okay tqhoang, that's great... again, your interest and help in this matter is just great. I think your right, a domain for each physical cpu sounds right! Thanks Danny
Hi, Unfortunately the system is worse again, it locks up within minutes of boot up. I don't believe it's the domains you declared, but the PLL Lock adjustment I requested. I guess the only way to prove that is to use the same _PSD as the last mod, but go back to 10us PLL and see what happens then. If it still locks up within minutes we know it's the domain issue. I'll let you have a think and speak next week. Thanks tqhoang
Yeah it just seems that the PLL lock time is the problem. In the past, I remember testing with the PLL lock time and 10us was the max I tried to get working for another guy with the Opteron 280 OSP. I even made a BIOS to keep the voltage at 1.25V and just changed the frequencies. Maybe we should try the PLL lock time at 100us? Maybe by Sunday night I can get you another BIOS to try. I'll keep the _PSD object in there since it seems like it makes sense with the 2 domains.
Hi, Ive spent some time trying to find information on them damn osp opterons, unfortunately it appears they were only an oem version hence amd never released any data. I then tried to find what pll lock rmclock was using but found no information either, the only mention was something about some mobile intel centrinos and a minimum of 30.72us. Although under the centrino section, it never mentioned pll lock anywhere in the rest of the document, so when it referred to the minimum of 32.72us as being the default, it may have meant the default for rmclock and not the default for centrinos. Unless you've already made one at 100us, i say try 30.72us. Thanks Danny
Unfortunately it's tells us nothing in fact it's worse than that because it seems to give wrong readings on the clock mulipliers and other settings. I've read more on the intel 90nm chips and they use pll lock of; min 0us - max 30us. Though on some models the the max was increased to 200+ but only for deep sleep states.
Deep sleep states (C-states) take much longer to recover. We can try the PLL lock time at 30us. Also I've been thinking about the VRM's and wonder if I need to change the PSD object again. This time I was thinking of making all 4 cores part of one domain, but the key difference being that the OS changes the p-state on all cores and not just any. I'm wondering if the problem is that the CPU's are getting out of sync and running at different p-states with different voltages. Let me know what you want to try first.
Damn it, still locks up. I had the burn it tests pushing the cpus just enough to switch between x5 & x9 scales only, still crashed after around 6 minutes. It's baffling me, why it would take 6 minutes to crash when everything seems stable up to then is beyond me.
I have a feeling that RMClock is disabling the OSPM and taking over. There's not much left to tweak...seems that whatever value we put in for the PLL lock time, doesn't really make things more or less stable.
Well, i'll take off rmclock and retry. Pll lock at 2us was terrible, 10us was definitely better. That said if thats the case i dont know why it wouldnt improve at 30us. I'll see what happens without rmclock.
I meant that it appears to do all the p-state stuff...or at least configure all the p-state stuff inside of Windows. Perhaps it's adjusting some timeouts that are not part of the ACPI spec. Have you tested any of my mods with only memory sticks in one bank?
First hi I just got a rackable unit from eBay and I install your mods but I get a code 12 now for one network port and the on board sata I've dumped the CMOS and reflashed with 1.18b with and with out raid I tried to get the bootable cd and followw instructions on page 4 but the olny thing in the iso is a readme file I got from the bulk file I tried to install win 7 and server 08 I even pulled the raid card to see if it was the problem I thing I messed up by not first install the first mod but I can't find the file
What is BIOS v1.18b? Did you mean v2.18b? What was the original BIOS version that you had on there to begin with? Some HDAMA-I owners have had HDAMA-G etched on their motherboard leading to the confusion. The bootable ISO image must be burned as an image. The free ImgBurn works very well for this. Make sure you're trying to boot with a PATA CD/DVD drive...not sure if the SATA ports support booting CD/DVD's.
Looking at sticker it a rev A When I look at the iso file with winrar and demon the olny thing in it is a text file no rom files or any thing When I go in to the .zip for 2.18b I don't see any rom files either is it a different file type like the rf4