The pictures are all shrinked down a bit. If you save the pic locally and look at it later it'll be bigger. You may also right-click on the image and choose "view image"
This is my first serious overclocking project. What I did before was pushing a P120 to 133 and after that I clocked my P200MMX to unbelievable 225MHz!! You get the idea of my experiences... I've already collected much knowledge in overclocking and hope that I'll be able to push my Athlon to somewhat around 850MHz...
CPU Before Overclocking
The Processor is an Athlon Classic 500. Produced in week 49 in 1999. The Case is marked as a 500MHz but the core is 650MHz, made in 0.25um process. The default voltage and speed are 1.65V and 500MHz.
Stripping the CPU
Many people say, to get rid of the plastic case of the Athlon you have to take a tiny screwdriver and wiggle around somewhere inside the case. I thought that's too risky. I took a pliers and "removed" the sides of the plastic case. After that, I could see the pins, where the cover is plugged onto. I took a small wirecutter and cut off the part of the plastic cover that was over the pin. After doing that on all four pins I could take the cover away. The Athlon never run with the thermal plate. I stripped it and screwed it onto my selfmade H/S.
The Motherboard I bought is an EPoX EP-7KXA. It uses the VIA KX133 chipset which allows to set the RAM clock to either default FSB clock or to FSB * 1.33.
Due to earlier overclocking attemps I soldered several things on the motherboard. The PLL-IC, a IC-Works (Cypress) w210, is able to provide clocks of up to 200MHz (what equals 400MHz DDR). The chip has a I2C interface. Using this two pins of the chip you'd be able to clock the EP-7KXA to 200MHz. The chip also has 4 pins which are used as inputs during power-up and are outputs after power-up. I first soldered 4 jumpers to these 4 pins. But I couldn't get more FSB settings than I had in the BIOS anyways. I decided to disconnect the I2C pins from the motherboard and try again. Nothing changed. I disconnected and reconnected the I2C pins several times and now I have either a short-circuit between these two (clock and data) pins or the pin's gotten loose. The pin is still on the case of the chip.... I'm waiting for a replace part at the moment.
Unfortunately I seem to be unable to get one of these parts, anywhere. Please help!! here
The EP-7KXA is a very stable motherboard. I like it...
AMD K7 500
1 stick of generic PC100 RAM
2 IBM DPTA GXP 20.5GB
I mentioned a selfmade H/S... I don't know where I could get an Alpha cooler for example. So I decided to build my own little cooler. I took a heatsink I had laying around and let a friend cut it straight. The heatsink has a size of approx. 16*12*4cm. I've held the CPU onto the heatsink and marked where to drill the holes to fasten the CPU. And I screwed the CPU onto the heatsink with 4 screws.
On the ends of the Athlon PCB I've put some paper between the PCB and heatsink to prevent the PCB from flexing and breaking if I screwed the parts together.
I mounted a 120mm 230VAC fan onto the heatsink. But somehow it was just too loud. I took two fan grills and put some filter material in between and screwed it all onto the fan. I sealed everything around the grill with this cool yellow tape to force the air trough the filter. All in all it was not as loud as before but still not quiet enough to sleep next to that comp. The filter also prevents the H/S from getting dusty. At least it takes longer until I have to clean the H/S again...
Recently, I removed the mains fan and mounted two 60mm fans instead. One of them is an extremely quiet 5V Micronel fan. The other is a 24V NMB fan. The NMB runs at 12V. At 24V the fan has an unbelievable troughput of air and all that air flowing trough the fan causes a lot of noise. At 12V it runs very quiet and blows enough air too. Both fans are a little pricey if you'd buy them. I had the luck to get them free in the company I work. Same for the HS and mains fans and.... many other things :-)
The pictures below show a few of the fan configurations I've had.
With the 230V fan the mobo/CPU were at around 27 / 34 degrees Celsius. With the two small fans the system came down to around 26 / 30 degrees Celsius. All that with open case and under full RC5 load. Now the case is almost closed and the temps are 36 / 37.
I had no GFD until now. And this part rocks! 3-way dip switches and no need to supply power to it. That's a dream GFD! It's size is 3.5 x 4cm
I mentioned this paper I used between the PCB and the H/S. I thought I'd rather make two small blocks to put in between the cache chips and the H/S. It took about 15mins to cut a sheet metal to fit and sand it to get a fine surface. The photo was taken as I checked if the blocks touch anything. They didn't...
I applied a ton of thermal paste and screwed it all back together.... ready to rock!!
The chip initially ran at 500MHz / 1,65V. I've set the GFD to 700 and 1,7V. Switched on, post, boot, Windows, Unreal Tournament... Cool, it seems stable. I went further to 750MHz: Switched on, post... and then this character. The HD began to start working and then I got that arrow pointing down. The corevoltage played no role... What the hell is that??
Well, I turned out that either the mobo or the CPU wanted to say that I have to take an other cache divider. At 700 the cache was at 350MHz. 50MHz over the 300MHz a 3,3ns chip can reach. Okay, I resoldered to 2/5 cache. and tried again: post, boot, windows... stable! I'm getting closer to the speed I wanted to reach: 850MHz.
I resoldered the cache to 1/3 and set the GFD to 800MHz. And this was stable too. But the USDM proggy, which reports fan-speeds, voltages and temperatures... told me, my 5V has gone down to 4,8V. In Windows the computer was stable at all. I started UT and as it came to the intro, I could hear the music but there was no picture on the screen. Okay, that's not stable enough for me. In fact, it's unstable! But the comp crunched RC5 during the night and was still alive at the morning. I updated the buffers and went to work. As I came home in the evening I saw, the comp hung and was running only ca. 30mins after I've left the house......
I went back to 750 and ran the CPU at this speed for a week or so. Cache was set to 2/5 by the K7L2DOS program. And I've let the comp crunch RC5 for that time, almost 24/7/365. With a few reboots, but stable...
For no reason, I thought I could try that 800 again. And: as I looked at the Voltage-meter I saw 5,009V. How that? I don't care about that fact, as far as the Voltage stays at this level. I'm running the comp at 800MHz at the moment with a cache divider of 2/5. The RAM runs at 133MHz and soon, I'll mod the board to be able to set the 3,3V... It's a little modification: You've got to solder a potentiometer onto the mobo... By increasing the 3,3V it could be possible to get the entire board to run at 133MHz..
So far my overclocking experiments. If I try for 850MHz I don't even get a post, regardless of the voltage. Stay tuned for a little article about my case and how I modified it last week.
tide [ mail ]
This page only reflects the author's personal opinion and assumes no responsibility whatsoever regarding any of the contents or any damages that may occur explicitly or implicitly from reading the contents of this site.
All contents of this site are protected by international copyright law. Reproduction even of parts of its contents is not allowed except after written permission by the author and referral to the source of the material.
All brands are trademarks of their respective owner.
|Copyright © 2000, Coolernation, All Rights Reserved || |