athlon xp 1800+

Micro-star international / 6712
October 6, 2011 at 13:40:44
Specs: Windows XP, Athlon xp1800+, model 8/768MB
athlon xp 1800+ model 8
mobo is MSI 6712
768 MB Kingston & Centron Ram

I was upgrading memory and ran into all kinds of problems, so I ran Memtest and Windows memory diagnostic and both were returning all kinds of errors, both with the original Kingston memory, and the new ACP-EP memory I had just purchased. I decided to go back and start all over, reinstall the Kingston memory which I had been using for about 7-8 years, and check the bios settings of both memory and the cpu. In spite of errors, the machine is stable with the old ram.

When I set up the machine all the timing defaults were set to Auto. Turns out the default FSB on the mobo is 100Mhz. I set it to 133 and the multiplier to 11.5. Then, I set the sdram frequency to 266 mhz, to match the cpu. The machine seems to be stable, but I haven't had it running for very long. Cpu temp is 47c. Both (MSI) FuzzyLogic and iObit's system information show the cpu is running at 2000Mhz. Why 2000Mhz? Why, if the ratio of cpu>memory is 1:1 am I getting so many errors in the diagnostics? What the heck am I doing?

Ok, so now with the FSB and multiplier set to 133*11.5, the system says I have an XP 2400+. I am confused. It's been so long ago I don't know exactly what I bought.

See More: athlon xp 1800+

Report •

October 6, 2011 at 13:52:16
I didn't quantify the memory (old and new):
PC 2700 DDR 333MHz

For the new memory I was attempting to upgrade to 2GB. BTW, I ran diagnostics on each stick one at at time, and in each slot on the board and received no errors! What the hey?

Report •

October 8, 2011 at 08:22:21
Ok, since no one seemed to have a clue or motive to jump in the thread, I’ll answer my own question, and maybe this will help someone else.

After installing the new memory the first problems I noticed were crashes in Google Chrome. A page would load and almost immediately crash. On reboot Windows XP (SP3) would call for CHKDSK every time. Windows explorer would crash, or not show any contents in a folder, or would report that a file was corrupted and could not be read/moved/deleted. While I suspected a HDD failure, this HDD was only a couple years old, and had only been used for maybe a year or so. I couldn’t ignore the fact that the problems only started after installing new memory.

So, I developed a schema to test each memory module, (1) one at a time, (2) in combinations, (3) and on each mobo slot. Next, I went back and reconfigured FSB and core/bus ratio on the mobo to match that of the processor, which at the time I thought was an Athlon XP 1800+. The mobo default FSB was set at 100MHz, and when I built the machine I had relied on SPD to detect and set the correct timings.

As it turns out when I reset the FSB to 133MHz and the multiplier to 11.5, the system returned saying I had a 2400+ processor. I also noticed that even though I had manually set the multiplier to 11.5, bios was returning 15x in MSI’s Fuzzy Logic, which is correct on a XP 2400+. I was not trying to overclock the processor, but just set it to it’s nominal speed, and match the speed of memory.

The nagging question was why the machine had been stable even though the original memory failed diagnostics. I recently read an article somewhere about memory timings. Since I had relied on SPD, timings seemed to be nothing more than a topic for discussion, because in the real world those settings were left to firmware. There are four main settings: (1) CAS latency, (2) RAS to CAS delay, (3) RAS precharge, and (4) RAS. I searched the web for the timings of the new modules I had purchased, and found nothing, not even from the manufacturer. Then, I found a program, “System Information for Windows”, that tells just about all you need to know about system software and hardware. And there is was, information about memory speed, supported frequencies, and timings at various MHz. The eureka moment came when I saw that the new modules do not support SPD. The bios was unable to AUTO set the timings.

I started diagnostics again after setting the timings according to the data stored on the module’s EEPROM. First on one, then two, and finally three modules in combination and received no errors. I had set up a 40 GB sacrificial HDD with a copy of Windows XP, and had unwired my main drives in case of another disaster. Everything seems ok, no crashes or instability after about 8 hours, whereas before it was just a matter of a couple of seconds when even the OS would simply crash and restart.

For me it was a lesson that many resellers don’t know diddly-squat about what they are selling, I didn’t know squat about what I was buying, and the limitations of bios and hardware when corners are cut to make a product cheaper. My new memory is working fine, and this old computer is zipping along nicely. But it would have been nice to know beforehand the challenges I had to face.

What I don't know is if the FSB and the multiplier had been set correctly at the outset, would I have had the same problems? I don't know and will leave that to someone else to figure out.

Report •

October 11, 2011 at 12:07:39
OK, here goes.

The MS-6712 (KT4AV) isn't a great board. The best socket A boards were based on the nvidia nForce2 or VIA KT600 chipsets. The nForce2 is the better of the 2.

The 1800+ runs at 1.53GHz (11.5 x 133MHz), not 2.0GHz. The vast majority of them had a locked multiplier, so either you're one of the lucky ones with an unlocked CPU or you don't have an 1800+. Is the core square or rectangular? If it's square, it's a Palomino. If it's rectangular, it's a Thoroughbred. If it has 512k L2 cache, it's a Barton. CPUZ will tell you what you have:

It's extremely important to maintain a 1:1 CPU:DRAM frequency ratio on the old socket A systems. That means that regardless of the fact that you have DDR333, you should be running it at 133MHz (DDR266) to match the CPU frequency. Anything other than 1:1 hurts performance. And you should be manually configuring the memory settings. Do NOT rely on AUTO or SPD. Make sure the voltage setting is correct (probably 2.5v) & manually set the timings. I don't recall what the ideal timings were for the VIA based boards, but for nForce2 boards it's 2.5-3-3-11.

BTW, which BIOS version are you running?

Report •

Related Solutions

October 12, 2011 at 09:29:03
Thanks, riider for the follow-up. Like I said, I had decided I have a 2400+ because the multiplier is locked at 15. That rules out a 1800+.

Yeah, I had figured out that SPD wasn't working. The cpu is clocking at 2.0. I have the timings set at 2-3-3-6, in bios, according to the memory EEPROM. But, cpuid shows 2.5-3-3-7 on the memory tab. Is that actual timing or recommended?

DRAM frequency shows
FSB:DRAM shows 1:1

Oh, and the bios is 7.00T. Thanks.

Report •

Ask Question