming
45 nm “
Here is what Mario had to say:
“I can tell you what I know. We have 45nm on the way. We will have initial samples also in January. I’m fairly confident that those puppies are going to boot, and then we can have a follow-up conference call and I’ll tell you, ‘The sucker is booting.’ The 45nm, we consider it Rev C of the device.”
This is the 45 nm redesign of the Barcelona/Agena
processor. It includes the fixes to the
TLB errata, as well as some other significant redesigns involving the caches
and the integrated memory controller. Mario said that initial samples should be available internally to AMD in
January, which means that the 45 nm design was taped out some time ago. It takes anywhere from 8 weeks to 24 weeks to
get a chip manufactured, depending on how the lot is handled. In “super hot lots” we can see silicon
returned in as little as 4 weeks or so. That is not terribly common, but because AMD is still working on their
45 nm process they can shuffle lots through very quickly as compared to hot
lots running through their 65 nm production line.
If these first samples boot up like they are supposed to, then AMD could really be in business. The chances of these processors booting from the first lot are good, mainly because of the excellent verification tools that CPU manufacturers have at their disposal. 95% of the debugging work happens well before metal is laid down on silicon, and for the past several years most first silicon comes out in working order. Usually it takes around a year from first silicon to when it is actually put into full scale production. We can imagine that the pressure on AMD right now could in fact shorten that timeline. If the company focuses the majority of its resources into getting 45 nm parts out in short order, we can speculate that production will start in mid summer with the first mass produced parts hitting the market in early October. Or November if AMD decides to build up some stock first. Unless of course the wheels fall off again, and we are looking at a December release.
Either way, AMD is slowly closing the process gap with Intel. While they would like to be 6 months behind Intel in releasing mass produced 45 nm parts, it looks more like 9 to 10 months. Still, it is a big jump from when Intel started mass production of 65 nm products to when AMD was able to finally release small lots of 65 nm parts, which was about 14 to 16 months, depending on how you count it.
Even saying that, the jump to 45 nm is not a given for AMD. They are using an immersion lithography component to their process, which does add some complexity to the line. Intel is still using a dry litho for their latest 45 nm process, and apparently they are having very good results with it. Depending on who you talk to though, there are a lot of differences in opinion about which is “better”. Whatever the argument is, Intel will eventually have to utilize an immersion litho technique in process nodes below 45 nm. Is it cost effective for AMD and its partners to develop this now, or later? Well, apparently by their actions they feel that now is the right time to develop immersion litho, before they are absolutely forced to develop it.
Mario goes on to say that the first production quality B3 revision processors will start coming off the line in January, but production verification has to be done. If that comes out ok, then B3 cores can be sent to the desktop market. Further verification by AMD’s high level customers will also have to be done on these new parts, so introduction into the server and workstation arena will be a bit later. So we can hopefully expect to see B3 desktop Phenoms hitting the shelves in very late January or February, but likely the majority of product will be availably by March.
Here is Mario’s line of thinking on that topic:
“What I can tell you is that the fix involves for the most part just the top layers. We held material that I believe is first metal. And then we are doing the fix with the layers that are left over, since we have eleven layers of metal, we have quite a bit of play. It takes us in the mask and it’s going to start running at the factory. We will get samples of the device in the January timeframe, but then we need to do our own validation because if we just assembled them and shipped them prematurely and there is another big bug, the tempest of press that we have right now would get even worse. We’re going to go through a very rigorous verification process and then hit the market with samples. And once customers validate them themselves, then we will pursue shipping in mass.”
AMD is still kicking and trying hard to rectify their mistakes. They made some doozies in the past two years, but the company keeps running. Phenom is their future whether we like it or not, and they are working hard to make the proverbial lemonade out of lemons. Getting a fixed Phenom out in short order is key, and creating a 45 nm product that not only capitalizes on the advantages of that process node but also adds performance per clock to the mix is key to AMD’s long term survival.