EDIT: (25 may 2011) I've posted on an amd forum all about my findings regarding my memory issues here: http://forums.amd.com/forum/messagev...hreadid=151625
EDIT: (28 may 2011) it's not my +5V VCCH being 3.6V, that's causing me memory issues after I've left my computer off for the night. read below the "EDIT:"
EDIT: (11 june 2011) it seems that the only way to "fix" my memory errors is to unplug computer and make sure it's drained (by clicking powen on button while it has no power) then plug it in again. Now, I don't know how much time it needs to run with memory errors before this fix will work, maybe it will work even after 1 minute, but now I just did it after 10 hours of it giving me memory errors and I tried a lot of variants by modifying bios settings with no luck. My preferred way of testing is to md5 sum check like 80gig of files like mp3 flac and stuff, of which I already had the md5 file with their sums and if they err, I then stop the check, so the file remains in memory cached, then make a copy of it and named it "bad" then reboot to clear cache from memory and then compare it to the original file and I notice that only 16 bytes differ, but mostly because the offset is wrong I think, that is, it possibly read from the wrong offset 16 bytes, because of all the times I've checked files like these, the same kind of offset they have, the 2 and 6 as the last digits in these offsets:
those are the offset ranges (16bytes) which represent locations in the files whose contents differ when crc is bad(due to memory)
hmm there's an E here too, so 2, 6 and E
Disabling chan interleave didn't work, I didn't md5 test back then, and memtest86 seems to rarely detect errors, the v4.2 one anyway
I am strongly suspecting it's the higher than normal NB and HT Link frequency, which is over 2000Mhz by 9Mhz, and sometimes it can be by 13Mhz and then I get errors easier, however not after my above "fix" with poweroff
I would not have resurrected this thread (I guess there are some rules against this but the reason as to why is not allowed escapes me, also I've zero forum experience) except that googling for "vcch" gives this thread the third result and so this might be useful in that respect.
Now, I have the same reading on HWMonitor 1.16.0
+5V VCCH 3.60V Min: 3.60V Max: 3.63V
and looks like this was read from ITE IT87 chip which is on my mobo Gigabyte GA-MA790FXT-UD5P with Bios version F8k (beta they say), so I figure it should be somewhat accurate. PSU is Corsair TX750W so 750W. I also drain some power for 4x2TB Western Digital green HDDs and 3x case fans and some dvd-rw , usb key/mouse, 1xLAN,
Now, the only issue that I find to be connected to this voltage is that when I shutdown my PC (without unplugging it from socket) and leave it to rest for 6 hours (or more) and then I turn it on (power button since my keyboard is USB won't go power on by password) then I run memtest86+ and I get memory errors in the 6144MB -> 6300MB range only on test #7 (random inversions or something) and only on bit 00000800. (8GB ram total). Now these only stop if I unplug my computer from the power socket (mobo lights will go off in a few seconds until the capacitors drain) and then after like 20seconds I plug it and start it again , memtest86+ reports no errors.
As I understand it VCCH voltage should be 5V (yeah might be badly reported) but this is the only (?) voltage active while computer is off, and when I turn it on, it seems like it didn't do it's job properly because if I switch computer off completely (thus loosing this VCCH power) and then on again and starting it works flawlessly. And after doing this, I can leave the computer on for months (literally; since this is how it was working for about almost a year: on full time w/o problems; and yeah those capacitor say 50,000 hours, else the normal 10,000 hours would have be maybe the cause of this behaviour) and it will have no memory issues, can even reboot , no problems, but if I turn it off and leave it for 6 hours (the least time I take to sleep) and then turn it on again ... memory errors in that area. Ok I swapped the RAM sticks(4 of them) between them and doing this required unplugging computer from socket and so each time when turning it back on it worked, I figured ok maybe small dust caused imperfect contacts or so, but then next day after sleep again. I haven't tried only with 1x ram stick or 2 or 3, always with 4. The memories are supposed to be the same type (they look the same too) but CPU-Z says that two of them have part number and no Week/year info and the other two have no part number but have week/year info so somehow they are different, oh and also those with part number have 5 JEDEC-s with 740Mhz and the other two don't, but I only use the prev JEDEC 666Mhz with same timings on all rams 9-9-9-24-33. Now the voltage reported on JEDEC in CPU-Z is 1.50V on all , but BIOS puts it on 1.60V. And thus VIN1 on HWMonitor is 1.60V (I read somewhere that it should be 1.65V instead).
And another thing, while in the mode where the ram gives errors (unless I unplug computer from socket) then if I go into bios and change dual channel to off (Channel Interleave to Disabled) then surprise no ram errors, as soon as I reboot and change it back to Enabled again , memory errors return. The Bank Interleaving option can be to either Enabled/Disable with no effect. Also the Unganged/Ganged of the memory controllers can be either with no effect (only on speed). And also i tried lower memory speeds down to 800Mhz (from 1333Mhz) and same errors occur. And so the only solution was to replug the computer and errors disappeared.
That is why I am so much into this +5V VCCH being onlu 3.60V and now it's working fine and it's this value (I don't yet know what's the value when the memory gives errors, since I didn't stop my computer(aka shutdown) since a while now).
So I'm guessing it's either the PSU or the mobo ... any ideas?
Anyway hope this gave some looking into the same issue some insights...
If I eventually find out what the issue is, I shall not forget to mention it here (if I don't find out I won't say anything
help is appreciated
PS: memory fail screenies or from bios settings can be posted if required
Also forgot to mention that I had bios F8c and then updated to F8k and both had same issues with memory errors after certain shutdown time elapsed (6 hours and 8 hours was tried)
EDIT: (28 may 2011) apparently I just now found out my issue was not related to the +5VCCH (which is maybe just reported wrongly? dno) but it's instead related to the processor's internal memory controller. Up until now my workaround was to keep computer running full time without turning it off (that is using 236Watt per hour which was calculated using total used for last 7 months)
AMD Phenom II X4 965
As I've seen here: AMD CPU / Memory Controller Limitationhttp://www.gskill.us/forum/showthread.php?t=2977
they direct you to that pdf: http://support.amd.com/us/Processor_TechDocs/41322.pdf
where you can see Product Erratas for this processor, and apparently, among other issues, issue 379 (which is currently at page 102/143) is:
379 DDR3-1333 Configurations with Two DIMMs per Channel May
Experience Unreliable Operation
In systems with more than one DDR3-1333 unbuffered DIMM on a channel, the processor memory
subsystem may exhibit unreliable operation over the allowable VDDIO voltage range.
This erratum does not apply to DDR3-1333 configurations when only one DIMM per channel is
Potential Effect on System
Memory system failure leading to unpredictable system behavior.
In a configuration where two unbuffered DDR3-1333 DIMMs are populated on one channel, BIOS
should derate DDR3-1333 system memory to 533 MHz operation (DDR3-1066) by setting the
DRAM Configuration High Register[MemClkFreq] (F2x[1, 0]94[2:0]) to 100b and adjusting
memory subsystem timing parameters accordingly.
Thing is, I remember trying this, lowering to the next below setting and still got errors (unless I don't remember something right, OR I did it in the same session(that is, without shutdown/poweroff) via a ctrl+alt+delete and entering bios and lowering then going memtest86 again and still got errors).
So far, I've set my dram to 1066 (x5.33 in bios) then I left PC off for little over 5 hours (it should've been 6 but I'll try again, if fails I'll update this, else assume it didn't and this works
then turned it on and memtested for 40min, no errors, seems to be working...
Memory is: CSX, two packs of 2x2GB, (8GB total) the packs are named: CSXO-CEC-DDR3 (part number: F6456U61F9333G ), and timing is 9-9-9-24, speed: 1333Mhz
Motherboard: Gigabyte GA-MA790FXT-UD5P specs here
Here's when downclocked to 1066: (8-8-8-20-27-1)
and here's normal at 1333: (9-9-9-24-33-2)
the memory fastness is basically the same, if you don't get this read this article:http://www.techreaction.net/forums/showthread.php?t=367
I also made sure it is so by testing with PerformanceTest 64bit , Advanced->Memory
So let us hope it's working with 1066, after leaving it off for the night(as I said it works so far after I left it for 5 hours). Since with 1333 doesn't.
EDIT(again; 2 june 2011 ): looks like if I leave computer off for at most 6 hours, I get no memory errors, but if I leave it like over 7 hours (or possible over 6 hours) then I get some memory errors because firefox crashes and then I go into memtest and mostly I cannot hit any errors, but when I do hit one it's only one and long time can pass and I still never hit an error again, and I don't have to turn off my computer to get the fix so errors are gone, so looks like, when I turn it on, firefox crashes after some minutes then I reboot and memtest (I may or may not see errors) then I reboot and back to windows again and seems to never err again. This time I disabled the Memory Bank Swizzle which was a hidden (must press Ctrl+F1 in bios to show it and others) option and was always enabled, but its effects can only be seen next time I turn computer off for like 7 hours and see if I still get memory errors.
Anyway here's the screenie from when I got only 1 error and after this (without turning computer off) I didn't get any errors and I just rebooted into windows and worked w/o errors.
So looks like doing the AMD workaround didn't work for me, well it worked better than without it, that's for sure, but do I really have to lower it even more for stability ? Btw, memory voltage is 1.600V reported in BIOS PC Health, and I cannot set it lower, (on memory's paper sheet is says 1.5 +/- 0.25 so, it is in range but I wonder if it should be just 1.5V exactly instead of 1.6V; on the timings table in CPU-Z all those JEDEC are seen and voltage is always 1.50V for each; but I don't pretend to know wth)
EDIT: the errors persisted (firefox crashing and some explorer stuff) but memtest wasn't detecting anything (and I didn't have the nerve to keep it running too long) so I set the Channel Interleave to Disabled (this is the evil workaround that also decreases speed but hey no errors when this is disabled, right?!) then I also set to chips to Ganged and left Bank Interleave to Enabled and also set memory to x6.66 aka 1333Mhz, and I have the same memory speed and hopefully no errors; so far so good. (previously was 1066Mhz and both interleaves enabled and Unganged)Edited by joojooo - 6/25/11 at 12:36am