Milo, thanks, a very interesting read indeed!
I am not as optimistic as this BlackMagicGuy, though. When he writes that "NW is a big disappointment. It is the P4 'done right', and it still sucks" I believe he underestimates the big progress Intel made with Northwood. Johan from Aces's Hardware showed some spectacular gains of NW in scientific applications. Many others claimed an advanced lead in good old Quake 3.
Some well-done, very in-depth reviews showed that the AXP 2000+ still keeps a tiny advantage over the Northwood 2200 but this is hardly the big picture which the general public gets. In the eyes of the general public Intel has regained the performance crown. Thereby, AMD lost a bit of its "cool" factor. Besides, Northwood is highly overclockable, making it attractive to many enthusiasts (the biggest AMD supporters so far). And last but not least, Northwood runs a lot cooler, that means less noisy systems.
Besides, that BlackMagicGuy is probably not right when he implies that this was Intel's last shot. He may underestimate the relevance of four points:
- Northwood may scale a lot better than TBred as Paul DeMone keeps predicting,
- Northwood will soon get a 533 MHz bus, giving it a clear performance boost,
- new, advanced fabs are coming online
- and there is still the "wildcard" of SMT for the desktop processor family.
OTOH, AMD is lucky that Intel's production problems seem to persist throughout Q1. As long as NW is not available in the screwdriver shops and the DIY market AMD's home market is not in danger. If really (and I mean really *if* <g>) NW is indeed not widely available now INSPITE OF ITS DELAYED INTRODUCTION (which should have given enough time to Intel to stockpile the product) then something definitely does not go right at Intel.
In contrast, the Athlon XP 2000+ seems to be a typical (!) high-volume launch for AMD. At least here in Germany it is widely available right now, be it in big computer chains (for example Vobis) and even in small scewdriver shops (see litec-computer.de - my favorite shop in Munich).
Continuing low availability of NW may even let some of the big OEMs turn back to AMD again. That may have been the reason for the recent e-machines design win. We may also see an increased presence of AMD based PCs in the US retail market. To me, it seems that there is a good chance that Q1 is more or less in the bag for AMD, mainly because of Intel's production problems.
For Q2 and beyond, the big questions are, IMHO:
- How well does AMD's transition to 0.13 micron go? Will AMD have less difficulties than Intel?
- When and in which quantities will the mobile TBred appear on the market?
- How well will TBred scale? Will AMD have an answer to the forthcoming 533 MHz NW?
- When will SOI mass production start? How much performance improvement will it give to TBred?
- Is there any chance of ever getting a better share in the corporate market?
- And of course the "Mother of all Questions": When will Hammer be launched?
- Finally: Will Microsoft support X86-64?
- Oh, and I almost forgot: When will flash finally recover?
If we get another "AMD will be roadkill soon" prediction by some analysts the stock may tank again. I do hope, however, that due to a positive Q4 earnings surprise and a very strong outlook at least for Q1 the stock should do quite well in the near future. (Let's hope that there won't be many "black days' like yesterday again...)
Any comments?
EDIT: I should mention that in my opinion the best move for AMD would be to release a TBred MP with 512 KB L2 cache (while leaving the ordinary TBred with 256 KB L2 cache) and launch it in rather high volume. By this, AMD would not only make big inroads into the multiprocessor market because (as Johan has shown) an increased cache helps a lot in scientific applications. AMD would also be able to claim a *clear* performance superiority over any competing chip (the "cool" factor, which is quite important for public mindshare).
Such a chip would attract all the enthusiasts (I would buy it in a heartbeat...). It should therefore sell really well and by that substantially increase ASP. The OEMs and the cost conscious buyers, OTOH, will be happy with the cheaper "ordinary" TBred.
Peter |
