X2 3800 overclocking problem

[velis]

My config:
ASUS A8N-SLI Deluxe
Athlon 64 X2 3800+ 939 (10x200 = 2GHz stock)
RAM Corsair CMX512-3200C2 x 2 for 1GB of RAM

Until now I've been running this computer at this:
CPU: 250 x 10 = 2.5GHz @ 1.3625V
RAM: 333 * 250 / 200 = 416MHz @ 2.5-3-3-5 1T

Today I moved the FSB to 270 and lowered the multiplier to 9.5. This should give me:
CPU: 270 * 9.5 = 2.565 GHz @1.4125
RAM: 333 * 270 / 200 = 450 MHz @ 2.5-3-3-5 1T

Both of these speeds are good for both my CPU and RAM. CPU will go to about 2.65 GHz and RAM will go to about 460MHz at those speeds. FSB will go over 300, but I never cared to test just how high. I know 300 works.

Now for the question:
Even though I select RAM speed 333 and the frequencies should be as I listed above, I get RAM speeds lower that calculated.
@FSB 250 RAM = 400MHz
@FSB 270 RAM = 427MHz

These values correspond to RAM speed setting of 320 and 317 respectively. But those two settings don't even exist in BIOS.

The speeds reported are the same with CPU-Z and Everest while NTune shows them very wrong.

I'm using bios 1017-004 (beta) (1016 did not obey HTT multiplier setting). I'd love to use 1017 final, but ASUS decided to ditch it and stop BIOS updates altogether.
I used to have a 3000+ Winchester in this mobo and there RAM speeds were good using the same BIOS version.

Why doesn't my RAM run at the speed that I set in the BIOS?

 

[aigomorla]

im thinkn it could be a voltage problem. Maybe you were running the ram at a slightly higher voltage the first time when it worked?

I remember having a simular issue like yours on a p4 asus board. I just had to up the voltage a tad bit and it worked fine.

 

[Brunnis]

Well, the RAM frequency for the 270MHz FSB looks correct. Here's how you calculate it:

A multiplier of 9.5 would give a CPU frequency of 1900MHz if the FSB was at 200MHz. The memory frequency of an Athlon 64 system is set by dividing the CPU frequency with an integer. Consider this equation:

1900/X=166.666... => X=11.4

As you can see, there is no integer that divides 1900MHz to get a result of 166MHz (or 333MHz DDR). The closest integer that provides a memory frequency not exceeding 166MHz is 12.

When you clock your CPU to 2565MHz, you get a memory frequency of 2565/12=213.75MHz. Since the memory is DDR, it runs at the effective frequency of 213.75*2=427.5MHz. So, that frequency is correct.

Now, with the CPU running at 10*250 and a 333MHz RAM your memory should run at a frequency of 417MHz. I have no idea why it runs at 400MHz instead. Sorry.

 

[velis]

Thanks for the suggestions. I will try upping the voltage although in the past that has always proved to be futile.
I went back to the 10 multiplier and just set the FSB to 257. This is 24 hours prime stable @ 1.4125V which is acceptable to me. The temps never exceed 58°C while running two instances of prime torture test (the second - maximum heat one). While I could go higher to 265 or even a bit more, I really don't like the idea of having CPU voltage @ ~1.45 and temps > 60. I'm not really sure what safe voltages and temps are for this processor, but I guess I'm just going to have to settle for this speed for now.

With this "new" X2 processor, my RAM can *almost* work 400MHz @ 2-3-3-5. The only problem is that MemTest86 tests 5 and 6 start reporting errors at about 99% through the test. Upping the voltage doesn't help. The previous processor had no chance @ CAS 2. But since this is still an "almost", I figured I'd try to maximize my ram in frequency if I couldn't do it in timings. If what brunnis says is correct I can't do that anyway since the processor simply doesn't go as high to allow my RAM running @450 MHz. Oh well....

I also find it quite strange that memtest reports memory speed a little below 2700 MB/s while with my previous processor it was almost double that. It's supposed to be double channel so I find this figure interesting to say the least.

Oh and I also forgot to mention that I'm running all this +8800GTS on an Antec TruePower II 380W. The PSU is really at it's upper limit with this setup. The 12V rail went from 11.93V (on my 6600GT) to 11.7 when I plugged in the 8800GTS. CPU voltage variations (while running prime) reported by Everest are as high as .03V which is one of the primary reasons I can't go higher with my CPU (at acceptable voltages + temps).

 

[Noubourne]

X2s have always had issues with half multipliers right?

You are testing that PSU voltage with a digital multimeter, and not just looking at some garbage software that gives you the wrong reading right?

60 is still a safe temp for the chip, but that doesn't mean the chip will like being run there. My Venice hated anything over 50C. My Opteron is much more forgiving. Sounds like your X2 is also forgiving of higher temps.

 

[velis]

Eh, yes, I'm using some garbage software. I'm not interested in *exact* voltages, just the variations. I don't care about the actual voltage as long as stuff works, but variations can show me that there's a problem in power supply.

I don't know about problems with half multipliers, but I gave up the idea that I could run the RAM @450. On the side note, I'm also thinking of selling my RAM and getting a decent PC4400 2GB kit which would help with the more RAM hungry applications that I use.

AMD site says my chip can go to 72°C or something near that, but I don't know how bad such temps are in regard to chip longevity. So I'll rather just leave it at <60.