Strange memory errors on GA-EP45-DS3P

[Paul Johnson]

Hi, I'm running Fedora 9 on a Core 2 Quad 2.4Ghz with 2x1GB memory, and I'm getting some odd things. The system boots fine and runs for a while. Then at some point (which may well be associated with backing up to my USB hard drive) the system locks up hard. If I immediately reboot with Memtest86+ then it shows a cluster of memory errors.

When I first saw this I tried testing each stick individually, but could not produce any errors. So I put them both back in and ran Memtest for a few hours, with no problems. So I rebooted Linux. Everything seemed fine, so I started my (now long overdue) backup. It hung after a few gigabytes, and rebooting with Memtest86+ showed the memory errors were back.

I think the problem is temporarily cleared by a power cycle. When I tried playing with the RAM timings in the BIOS I noticed that the power cycled automatically when I exited the BIOS screen. This leads me to suspect that something is inadvertently tweaking the RAM parameters in the BIOS. One possibility is a kernel bug in the Linux USB driver.

I'm using Corsair RAM. On it is printed 1066MHz 5-5-5-15, 2.2v. This matches the BIOS settings except for the voltage, which the BIOS is setting automatically at 1.8v. There are big red warnings about over-voltage, so I'm nervous. When Memtest was giving errors I pushed it up to 1.9v, but that seemed to make the problem worse rather than better.

So: some questions:

1: Has anyone seen anything like this before?

2: Should I ignore the warnings and set the RAM voltage to 2.2? If I do so and it is wrong, could this damage anything other than the RAM?

3: Is it possible for an OS kernel to modify the RAM parameters "on the fly"? If so then I'll pursue this option.

[mxcrowe]

Re: Strange memory errors on GA-EP45-DS3P

I think you'll find many, many people with similar configurations to yours who have had similar problems when memory voltages and timings are not set correctly. It's very common for the system to be more forgiving with one stick of RAM, but not two. I would manually set the timings and voltages to the specs on your RAM. The auto settings rarely work with these boards.
I don't see any connection to the Linux kernel modifying the BIOS settings. Just set your DRAM voltage to 2.2v and see if that fixes it. You could add 0.1-0.2v to MCH Core, but I've found that auto lifts that up correctly anyway. Beyond that, LSD will help you tweak your memory settings further, if you want.

[VorLonUK]

Re: Strange memory errors on GA-EP45-DS3P

Before changing any voltages, I would recommend that if you have the setting "Performance Enhance" in the M.I.T. settings screen of the bios, that you set it to "Standard", then run memtest.

There is no point pushing your Ram straight to the maximum allowable voltage either. Checkout the corsair forums and/or leave a message for other members there or the "Ram guy" to advise you on the best settings for your board and the specific corsair memory you have.
A search on there though often provides the information you require.

[mxcrowe]

Re: Strange memory errors on GA-EP45-DS3P

For completeness, grab the BIOS template from another post around here and post your full specs and BIOS settings. That will make it much easier for people to help you.

[Lsdmeasap]

Re: Strange memory errors on GA-EP45-DS3P

Yes, I would manually set timings, and ddr2 voltage to 2.1V. 2.2 would likely only be needed if you overclock the ram, and is the MAX allowed under warranty. Some ram will need the spec to run stated speeds, but not generally and 2.2 is hot, so I would try 2.1 first.

1.9V may have only made things worse for a few reasons, Performance enhance was left on Turbo (Default) or your ram speed went up to 1066 which is would not be running at with only 1.8V so things likely just became all out of whack without all things being properly set

And as said MCH needs 1.25-1.3 for 2x2GB to be stable.

If you would like further help or suggestions please post back your full settings >>>

This is Just a template I use to ask people their settings, pay no attention to the numbers in it, replace with yours

MB Intelligent Tweaker(M.I.T.)
Robust Graphics Booster ...............:
CPU Clock Ratio ..........................:
Fine CPU Clock Ratio.....................:.0
CPU Frequency ...........................:


Clock Chip Control
Standard Clock Control
CPU Host Clock Control..................: [Enabled]
CPU Host Frequency (Mhz) ............: Mhz
PCI Express Frequency (Mhz) .........:Set to 100


Advanced Clock Control [Press Enter]
CPU Clock Drive...........................: <<<
PCI Express Clock Drive.................: <<<
CPU Clock Skew (ps)....................: <<<
MCH Clock Skew (ps)...................: <<<


DRAM Performance Control
Performance Enhance...................: [STANDARD]
Extreme Memory Profile (X.M.P.).....: Disabled
(G)MCH Frequency Latch...............: [Auto]
System Memory Multiplier ..............:
Memory Frequency (Mhz) ..............:
DRAM Timing Selectable ................: [Manual]

Standard Timing Control
CAS Latency Time........................
tRCD .........................................
tRP'...........................................
tRAS..........................................

Advanced Timing Control
tRRD...........................................
tWTR..........................................
tWR............................................
tRFC...........................................
tRTP...........................................
Command Rate (CMD) ....................:2

Channel A
Static tRead Value.........................:7-10
tRD Phase0 Adjustment...................:<<< Leave for all below as well
tRD Phase1 Adjustment...................:Auto
tRD Phase2 Adjustment .................:Auto
tRD Phase3 Adjustment..................:Auto
Trd2rd(Different Rank)....................:Auto
Twr2wr(Different Rank)...................:Auto
Twr2rd(Different Rank)...................:Auto
Trd2wr(Same/Diff Rank)..................:Auto
Dimm1 Clock Skew Control...............:Auto ps
Dimm2 Clock Skew Control...............:Auto ps

Channel B
Static tRead Value.........................:7-10
tRD Phase0 Adjustment...................:<<< Leave for all below as well
tRD Phase1 Adjustment...................:Auto
tRD Phase2 Adjustment .................:Auto
tRD Phase3 Adjustment..................:Auto
Trd2rd(Different Rank)....................:Auto
Twr2wr(Different Rank)...................:Auto
Twr2rd(Different Rank)...................:Auto
Trd2wr(Same/Diff Rank)..................:Auto
Dimm1 Clock Skew Control...............:Auto ps
Dimm2 Clock Skew Control...............:Auto ps

Motherboard Voltage Control
Voltage Type.………... Manual
CPU
CPU Vcore….……….........................:
CPU Termination…..... 1.200V*.......: <<< Leave
CPU PLL…………………....1.500V*.......: <<< Leave
CPU Referen.…………....0.755V*.......: <<< Leave
CPU Reference2………,,0.800V*.......:<<< Leave

MCH/ICH
MCH Core…………….....1.100V...........: <<
MCH Reference….…….0.800V...........; <<< Leave
MCH/DRAM Ref.…......0.900V...........: <<< Leave
ICH I/O……………….....1.550V............: <<< Leave
ICH Core…………...……1.100V............: <<< Leave

DRAM
DRAM Voltage ……....1.800V............:
DRAM Termination .…0.900V............: <<< Leave
Channel A Reference 0.900V............: <<< Leave
Channel B Reference 0.900V............: <<< Leave

Advanced Settings
Limit CPUID Max. to 3.....................: [Disabled]
No-Execute Memory Protect............: [Enabled]
CPU Enhanced Halt (C1E)................: [Disabled]
C2/C2E State Support....................: [Disabled]
x C4/C4E State Support..................: [Disabled]
CPU Thermal Monitor 2(TM2) ...........: [Enabled]
CPU EIST Function.........................: [Disabled]
Virtualization Technology................: [Disabled]

[Paul Johnson]

Re: Strange memory errors on GA-EP45-DS3P

Yes, I would manually set timings, and ddr2 voltage to 2.1V. 2.2 would likely only be needed if you overclock the ram, and is the MAX allowed under warranty. Some ram will need the spec to run stated speeds, but not generally and 2.2 is hot, so I would try 2.1 first.

I tried that. It didn't help. However I tried turning down the RAM timings, and my computer seemed to be stable, which is why I didn't get back to this before. But now its acting up again. This intermittent nature is a real pig: if only it would just fail consistently I could get it figured out. :-/

And as said MCH needs 1.25-1.3 for 2x2GB to be stable.

I'll give that one a go. In the meantime, here are my settings.

MB Intelligent Tweaker(M.I.T.)
Robust Graphics Booster ...............: [Auto]
CPU Clock Ratio ..........................: [9x]
Fine CPU Clock Ratio.....................:. {not on my BIOS screen}
CPU Frequency ...........................: 2.40GHz (266x9)


Clock Chip Control
Standard Clock Control
CPU Host Clock Control..................: [Disabled]
CPU Host Frequency (Mhz) ............: 266 Mhz
PCI Express Frequency (Mhz) .........: [Auto]
C.I.A.2 [Disabled]

Advanced Clock Control [Press Enter]

{I didn't check these, but I haven't touched them}


DRAM Performance Control
Performance Enhance...................: [Turbo]
Extreme Memory Profile (X.M.P.).....: [Auto]
(G)MCH Frequency Latch...............: [Auto]
System Memory Multiplier ..............: [Auto]
Memory Frequency (Mhz) ..............: 1066
DRAM Timing Selectable ................: [Manual]

Standard Timing Control
CAS Latency Time........................ 5 [6]
tRCD ......................................... 5 [6]
tRP'........................................... 5 [6]
tRAS..........................................15 [20]

Advanced Timing Control
{Again, I didn't check this, but I haven't touched it}

Motherboard Voltage ControlCPUMCH/ICHAdvanced Settings
{I have neither checked nor fiddled with these}


Finally, here are some numbers from Memtest86+ v2.01:

MemMap e820-Std

Failing addresses: 0007c0ffc84, 0007f1f689c, 7f9f689c, 0007fc9a89c, 0007fd9289c, 7fdf4a8c.

The error bit patterns were 00001000, 00008000 and 00000800.

Memory is running in interleaved mode, with 1GB in slot 1 and 1GB in slot 3. Are all these addresses from one slot? If so, which one?

Thanks,

Paul.

[VorLonUK]

Re: Strange memory errors on GA-EP45-DS3P

I would put the MCH value back to "Auto", change Performance Enhance to Standard and increment the DDR2 voltage by 0.1volts (so it at 1.9volts) as indicated on your PC health Bios Screen. Also set the main memory timings to 5-5-5-18 or 6-6-6-18

Then run memtest,

If it still fails, increment DDR2 voltage by 0.2volts (so its at 2.0volts)

Then run memtest again.

[Lsdmeasap]

Re: Strange memory errors on GA-EP45-DS3P

The only way to know for sure about memtest is to test one stick at a time, and if it errors always in the same spot you have DOA/dying ram.

Ya, I was going to say >>>
DRAM Performance Control
Performance Enhance...................: [Turbo] << Needs to be standard probably to run overclock to 1066 for your ram speed

And you need to likely set the ram to spec or close, 1.9 is just not going to work at 1066 speeds. I would use 2.05 +/- or 2.1