Professor PTNEWELL says....
1st Post:
Yep, in the 18 months I've been closely following the DRAM wars, predictions of
victory for one's own side and defeat for the other have not amounted to much.
RDRAM and DDR both continue to ramp, neither as fast as its adherents hope, but
faster than the other side figures it will.
I seldom read S.I. (although it is gratifying that to discover Bilow feels obliged to keep
such excellent accounts of my YHOO postings -- which cannot easily be recovered by
searches). Of course by now Bilow, despite her non-technical background, should be
aware that my predictions of electromagnetic noise problems associated with DDR 266
proved prescient. However they also proved exaggerated, in that the work-arounds
took only a few months. Specifically remind Bilow that even the cream-of-crop, the
AMD 760 experienced the problems I predicted on basic physics grounds:
techweb.com
"When Micron first started evaluating the systems, the first AMD-760
boards used in the Millennia XP were essentially samples cobbled
together by hand, the spokesman said. But when Micron started
ramping volume shipments, the company found that the boards were
unable to meet the noise tolerances required by the faster 266-MHz
front-side bus speed used by the AMD-760 chipset. Additional
hardware filters were required, and a new six-layer motherboard with
improved tolerances was approved Tuesday."
So the noise problems were real. Those, like Bilow, who argued that the SDRAM-->
DDR transition would be effortless, were wrong. It took longer, cost more, and had
noise problems. These problems cost AMD, VIA, and others some embarassment and
credibility:
dailynews.yahoo.com
.html
theregister.co.uk
As originally reported by The Register over a month ago (see AMD 760 DDR chipset
delayed), Jan Gütter, an AMD spokesperson in Germany has confirmed that Micron PC
is delaying shipment of its 760-based DDR systems with a 266MHz FSB.
The problem stems from noise and the spokesman said it could be avoided by 'very
minor' modifications to mobos. Although he declined to give specifics of
these mods, they probably involve either filters and buffers and possibly even the
removal of a DIMM slot - precisely what happened to Intel's RDRAM
CaminoGate VC820 mobo.
"Systems running at 200MHz FSB are unaffected, said the spokesman, who
couldn't tell German site Tec Channel where 760-equipped PCs or mobos could
be purchased in Germany, adding that there would not be any available before
January, despite the company previously claiming they had been available
since October."
In fact the DDR noise problems threatened to turn AMD and DDR into laughing stocks:
heise.de
Speaking of the Via’s KM133 for the Athlon:
"In Germany, neither the boards nor the chipsets are available; and nobody can give any
definite date. This follows the recent trend of AMD and Via for "virtual product
releases". Likewise, neither the AMD systems with DDR, announced several weeks ago,
nor the supposedly "easy, problem free" DDR modules themselves are anywhere to be
seen."
Yes, they did solve these problems. It is still true that most boards are avoiding the use
of more than 2 DIMMs, and that the noise problem is permanent, and can pop up again
and again in unexpected ways. But sure, it clearly will not be a s
===================================
2nd Post:
Hmm, part of my message was obliterated. Anyway, after documenting that the
predicted noise problems with 266 DDR were real, and pointing out that such
electromagnetic noise problems tend to crop up again unexpectedly (i.e., when adding
DIMMs, or upgrading processors), I admitted that they will not be the show-stopper
for DDR I hoped they would be.
Meanwhile, Ms. Bilow might wish to review the accuracy of her own predictions:
(1) In Spring 2000, when RDRAM production for desktops CPUs stopped at around
1 million/month, she started a "Rambus is toast" campaign. She thought it was all over
then, and RDRAM was not long for the world. In fact RDRAM soon thereafter
captured most of the workstation market, and has continued ramping up ever since.
(2) Bilow predicted that DDR would be established as the big winner in desktop PCs
by summer 2000. Oops. DDR managed to make it into exactly zero PCs by summer
2000, or even fall 2000, despite many vaporware announcements. You cannot be
further wrong than Bilow was there, unless she redefines "winner" to mean "zero sales."
Those noise problems proved real, making the design much harder than the
technologically illiterate, such as Bilow, supposed. And DDR has continued to
disappoint everyone who supports it, from motherboard makers to the disgruntled
techies who wonder why the bandwidth in reviews by previously favorable sites like
Pabst and Annand falls so far short of what RDRAM delivers.
The biggest acheivement AMD has had with DDR is with their slogan:
"Powered by DDR" which, at least for now, has become the world's most seldom seen warning label. |
