Somebody must have the answer

I posted a message last night under the heading:

IC7-MAX3 SPD Problem

So far there have been zero replies. Somebody out there must have some insight as to why my IC7-MAX3 (2.4c P4) is exceptionally well behaved (stable & fast) with 512MB of memory and a 267 MHz FSB (5:4) but will not run at anything but a 1:1 multiplier when I add two addition sticks of memory for a total of 1 GB. This problem severly limits the performance of this machine because I have been forced to reduce the FSB from 267 MHz to less than 214 MHz just to get the thing to POST. This is a significant performance hit so I am very anxious to resolve this problem. I can revert back to 512Mb but that means I wasted a lot of money on the new memory, and worse, I cannot use it to improve the performance of the video capture/editing software I am using.

Help would really be appreciated.

What kind of memory do you have? What vDIMM are you using?
Using two sticks of memory will enable dual-channel mode, versus one stick and single channel mode.
Post some additional info regarding your memory and maybe we'll find the cause.

mattspalace, thanks for the reply.

I have been running the IC7-MAX3 for a number of months now. Using the following settings the machine has been completely stable and runs both Memtest386 and Prime without errors.

2.4c P4 @ 267MHz FSB (3.2 GHz)
512 Mb Mushkin PC3500 Black Level II (two 256Mb sticks)
5:4 memory multiplier / 2.7 volts
DRAM Timing Manual 2.5/6/3/3
BIOS Ver 13 10/30/03 (This is the latest BIOS)

I just increased the memory to 1Gb by adding two 256 Mb sticks of PC3500 Mushkin Black Level I. Mushkin no longer sells PC3500 Black Level II DDRAM. However, prior to buying this memory, Mushkin assured me that mixing Level I and Level II would not present a problem because they are in different banks.

Now, with both memory banks fully populated for a total of 1 Gb, the bios insists on defaulting to "settings by SPD" no matter how I manually set the DRAM timing multiplier and latencies. In other words, the bios will now only use a 1:1 CPUDRAM ratio, severly limiting the maximum FSB speed.

This should provide the information you are looking for. Thanks for the help.

What BIOS are you running?

Try _all_ settings I've described on this page:
http://www.ocforums.com/showthread.php?s=&threadid=258345

crotale, you solved the problem via the link in your post. I noticed in that link that the "N/B Strap CPU As" was set to "PSB800". Since I never understood this particular paramenter, I always left it at the default setting. Just to see what would happen, I changed it to PSB800. To my very pleasant surprise, that instantly resolved the problem I described.

Can you explain what this particular parameter does? It is obviously important because it made a huge difference in how the bios behaves.

Previously, with only 512 Mb (2-256Mb sticks), I was able to achieve 267Mhz FSB with 100% stability. Now, with 1 Gig, I was forced to drop to 260 MHz, a not completely unexpected but certainly acceptable price to pay for the additional memory. The machine seems to run fine at 267 MHz but Memtest86 shows errors at this speed no matter how I set the timing latencies. I place a much higher priority on stability and reliability than speed so I feel much more comfortable at the slightly slower 260 MHz FSB.

Thanks for the help.

That's nice to hear!

I've also had trouble with instabillity when not setting the N/B Strap and specify the timings. I don't know exactly what they do more than that they give you diffrent options under DRAM Ratios.
Some people also report that with certain BIOSes, the 667 "strap" gives you 1:1 even when you choose 5:4, but I guess they solved that in recent BIOSes...

By the way, increase vdimm to 2.8, the RAM handles it without problem, and you might be able to return to your previous overclock.
Good Luck!

I forgot to say that I am at 2.8 volts already. I raised it to 2.8 when trying to troubleshoot thge problem. As soon as I have some time I do plan to try lowering it as I am not comfortable at this voltage. With all four memory slots filled, the sticks are almost touching one another. That means that the inner sticks and inner sides of the outer sticks are not doing a very good job of rejecting heat. If I can lower the voltage at least I can reduce the amount of energy that has to be removed and lower temps to some extent.

Thanks again for the help.