I spent a good 6 hours today learning about and testing (mostly the latter, reboot after reboot) overclocking. Experience has always been my best teacher – although I would not say it is always the most fun of instructions.

For starters:

After reading a slew of google hits on overclocking and sometimes using the terms above, I learned a few things.

  1. Heat is bad.
  2. Good to have the RAM-to-FSB ratio by 1:1
  3. Need more Core Voltage for the CPU to reach the really high speeds
  4. Sometimes as the FSB increases, one has to make the CAS timings “looser”
  5. Using tools like TAT, CoreTemp, CPU-Z and Orthos seem to be common to the overclocker crowd

So first I had to think about heat and airflow. I did a bit of research in the days prior and found a nice cooler that I like that will eventually cost me $35 ($10 discount + $15 mail-in-rebate). I am a bit cynical on how much heat it will actually displace. I got to thinking about convection and heat transfer and air flow. My case currently has two fans that force air out of the box on two non-opposing sides, a large vent right over the CPU that allows air to come in, and a stock fan that whirls that incoming air onto the heat sink and aluminum fins. If the outgoing vents get clogged up with dust (which happens to a greater degree than I imagined, than you have practically zilch new cold air coming in. Which is bad. Not only does my CPU generate heat, but I have four hungry harddrives (two of which are quite warm), and gas guzzling video card and memory sticks. These all contribute to a general state of “not cold”. All that heat building up inside the case makes the heat fins and CPU stock fan rather useless. So I took off a side panel and suddenly the average CPU temps dropped 6 degrees Celsius. How is that for a heat sink! 🙂 I also reduced the case fan speeds to their lowest settings and am thinking about just removing them completely (reduction of noise).

Most of my time was spent tweaking one of the many parameters in my BIOS and rebooting to see what kind of effect it had. Sometimes my computer would not boot, sometimes it crashed while loading Windows, sometimes it crashed shortly after loading Windows, and sometimes crashed during a test of Orthos. Very few times everything worked smoothly. Specifically at the very beginning before tweaking, once in the middle during an “AHA!” moment, and then at the end when I finally gave up and took some screenshots of where I am stopping today.

It doesn’t help that it was incredibly hard to match up my BIOS with things I was hearing online and in the forums. For instance, there is no setting for the RAM:FSB ratio (and to be even more accurate, I believe it is the FSB:DRAM ratio). On top of that, my BIOS has a setting for the DRAM clock speed, but the absolute lowest value is twice that of the FSB. So I was extremely confused. During the first half of my testing, I was noticing that my FSB:DRAM ratio was consistently 3:5 if I left the BIOS setting as AUTO. And then I finally stumbled upon an article that explained how the Ratio halves the value that is specified in the BIOS. Well, if only the BIOS writers could make that a bit more clear!! So I finally found that if I set my DRAM clock to FSB x 2, the sun shone upon me.

That was my half way point; I had the FSB up to 320 MHz, 60% above the base clock speed of 200 MHz, taking the CPU clock from 1.8 GHz to about 2.88 GHz. Not too shabby at all. But I wanted to push 3 GHz. Especially if I have a nice fan coming in, I should be able to push the envelope a little and find out what happens.

I quickly hit a wall, though. I could do a FSB of 333 MHz (which comes in at just shy of 3 GHz), but a FSB of 343 MHz just gave me all kinds of problems. It was frustrating. I could do the FSB at 333 and alter the DRAM to 666 and leave everything else at auto and have it working. But 343…. no way, man. I tried to change the CAS latencies, I tried to add more power to the CPU, I tried to play with things that I had no idea what they were (FSB Termination Point Voltage, Row Refresh Cycle, PCIE speed). I rebooted, tested, rebooted, crashed, rebooted, didn’t POST, rebooted…. again and again. After a while I was cursing Orthos for not telling me why it was crashing (or windows for that matter). The online documents told me that I should add more power to the memory, but my BIOS has locked it at Auto, and no amount of Googling or phone calling could land me a practical way to unlock it (tried, Ctl, Alt and Shift F1 in many different combinations).


So I called it quits with 333 MHz. I am not displeased with that. I just want more. MOAR!!! 🙂


Orthos had me up to 62 Celsius after a few minutes and I don’t want that to happen often. If my fan does not keep it down closer to 55 under load, I am going to return. Or throw it out the window.

I am rather convinced CPU-Z has errors; Allendale? 1.288V??
I am glad I was able to keep the timings "tight" at 333 MHz
62 degrees! Not awful, but not something I want long term