Elmer: Sure fyodor. You must be right. The foils say so.
Stop playing stupid.
This has nothing to do with foils and everything to do with facts.
Name a product transition in the last 10 years where AMD had as much cash on hand?
Name a product transition in the last 10 years where AMD had as competitive a product?
Name a product transition in the last 10 years where AMD had as exciting, proven technology? (HyperTransport, which in case you hadn't noticed has sold in millions of chips now).
So, this has nothing to do with Hammer and everything to do with AMD (as compared to Intel).
As for product performance, I think you need to take a look at where we are right now. P42200 vs XP2100+
Application P4-2200 XP-2100+
Quake3Arena (demo1) 272.2 fps 256.9 fps P4 wins by 6%
Quake3Arena (NV15) 60.5 fps 62.9 fps XP wins by 4%
3DMark2000 10889 11749 XP wins by 8%
3DMark2001 8411 8774 XP wins by 4%
Lame MP3 encoding 136 s 109 s XP wins by 20%
XMPEG DivX 34.86 fps 33.95 fps P4 wins by 3%
Sisoft Sandra2002
Dhrystone 2720 2412 P4 wins by 13%
Whetstone 4236 4816 XP wins by 14%
MultiMedia INT 10795 11123 XP wins by 3%
MultiMedia FP 8751 9459 XP wins by 8%
Memory INT 2508 2009 P4 wins by 25%
Memory FP 2518 2076 P4 wins by 21%
Lightwave 7b 257.7 s 362.0 s P4 wins by 29%
Cinema4D 247 s 221 s XP wins by 11%
3DStudio MAX 165 s 154 s XP wins by 7%
SPECviewperf 10.31 fps 10.74 s XP wins by 4%
WinACE archiver 210 s 218 s P4 wins by 4%
MPEG2 PinnacleStudio 96.1 s 93.7 s XP wins by 2%
SysMark2002 216 188 P4 wins by 15% AthlonXP 2100+ wins 11 of 19 (or, alternatively, 5 of 9 if you discount the synthetics).
Not too hard to understand Tom's Hardware Guide's conclusion:
The Athlon XP 2100+ Outclasses the Pentium 4/2200
Considering that AMD is more than competitive right now (and this is with Intel on .13µm and AMD still on .18µm, even if you do claim that it's a hybrid), AMD seems to be a very solid position.
Even better, if you look at the results above a bit closer, you'll notice that all but one of AMD's losses (the one being the WinACE archiving test) are do to either lack of SSE2 or sufficient memory bandwidth (or a combination).
If AMD really believed there was a problem with Hammer, they could just work on solving these two issues.
Number 2 is easily solved by equipping the Athlon core with a HyperTransport interface. This is proven technology, with millions of chips shipped (primarily XBox) and chipsets would be very easy to do as well (a slightly modified NVIDIA 415 (or 615) would do the trick, for example). Another upshot (in addition to the lower latency and higher memory bandwidth) would be that motherboards would be cheaper, do to fewer traces and lower pin-counts.
SSE2 may or may not be so easy to do, at least in such a way that it is very fast. The floating point unit on the Athlon is more than capable, but SSE2 registers would need to be added, along with changes to decoders and such. However, considering all the work done implementing SSE2 on Hammer, it's quite likely that this could be accomplished relatively quickly for the Athlon.
Add another 128kB L2 cache (to roughly match Northwoods total effective cache) and AMD would be well set for quite a while.
So, you see, AMD doesn't need Hammer… At least not nearly as desperately as you claim ;-)
-fyo |
