my replacement memory to replace my failed memory failed [update: or not...?]

[hitbyambulance]

This is regarding my PC build in this thread (actual build here: http://www.gamingtrend.com/forums/index.php?topic=38662.msg682681#msg682681).

I purchased an additional set of the RAM I had already (A-DATA Gaming Series 4GB (2 x 2GB) 240-Pin DDR3 SDRAM DDR3 1600 (PC3 12800) Desktop Memory Model AX3U1600GB2G9-2G).

This set, when installed, caused Win7 to BSOD immediately upon booting. MemTest86+ started spewing errors immediately upon running. I RMA'd the RAM with newegg and got a replacement set. The replacements seemed to work OK for a while and displayed no errors in MemTest. However, when I started playing games, I would get interesting issues such as King's Bounty crashing to desktop (a game which had never exhibited any sort of stability problem before this). When I had left the system off for a few days and powered it back on, it would BSOD a few minutes after booting into Windows. A second MemTest displayed nothing, but Windows Memory Test displayed error messages. A *third* run through MemTest finally started displaying errors.

Since I'm now outside of newegg's 30-day return window, I just contacted A-DATA on their website to see if I can get support through them.

I have removed my older set of memory and installed this set into the first two banks of the motherboard by itself - still displays errors. I do wish to mention that I have an older set of this memory purchased back in April that has given me no problems, displays no errors in MemTest86+ running overnight several times, and has worked well so far.

I take great care in installing the RAM, like using an anti-static wristband, not handing the RAM by the contacts or dropping it and supporting the mobo so it doesn't flex too much when the RAM is installed. Is it outside the realm of possibility that the second set of memory banks in the motherboard itself is damaging the RAM? If/when A-DATA sends me a new set, I'm going to install that into the first set of banks and my original trouble-free RAM into the second set of banks. I'm going to be *highly* annoyed if that could be the issue...

[TheAtomicKid]

Check bios to insure proper voltages and speeds? A lot of ram seems to exhibit issues with not defaulting properly.

In particular, if it IS defaulting to the proper speed/voltage... try slowing it down and see if the errors go away. At which point you can call A-Data and howl at them about selling you faulty product that doesn't perform as advertised, as opposed to a situation where they tell you you nuked the ram installing it :/

Atomic

[hitbyambulance]

Check bios to insure proper voltages and speeds? A lot of ram seems to exhibit issues with not defaulting properly.

In particular, if it IS defaulting to the proper speed/voltage... try slowing it down and see if the errors go away. At which point you can call A-Data and howl at them about selling you faulty product that doesn't perform as advertised, as opposed to a situation where they tell you you nuked the ram installing it :/

Atomic

they currently *are* underclocked (running at 1333 instead of the spec'd 1600) - i can run 4GB at 1600 (and weirdly enough, no errors in MemTest are displayed with the faulty memory when run at this speed), but when i try 8GB i get a BIOS warning message stating that the boot failed due to overclocking. i then attempted to adjust the timings from Auto to 8-8-8-24 and boosting the voltage all the way up to 1.64, but the same error appears. if i use those new settings and drop the timing back to 1333, then it boots, but still shows errors in MemTest. frustrating.

[hitbyambulance]

update: i'm starting to wonder if the problem resides with ... the CPU?

http://forum.giga-byte.co.uk/index.php?topic=1930.15

[Thin_J]

You know, since the memory controller is on the CPU that actually makes perfect sense.

[TheAtomicKid]

After reading the responses, that was the first thing that came into my head. You might also check out things like running with 4 dimms instead of 2, etc. That was an old AMD limitation that I thought had long since been overcome... perhaps not?

Try things like testing the memory 1 stick at a time, in dimm 0 or dimm 1, whichever one is 'first' from the cpu. Aka, the closest dimm to the actual cpu location. That will be the one closest to the cpu electrically, as well.

If the memory works at speed, but only 1 or 2 dimms at a time, aka 1 dimm per channel... then the failure isn't with the memory itself, per se.

Atomic

[Knightshade Dragon]

I had similar problems and it ended up being my board.  The memory would test out perfectly but paired I'd only get half speed.  It's hard to see the forest for the trees when it's your board...

[hitbyambulance]

Update:

I wrote AMD with a description of my problem. Here is the response:

Your service request has been reviewed and updated.

Response and Service Request History:

Thankyou for contacting AMD.

According to the specifications for the Phenom II line found here: http://www.amd.com/us/products/desktop/processors/phenom-ii/Pages/phenom-ii-key-architectural-features.aspx DDR3-1333 is the highest supported memory type.  The issues that you are experiencing are due to a damaged memory controller on the processor from running it with DDR3-1600.  This is why going to 1333 RAM isn't showing any improved results.

Your best option at this point is to return the processor to your retailer for a replacement.  If they are unable to assist you, you can contact us again for an RMA, so long as you have a valid warranty.

OK, i had been completely and utterly unaware that running the 1600 speed memory at 1600 speeds on a Phenom II WOULD FREAKIN DAMAGE THE CPU'S MEMORY CONTROLLER! nothing in my investigatings and queryings (including from here) had revealed that this was at all a risk, nor was this ever pointed out to me. is AMD feeding me a line, or did the interwebs fail me?

[TheAtomicKid]

If it's true, some simple testing with a replacement cpu at 1333 speeds, will verify their claim...

Atomic

edit: tbh, I've never heard of someone _damaging_ the onboard memory controller by trying to run it at 1600 mhz... it's theoretically possible... the part isn't spec'd for more than 1333. But it'd definitely be a first.

[disarm]

i don't know if AMD is feeding you a line or not, but i would take them up on the replacement offer if you're still covered under warranty.  if there's any chance that the CPU is no longer functioning correctly, you might as get it replaced for free...can't hurt and might help to fix the problem.

[hitbyambulance]

AMD finally issued an RMA for my processor and it's now packed, boxed and in a shippable state - will drop off at FedEx on Monday during my lunch break. they were nice enough to spring for postage both ways.

i'm thinking that since i'm going to have to re-install the processor, re-apply thermal compound and re-install the fan, i might want to replace the stock CPU cooler i had been using with a more powerful fan. is this even necessary, and if so, what would be a decent choice for an AMD Phenom II X4 925?

[hitbyambulance]

futher update: one of the replacement sticks of RAM was bad after all. currently requesting RMA from A-DATA.

totally ridiculous...i will not recommend A-DATA DRAM to anyone at this point..

[ravenvii]

This is why I don't bother with custom-built PCs those days. Games are no longer teetering on the cutting edge of technology anymore, and manufactured PCs are extremely cheap those days. Just grab a pre-built one, and upgrade the RAM and graphics card yourself. Easy as pie and you don't spend that much more money either.

[hitbyambulance]

i'm happy to report the RMA'd memory fixed my crashing problems.