A little help please with a barebones kit. I got an Asus M4A88T-M barebones kit with AMD PhenomII 1045T X6 cpu from TigerDirect, and it's been intermittently unstable, especially with any install process (OS or other). On installing Windows (any version), the system experienced instability, various system errors, & blue screen crashes, often (~50%) IRQL_NOT_LESS_OR_EQUAL.
barebones kit consisted of:
ASUS M4A88T-M AMD 880G Socket AM3 Motherboard
AMD PHENOM II X6 1045T THUBAN 2.7GHZ AM3 CPU
Corsair XMS3 4GB PC10666 DDR3 1333MHz (2 X)
Thermaltake V2 Mid Tower ATX Case with 450W PSU
Seagate 1.5TB LP Serial ATA HD 5900/64MB/SATA-6G
Lite ON 24X DVDRW SATA OEM
Corsair GS600 600 watt Power Supply (from Best Buy, to replace Thermaltake 450W PSU)
Installed Windows 7 (32) on first day, and ran overnight? (Thursday 4/12) without problems. After applying Asus overclocking utility, started experiencing multiple, various BSOD & system errors. Was unable to re-install Windows 7, or other versions without errors. Reset BIOS, then after multiple re-tries, managed to install Windows 7, updates and base applications. Ran Chkdsk /r/f C: twice now, with no problems. Ran Microsoft memory diagnostics, no problems. Ran Hot CPU Tester Pro to do live overclock diagnostic testing. The diagnostic ran all 6 cores at 100%, overclocked up to 10%(? - "Rocket") for 1 hour at each of 3 overclock settings. The diagnostic reported HD test failures, "file inaccessible or corrupt" during the test runs, but stopped that test module, yet reported the test complete when finished. On extended 6 hour test, the video went random (image hashed), and system had to be reset. Then the diagnostic ran the full 6 hours at normal not-overclocked setting without problem.When I went to install Windows 7 (64 bit), it kept getting system errors and BSOD during OS install.
I read somewhere that someone had to set their VCORE voltage to 1.3V, so I adjusted that to 1.34 in Bios so that the HW monitor would read 1.30. When running Windows, it would bounce around from 1.17 to 1.30 to 1.41. I've set it as high as 1.35 from BIOS. The Asus overclocking utility says it sets it to 1.375 for any of the overclock profiles.
Are there other BIOS settings I should try tweaking, possibly chipset voltage, and to what value, or should I just send the mobo back to Asus for replacement?
I'd like some guidance on what to do from here. Right now, I'm unable to install an OS.
I had this running again (Win 7-64) with memory Channel Interleave disabled, and DCT Unganged set to Auto. Then I left it in sleep mode overnight, restarted this morning, and it BSOD'd when I went to install the next round of updates, then again when I tried system repair. It appears to be temperature related. I've been reading some reports of a systemic design flaw in the Phenom memory controller with two or more 1333 memory modules. I'm going to try boosting memory voltage up a step (0.015) at a time up to 1.650 or until it keeps working all the way. If that doesn't work, the last resort suggested solution is to back the mem frequency to 1066.
By Jose Vilches, TechSpot.com
Published: February 13, 2009, 6:03 PM EST
Barely a week after the introduction of new Phenom II processors for the AM3 socket, reports are beginning to emerge about an alleged bug within the DDR3 memory sub-system. Specifically, a revision guide document for the processor mentions that on machines with more than one DDR3-1333 (or greater) memory modules installed per channel, users may experience “unreliable operation.”
Without getting into the specifics of the symptoms, AMD’s Damon Muzny has acknowledged this issue, and also took the time to clarify the situation. According to him, internal testing showed that with certain lower quality memory modules and all four slots populated (2 modules per channel), inter-component signaling could become erratic and result in system instability. Muzny says they were faced with a choice: to either drop listing support for DDR3-1333 altogether or design the system to down clock the memory to 1066 – they took the most conservative approach and went with the latter to ensure stability.
He went on to suggest how users could manually configure memory to compensate for the slight drop in performance and assured customers that the issue is of a software nature and that it will be fixed shortly.
06-02-2011, 11:30 PM http://tech.icrontic.com/news/amd-explains-phenom-ii-me...
Damon Muzny, a PR representative at AMD, took the time to explain the situation. Yes, the Phenom II doesn’t “officially” support four sticks of DDR3-1333 or DDR2-1066. Internal testing showed that four sticks triggered instability in multiple test platforms when using standard JEDEC voltages and speeds. Two high-speed sticks on a single channel can push the memory controller’s limits; combine that with subtle variations in motherboard architectures, and inter-component signaling can become erratic.
“We had to make a choice,” Muzny said. “It was enough of a problem that we had to make a choice.”
Overclockers know that a bump in voltage is the easiest way to fix memory instability. A little extra juice in the system keeps things working correctly and will allow users to fill all four memory slots on their motherboards.
According to Muzny, AMD had three options for rectifying the Phenom II’s memory situation:
No official support of DDR3-1333 and DDR2-1066 in any capacities.
Require motherboards to run outside of JEDEC voltage specifications when they detect two memory sticks on a single channel.
Clock memory down to fail-safe speeds.
Muzny said not supporting DDR3-1333 and DDR2-1066 would be a terrible choice because both speeds work, and requiring motherboards to run out of official specifications still left room for instability or legal troubles. Downclocking was the safest bet catering to the lowest common denominator and, as such, motherboard manufacturers were instructed to have the BIOS downclock DDR3-1333 to DDR3-1066 and DDR2-1066 to DDR2-800 to ensure stability.
“We have to take the most conservative approach,” Muzny said.
Overclockers and enthusiasts are still encouraged to manually set memory parameters for stable operation. After all, the most popular overclocking RAM already runs well outside specification. In the end, it’s nothing new. It’s just AMD playing things safe.
According to Muzny, future chip revisions could address the issue, but the problems aren’t just caused by the Phenom II’s memory controller. Motherboard manufacturers will need to use quality components to ensure clean, clear signals if they want to do their part.
For whatever it's worth... I've run this board in all kinds of settings, memory sticks in different slots, VCORE up to 1.35 (I think), VDIMM up to 1.65, memory frequency downgraded to 1067, bank and channel interleaving turned off, unganged forced on, and drive in all 6 SATA slots. Every time, doing installs or coming out of sleep into a cold processor has chance of causing system error or BSOD. I actually made a big table of all the combinations I've tried.
Now, for the first time ever, I ran with only one 4GB memory stick, all default BIOS settings, and it did a full base install: full Win 7 64 install, all upgrades, browsers and Adobe's, and has been running stress diagnostic for 2 hours now. Including multiple cold boots. No errors whatsoever.
Apparently there's more to this buggy Phenom memory controller than AMD is willing to admit. I guess I can run with one 4G stick for now, and keep the other stick in case some solution for underclocking the problem away turns up.