PDA

View Full Version : How I got my computer to crunch WUs faster.



Swing
2nd July 2000, 18:57
First of all, I want to thank all of you who took the time to write posts to help me here: http://forums.murc.ws/ubb/Forum12/HTML/000120.html

This is a repost from that thread as it is easier to find one post than it is to sort through a lot of posts in a topic.

After all of the help you have provided me, I have implemented something that may help others.

Someone said that SETI@home can operate within a 1MB cache. This proves the importance of system RAM speed for those of us using CPUs with much less L-2 cache than 1MB.

I am using an early stepping PIII 600E at 800MHz (133MHz FSB). I have a 128MB stick of high quality PC133 SDRAM that is stable up to or beyond a FSB of 144MHz and RAM timing of 2,2,2 (that is as fast as I have been able to get the PIII).

The other stick of RAM is high quality PC100 that I have had for some time (two years or so). The PC100 runs at 133MHz, but I forgot that I had increased latency from 2,2,2 to 3,2,2.

I wanted to keep 256MB of RAM because of image work that I sometimes do and I cannot afford to buy another stick of high quality PC133 yet. Dropping the speed a little had no visible impact upon manipulating images.

However, SETI@home is greatly effected by RAM speed. This is what I did to permit me to keep the PC100 RAM and increase speed (drop the time to complete a WU.

As you advised, I switched from SETI@home for Windows to the command line version. This helped, but I was still averaging around 7 1/4 hours per WU.

This is the important part. Voltage I/O is 3.3 volts. If the VI/O is increased slightly to 3.65 volts, circuits that are pushed to the limits become more stable (including RAM).

I cut a section from an old OEM PII heat sink (tall), cleaned it up and used silver thermal epoxy to attach it to the clock generator on the motherboard. This provided effective cooling for high FSB. The next thing I did was to increase VI/O from 3.3 volts to 3.65 volts.

I then entered BIOS and reduced the system RAM timing from 3,2,2,5 to 2,2,2,5 (as fast as the BIOS allowed). After benchmarking, I found that the system, including RAM was very stable and faster with no increase in CPU speed.

Upon finishing the above I ran SETI@home. While crunching the next WU, I did a lot of other work, as I only have one computer. I completed the WU in 6 hours 2 minutes. The next WU was the first that I compledted in less than 6 hours, in fact my time was 5 hours 24 minutes. That WU was completed while I slept. The third WU was completed in 5 hours and 43 minutes (or somewhere in that area). The fourth WU was completed as I write this, time to complete was 5 hours and 39 minutes.

The first and third WU was performed while I wrote posts on the Internet, letters in MS Word, wrote e-mail and ran PhotoImpact. I used the computer during the fourth WU, but not as much as the first and third. If these times are now average for me, this is a great improvement.

I have not gone to Safe Mode to complete WUs... yet.

For those of you who are stuck with a RAM timing of 3,2,2 or higher (3,3,3), this is good news. In fact it is so good that I may repost this information in a new topic, because this one is getting so many posts, some people may not read all of them.

My motherboard is an ASUS P3V4X. The jumper location to change voltage I/O is located next to the CPU fan header. Cooling the clock generator by using a tall heatsink (or heatsink and fan) can help in overclocking the CPU, if the CPU can go faster than the motherboard is able.

I do not recommend using silver thermal epoxy unless you are expert at using it, use thermal epoxy instead. For those of you who do not know where to get thermal epoxy, one good source is here: www.chemtronics.com (http://www.chemtronics.com) I now buy all of my heatsink (thermal) compounds from them.

Thanks again.

Swing