Jump to content
Corsair LogoCorsair Logo
Corsair Community
CORSAIR COMMUNITY Survey, we want to hear your feedback!

Corsair TwinX1024-3200C2PT 2x512 MB

monsterbuilder

Recommended Posts

monsterbuilder

monsterbuilder

Posted
Posted

Hi,

 

I have a pair of sticks (Corsair TwinX1024-3200C2PT 2x512 MB) installed in a Chaintech 9CJS Zenith board. Proc is a P4 3.0E 800 FSB 1 MB L2. It is the second of two identical machines that I use exclusively for Stanford's Folding@Home project.

 

Normal use, i.e. accessing internet, causes no apparent problem. However, when Folding, I experience multiple Folding core crashes (i.e. Early_Unit_End). Having swapped components between the two machines, I think I've narrowed it to the RAM, so...

 

Ran MemTest - multiple errors. Let it run for 29 passes, had failing addresses on Pass 0, 2, 8, 23, 23, and 24 - all on Test 7. I've recorded the details - if you need to see them, let me know and I'll followup in this post.

 

I haven't tried it with one stick and then the other to narrow it to a specific stick... but since I need a matched pair to run dual channel, I don't know that it matters if only one is failing, since I am assuming that both might have to be replaced to get me back to a matched pair again.

 

The machine is set to stock settings - no OC on the Folding Boxes - and the RAM timings are set automatically (SPD) in the BIOS - they match those recommended by Corsair.

 

What do I need to do from here?

 

Thanks!

 

M1

Link to comment
Share on other sites
  • Corsair Employees
RAM GUY

RAM GUY

Posted
  • Corsair Employees
Posted
Please set the Dim Voltage to 2.7X volts and then set the timings manually to the tested settings for the specific module you have (Cass 2-3-2-6 with the Rass to Cass set to "3"), and then test the module/modules one at a time with www.memtest.org! If you still get errors, please follow the link in my signature “I think I have a bad part!” and we will be happy to replace them or it! However, if you get errors with both modules that would suggest some other problem and I would test them in another system or MB to be sure. In addition, with this MB you have to disable legacy USB in the bios when running any memory test.
Link to comment
Share on other sites
monsterbuilder

monsterbuilder

Posted
  • Author
Posted

Running dual channel (as you know) on the 9CJS (intel) is pretty much straightforward: My pair is/was installed in DIMM's 1 and 3.

 

I pulled the original stick from DIMM3, leaving DIMM1 in its original slot, then went into the BIOS. I bumped the voltage up as you recommended, and set the timings manually per your instructions. I didn't see a legacy USB controller in the BIOS, but I did find <USB Controller> and <USB 2.0 Controller>. The former is probably the legacy controller, whereas the latter probably supports only USB 2.0. To be on the safe side, I disabled both.

 

Rebooted into MemTest; after roughly 1:25:00 hours:minutes:seconds, a number of errors appeared on the DIMM1 stick:

 

Test Pass Failing Add'y Good Bad Err-Bits Count

4 2 00018e540b8 - 393.2MB 53a39c0c 553a3bc0c 00032000 13

 

I pulled the DIMM1 stick, put the DIMM3 stick back into the DIMM3 slot, and rebooted into MemTest.

 

After roughly 0:35:00 hours:minutes:seconds, a couple of errors appeared, each one during the second pass on Test 7... there was no point in continuing MemTest once these errors appeared, so I took the machine down.

 

To try and rule out the memory being the problem, I've swapped the identical sticks between the two 9CJS machines and set-up MemTest to run on both through the night.

 

I'll post the results here sometime during the morning of 12.13.2005 (Tuesday - which is now today).

 

Thanks again for your help on this.

 

M1

Link to comment
Share on other sites
monsterbuilder

monsterbuilder

Posted
  • Author
Posted

Hi Ram Guy,

 

UPDATE TIME:

I might have opened Pandora's Box here.

 

Per my previous post, I pulled the identical sticks from the twin Chaintech 9CJS box and swapped them with the two sticks from the problem 9CJS box.

 

I modified the settings of the twin 9CJS to match those you recommended for testing the problem 9CJS.

 

In each case, slots used are DIMM1 and DIMM3.

 

Started MemTest on both machines and left it overnight.

 

PROBLEM BOX w/sticks from TWIN BOX:

 

No errors reported until the 12th pass; from 12th pass to the 20th pass, there are 26 errors; 22 errors in Test #6, 4 errors in Test #7.

 

Interestingly enough, one particular failing address reappears multiple times, and the numbers appearing in the <GOOD> and <BAD> columns is the same regardless of the failing address, at least in Test #6. I cannot scroll down, even with Scroll Lock unlocked...

 

These results are disconcerting - the twin box from where I swapped the sticks is a Stanford Folding@Home dedicated machine that has run the memory-intensive QMD core (uses roughly 327 MB of RAM while folding a QMD work unit) for months without a single EARLY_UNIT_END error...

 

I would have assumed by that result alone that there were no problems with the RAM, but...

 

TWIN BOX w/sticks from PROBLEM BOX:

 

There are 5 total errors after 21 passes. Four appear at the same failing address on Pass 0 under Test #4. The fifth error shows up on Pass 15, also under Test #4, at a different failing address.

 

**********

 

Wow... what to do next? At the moment (and I'm pipe-dreaming as I type this), I'm wishing I'd stayed up with the machines during the night... if I'd let both run 14 passes and then stopped the tests, then we'd only see the Pass 0 errors in the Twin box running MemTest on the sticks from the Problem box. MemTest on the Problem box running the sticks from the Twin box wouldn't have yet shown an error.:confused:

 

What can be concluded from these results?

 

1. The sticks from the Twin box have some obscure errors that only appear after multiple passes of MemTest?

 

2. The sticks from the Problem box have significant problems that appear as errors on Pass 0 (and subsequent passes) regardless of which box they are installed in when running MemTest?

 

I'm scratching my head here... what next? Retesting in DIMM2 & Dimm4? Retesting the swapped sticks one at the time in both boxes?

 

M1

 

(P.S. Heading back to work - shutting down both boxes until I return - if further testing is required, please advise what it should be and I'll run them when I return home this evening. Thanks again!)

 

LATE NIGHT UPDATE

 

More results to ponder...

 

For the moment, I've restarted each Chaintech 9CJS Zenith box. The Problem Box still holds the RAM from the Twin Box and the Twin Box still holds the RAM from the Problem Box. SO...

 

PROBLEM BOX w/sticks from TWIN BOX:

 

I have Folding@Home set-up on both machines to run as a service on startup, so when I restarted the machine (retaining the settings you recommended for MemTest), it promptly resumed EARLY_UNIT_END errors of the QMD core. Seeing as I'm now running the RAM from the Twin (where there have been zero crashes), I'm of the opinion that despite the errors found using MemTest, the problem isn't the RAM. After 4 core crashes, I killed Folding@Home as a service on the machine and restarted it as a console (i.e. running at a command line level) so I could observe it running in realtime rather than have to constantly access the logfile to see what was going on. Get this: It has run perfectly stable going on 5 hours now.

 

I was mystified before... now I'm completely baffled.

 

I will let this machine run through the night and see how it progresses. If it completes the current work unit without a crash and proceeds to complete a couple more without crashing, then I'm inclined to think that the original problem that led me to think I had a RAM issue is actually software-based... perhaps some kind of glitch in the set-up of XP or Folding@Home.

 

TWIN BOX w/sticks from PROBLEM BOX:

 

Restarted normally, keeping the settings you recommended for MemTest... machine picked up where it left off a couple of days ago. Perfect.

 

Folding@Home is running as a service without a hint of instability, despite the fact that it's using the RAM from the Problem Box, and that MemTest returned errors for these sticks as well.

 

WHAT NEXT?

 

Do you see a need to RMA any of these four sticks? Frankly, at this point, I don't think the RAM is a problem... but I can't explain away the consistent appearance of errors under MemTest.

 

I think I read somewhere that MemTest isn't 100% perfect, i.e. no errors doesn't always mean the RAM is good, nor does the appearance of errors always mean the RAM is bad. Is this correct?

 

I guess if you think I ought to send them in, I'll do so... I'll update this post again in the morning. Hopefully the console running on the Problem Box will prove to be stable.

Link to comment
Share on other sites
monsterbuilder

monsterbuilder

Posted
  • Author
Posted

Sure thing - I won't have access to the machines until later this evening, so it'll be toward midnight before I can post the configs.

 

I tend to agree with you, though - the problem I am seeing in the Folding@Home program didn't follow the swapped RAM... I'm inclined to conclude that the multiple errors in MemTest do not mean errors in the RAM this time...

 

More later!

Link to comment
Share on other sites
monsterbuilder

monsterbuilder

Posted
  • Author
Posted

Hi Ram Guy,

 

It's late out East, and I'm about to drop, so if I fail to provide something you require regarding system info, let me know and I'll dig up missing data tomorrow evening (12.15.2005).

 

PROBLEM BOX w/sticks from TWIN BOX:

 

Mainboard Chaintech 9CJS Zenith

Chipset i875P

Southbridge 82801EB (ICH5)

CPU P4 3.0E Prescott 800 FSB 1 MB L2

CPU Core Speed 3042.5 MHz

Multiplier x15

FSB 202.8 MHz (slight OC due to selecting preformance option in BIOS)

HT enabled

Memory Corsair TwinX1024-3200C2PT 2x512 MB Dual Channel Perf. Mode Enabled

Frequency 202.8 MHz

FSB:DRAM 1:1

CAS# Latency 2.0

RAS# to CAS# Delay 3

RAS# Precharge 2

Cycle Time (Tras) 6

PSU Enermax EG365P-VE (350 W)

Phoenix BIOS 6.00 PG dated 12.29.2004

 

TWIN BOX w/sticks from PROBLEM BOX:

 

Mainboard Chaintech 9CJS Zenith

Chipset i875P

Southbridge 82801EB (ICH5)

CPU P4 3.0E Prescott 800 FSB 1 MB L2

CPU Core Speed 3042.5 MHz

Multiplier x15

FSB 202.8 MHz (slight OC due to selecting preformance option in BIOS)

HT enabled

Memory Corsair TwinX1024-3200C2PT 2x512 MB Dual Channel Perf. Mode Enabled

Frequency 202.8 MHz

FSB:DRAM 1:1

CAS# Latency 2.0

RAS# to CAS# Delay 3

RAS# Precharge 2

Cycle Time (Tras) 6

PSU Cooler Master RS-350-AMSR

Phoenix BIOS dated ??? is most recent flash offered by Chaintech

 

UPDATE:

 

Twin Box running without incident since reboot into Windows. Folding@Home core shows no sign of instability.

 

Problem Box running without incident since reboot into Windows. I left Folding@Home running this morning without incident from the command line, however at some point during the day, the console version shut down spontaneously. The automatic service start took over at that point and restarted F@H as a service. It spontaneously shut down within a couple of hours, but since it is set to automatic (as a service), it immediately restarted, and here 7 hours later it has run without incident.

 

The more I test, the less this seems like a RAM problem to me.

 

I'm considering taking down the problem box and rebuilding it from the ground up. This problem might turn out to be nothing more than a loose component...

 

In the meantime:

 

1. Any further testing suggestions?

 

2. These machines are only mildly OC'd... would you recommend going permanently to different settings than what SPD will assign as optimized performance defaults?

 

Thanks again!

 

M1

Link to comment
Share on other sites
monsterbuilder

monsterbuilder

Posted
  • Author
Posted

At Cas 2-3-3-6, would you keep the voltage at 2.7V or should I dial it back to 2.5V?

 

Sorry I omitted the video card information - these are not gaming machines, they are primarily Folding boxes, hence I went with cheap video card solutions. The Problem Box has a 32mb GeForce 2 video card; the Twin has a 64 mb card... nothing better than GeForce 3.

 

I didn't figure these to be power hogs....

 

I'll reset the timings and disable performance mode tonight, restart, and see what happens. I'll post back with results late.

Link to comment
Share on other sites
monsterbuilder

monsterbuilder

Posted
  • Author
Posted

Hi Ram Guy,

 

Sorry I've been wrapped up elsewhere (Christmas shopping, etc) and wasn't able to get back to the problem until now.

 

Checking in the BIOS (AwardBIOS v6.00PG), specifically JS79BC0CC-00 dated 12/29/2004 (latest offered by Chaintech), I see that I can adjust the DRAM voltage in 0.1V increments, so I can't get to 2.75V. I'll stick with 2.7V for now.

 

I did change the Precharge from 2 to 3, so timings are now 2-3-3-6 per your latest recommendation. I didn't see an option for performance mode (disable or auto)... was this for the machine as a whole, or just for the RAM?

 

As for a 450W PSU... I have an Antec Neo 480W awaiting my next build, but that's weeks away, so I'll have a go at swapping and see if this helps.

 

Some further observations on the problem box:

 

I was using this specific machine to surf the 'Net last night when suddenly "BANG!" Well, surprise, spontaneous reboot. Looking back and seeing in the Folding logfile the number of times the F@H program has restarted (it is setup as a service) without any notation indicating why the session terminated, I'm starting to think the problem might be in one of two areas - 1) CPU overheating... I'm running a stock cooler in this machine, and it IS a prescott CPU, although CPU temp doesn't seem high, and the side of the case has never been installed. The Twin box is running a ZalMan CNPS7700-Cu. 2) This box is loaded with XP, but it began life as SP1. It's fully updated via the Microsoft website to SP2... but perhaps there are some gremlins in there somewhere... ALL my other boxes are SP2 right off the MS CD-ROMs with the exception of a Debian box.

 

My thinking again is that the RAM is probably ok, but that I have either an overheat problem or the torturous update path from SP1 to SP2 might be the source of the problem.

 

SO.... I'm going to let it run with the new settings you offered, then tear the machine down to the mobo standoffs and rebuild... this time including a Zalman HSF identical to the Twin box AND going with a fresh load of SP2.

 

If at that point I'm still crashing, I'll take you up on the RMA offer... but if I can avoid that, I'd sure like to!

 

Again, thanks for your assistance. I'll post back as soon as I get all this taken care of.

 

Assuming I wind up going the RMA route, what days during the Holiday week will it be impossible to get an RMA number?

 

Merry Christmas!

 

M1

Link to comment
Share on other sites
  • Corsair Employees
RAM GUY

RAM GUY

Posted
  • Corsair Employees
Posted

Merry Christmas,

We will be here thru this friday and we will not be here the day after christmas. But you can submit the request if needed any time and we will process it ASAP. You can follow the link in my signature, but I would suggest TSXpress for faster service.

Link to comment
Share on other sites
monsterbuilder

monsterbuilder

Posted
  • Author
Posted

Don't give up yet!:):

 

I've replaced the PSU with the Antec Neo 480W I mentioned in the previous post - also did the rebuild.

 

No change in the behavior of the machine.

 

I'm hoping at this point that I have an OS problem (overheating doesn't seem to be it - my cpu temps are staying under 50 dec C)

 

I'm going to try a reinstall of the OS, this time starting with XP SP2 rather than the old SP1 disk.

 

I hope that'll do it... Seeing that I swapped the RAM between identical machines, and the problems stayed with the machine rather than following the ram, it seems a stretch that the ram is bad...

 

I'll update again once I complete the reinstall.

 

Merry Christmas!

Link to comment
Share on other sites
monsterbuilder

monsterbuilder

Posted
  • Author
Posted

Well, I'm stumped. Reloaded OS, this time with XP SP2 disk - installed anew all chipset drivers, etc, downloaded new Folding@Home console file & QMD core... and still came up with an EARLY_UNIT_END after only 5 frames (out of 100).

 

Thanks for sticking with my slow motion analysis as long as you have - I'm really not trying to drag this out... but I just can't get past the fact that the failure didn't follow the ram sticks when I swapped between Problem & Twin...

 

Granted, MemTest found numerous errors on all four sticks... but that ram performs fine in Twin, even on the QMD workunits.

 

:confused: Questions For You: When MemTest reports errors, is the problem always the ram, or is it possible that the problem is in the memory slots themselves? Perhaps in the memory controller? I'm just guessing here, since the only thing I can't swap is the motherboard, and I cannot duplicate the failures on the Twin box. What do you think?

 

I guess I'll go ahead and RMA all four sticks - at least that'll give you a chance to test them yourself and conclusively rule them out if possible.

 

I'll try to send later this week.

 

Thanks for all your assistance.

 

M1

Link to comment
Share on other sites
monsterbuilder

monsterbuilder

Posted
  • Author
Posted
Yes, it can indicate something other than the memory.

 

I'm leaning toward a half-cooked northbridge myself - it only shows up when running memory-intensive applications, i.e. a QMD work unit for Folding@Home, which uses approx 327 MB, or MemTest.

 

The trick will be convincing Chaintech. I reckon I'll have to go through this all over again with them.

Link to comment
Share on other sites

Archived

This topic is now archived and is closed to further replies.

Go to topic listing
×
×
  • Create New...