Ye Olde POS crazy behaviours on OC...

[mitch074]

Hi all,

short-time member but long-time tinkerer, one thing still baffles me about my old rig...
Description: MSI K7T266Pro2-A + Duron 950, 2x256 Mb of value DDR333 + MSI Geforce4 ti 4200.

CPU: 7.5*133 MHz (1GHz), RAM: 2-2-2-6-1 @ 133(DDR/266) - multipliers unlocked using the old pencil trick.
VID: Core 250@280, Ram: 500@560, Nvidia generic 56.72 drivers.

Do mind that I pushed this ol' lady enough to have it play Doom 3 on 'low' @ 800x600 (no other vid details disabled), and it was... well... I managed to finish the demo. I also managed to score a whopping 16950 in Aquamark 3 once with this. Using good timings but no OC, score was 14500-15000.

The problem: if I tried to reboot the system with the CPU overclocked, it would hang before POST - it needed full power-off then back on to restart. However it wouldn't hang for days on end doing DVD to DivX backups, it wouldn't crash and, actually, it still works fine.

Any ideas?

For those thinking I can't afford a better rig and would start to flame, let's just say that this is my OLD system...

 

[ChipDeath]

I would suggest a VIA chipset bug.

They had quite a few back then.

 

[mitch074]

...I hadn't considered that, but it is, indeed, possible.

My mind was more along a problem with the pencil trick: upon reboot, if the processor was still under tension, the 1V signal used to check the CPU's frequency was just too off to register, while if the signal was sent at start (while the CPU was powering up) it would register - and allow boot.

Why the measly 50 MHz increase? I'll just say that I actually pushed the FSB as high as 150 MHz and the CPU speed as high as 1150 - but then it would be too unstable for my liking, for a very small performance increase over 7.5*133. And yes, I tried adjusting voltages and timings.

 

[ChipDeath]

I suppose you could be right about the pencil thing, but I would expect it to be a bit less repeatable if that was the case... more of a random occurance thing.

But I've seen, heard of, and played around with countless VIA chipsets of that era, and they almost always had some problem or other, and never the same problem twice :lol:

Well... It's not like it matters much eh? It's not exactly a huge problem, and it's not even your main rig anymore... Reminds me of a PC my brother had (years ago... it was a Pentium 100Mhz or 90Mhz IIRC), which from a cold start would just hang before the post screen. You had to leave it to warm up for a couple of minutes, and then it would just spring to life and be fine... Can't honestly remember much about that PC, since I didn't know anything back then...

 

[mitch074]

Funny you'd mention it, my bro's rig had a similar bug: cold-started, it would crash after a few minutes, and once rebooted remain rock solid for days...

Another bug, that whe could isolate was located in a socket 7 Shuttle mobo's BIOS chip, was my CDROM not being recognized on first boot: on reboot, it would work ok, but before that, nothing - and we knew it was located in the BIOS chip because, having misflashed my BIOS once, I was using my brother's chip (he had the same mobo), we booted my mobo with his chip - and the CDROM was seen. Then we swapped the chips (hot swap-eugh!) flashed my chip, powered off, reset CMOS and rebooted: CDROM not recognized.

Good ol' days...

 

[durbanpoison]

someone explain what the pencil trick is?

thanks lol

 

[mitch074]

on the older ceramic packaged Athlons and Durons, multipliers were determined by 5 small bridges on the CPU top; out the fab, some of those bridges were laser-cut, and depending how the cuts were made, the multiplier was 6, 7.5 ... 15. If no bridges were cut (engineering samples), then the multiplier could be set in the BIOS.

Ceramic is non-conductible, and the contacts were made of copper; using a pen (which contains conductible metals such as lead), you can re-bridge those contacts so as to unlock the multiplier. For a better result of course, you could use a special pen used to draw printed circuits, but your run of the mill pencil worked great enough that, provided it was sharp enough, you could do it in a minute and be happy for 600-1000 boots.