The Celery Report: Issue #4

This week, despite my misgivings about the overclockability of the CPU, I just couldn't fight off the urge to grab a 400A and give it a workout. This was the Slot version (of course), OEM, COMPEQ, normal PCB color with a PCB date of 9847 and label code of 08511565 (Week 51). The sSPEC number on these CPUs is SL39Z. This chip is clock-locked at a multiplier of 6, (6x66=400).

The processor is visually identical to the 300A. The silicon is the same, the square substrate is the same and the PCB is the same. The only way to tell the difference is the white label on the back. I ran my tests with the following setup:

Windows 95 OSR2.1
Trusty test BH6, flashed to HN BIOS to support the processor
Single stick 64MB Samsung GH memory
Trident 975 4MB AGP video card
Junky old 270MB Quantum hard drive
(Side note on these drives: I have three of these old geezers that I use for testing these setups and I can't tell you how much I've abused them and how many times I've trashed them. But, they keep coming back for more, even at 75Mhz, 83Mhz and even 112Mhz!).

I put the chip thru a quick round of Business Winstone at the default of 400Mhz, just to insure basic functionality. No problems here.

I then set the FSB (motherboard speed) at 75Mhz with the core voltage set at default of 2. This produces a CPU speed of 450Mhz (6x75). She breezed thru 6 rounds of Winstone without a glitch. Okay, I thought, so what? A 300A with a good core would do exactly the same thing.

Cranked up the FSB to 83Mhz, again leaving the core at 2v. We're now at 500Mhz (6x83). She booted lickity-split right in to Windows. I'd be lying if I said I wasn't getting a little excited at this point! She went thru the first round of Winstone without incident, but crashed half way thru the second round. After a reboot, it locked up at the desktop.

Must be heat, I thought. So I pulled the "cartridge" heatsink/fan that I use open, and removed the tape covering the thermal transfer material. Slapped it back together and booted her back up, still at default voltage. After that the little puppy zipped right thru 6 more rounds of Winstone with nary an error.

Now, before we all start jumping for joy, I would point out that a 300A with a superior core will do exactly the same thing. I would also note that, although the CPU was fine at 500Mhz, this was done at 83Mhz FSB. This bus speed puts the PCI bus at 41.5Mhz, which is substantially overclocked, and the AGP bus at 55Mhz (2/3 of 83), which is underclocked. Some PCI peripherals simply won't handle this, and an AGP video card would not be operating up to it's full potential.

Okay, now for the real test. With heart pounding, and hands shaking, I set the bus at 100Mhz (6x100), 2v. Fired her up; black screen, not even a video signal. Upped the voltage to 2.1, same result. Upped the voltage to 2.2, still staring at a blank monitor with no video signal.

I moved the voltage up to 2.3 and finally, a little action. Very little. I was able to get a signal to the monitor and a brief glimpse of the video card logo, before she died. No beep, though. I disabled L2 cache and tried again, same result. Then I also disabled L1 cache, still same result.

I put the CPU outside for a little while at 35 to 40f. Brought it back in and got a beep this time, but then blank. It looks like the core, at least on this chip, is simply not of a quality high enough to run at 600Mhz without overheating very quickly. It is my opinion that this will be the case with most 400A's. Frankly I'd be surprised if more than 5% will run at 600Mhz without the aid of a Kryotech-type system.

However, I do have a Glacier 4500C that'll be coming in sometime this week. This heatsink/fan combo is a 1.2 pound monster. Being the stubborn Geek I am, I will be trying again!