Other ideas after successful test?

[RichS]

I finally got around to adding another 16GB to my machine, so 32GB total, as memory has become quite cheap. But ever since I installed the memory, the machine crashes every 12 hours or so. It just shuts down with no warnings and no other symptoms - just power off.

I ran memtest86 three times - about 5 hours each time - with no errors reported. The only issue I saw was selecting the Test 14 (iirc) marked as in beta test, one one other test cycle. My machine froze at 52 minutes, so about when it would be running the end of the first cycle. It didn't shut down, just froze on the screen. The timer was stuck at 52 minutes and the system clock was frozen, about 2 hours earlier than the actual time. Keyboard, mouse, etc. did nothing - I had to hard reset.

Attached is the report and log file (can I really not upload the default html version?) The only thing that looks even remotely odd is that the memory JEDEC and SMBIOS Profile says either 1.5V or 1.35-1.5V but the "RAM Configuration" says 1.185V. But I've never overclocked, over-volted, under-volted, etc. - it's configured just as I bought it years ago.

I tried swapping memory slots - moving them in pairs to the alternate positions, but it made no difference - crashed again in a few hours.

Any ideas, other than tossing $25 of memory, sticking with 16GB, and a lesson learned?

[David (PassMark)]

Test #14 (DMA) writes a lot of data to the USB drive that you used for booting.
However we found that USB drives are way less reliable than RAM. Way way less. So more often than not, faults in the USB drive were exposed before RAM faults were exposed.
So that might be the case here as well, you might have a bad USB drive.

[leclod]

I'd try to run the new kit only, to see if it maybe crashes on its own.
(Care to share some specs ?)

[RichS]

I've pulled the new memory and the old kit has run fine for a bit over 3 days, so it doesn't appear that something broke somewhere else while adding the memory.

I'll swap to the new kit at the end of the day and see if it runs on its own also.

The USB drive is probably at least 10 years old, so if that's a potential weak point, then it's likely what happened, so I'll ignore that failure.


I'm not sure I have more specs than what is in the log/report - what else is useful?

2 sticks of each - Samsung is the original, Crucial is the new - part number specified on their website.

SPD #0	8GB DDR3 2Rx8 PC3-12800
Vendor Part Info	Samsung / M378B1G73DB0-CK0 / 23DA187C
JEDEC Profile	1600MT/s 11-11-11-28 1.5V
SPD #1	8GB DDR3 2Rx8 PC3-12800
Vendor Part Info	Crucial Technology / CT102464BD160B.M16 / E9000022
JEDEC Profile	1600MT/s 11-11-11-28 1.5V, 1.35V

Same with PC specs:

System
Manufacturer	ASUSTeK COMPUTER INC.
Product Name	G10AJ
Version	System Version
Serial Number	F2PDCG0017YE
BIOS
Vendor	American Megatrends Inc.
Version	0304
Release Date	06/05/2014
Baseboard
Manufacturer	ASUSTeK COMPUTER INC.
Product Name	G10AJ
Version	Rev X.0x
Serial Number	150137396500329
CPU Type	Intel Core i7-4790 @ 3.60GHz
CPU Clock	3592 MHz [Turbo: 3791.2 MHz]
# Logical Processors	8 (4 enabled for testing)
L1 Cache	4 x 64K (183844 MB/s)
L2 Cache	4 x 256K (57071 MB/s)
L3 Cache	8192K (41465 MB/s)
Memory	32728M (19077 MB/s)
RAM Configuration	DDR3 1600MT/s / 1.185V

[RichS]

New memory also has run fine for 36+ hours - I never got more than 10-12 hours without crashing with all 4 sticks in.

I'll try interleaving slots, but have little faith that will make a difference. An ideas? Or do I just throw the new memory away/in a drawer?

[leclod]

I'll try interleaving slots, but have little faith that will make a difference. An ideas? Or do I just throw the new memory away/in a drawer?

What do you mean interleaving ? of course you should interleave those 2 kits

[David (PassMark)]

Mixing different sticks of RAM with different specs is never ideal. There are a lot of other timing values than those advertised on the packet.
See RAMMon for a complete dump of the RAM specs.

The BIOS is forced to try and find a collection of settings / voltage / timings that will work with both sticks. Sometimes this results in a marginal result. At the very least a situation that was never tested by the motherboard and CPU vendor.

Maybe it is time to start bumping up the voltage in BIOS?

Or buy two more of the exact same Crucial sticks?

Hard to know if the problem is with the motherboard not being able to deal with 4 sticks (in any combination) or the problem is due to slightly different RAM specs between the sticks. Needs more testing.

[RichS]

Well, it's been a week with the mixed positions so I'm going to call it good. PassMark PerformanceTest numbers are about identical to a year ago (other than Available RAM) and BurnInTest has run for billions/trillions of transactions, and several hours.

My engineering brain is tempted to swap them again to validate, but I'll just leave it as is.

Weird.

[leclod]

Well, it's been a week with the mixed positions so I'm going to call it good.

What do you mean with mixed ?
You have 2 channels, one in slot 1 and 3, the second in slot 2 and 4.
You should populate those 2 channels identically.

[RichS]

By interleave I mean do not put the original memory in slots 1 and 3, where they originally were, and put the new memory in slots 2 and 4, as would be logical, but is also what caused it to crash regularly.

Instead, put the original memory in slots 1 and 2 and the new memory in slots 3 and 4, so each channel has one stick each of old and new memory (i.e. "mixed").

2+ weeks now and all is fine, as was all bench-marking and stress testing, as previously noted.