HyperX PC3000 512 - AT7 Max Stability Issues

I just got in a 512 mb stick of PC3000 HyperX memory. I put it in and ran a few benchmarks, which showed a big improvement over my pc2100 PNY memory. Unfortunately, my system is now very unstable. There is no predictability, but I have never had blue screens with this system, and I have had three tonight already. I raised it to cas 2.5, even though it is cas 2 memory, we'll see if this helps. I am not even running it at its rated speed, I am running it at PC2700 speed.

Please help me out if you have any advice. This is driving me nuts.

Motherboard is an AT7 Max

Power supply maybe? If you go to the monitoring page on the bios, what are the readings on the 3.3V, 5V and 12V rails?

Power Supply

I've got

VIO: 3.49
+5V: 4.76-4.81
+12V: 12.23

I think I am able to change the IO voltage in the BIOS, is this +5V bad? I have the 420 W power supply that came with my chieftec dragon case. I know it's not the greatest, but it is quiet, and I don't have that much hooked up that pulls a lot of power.

Thanks,
-Ken

The voltage regulation on your power supply seems not too good. The 3.3V rail is over by more than 5% and the 5V rail is under by almost 5%. But if the motherboard has good voltage regulation, it might still be okay.

If your motherboard allows adjustment of memory voltage, you can try increasing that.

Otherwise, you mentioned you're now running CAS2.5 instead of CAS2. Is the system more stable now?

The IO voltage in the BIOS applies only to the CPU. (The core voltage and IO voltage settings in the BIOS basically controls the CPU voltage regulator module.) I don't think it'll affect memory.

I ran cas 2.5 and other memory settings relaxed for a bit, but it crashed just as frequently. A couple days ago I decided to run the settings kingston reccomended again, figuring that if it was going to crash like mad, might as well have it be fast while it does it. My box hadn't frozen in a couple days at Cas 2 3-3-1T. It just froze a few minutes ago, so I guess it didn't get better. I should have gotten a good power supply with the case. I didn't think it made a difference, and my box was rock stable with the PNY pc 2100 stick in there. I would just buy a new PS, but this project is already getting way out of budget. I upped the DDR voltage to 2.65, hopefully that will help.

Can you give me a quick rundown on what VIO does for you? I know it's not usually necessary to mess with it, but I want to know what it does. Will it increase CPU stability? What about CPU temps? If I had to guess what it did, I would guess that it's increasing voltage to the chipset, but I have no reason to think this, so please clue me in.

The one thing I did that I can directly trace to increased stability is lowering the CPU speed from 166*12.5 to 166*12. This makes no sense to me, since at 166*12 it was rock stable with the pc2100 stick, running prime95 endlessly with low temps. Even with the new stick, the temps were very low. When things got unstable, I lowered the mult to *12, and it seemed to help. I will probly try 12.5 again later this week and see if this was just coincidental. The whole reason I got this memory was so I could run the FSB in synch with the DDR. Before I did this my box was rock solid, it had never crashed (aside from when I forgot to release memory in a program I was writing... my own damn fault).

Thanks for the help!
-Ken

Here's my limited understanding from various readings on the Internet over time. You or anyone else, feel free to correct if I'm wrong:

Theoretically, faster RAM draws more current, because I=C dV/dt, so dV/dt = I/C. So in order for voltage to change fast, you have to either increase current, or decrease capacitance. Here's a good article on how much current DDR SDRAM can draw, which can be quite significant:

Overclocks article 696

So this is a possible hypothesis of why your system was rock solid when you ran it at DDR266 but unstable when run at DDR333 (because DDR333 draws more current than DDR266.)

As for VIO, my understanding is: CPUs these days run their internal circuits at a Vcore that's lower than 3.3V to minimize heat generated (eg. the Athlon TBred B runs at 1.6V Vcore.) However, other components eg. chipset, memory, are manufactured independently so there needs to be a common voltage for communication between the CPU and these components. VIO = 3.3V is the common voltage that the CPU uses to send signals to other components. So, basically, within a CPU, the internal circuits operate at Vcore, while the signals that travel in and out of the CPU (thus the name VIO) operate at VIO.

There's a component on the motherboard called the Voltage Regulator Module (VRM) that converts either the 5V rail (for most Athlon motherboards), or the 12V rail (for all P4 motherboards) to the two voltages Vcore and VIO required by the CPU. Thus there are actually two stages of voltage regulation between the power supply and the CPU, one at the power supply unit itself, the other at the VRM. (This is the same VRM that you need to buy when you add a 2nd CPU to dual processor boards.)

You said that lowering your CPU multiplier from 12.5 to 12 seemed to improve stability. One possible hypothesis is maybe this lowers the current drawn by the CPU, and thus allowing the DDR memory to draw the amount of current it needs to achieve stability. This is because on Athlon motherboards, the CPU and the DDR memory share the same 5V rail. If it was a P4 system, the CPU would use the 12V rail, and won't have to share power with other components that use the 5V rail.

In terms of overclocking, increasing Vcore or VIO both might increase stability, although Vcore more so than VIO I think.

So I would suggest trying another power supply and see if it helps. By the way, you don't want to increase the voltage of the HyperX more than 2.8V. Someone over at Anandtech forums reported burning their RAM at 2.8V:

Anandtech forum thread on PC3000 HyperX memory

Cheers,
Dave