Epox and Overclocking

[Sol]

I am having a problem with an Epox board, its a 8KH+ using the VIA 266A chipset and I'm running an Athlon 1700+
Its multiplier unlocked and runs fine at 1533(1800+) but for some bizzare reason whenever I increase the multipler beyond 11.5 it goes back to running at 1467. If I keep upping the multiplier it will go back to 1533 at some point but then goes back to 1467 on the next incriment. I know it can run at at least 1900+ because I managed that by inreasing the FSB. The problem is that I can't go very far overclocking that way because my GF3 is a bit finicky about the bus speed if I increase it more than a couple of Mhz. I'm pretty sure that cooling isn't a problem because I just installed a water cooling system and even at 1900+ the temperature only just hits 36C under load for about an hour. With this kind of cooling I really think I shoul be able to get this thing over 2000+. I may be dreaming but I am pretty sure that somthing is wrong here.

 

[Tannin]

Hi Sol, welcome aboard.

The first question I'd ask is this: is the clockspeed actually dropping with increased multiplier settings, or is it just that the BIOS mis-reports it? If you haven't already done this, I should imagine that most any simple brute-force CPU speed test software should be able to answer that. In the old days I always used to use Norton Sysinfo. (i.e., si.exe from the Norton Utilities for DOS). Not sure how that would cope with a ~2000MHz chip though, it was designed to tell you if your Cyrix 486 was faster or slower than a DX-33!

 

[Mercutio]

My game machine runs on a Tbird 1.33GHz running @ 1.5GHz, on an Epox 8KHA (no RAID, same board). I've pushed it up another half-step without difficulty, BIOS reports fine. I'll try it again, just to make sure, though.

 

[Sol]

Yeah I have tested it with a benchmark and it is actually running slower on the higher speed setting, it also reports the multiplier as 11 and the Mhz as 1467 so it is definatly not just a misreport- when I go into bios though it does keep the higher setting recorded, it just doesn't seem to use it except for 11.5.

 

[James]

Make sure you're running the latest BIOS, for a start.

Also check any DIP switches or jumpers on the board that may be stopping you.

I've heard that some boards don't support multipliers above 11.5, whether this is a BIOS or a design issue I'm afraid I don't know.

Hey, Sol, since you're in Ballarat you can go and see Tony (Tannin) in his shop.

 

[Tannin]

The problem, James, is not getting Sol into my shop. The problem is getting the bugger out of it!

 

[Tannin]

Oh, and it's no use telling him to ask me to help with it. When it comes to video cards and cooling and overclocking, Sol is much better informed than I am. I just take his money, order arcane bits from around the country for him, and occasionally remember to offer him a coffee.

We could try a different main board, perhaps? I have Solteks and an MSI in stock, plus the new KT-333 Soltek arriving any day now.

I keep wondering if there is some kind of multiplier stepping happening here - i.e., that either the board or the CPU interprets multiplier increments according to a repeating table. Remember Socket 7 boards? There, the jumper settings were quite limited, only two or three jumpers, and the CPU was responsible for "knowing" if setting X was supposed to mean 1.5x or 3.5x. If this is what is going on, it would explain why, when Sol bumps his multiplier up, the actual multiplier drops.

Of course, there is a simpler solution: seeing as a higher FSB speed is a problem, just get rid of that obsolete old video card and put something modern in. :mrgrn:

 

[Sol]

Right as usual Tony, but since the TI 500s don't look like the best value for money and Radeons still seem to have those little driver issues I might have to wait for the GForce 4s to be released before I give the old Phrophet 3 the flick. Meanwhile the KT-333 sounds promising. With the RAM running at that speed it might well be on a completly different clock to the AGP and so that ancient video card might be able to run at the speed it likes with the RAM and CPU wherever I'd like.
Hmmm sounds dangerous too me.
I like it.

 

[flagreen]

Just a wild idea which is probably way off base Sol, but have you noted your onboard voltages when this is happening? Any significant dips in either the + or the - readings when it slows down?

 

[Sol]

I'll have to look in to that. I did try pumping the voltage up as far as it would go before doing it, but now that I think of it I never actually checked what it was afterwards.

 

[LiamC]

Bastich! How did you get an unlocked AMD chip? If you unlocked it yourself, you may have a slight short that keeps jumbling up the FID pins. I'd try redoing your bridge shorting.

On pencil unlocked Tbirds this was a common problem if it wasn't done right. You'd get one or two multi either side of the default and then the chip would revert to slower settings.

You could test this theory out with an AthlonMP chip if you know someone with one.

If you have a factory unlocked sample and you're unhappy with it, I'll take it off of your hands for say half price?

 

[Sol]

Actually I unlocked it with silver paint and some glue - pretty much the only way with XPs. The shorting thing was probably the case anyway because when I redid the unlocking it worked fine. Of course that means a total speed increase of 24Mhz without crahes.