Asus A8N32 SLI (AMD) Overclocking Problem

[Sunday Ironfoot]

I'm having an issue with my Asus A8N32 SLI with overclocking an X2 3800 which I think is related to my Motherboard.

OK, first of all under default settings, no overclocking etc. it's 100% stable, did Prime95 for couple of hours. Now I have a 3800 X2 that I'm trying to push to 2.4Ghz. I raised the FSB to 240, set multiplier at 10x, memory divider to 166 (to keep memory speed at 200Mhz - DDR400), vcore set to 1.4v, and HT multi set to 4x to keep HT speed under 1000Mhz. However, I can't run Prime95 for more than 10-20 minutes without it reporting errors (using two instances of Prime95 to test both cores).

Now I've tried every setting in the BIOS to get it stable, tried relaxing timings on memory, increasing vcore on CPU, setting HT multiplier to 3x, but it still continues to report errors in Prime95. So out of desperation I tried lowering the CPU multiplier to 8x, lower memory speed divider to 133, but keeping the motherboards FSB speed at 240Mhz. So basically my CPU and memory are UNDERCLOCKED, but Prime95 stills fails after a short while. It's only when I raise just the FSB and nothing else that I get instability!!

What's the problem here? Is there something I've missed? Do I have a defective motherboard? Any help much appreciated.

 

[Sunday Ironfoot]

I'm doing overclocking on the mobos FSB right now (keeping CPU and memory underclocked) to test the motherboards max FSB speed. So far it's stable at 230Mhz....gonna try 240Mhz now.

 

[keldog7]

The approach sounds right - isolate each component for overclocking separately. There's a really good FAQ somewhere which explains this approach (?DFI forums) in great detail. Personally, I Prime95-ed first, with a 10x multi and FSB set at 230, then 240, then 250, and the RAM set to a 100 limit (i.e. 100/200 x FSB = 125 MHz ram...WAY underclocked). She will only hold the 2.5GHz if I tweak the CPU voltage to 1.425 V. I happen to know its unstable after about 15 minutes at 1.40V - but my stepping is not particularly good for low-voltage overclocking. As we speak (since my BIOS 1205 upgrade), I'm double-checking to se if she's stable at 1.4125V - but I'm not too hopeful. [8 phase power enabled via 200mV overvolt on top of 1.2125, and C'n'Q disabled).

As far as my RAM goes - never had any problems reaching 250MHz - memtest verified at 12 hours, by overclocking FSB to 250, RAM divider of 200/200 (1:1) 2.5-3-3-7 1T timing. Of course, to verify this, I lowered the CPU multi to 6X, to avoid CPU errors confounding my testing.

-A

 

[Sunday Ironfoot]

After days of testing I've finally found the cause. Seems it really doesn't like running the memory at 1T command rate when I overclock the FSB past 230Mhz, but at 2T it runs perfectly at 240Mhz, even with low latency settings on the RAM and with CPU at 2.4Ghz....100% stable now.

As an aside, does anyone know how much the performance is affected by running at 2T command rate compared to 1T?

 

[lopri]

First of all, try different CPU multipliers. I don't know which one's responsible (whether the CPU or the board), you will probably see the difference. For instance, it doesn't do 9x250 but it does 10x250, etc. Weird, but that's what my experience says.
Second, with this board, sadly, after a certain point you have to raise the FSP by baby steps. This will also vary with other components (CPU/RAM) In my case, past 300HTT I had to raise it by like 5/10HTT incrementals. Otherwise it will fail. Funny thing is, changing from 300HTT to 200HTT will give "OC Failed" message. Basically this board doesn't like a radical change in configuration. But the potential is there.

425 HTT (This was done with command rate 1T)

And I wouldn't worry about 1T/2T unless your main day-to-day app happen to be Sandra/Super PI.

 

[lopri]

Ah also, you should try playing with the SB-to-NB multiplier, too. Try X4 or X3.

 

[Sunday Ironfoot]

OK, I now have it running at 2.6GHz with vcore at 1.55v. I can just about get it to 2.65Ghz with 1.6v but the extra volts aren't worth it. Passed 12 hour Prime95 (two instance) torture test at 2.6Ghz. Now I'm happy

 

[One43637]

glad you found a solution to your problem. what kind of ram are you running and what temps are you getting (CPU & motherboard) when you have it overclocked to your present speeds?

 

[spurdy1]

Originally posted by: keldog7
I'm double-checking to se if she's stable at 1.4125V - but I'm not too hopeful. [8 phase power enabled via 200mV overvolt on top of 1.2125, and C'n'Q disabled).

I've seen a few references to this (the idea that the 8-phase power is only used if the CPU overvolt is enabled). Is there data to support that this is something other than a rumor? It seems odd that they'd go to the trouble to implement something like that, market it heavily, and then not have it used all the time.

I just double-checked the manual PDF and I didn't see any references to the 8-phase power ever being disabled.

 

[keldog7]

Frankly, I'm in the same boat as you...didn't know if I was running 8 phase or 4 phase. Using the overvolt setting, I haven't noticed *any* improvement in temps / stability etc. , though I haven't really tracked/graphed the voltage supplied to the core during a variety of activities. For all I know, this could just be "computer voodoo"... If you stumble across a more authoritative source, please let me know.
-A

 

[oldben]

Ironfoot would you mind sharing the BIOS settings for the overclock you did. I've the same config MB/CPU... and I can't get over 2.4 stable... since I am pretty new to overclocking I am looking for any help at all..


Thanks,