OPTI was the mfg of choice for the 5x86 and 6x86, and they made decent enough
chipsets for both the Socket3 and Socket5 markets. UMC made decent enough
chipsets supporting the 5x86 as well. You're right, of course, in that you say
the VLB is a '486-era concept, though it was applied to late '386's. I've
found
that the boards, be they socket3, or socket5, that have VLB on them whether they
have PCI as well or not, have all been MUCH more dependable than the PCI-bearing
socket7 and later boards with which I've had experience.
Now, just because you didn't like the 10 or 20 boards that you had doesn't mean
that they were all rubbish. There were so many one couldn't possibly try them
all, even with Bill Gates' money.
I don't see the point in your comment regarding the 6x86 boards. I clearly
stated that the 6x86's had no VLB, and that was the point of my remark, since
the 5x86's had both PCI and VLB and yet didn't suffer much from the presence of
both interfaces. They were always quite impressive for their cost. I recall
that the last time I bought a dozen or so, I paid under $100 for VIP boards with
all the common peripherals on the motherboard plus one VLB slot, 3 ISA's and 3
PCI. These had a 5x86/133 with 256KB cache and 16MB of EDO RAM. Of that last
batch, all are still in daily service. I subsequently bought other 5x86 mobo's
without the VLB slots but with 4 PCI, and found them not one bit faster than the
boards with VLB on them, though of the 6 or 8 that I bought back then, only one
is still functional. Though my own sample isn't sufficiently large to verify a
trend, but with the evolution of the PCI, the failure rate I observe among
motherboards is about 10-20x what it was back in the VLB and earlier ISA-only
days. I've got dozens of dead PCI mothers, yet almost all the VLB and earlier
mothers I've owned or used are still functional, though many have had the cache
srams scavenged for embedded applications.
Sooner or later they'll build motherboards with the entire virtual address space
filled with nonvolatile RAM, with built in fiber or 10GB ethernet, and with all
the other stuff we need built on as well, and then the circuit can be designed
optimally for the, by then 10^100 Hz CPU's. In the meantime, we're at the mercy
of the geniuses who design the bus interfaces, particularly the software.
Dick
----- Original Message -----
From: <jpero(a)sympatico.ca>
To: <classiccmp(a)classiccmp.org>
Sent: Saturday, October 27, 2001 6:08 PM
Subject: Re: VLB SCSI?
> I'm not sure what you mean by slower. I found that no matter how fast the
mfg
> claims the CPU is running, it does DOS tasks
about as fast as a 133 MHz
Pentium,
> and I also found that the difference in
performance between these VIP boards
at
> 133 MHz (not for comparison, but just for
reference) and the 150 MHz 6x86's
that
I was using
back in '92..'94, was not perceptible. The 6x86 boards were
equipped with ISA and PCI, but no VLB, BTW.
Hello? Knock knock...
There was pentium VLB w/ ISA, other pentium has ISA ONLY, and from
what others have seen had or have all three (PCI, ISA and VLB) but
keep in mind, VLB is strictly 486 signal and to some degree 386dx,
pentium pinout isn't compatiable w/ it therefore has to go through
chipsets to work properly. Many of these early rubbish pentium
boards has these features and works w/ one 72 pins to keep cost down.
Remember, at that time cost was expensive for a 1~8MB 72pin
simm (93-96 era).
(Feb 96, paid 700 CDN for pair of EDO 8MB simms)
But these sane quality designers don't use these mish-mash that
chinese cloners (5V pentium era) tend to do in that early times.
Pentium w/ ISA and PCI, while 486 got VLB and ISA later on 486 late
ones got PCI (better).
Important thing, OPTI is ok but that's unknown chipset and most of
cloners didn't have resources to optimize the performance and
reliablity, compatiblity so that OPTI got black eyes, ditto to good
chipsets by SiS (460, 461, 471, 496/497), even intel so on!
> I suppose someone will eventually build a motherboard that runs fast enough
so
> that when the PCI peripheral interface fails it
still saves time to power
down
the system,
swap the offending device, and restart, but that time isn't here
yet.
Yes, already here now. Look to Serverworks stuff, PCI is already hot
swappable and available w/ 66MHz and 64bit. Also microPCI is hot
swappable as well, primarily used in racks and true server machines.
> It's true that it's possible to build a faster interface on PCI than on
VLB.
> That doesn't mean that the makers of a given circuit did that, however. I
doubt
> that a motherboard designed with both VLB and PCI
is necessarily slower than
one
with PCI only.
Good board makers can make without performance losses on both PCI and
VLB on board and still give top performance w/ 486 type processors.
We had Asus boards of those.
Dick
Cheers,
Wizard