F3-12800CL7Q-8GBXM with F1A75-M PRO - power cycle makes it work

[Kooboontoo]

I'm hoping somebody has some helpful idea or hint that might make this combo work. I am beyond frustration with this combo, and don't like to process an RMA for parts that may be working just fine.

EQUIPMENT:
ASUS F1A75-M PRO (BIOS 2206 - 8/31/2012)
A8-3850
F3-12800CL7Q-8GBXM
Seasonic M12II - 520W power supply (have also swapped with another known-good supply)
Other equipment: irrelevant, as it's all been disconnected for troubleshooting

Memory timings:
DDR3-1600
2048x4
7-8-7-24
1.6v
These sticks are not specifically on the ASUS QVL, but there are other sticks with the same timings present, so it should be possible to run them without error. I should note that ASUS does list a 6-stick kit on the QVL, and the board has only 4 slots available, so I'm not certain how trustworthy the QVL really is.

Background:
I don't actually care if the RAM runs at the specified 7-8-7-24; at this point I would be happy to get these sticks to run at any speed and do so consistently. I have already tried multiple combinations of timings, bumping voltages, etc. Each time I change timings I do so only after completely clearing the mobo settings using the onboard jumper. I have even tried running the sticks at the default 1333 8-8-8-24 speed, changing no other settings on the board, with the exception of enabling Q-Fan so I don't have to listen to the fans whine as a memory test is running. This has been a four-month (!!!) adventure so far, and I'm essentially getting nowhere.

Test procedure:

1. Clear all board settings using the onboard jumper
2. Set the testing settings on the board. Have tried using full Auto settings, D.O.C.P. setting, and full manual hand-entered settings.
3. Test the memory using Memtest86+ 4.20, as well as Prime95.
4. Watch the memory fail, horribly, with lots of evil, red error messages (or test threads dropping out one by one)

To fix the problem:
This is the part that is causing me the most grief.

1. Turn the machine off using the power button.
2. Turn the machine back on using the power button.
3. Rerun Memtest, Prime95, or any other test you can think of -- a fresh installation of the Full Desktop of Ubuntu or Kubuntu works well for this step.
4. No errors. Not one. Ever. Even after multiple hours, up to multiple days, of running tests.

To break the system again:

1. Turn the machine off and leave it off for at least 30 minutes.
2. Turn the machine back on and experience more errors.
3. To fix, just power cycle again.

The BIOS has been updated with each new revision from ASUS. To be on the safe side, I've even tried a re-flash of the BIOS, but with each revision there's been no difference.

I've tested the RAM individually and in combos, with the same results. I've tried bumping up voltages in the smallest increments possible trying to get a stable result. I've eliminated the power supply as a possible problem (and it's new, at that). The most irritating factor is that all that's required is to power cycle the machine and it runs perfectly, but let it go into sleep mode or get turned off and it will have memory errors again until power cycled. That's unacceptable for a real world machine.

Anybody have any suggestions? Should I be looking at a possibly faulty APU here?

[Tradesman]

On your comment " I should note that ASUS does list a 6-stick kit on the QVL, and the board has only 4 slots available, so I'm not certain how trustworthy the QVL really is"

that's a pet peeve of mine - they 'test' with whatever is available - might check my article here:

http://benchmarkreviews.com/index.ph...20387&Itemid=8

but on to your system - might try at 1600 8-8-8-24 (think CL7 might be to tight for your MC (memory Controller)), and add 0.05 to the DRAM voltage if needed. Also may want to try about the same to the CPU/NB voltage. Also check vCore (CPU voltage) and possibly add + .03....If still problems might want to check the CPU and socket, make sure no thermal compound in there and no bent/broken pins

[Kooboontoo]

Thanks for the incredibly fast reply!

DDR3-1333, 8-8-8-24 is the default boot parameters for this board with these sticks, and I had already tried bumping both the DRAM and CPU/NB voltages. I hadn't touched the vCore, so I'm attempting that right now, following having the machine powered off overnight. I decided to leave at DDR3-1600 instead of 1333, as that is the target speed, but am testing with 8-8-8-24 and the higher voltages.

As for bent pins, unlikely. They weren't bent out of the box, and the processor dropped smoothly into the socket on initial install (unlike some I've encountered that require some wriggling to drop into place). It hasn't been removed since. Thermal paste is the factory-applied paste from the boxed heatsink, which was never even uncovered until the APU had been firmly fixed into the socket. I'm loathe to remove the heatsink, as this will require re-application of fresh paste, and I'm entirely certain neither I, nor any other human, can do it as well as the factory process.

I'll post updates later tonight after the system has time to run through the normal tests.

[Tradesman]

Will be looking for them

[Kooboontoo]

Just for the fun of it, I decided to check the APU installation and pins. All perfect, but reseating the thing hardly ever hurts and sometimes helps, so probably worth it. This also meant removing the RAM, and I made sure to rearrange the slot order when I reseated the sticks. Errors still appear in exactly the same memory addresses, which is why I'm reasonably certain that the sticks themselves are not at fault.

As for playing with voltages, it doesn't seem to matter what voltage I adjust. The results are exactly the same as leaving at stock. Start the system completely cold and there are errors. Power off. Restart and it will run completely error free.

Have tried timings of 7-8-7-24, 8-8-8-24, 9-9-9-24. FWIW, the default boot timings read as 1333@9-9-9-24, and still produce exactly the same symptoms. I also checked the Asus SPD Information section in the EFI to grab the rest of the memory timings so they can be manually set. Some of the Auto settings do appear to be rather too aggressive, but the end result is exactly the same. Completely cold boot: errors. Power cycle: trouble-free.

If I weren't able to produce the same sort of symptoms using more than one test, I might even pronounce this a bug in the test itself.

This wouldn't be a problem at all (or maybe only a slight annoyance) if this were going to be a 24/7 on system, but it would be nice to have the option of turning it off at least occasionally.

[Tradesman]

What's the CPU/NB at? If less than 1.3, give 1.3 a try. If that works, can try lowering it a bit at a time - if not can you post your BIOS settings for DRAM both basic and advanced timings and all system voltages.

[GSKILL TECH]

If you continue to have a problem, feel free to send the memory in for RMA to see if a replacement works better.

http://www.gskill.com/rma.php

Thank you
GSKILL TECH

[Kooboontoo]

Improvement (but not solved )

Giving VDDNB Offset Voltage a bump of 0.15 for a final 1.3V has eliminated part of the memory errors. I previously got errors at the "beginning" of stick 3 and the "end" of stick 4. The middle errors (around 4200MB) are now gone, but the upper end errors persist.

Therefore, any suggestions for bumping anything else? I did try give a small additional bump to VDDNB but it made no additional improvement, so I've dropped back down to 1.3.


VOLTAGES

• EPU Power saving mode Disabled
• CPU Offset Voltage 1.412V (Auto)
• VDDNB Offset Voltage 1.300V (0.15000)
• DRAM Voltage 1.600V (Auto)
• SB 1.1V Voltage 1.100V (Auto)
• 1.1Vsb Voltage 1.100V (Auto)
• APU1.2V Voltage 1.200V (Auto)
• VDDA Voltage 2.500V (Auto)
• CPU Load Line Calibration Auto
• CPU/NB Load Line Calibration Auto
• CPU Current Capability 100%
• CPU/NB Current Capability 100%
• CPU Power Phase Control Optimized
• CPU Voltage Frequency Auto
• VRM Spread Spectrum Disabled
• CPU Power Duty Control T.Probe
• APU Spread Spectrum Auto

TIMINGS

• Ai Overclock Tuner Manual (D.O.C.P. works well for this also)
• Memory Frequency DDR3-1600MHz
• CAS# Latency 8 Slowed things down a bit until I can achieve a stable system
• RAS# to CAS# Delay 8
• RAS# PRE Time 8
• RAS# ACT Time 24
• COMMAND Mode 2
  
• Row Cycle time 34 Starting here, values were obtained from Asus SPD Info
• READ to PRE time 6
• RAS# to RAS# Delay 6
• WRITE to READ delay 6
• WRITE Recovery Time 10
• FOUR ACT WIN time 30
• REF Cycle Time 110ns
  
• CKE drive strength 1.50 (auto) Starting here, haven't touched these values
• CS/ODT drive strength 1.50 (auto)
• address/command drive strength 2.00 (auto)
• MEMCLK drive strength 1.50 (auto)
• data drive strength 1.25 (auto)
• DQS drive strength 1.25 (auto)
• processor on-die termination 60 (auto)
• CKE setup time 1 (auto)
• CKE fine delay 19 (auto)
• CS/ODT setup time 1/2 (auto)
• CS/ODT fine delay 0 (auto)
• address/command setup time 1/2 (auto)
• address/command fine delay 0 (auto)

[Tradesman]

Could try the DRAM at 1.64 - 1.65 (yes it's safe) and when testing w/ Memtest it's alwasy better to test each stick individually.