facebook rss twitter

Review: Pushing the Celeron 400 into a new era of Speed

by David Ross on 15 May 2000, 00:00

Tags: Intel (NASDAQ:INTC)

Quick Link: HEXUS.net/qa3

Add to My Vault: x

Overclocking

It was a few months later, I remember the Celeron had originally P.O.S.T.E.D at 600MHz. Could I repeat this? Could I make Windows boot without totaling the registry? I spent a few days trying, but it was no use. Although Linux would boot and distributed.net would run for hours, the X window system crashed instantly. Windows totaled the registry on just about every boot (lots and lots of re-installs). There was one day, when I managed to do almost an entire Quake2 time demo before it crashed. "OK, Dammit. I am gunna get this punk to 600MHz". I knew the CPU is capable, just not being given enough current. How did I know? Well, distributed.net uses very little memory access. It all runs from the L1 cache => the power available at 2.05v (the board is always over by 0.05v) is enough. As soon as something memory intensive starts, or that uses all sorts of bits of the processor, it all went rather Pete Tong.

I thought I had a cunning plan though. I tried a slocket with voltage adjustment, rather than dump the chip straight on the board and let it auto-detect the voltage. Ahh yes, clever in theory. I tried several slocket, and not one of them was as stable as the on-board socket, even with extra juice (I went up to 2.4v to try to prove a point). I think that's why they put a Socket 370 on the motherboard, as well as a Slot 1. To be fair, it's not the Slocket that's bad, its the Socket370 on board that was very stable. I certainly wont bother with slocket in the future.

I was so convinced of the CPUs ability, I went and threw even more money at it. Now people, throwing good money after bad does not fix things, except in this case when I was right and the saying was wrong. So there :P. I purchased a shiny new Elitegroup Motherboard (VIA Apollo Pro 133 Chipset, all the way to 150MHz FSB) and a new case (the tiny AT case just didn't cut it at LANs, I felt left out). In one hour, it was all built up and ready to roll.


Leaving the chip at 400 to start, tentatively I pushed the starter. It fired up, hard drive humming sweetly. I eased it into Windows, and put my foot down by running distributed.net for a few hours. So far so good. Huh? What? This isn't Top Gear? I haven't blagged a Cerbera Speed Twelve? Oh, OK, I'll behave ...

Your best tactic is normally to bring the CPU speed up a notch at a time and let it run for a few hours at 100% to let it settle. However, 75 and 83MHz disagreed with the Savage4 rather violently, so I had no choice but to go for 600Mhz, straight off. A little extra juice was applied (2.15v) and I fired it up.

*BANG*. No, not the PC exploding but my jaw hitting the desk. The PC rocketed though the Windows boot, ripped through distributed.net and promptly ran everything I threw at it. Well, everything apart from IE5. Internet Explorer has been known to crash a lot on some systems, but this was every other website. Damned Microsoft products (grumble, grumble). So close, and yet, still totally unusable. I braced myself for a small inferno and, a few BIOS settings later, 2.25v was on the cards. Sorted. No more IE crashes :)

Well, it ran perfectly for around two hours anyway, then the alarm bells starting ringing. Literally. The GL520m sensor utility probe had been quietly measuring the core temp all along, and had now decided that a 50 degree core was too much. Cursing, I shut it all down and tried to think for a bit. "What could I use to get more air through this HSF?". I spied some old CPU fans lying up on a shelf. "Perfect" I though, as I taped them in various positions around the HSF. Oh, tip for the day. If you put insulating tape down on the board, where you want the fans, you can then stick fans on the board, parallel with the HSF using double sided tape. Sadly, it was almost but not quite enough. The poor HSF really wasn't designed to dissipate the kind of heat, even with all those fans :(. The core temp shot up to around 42 degrees C and refused to drop. Now, this is OK, but not really what I was after for a PC on 24 hours a day.

With a sigh, I paid out yet even more money for a new HSF. After much deliberation in So here I am. Sweltering weather, a 35 degree core and not a single CPU-related crash. The Savage4 however .... look for another article in the future called 'How to destroy a Savage 4 in a fit of Rage', and the follow up 'Barbequing S3 and VIA driver technicians'.