KR> well i've given up, there seems no way that an amd5x86-133 will run
KR> with my motherboard. i've tried all combinations that i can think of,
KR> but it still hangs on boot with the chip set to a quadrupler. set to
KR> a tripler, it runs as you would expect exactly at the speed of a
KR> dx-100. i stiil dont see why, when a multiplier chip, afaik, should
KR> look just like a dx-33 externally. ah well such is life. ):

BG> On boards which do not specifically support the 5x86, the CPU will
BG> (or should) still run as a tripler, as long as your board is jumpered
BG> for an AMD DX4/100.  Later boards mention the AMD Enhanced CPU, and
BG> SOME of these handle the 5x86 as a quadrupler, as long as there is a
BG> jumper which physically switches the CLKMUL pin of the CPU socket to
BG> a low state (i.e. shorts it to the Vss plane).

BG> Pin 17 of the 5x86 is the CLKMUL (Clock Multiplier), and if left
BG> floating (or high), the chip runs as a doubler or tripler (not
BG> exactly clear from the AMD docs for the 5x86, but this depends on
BG> other jumpered inputs).  However, when physically pulled low, the CPU
BG> will run at 4x CLK, even on boards which don't support this chip or
BG> mode.

KR> i dont think that i have a problem getting it to quadruple,
KR> there is one link on the mb which put in one position causes
KR> the 5x86 to give identical performance to a dx4-100 (its doing
KR> it now to produce this). if i move it to the other position (it's
KR> function is not explained in the mb manual) the whole thing collapses,
KR> and locks on boot. using a dx4-100 the same positions make it run at
KR> 66mhz and 100mhz respectively. the problem is just that the rest of
KR> the board cant take the strain of the cpu running at 133mhz.

Sounds rather unlikely its actually the 'strain' of the cpu running
internally at 133MHz since that speed is entirely INTERNAL.

KR> ps. i tried another motherboard of identical model, jumpered as far
KR> as i could tell exactly the same. it still hung in exactly the same
KR> way with the chip set as a quad, but with it set as a tripler, it
KR> gave a landmark of 270 instead of 475 with my motherboard ??????

BG> Randall has the AMD doco for the 5x86 (it's around 2Mb I believe),
BG> which is necessary to figure out which is pin 17, and also which are
BG> the Vss pins (around 25-odd in all), and the idea is to use a tiny
BG> bit of wire on the CPU itself, before mounting it in its socket, to
BG> short pin 17 to the nearest Vss pin (makes it semi-permanent though).
BG>  The CPU will then run as a quadrupler, although the BIOS will
BG> generally report something different, as it knows fuck all about
BG> quadruplers.

BG> Randall tried this on the weekend, and although the BIOS said it was
BG> an AMD 486DX2-130 (must be as high as that particluar BIOS
BG> recognised), Landmark proved that it was indeed running as a 4x

{at} 160MHz (and reliably so as well).  If you want to try this yourself,

BG> I'd suggest you drop Randall Lane a message in either Z3_Tech or
BG> Oz_Modems (that should please Ian no end :)) and ask him for further
BG> details (like how to recognise pin 17, for a start).

KR> like i said, i dont think that this is my problem, i think that
KR> the chip is running at 133mhz internally, its just that everything
KR> locks up either just after the post (i suspect the aha2842)

It never sees anything of the 133Mhz internal speed tho does it ?
It only sees the external clock speed, which doesnt change.

And you can obviously check that to see if its relevant
just by leaving it out and see if it does atleast boot
up enough to start looking for system disks.

KR> or when the system box is displayed.

Cant see why it should lock up due to that. You do see some
motherboards have a bit of a brain fart about what they report
on the cpu speed in the display of the basic system params on the
bios status display while they are booting, but they boot up fine.

It does sound more like it just doesnt drive the pins on the
cpu properly in that mode and the cpu has a brain fart itself
when its allegedly jumpered correctly but isnt for 4X mode.

BG> The bottom line is that we think the 5x86 can be made
BG> to run as a quadrupler on the vast majority of 486
BG> boards which also support AMD DX2 or DX4 CPUs, and at
BG> just $60 ex tax for the chip, that makes them great value IMO.

KR> i'd agree that it is great value if i could just get the thing to run,
KR> i'd buy a new motherboard for $120 that i know will work, but that
KR> would mean getting a new video, and scsi board coz the mb is pci. ):

Yeah, bit of a problem. Looks like VLB is finally going down the tubes now.
@EOT:

---