Where Is The Problem? EP45-UD3R + F2-8500CL5D-4GBPK

[spiffae]

In order to troubleshoot a breaking-down old build I've got myself a nice new system:

Motherboard: Gigabyte GA-EP45C-UD3R
CPU: Intel Dual Core E8500
Memory: G.Skill F2-8500CL5D (2x2GB)
GPU: EVGA GeForce 9800GTX+ (256MB)

The HDDs, Optical Drives, and PSU are a mixed bunch of hardware between 1-3 years old.

This MAY be important: I didn't upgrade the PSU - it is a 3 year old cooler master 450W PSU. Here is a picture of the side panel: http://img222.imageshack.us/img222/3236/1717100103.jpg

First things first - I put in the RAM (slots 1 and 3) and booted up the system without even looking at the BIOS, and it Blue Screened while I was installing Windows. A second try at installing got me into the OS but I got another BSOD while adjusting desktop resolution moments after I had booted up.

I popped in my Memtest 86 boot disk, and rebooted. At the first test (a couple seconds after the machine starts) the computer hard reboots. This happens in slots 1-3 and 2-4 - at exactly the same time in both slots. It happens if the ram is clocked at 1066 and 800. It happens with stock settings or tweaked settings.

I took one stick out and boom, no problems. I swapped sticks into slot 1, and the other one also tested without problems. (I am writing this on a computer with only 2 gigs of memory )

At this point I did some research and came across this forum. I read the messages carefully and worked with the suggestions in this thread: http://gskill.us/forum/showthread.php?t=1169 - I followed the recommendations and there was no change. The settings made no difference.

I know the memory is good, but I'm concerned. Occaisonally the system POSTS without giving me a RAM count, and once or twice when I had an internal card reader connected the computer hung as it POSTED with only "Me" of "Memory allocated ..." or whatever it says before it counts the RAM on the screen. After about five second hung with the "Me " on the screen, it would reboot and do it again. I disconnected the card reader and that hasn't happened again, but I'm obviously nervous. Yesterday, for the first time I got an "IRQL_LESS_THAN_OR_EQUAL" BSOD while using windows with the once stick.

I realize this is probably the motherboard and that will be the end of it. That said, I am concerned about the PSU - I think it has been making intermittent clicking noises. Could the PSU be the cause of all this? My gut tells me no, but I don't know.

I'm ready to try anything. I'm going to borrow a couple sticks of known working DDR2 800 ram later today and try to run them dual channel to see what happens. I'm just completely stumped and am dying for a stable system I can count on.

[spiffae]

A disconcerting update: I threw in two sticks (2x1GB) of Corsair XMS DDR800 from an older build I had. The system so far has been rock solid in dual channel, and ran twice through memtest with not a peep.

So... is it the BIOS settings with the G.Skill? What am I doing wrong?

[Tradesman]

Couple of things to look for, 1st which actual set do you have (after the CL5D in your model # there should be something like '-4GB**' ....the ** in there actually identifies the specific set within the 8500/CL5D series, there are a number of sets with different timings and voltage settings..

You could start by trying the sticks in 2-4....I know, not what manual says, but with this board and their UD3P, sometimes the secondary sockets some sticks better than the first, have seen where say GSkill is fine in 1-3, but Mushkin won't even boot, but is fine in 2-4 (just background)

Anyway, if that's a no go, then with your CPU/mobo combo best bet is to enter the timings per spec manually, you'll want command rate at 2T, voltage to start at the high end of spec (i.e. if your set states 2.0-2.1 then start at 2.1, if it only shows 1.8 then go 1.8 (you can always reduce it once stable) and check settings on your advanced timings, if all are on auto (probably), jot down the values, if any need to be changed, it will probably be tRFC (maybe tRC).

Other voltages are prob good at auto unless you have an OC going (which you may well have, I would with the 8500, with a decent HS (thirdparty, you can easily run 3.8-4+ GHz on this mobo, though you will want a bigger PSU, I'd say at least 600 watt to allow for future upgrades (i.e. to 8GB or higher GPU, because this could esily last you a year or two...or more).

If the above doesn't work let me know the voltages and other timing setings, I can prob throw an 8500 in one of my UD3Rs and see what happens, (may even have the same sticks your using.)

[Tradesman]

Sorry bout that you have the PK set, just saw it in the title, was reading the thread, so timings are 5-5-5-15 volts at 2.1 to start.

[genetix]

Curious, 'boom'. Would also like to know what FSB the computer is in. Doesn't exactly sound good to me since you've tested 5-5-5-15 2.1v already as suggested above. (as you say you've read the G.Skill post on this board).

This memory however should run straight from the box. Only real thing which comes to my mind is boards sometimes saves some configs solid to settings. I'd try CMOS reset of board and then add the sticks in (probably loading the BIOS defaults would also do, but nothing beats old good CMOS reset). Other than that check the latest BIOS (which you've already probably done).

-edit-

I know the memory is good, but I'm concerned. Occaisonally the system POSTS without giving me a RAM count, and once or twice when I had an internal card reader connected the computer hung as it POSTED with only "Me" of "Memory allocated ..." or whatever it says before it counts the RAM on the screen. After about five second hung with the "Me " on the screen, it would reboot and do it again. I disconnected the card reader and that hasn't happened again, but I'm obviously nervous. Yesterday, for the first time I got an "IRQL_LESS_THAN_OR_EQUAL" BSOD while using windows with the once stick.

The freeze I could pretty much explain by North bridge / MCH voltages. Too low RAM voltage could cause this, but even with way too low RAM voltage board will POST clean with error, if the NB / MCH are good. Same goes to that Microsoft false BSOD. "It literally in english" means "We don't know what the hell is wrong, but we won't allow it..".

-edit #02-

Hmm, will be easier to help, if you tell me on 1 stick in the board(what ever you can boot). What does the MCH at 1.4v, MCH Reference and MCH/DRAM Reference say voltage to be at BIOS. Also, check the values on DRAM Voltage and DRAM Termination while at it at DRAM Voltage on 2.1v.

[spiffae]

Thanks for all the help guys - I'll get a chance to poke around with all these suggestions tomorrow and I'll let you know what happens.

Tradesman, I initially suspected a bad DIMM slot, and tried 2-4 immediately. Identical crash and reboot in memtest86 (at exactly the same moment).

[Tradesman]

spiffae,
Glad to know smart minds think alike on trying the other slots, haha, but do let us know what happens, I'm normally in and out late afternoons and evenings, though I try and check in in the early AM before I start working, so will be looking. GSkill is pretty much office type hours, but thankfully also jumps in at unexpected times and genetix is in an out sort of like me, at all kinds or weird hours

[spiffae]

So... big mistake. When I went to update the BIOS the first time around my internet connection blocked the BIOS version checker, leading it to falsely claim my BIOS was up to date. I've just found out I'm using F6, and they are up to F11. Flashing now, will update.

/stupid

[spiffae]

<big sigh> that wasn't it

BIOS flashed, system booted up, I gave it the benefit of the doubt and sat in windows for a few minutes. Nothing happened and I became hopeful. I use a little windows app called memtest, and ran it. Almost immediately I was getting messages like this:



clicking OK would pop up another message with the next address (they were often one digit apart, but seemed to skip sometimes. I clicked about 50 of them and then at about 0.9% the computer BSOD'ed with a message like "The system attempted to access memory at an illegal address" or something. I rebooted and booted Memtest86+ and lo and behold... it reboots during the first test. Same as ever. I have checked all the BIOS settings, and I am running at 2.1v.

I will run memtest (in windows) with the one stick and see what happens, and with the corsair 2x1GBs.

EDIT: Ran memtest with the single stick of G.Skill and no problems. My machine is rock solid with just one stick (either one)

[spiffae]

just ran the Gigabyte motherboard tuner/monitor and noticed this:



Is that normal?

[spiffae]

Hmm, will be easier to help, if you tell me on 1 stick in the board(what ever you can boot). What does the MCH at 1.4v, MCH Reference and MCH/DRAM Reference say voltage to be at BIOS. Also, check the values on DRAM Voltage and DRAM Termination while at it at DRAM Voltage on 2.1v.

FYI I can boot with two sticks, but it is unstable. With the one stick this is what I have in BIOS

MCH Core 1.100V [Auto]
MCH Reference 0.760V [Auto]
MCH/DRAM Reference 0.900V [Auto]

(they all say "auto" after that - do you want me to change core to 1.4?)

The DRAM looks like this:

DRAM Voltage 1.800V [2.100V]
DRAM Termination 0.900V [Auto]
and all following are 0.900V and [Auto]

[genetix]

(On screenshot those are correct although showing only JEDEC standard so that's why it's so.)

---

Well, since I started helping gonna help this one..

Aim to these voltages:
MCH Core: 1.200v
MCH Reference 0,830v
MCH/DRAM Reference 0.982v

(You can calculate these again hmm, little faulty method would be:
0,760 / 1.100(original) * 1.300(new) = MCH Reference for 1.3v MCH Core
0,900 / 1.100(original) * 1.300(new) = MCH/DRAM Reference for 1.3v MCH Core
-as said this is faulty but accurate enough to see, if the system stabilizes with higher MCH Core "and all this only if the board doesn't automatically calculate with 'Auto' when MCH Core is changed.)

DRAM Termination <- this should be -0.016v from 1/2 of DRAM voltage. Don't understand why it's not set automatically. = on 2.1v DRAM, "DRAM Termination" should be near 1.034-1.050v, if it's not aim to that range. (do not overvolt it need to be +/- 0.02v absolute maximum error preferable lower.)

Should be enough to stabilize the machine. I don't think 1.4v would be nessessary nor required. It's just an minor instability.

[spiffae]

thanks for the reply genetix - I'll be out of town until sunday night so I'll plug those settings in then... I don't really understand why this would be happening though, even at DDR800. Any ideas?

[genetix]

Have no clue why this would happen kinda weird actually. I just know where to start looking with explained issue. This could/should even correct with simply BIOS update, but since it didn't trying something else to compensate the board stability.

-edit-

Fail to mention above dunno, if you understood correctly. When you modify MCH Core or DRAM Voltage don't change the references voltages if the 'Auto' changes those automatically. 'Auto' on those would be more accurate than can be added, however, like in DRAM Terminate seems board doesn't wanna automate those. I'd also suggest keeping MCH core always below 1.36v. Usual would be 1.20-1.30v even while OC.

[spiffae]

Well I've got to say I'm completely stumped and pretty disheartened. Setting the MCH levels to what you recommended made my problem WORSE. I booted into windows (using the W7 developer preview FYI) and instantly got the PFN_LIST_CORRUPT BSOD. Rebooted, and got the BAD_POOL_HEADER before I even saw the desktop. I shut down, put the Corsair back in, set everything to default and here I am.

What can I possibly do?

One other question: Is it at all possible that this could have something to do with the power supply? Last night I had a hard lockup almost identical to the ones I was getting on my last build, and the only thing I didn't change from that build to this one is the hard drives and PSU. Could this be the root of it all?