For the HDAMB, I only modded the DSDT to support the Opteron 280 (what I have sitting around). What CPU do you have?
I have two Opteron 250. I remember some years ago I made a DSDT file for loading into the Linux kernel. I will try to find it on my hard disks and backup cds My hope was to find a BIOS somewhere that I can just flash, because for using a DSDT-file I must always recompile the Linux-Kernel after every kernel-update :-(
Specifically which Opteron 250's do you have? OSA250CEP5AU (Rev CG) OSA250FAA5BL (Rev E4) OSP250FAA5BL (Rev E4) <-- These have been problematic since they don't have information in the AMD datasheets
This would be great! Today I tried 10hours to compile a running kernel but without any success :-( Some years ago it was much easier because of less kernel restrictions. Thanks for your help in advance.
Flashing your "V2.03_DSDT" BIOS-version was successful. The "powernow-k8" driver is working now. As long as I select the cpu frequency manually, the computer runs stable. If I allow the "ondemand" governor to choose the frequency dynamically, the system freezes after some minutes. But this is not a big issue for me. I already know this behaviour very well with the separately loaded DSDT-file at boot some years ago. Thanks again! You saved my weekend and a lot of hours for kernel compilation in future!
These boards are very weird. Sometimes it's stable and sometimes it's not. Mine works great with my Opteron 280's (OSA280FAA6CB) and the CPU scaling works with Windows 7 & 10 just fine. Also CentOS Linux was very stable for me too. I could change a Ramp Voltage Offset (RVO) variable if you want to test another BIOS.
I assume that the power consumption of the cpus (85.3W) are very close to the max allowed power consumption of the board (max. TDP90W). While the frequency is constant there is obviously no problem. But I assume that frequency switching causes peaks with higher consumption > TDP90W. When I contaced the rioworks/arima support some years ago, they told me that the board is not designed for cpus > TDP90W. Maybe this the power consumption of the RAM (16GB DDR400 4x20W) has a side effect. Obviously you have got good Opteron 280's samples (OSA280FAA6CB) with a much lower consumption in reality than the official TDP95W or the version of your board is a bit more powerful. If you change the Ramp Voltage Offset (RVO) to lower voltages, I will try it. If only higher voltages fixed the problem, the risk to break the board or the cpus would be to high for me. (Dynamic CPU frequency scaling would be only a nice to have feature for me)
Yes, only 50mV over-voltage sounds safe enough. Could you try / flash this modified BIOS on your HDAMB-board first? Just to make sure that the modified RVO itself is working as expected.
I have tried an RVO=2 (50mV) on my HDAMB...I think I'm still running it right now actually. I didn't notice anything different at all when running CPU-Z. I didn't have time tonight to generate the new BIOS, but I will get to it by tomorrow.
Yesterday I flashed the RVO=2 BIOS but I had the same problem. After round about 5 Minutes the screen freezed. I did not measure exactly the time until freezing but the time is always almost the same. But round about 5 Minutes is more than enought time for setting manually a fixed frequency.
I switched back to your first "V2.03_DSDT" BIOS and I found solution for the instability. When I reduce the allowed frequency range from 1000<->2400MHz to 1000<->2200MHz the system is running stable. sudo cpufreq-set -c 0 --min 1000MHz --max 2200MHz sudo cpufreq-set -c 1 --min 1000MHz --max 2200MHz Obviously the direct switching from 1000MHz to 2400MHz or from 2400MHz to 1000MHz leads to the instability
I wonder if the Opteron 250 E4 CPU's really don't like the Max VID 1.40v then. You should be able to safely use the HDAMB BIOS that I have for the Operton 280 (E6). It has the same P-state frequencies and voltages as the Opteron 250 E4, with the exception that the Max VID at 2400MHz is 1.35v instead of 1.40v. Let me know if that works.