| This week:
Monday
Tuesday
Wednesday
Thursday
Friday
Saturday
Sunday
TOP |
Sunday,
21 May 2000 An astonishing story
about cloning in mail.
Clearing up a confusion:
I was inattentive I guess: for some reason I got the idea that Socket
"Super Seven" and Socket 370 were the same. They are not, and
here is Bob Thompson on the subject:
No. Super Socket Seven is just a minor enhancement
of Socket 7. It's more a marketing term than a technical one. Super Socket
7 was basically an AMD initiative designed to support their K6-* series of
processors better than the older Socket 7 boards did. Super 7 boards
support 100 MHz bus (which some plain old Socket 7 boards do, too),
support AGP, have recent BIOSs that support large hard disks, and so on.
But SS7 uses the same basic electrical protocols, pinouts, and so on that
Socket 7 does.
Socket 370 is basically a socketed implementation of
Slot 1, uses the GTL+ protocol, and supports PPGA Celeron processors in
its first variant. The most recent version of Socket 370 is backwards
compatible with the original Socket 370 (that is, it supports PPGA
Celerons) but it also supports Socket 370 FC-PGA Pentium III and FC-PGA
Celerons (the "Celeron II"), which the original Socket 370
boards do not. There are actual pinout differences between the original
Socket 370 boards and the current versions.
Which should take care of that confusion. In case
it doesn't:
Actually, the boards you mention support SC242 (Slot
1) and Socket 370, rather than Socket 7. The Intel 440BX chipset indeed
supports 100 MHz FSB, UDMA/33, and AGP 2X. There are two similar VIA
chipsets at issue. The Apollo Pro133 supports 133 MHz FSB, UDMA/66, and
AGP 2X. The Pro133A is the same, but adds support for AGP4X. All of this
goes to show the danger of buying on numbers, however.
Nearly all benchmarks show that Apollo Pro133/133A
motherboards using a 133 MHz FSB are actually slower than a 440BX-based
board using 100 MHz FSB. When running a 100 MHz FSB processor (or a 66 MHz
FSB Celeron), the VIA boards are much slower than the 440BX boards.
UDMA/66 shows little or no throughput advantage
relative to UDMA/66 with even the fastest of the current-generation ATA
drives. The maximum throughput of a drive depends on whether data is being
transferred from the inner or outer tracks. Drives like the Maxtor
DiamondMax Plus 40 barely exceed 30 MB/s at their fastest, and typically
only provide 16 MB/s at their slowest, with perhaps a 22 MB/s average.
That's well below even UDMA/33. My best guess is that we're unlikely to
see drives that can actually benefit from UDMA/66 for at least a year or
two. Even when we do, the impact of UDMA/66 versus UDMA/33 will be a
relatively minor factor in overall system performance.
AGP 2X versus 4X is another non-issue. Current
systems and video cards typically can't come close to saturating even AGP
2X. AGP 4X shows no discernable real-world advantage in any situation I
know of. And in fact, the only time that AGP has any advantage whatsoever
over even PCI is when the video card uses system memory as virtual video
memory. You'll notice that most high-end video cards (the only kind that
could even theoretically benefit from AGP) usually come with 32 MB, 64 MB,
or more of physical memory installed on the card. That's because local
video memory is even faster than shared system memory accessed via AGP. So
the truth is that any flavor of AGP has no real-world benefit whatsoever,
and downgrading a motherboard because it supports only AGP 2X instead of
AGP 4X is ridiculous. And, yes, my opinion of VIA chipsets holds true
across the board. Given the recent news articles about Intel's chipset
problems, it's easy to forget that historically Intel chipsets have been
rock-solid, fast, and bug-free compared to competing chipsets from VIA,
which have always been a step behind Intel, particularly in compatibility
and stability. And VIA's problems extend beyond Intel processors. For
example, it was recently disclosed that problems with the VIA KX133-based
Athlon boards may have been responsible for AMD's decision not to release
their new Thunderbird-core Athlons in Slot A, except for limited OEM
distribution. Why? Because the KX133 has timing problems with Slot A and
the Thunderbird core. Even the elderly AMD-750 chipset can support
Thunderbirds in either Slot A or Socket A, so how did VIA, which
apparently based the KX133 on the AMD-750, end up producing a chipset that
can't run the Thunderbird in Slot A?
All of that said, I don't think that VIA chipsets
are junk, but I'd always choose a board based on an Intel chipset if I had
that choice. In my opinion, you'd be better off going with a 440BX based
model. It'll be faster and more stable, and you won't miss those
marketing-speak features that appear to give VIA the advantage. Although
it is elderly and being ramped down, the 440BX is still the best chipset
that Intel has ever made, and that's very good indeed.
-- Robert Bruce Thompson
thompson@ttgnet.com
http://www.ttgnet.com
Much ado about cloning article. See
Mail. But it was probably all a hoax. Sigh.
And things are coming apart in Rhodesia now known as Zimbabwe...
Link: www.sunday-times.co.uk/news/pages/Sunday-Times/frontpage.html?999
You have to find the WORLD index on the left side, then the Zimbabwe
story.
TOP
Current
View
Current Mail
|
CLICK HERE for
a Special Request. 
