Welcome to Computing on Demand 0
0
0
0 0 July 24th, 2008 0 0

Dual Xeon Sweetness Unwrapped, The Asus PC-DL Deluxe

Date: Wednesday, December 15, 2004
Author: Robert E. Bailey aka: Rbreb13
Page: 3 of 4


Now we're beginning to see some real improvements in the scores. A 700 point increase in Memory Bandwidth is real nice.

The next set of tests will be with a FSB of 150. I tried to get more but it just wasn't stable at stock voltages. I could get into Windows at 155 but it would freeze and BSOD quite frequently. I used a Distributed Computing program called FaD (Find-a-Drug) and Prime95 for stability testing. If you can't use it 24/7 any overclock is useless.

CPU: 2.40 GHz @ 16*150 FSB
RAM: 150 MHz for a DDR300 setting and timings of 2.5-3-3-6
Here I also had to change the memory FSB so it was running 1:1. It wouldn't post with the dividers set.


SiSoft Sandra CPU Arithmetic: SiSoft Sandra Memory Bandwidth: PCMark 2002:

A decent improvement is seen at 150. An 800 MHz overclock is nothing to sneeze at that's for sure. Memory scores didn't improve much because of the settings I had to use to keep it stable. As you can see this board is a pretty good overclocker even with the lousy BIOS it comes with. Having to use jumpers for settings is so old school.

At this point I was stuck. I might have been able to get another MHz or two out of it but why bother. I wanted much more than 1-2 MHz. There is no way of knowing for sure if it was a lack of voltage on the CPU's or the memory. This board undervolts them both. MBM5 was reading the CPU voltage as 1.24-1.28 volts. With a multimeter the memory was reading 2.46 volts. So it was time to do a few physical modifications to the board to increase some voltages. First one I did is by far the easiest.

Here I connected a couple solder points on the board to increase the memory voltage by 0.08 volts. Yeah, I know. Woop Dee Do a whole 0.08 volts but that got me up to 2.54 volts which was within the specs for the KingMax RAM. The link to this mod is at the PC-DL Users Site.

Next I wanted to up the voltage on the CPU's. This mod is a little more involved and not real easy but doable. I managed to get it done after 3 or 4 tries and I have fat fingers. To change the voltage on the CPU's you have to insert U shaped wires into the socket holes in the proper place. The placement depends on the CPU you're using. It's different for each type. If you're going to try this, be sure to do your research and double/triple check your work.

For the wires I used telephone wire. Speaker wire also works well from what I've read. It needs to be a very thin wire as it has to share the socket hole with the pins. For inserting the wires I used a pair of forceps that have a very thin nose. Most people are getting this done with tweezers. Don't get discouraged doing this mod. It's not real easy but after 6 tries on the first hole it became easier. Also if you're an old fart like me, get a good magnifying glass!



For a complete overview on the VID mod CTechNet has a good article written by HFU on this and other things.

All right! Mods are done!

Booted up the system and sure enough it showed 1.57 volts for the CPU voltage in the BIOS. Now it's time to get serious!

At this point I rebooted and went into the BIOS. From here I set the memory to "Auto" and dropped the multiplier to the lowest one available. For the 1.6 Xeon's this was 12. Yes Xeon's allow multiplier changes. Only down though not up. I then exited and saved the BIOS but did not allow it to reboot. I did a hard shutdown during boot. Then I moved another jumper to get a 200 MHz FSB. This gives you a FSB range of 200-233 in the BIOS but nothing lower or higher. I have seen where some peoples systems won't boot at this setting. It will have a lot to do with the CPU's you're using. Some say you might need to test each CPU separately in Socket 1 to see which one will boot at 200 FSB or which one will get the highest FSB. Then use that CPU in Socket 1. I didn't do this but may try it at some future time. As it was I powered back up and much to my enjoyment it booted into Windows. 2.4 GHz at 12*200 FSB! WOW

I ran a few stability tests and everything was fine. So I rebooted and upped the multiplier to 13= 2.6 GHz @ 13*200. I ran some more tests and everything was still fine. So I tried a multiplier of 14= 2.8 GHz @ 14*200. It booted into Windows! All right I thought! Wrong! Occasional freezes and BSOD's were happening whether I was stressing the system or not. So I went back to a multiplier of 13. I'm reasonably sure it's my RAM causing the problems. It just needs more voltage than 2.54 at these settings. I don't have the tools to do the Pomona mod so until I get what I need it will have to wait.

[Previous] 1  2  3  4  [Next]
0
0
0 0 0 0 0
   Graphical Design by: Joe DiFiglia Copyright 2000-2007, Computing on Demand http://www.computingondemand.com