ASUS P5W DH Deluxe - 4 month replacement intervals for Twin2X2048-8500C5?

[Ambress]

Hello,

 

Since building a new PC in Sept. '06, with 4GB of Twin2X2048-8500C5 Corsair memory installed in an ASUS P5W DH Deluxe, I have seen memory modules failing at nearly 4-5 month intervals. The first two failures were each of single modules from different matched pairs, and in both cases Corsair replaced the matched pairs. I now find myself with yet a 3rd failure, but this time of 2 modules from the same matched pair. I have no idea why this motherboard seems to be "eating" this 8500C5 memory since I am running the memory below spec performance levels and below peak voltage ratings.

 

I originally built my system with overclocking in mind, and with plans to run the 8500C5 memory at DDR2-1066, only to learn through trial and error that this particular memory and motherboard combo would never permit stable operation at more than a substandard performance of DDR2-720. Needless to say, I was disappointed at overspending on memory where 6400C4 would've saved some money and worked as well or even better, from what I've read since.

 

For my current stable operation at 3.5GHz using a 1:1 FSB:DDR2 ratio, the relevant BIOS settings are:

 

AI Overclocking: Manual

CPU (FSB) Frequency: 350 MHz

DRAM Frequency: DDR2-700 MHz

Performance Mode: Auto

Memory Voltage: 2.05V

CPU VCore Voltage: 1.375V

FSB Termination Voltage: Auto

MCH Chipset Voltage: Auto

ICH Chipset Voltage: Auto

Ratio CMOS Setting: 10

Enhanced C1 Control: Disabled

Configure DRAM Timing by SPD: Disabled

DRAM CAS# Latency: 3 Clocks

DRAM RAS# Precharge: 3 Clocks

DRAM RAS# to CAS# Delay: 3 Clocks

DRAM RAS# Activate to Precharge: 9 Clocks

DRAM Write Recovery Time: 4 Clocks

DRAM ECC Mode: Disabled

Hyper Path 3: Disabled

DRAM Throttling Threshhold: Auto

 

I have verified that I have two failed modules, by way of confirming neither will allow my system to POST, nor would another compatible PC, so it is clear they must be replaced. But, before doing so, I would appreciate advice on whether it would be better to have all 4 modules replaced with 6400C4? If so, and given that is less expensive memory, would Corsair permit a replacement with at least the TWIN2X2048-6500C4DHX modules? If I had to stay with the 8500C5 memory, can you suggest what settings would ensure peak performance and lifespan from them in an overclocked configuration? Also, if a replacement of all 4 modules for 6500C4DHX was permitted, would Corsair allow me to return the 2 bad modules and keep my 2 working modules until the replacements arrive, completing the process by then returning the remaining 2 good modules? Obviously the goal is to keep my PC running and usable if possible while pursuing the memory replacement.

 

Thank you!

[RAM GUY]

Have you tried testing the modules individually with your cpu settings at default? Also I would suggest testing these modules at 2.2 volts even if you are unable to set the memory to 1066MHz.

[Ambress]

Well, I feel a bit stupid...I had not thought to try raising the voltage on the memory, since it had been working at lower voltages.

 

With the prior modules that had failed and been replaced, I had been running them at 2.15v, but in more recent efforts to better cool yet quieten my system, I tested it at lower memory and CPU voltages and found a 2.05v memory still gave stable operation. But, perhaps that was the culprit.

 

I had tested at BIOS/CPU defaults and not had any luck POSTING until removing the 2 suspect modules. But, I had not raised the memory voltage. Doing so, setting it at the 2.2v level you suggested did the trick and I'm now back to running at 3.5GHz, DDR2-700, 1:1, 3-3-3-9-4.

 

I was very surprised to see the prior 2 failed memory modules, but a pair failing now seemed even more irregular and I'm glad to see that they are working.

 

Thanks!

[RAM GUY]

I'm glad to hear you are back up and running! 2.2v is what the modules are tested at and there is definitely no harm in keeping the voltage at 2.2v!

[ChasR]

In my experience with the P5W DH, you must increase MCH voltage to 1.55 or 1.65 v to get the 8500 C5 to run near spec.

[Ambress]

Chas,

 

Thanks...I've never tweaked the MCH voltage any as I recall, so I'll research that further and give it some consideration. I'm pretty happy running at 3.5GHz though.

 

Regards,

 

Daryl

[RAM GUY]

Thank you for letting us know!

[Ambress]

Rather than post a new thread, I thought I'd "reawaken" this one, since my system is the same as I've previously posted, and part of my question is at least somewhat related to the change in memory voltage discusse here.

 

Typically, I leave my system running 24/7, rebooting only for changes prompted by software updates or to change the active O/S for the multiboot configuration I run. For the recent holiday period where I was going to be away from home for 2 weeks, I chose to shut down my PC. Upon returning home and booting the PC, it would not boot. All I observed was that the fans and drives all spun up, but there was no display of the BIOS startup screen at all. My thoughts immediately went to memory issues, based upon my past experiences with the 8500C5 memory on this P5WDH Deluxe motherboard.

 

Long story short, I ran into problems where I could never boot my system until I pulled a DDR2-667 memory module out of a PC at my office and used that in my system. Prior to that, my system would not boot from any memory slot with any of the 4 1GB modules of 8500C5 memory, even after fully clearing the BIOS and resetting the defaults. With the successful boot from the DDR2-667 memory, I entered the BIOS, overrode the SPD settings, and set memory voltage to 2.2v, DDRAM Freq. to DDR2-700, timing to 5-5-5-15-5, and enabled the memory remap for 4GB of installed memory. Then, removing the DDR2-667 memory and installing one module of the 8500C5 memory, the system booted. Adding another module, a reboot failed with no BIOS POST screen. Repeating this, I eventually found another module which my system would boot from, and curiously I found that the 2 "good" modules were 1 from each of 2 matched pairs. So, I then took both of the good modules and verified I could boot from them whether in a single or dual-channel configuration. While in the single channel mode, I tried again installing one of the 2 remaining modules, and my system failed to boot but did at least show the POST screen with the memory total indicated as 2858MB...obviously wrong (3072MB expected). Swapping that 3rd module out with the other still loose, again the system completely failed to boot, suggesting that 4th module was truly a bad one. Reinstalling the module where the erroneous total was observed, I now obtained a valid result of 3072MB and the system booted properly! My only conclusion amidst all this odd behavior, was that at least 1 module of the 8500C5 memory was bad. So, I expected to be requesting an RMA for pair of modules it was part of.

 

BUT...I then remembered this thread and how running the memory at the 2.2v spec recommended by RAM Guy, resolved earlier boot issues that arose after a period of time where I'd successfully been using a lower memory voltage. This left me wondering if the bad module I'd identified might now be needing a higher memory voltage. So, I removed all memory modules, installed the "bad" one, and set the memory voltage to 2.25v, the next higher increment. To my surprise, the system POSTed properly, so I shut it down before fully booting on into WinXP, reset the voltage to 2.2v and rebooted....Success! I then installed all 4 modules, success once more. I then restored my settings to those previously used for my overclock setup and again, success.

 

Success is good....but, why? Why did it take an extra 0.05v "kick in the pants" to get this 8500C5 memory module to work properly, especially considering that it continued to work at the spec voltage where it had been failing?

 

I appreciate any insight on this issue.

 

Daryl

[RAM GUY]

Not sure with out seeing the modules but that would suspect something has changed on the MB or the module, can you test them in another MB to be sure?

[Ambress]

Thanks for the reply, but no, testing the modules in another motherboard isn't practical for me. As I've got my system running for now, I'll just watch for future problems to arise. At least this "kick in the pants" temporary boosting of the memory voltage gives me another idea to keep in mind should the problem arise again. As for changes to the motheboard or modules, there have been none that I've made. This is just an oddity for now I guess.

[RAM GUY]

I would recommend testing the current configuration with http://www.memtest.org to be sure there are no errors and the system is stable. If you get errors then we will want to get the set replaced for you.

[Ambress]

Thanks...that's a good idea, and I'll take care of it soon.

[RAM GUY]

NP Please let us know!

[Ambress]

Here are my test results from Memtest 86+:

 

Memory: Corsair TWIN2X2048-8500C5 memory as 4x1GB in ASUS P5WDH Deluxe Motherboard with Intel Core 2 Extreme CPU.

 

ECC Off, Standard Test. DRAM Throttling disabled on last test case.

 

First test case was using CAS timings slightly looser than normal overclock configuration I run my system at, which are the timings in the 2nd test case. Test cases 2-5 were all just to get a quick look at whether any errors were reported within a short run time where not even a full pass through all tests was performed. Although 3.5GHz system running at DRAM Freq of DDR2-875 might prove to be stable over a longer test period, it is not a 1:1 FSB:DRAM ratio, which I understand (right or wrong?) may actually be more efficient although the DDRAM Frequency has been reduced. Based upon the 1st longer test run, it appears all 4 modules are good, so the puzzle remains as to why I had the recent problems with power-up of my PC after a 2-week period of being shut down.

 

WallTime...Speed....FSB Freq...DDRAM Freq....CAS Timing....Pass...Errors

7:29:05....3.5GHz....350MHz....DDR2-700.......4-4-4-12-4......8.....0

0:23:54....3.5GHz....350MHz....DDR2-700.......3-3-3-9-4.......0.....0

0:06:45....3.5GHz....350MHz....DDR2-875.......4-4-4-12-4......0.....0

0:00:05....3.3GHz....330MHz....DDR2-1000......5-5-5-15-5......0.....248

0:11:58....3.3GHz....300MHz....DDR2-900.......5-5-5-15-5......0.....0

[RAM GUY]

That would suggest its not the memory try and lower your over clock.

[Ambress]

Please just disregard this post. I saw from another thread that Corsair will not replace a bad product with anything but more of the same, which is what the 8500C5 modules appear to be. I'm leaving this post for the benefit of others who might encounter similar problems. I'll just switch to another manufacturer.

 

Well, I guess it never hurts to ask first...

 

This is largely recapping what I've previously reported in Post #325558 regarding my 8500C5 modules, only perhaps I could say a bit worse, because they now seem to need a bigger "kick in the pants". The details:

 

Overall, when my system is up and running at 3.5GHz, CPU Freq. 350 & DDR Freq. 700 for 1:1 ratio, with CAS timings of 3-3-3-9-4 or looser, and Vmem of 2.2v, all is fine and working well. BUT, if I power down my PC for more than 20 seconds or so, and more notably if I power it off via the P/S switch, then it will not POST at all when powered back on. My prior experiences found that 2.25v was needed on the Corsiar 8500C5 modules in order for my PC to boot. That has now increased to 2.3v, for reasons I can't explain, given no changes to my system. I did recently upgrade the AMI BIOS to see if it might possibly yield any different result, but it made no difference as I expected. So, here is my current drill:

 

1. Power down > 15-20 seconds

 

2. Learn system doesn't POST on power up, just blank screen.

 

3. Power off and remove all 4x1GB 8500C5 modules

 

4. Install single KVR667D2/1GB PC5300 module

 

5. Power up, POST is good....enter BIOS, load default settings.

 

6. Given prior efforts to boot with 8500C5 using BIOS default voltage of AUTO will not work, nor if manually set to 2.2v, and now not at 2.25, set Vmem to 2.3v.

 

7. Power off, remove PC5300 module, reinstall 4x1GB 8500C5 modules.

 

8. Power on, POST is successful...enter BIOS quickly and reset Vmem to 2.2v since 2.3v is over spec and presumably invalidates warranty.

 

9. Save settings and reboot...system boots normally and runs well until next full power off.

 

From all of this, and notably how the system will not boot if the Corsair 8500C5 modules are running with the BIOS defaults yet Vmem manually set to 2.2v, it seems to me this is not an overclock issue at all, but rather something else quite awry. If a motherboard issue, what could be going wrong that would affect the memory in this way? Given that the PC booted without problems using the KVR667D2 module and also with 2x2GB PC6400 kits I've now tried in both PDC24G6400ELK and Corsair TWIN2X4096-6400C5DHX, although recognizing these modules run at lower voltages, it seems the motherboard is working properly. Adding to this, I'd previously been running the 8500C5 modules succesfully for 16 months at 2.05-2.2v, so why would they start requiring a voltage "kick" at power up, not to mention one that has increased from 2.25v to 2.3v?

 

Obviously, going through this drill any time I shut my system off is quite a hassle. I could just let my system run at Vmem = 2.3v, but if I've not already invalidated the Corsair warranty by using that setting briefly to enter the BIOS at POST, then I've just been concerned that running full-time at that setting would surely invalidate the warranty. The memory is a heck of a lot cheaper than the $800 initially paid, so maybe the warranty is no big deal, but if Corsair is going to support me by replacing failing modules under warranty, then money saved is money saved, regardless.

 

Seeing that both the PDC24G6400ELK and Corsair memory work in 2x2GB kits, but are lesser performance PC6400C5 modules than my PC8500C5, my plan is to move to lower latency C4 modules. My understanding is that in a 2x2 configuration, the PC6400C4 should have performance largely akin to PC8500C5 in a 4x1 configuration. So, I'm approaching Corsair here with a question of whether they would, under warranty, be willing to replace my 4x1GB of TWIN2X2048-8500C5 modules with 2x2GB of TWIN2X4086-6400C4DHX memory? And, if so, would they do it by way of shipping the modules to me in advance, with my returning the 8500C5 modules after the new kit is received?

 

If not, I'll just go buy some 2x2GB kits from one of 3 other manufacturers, chalking this up as a learning experience with hope that good reports on these other kits I've looked at do not lead me down the same path that so many good reports on Corsair memory did.

 

If Corsair does agree to replace my memory in what seems a rather unique situation, please advise how best to proceed with an RMA.

 

Thanks,

 

Daryl

[RAM GUY]

Do the modules post individually at default or 2.2v? If all four of them are having the same issues individually then there may be some other issue, such as a voltage regulation issue with the board. Have you ever measured the voltage with a DMM (Digital Multi Meter) to verify that the board is actually providing the power that it says in the BIOS? We would be more than happy to replace the modules for you, but you are right, normally we can only replace modules for the same exact part number. However you can contact our customer service at 888-222-4346 and dial "0", (510) 657-8747 or email rmaservice@corsairmemory.com to see if there are any options for you.

[Ambress]

To tell you the truth, I don't recall if all modules failed to boot if tested individually at 2.2v or not. Certainly I can try that; given the workaround and that my system would run OK once I went through my "2.3v drill", I just dealt with the issues as is. But, after I had recent reasons to power down a few times, this just became enough a nuisance that I camer here with my questions again.

 

I'll explore the single module booting at 2.2v and up, to see how each module behaves. Can you tell me if my brief 2.3v "kicks" to get into the BIOS and reset to 2.2v are going to invalidate my warranty?

 

With regard to testing the motherboard voltages on the slots, I'm not quite sure how to go about that, but I'll look for pinout info and see what I can do. I assume the process would be to boot with a module in one slot & measure the voltages off the other 3, then reboot with a module in a tested slot and then test then remaining slot that is now freed up?

 

I've already ordered a 2x2GB kit that was inexpensive yet has received good reviews, so if it gives me the performance I had with the 4x1GB configuration, I may just go with that.

 

Thanks,

 

Daryl

[RAM GUY]

Can you tell me if my brief 2.3v "kicks" to get into the BIOS and reset to 2.2v are going to invalidate my warranty?

This will not invalidate your warranty. If the modules truly require 2.3 volts to operate normally then they are not working at our tested specifications, and we would want to replace them.

 

With regard to testing the motherboard voltages on the slots, I'm not quite sure how to go about that...

There should be a MOSFET on the board near the memory which will show you the vDIMM when you measure it with a DMM. You should contact ASUS to find out exactly where it is, and what procedure you should follow to measure it correctly.

 

I've already ordered a 2x2GB kit that was inexpensive yet has received good reviews, so if it gives me the performance I had with the 4x1GB configuration, I may just go with that.

 

If our modules are faulty then we will want to get them replaced, and you will always have the option since we offer a lifetime warranty on all of our memory. If possible you may even want to test them in a different system to be sure that there is not an issue with the board.

[Ambress]

Given my earlier MemTest results and even running successfully with tighter 3-3-3-3-9-4 CAS timings on my 8500C5 modules, it does seem odd to think the memory is the source of my problems, although a switch to different modules may provide a cheap fix...but only a short term one, or permanent? Only time would tell.

 

The more I research the problem, I'm finding others who have reported various issues that might be related to mine, with the ASUS P5W DH Deluxe motherboard. In the ASUS forum, I was encouraged to try raising vMCH rather than vDIMM, as the Intel 975X Northbridge of this motherboard is said to be its weak spot; similarly I've read raising vMCH may allow one to break past the DDR2 frequency wall that seems to lie around the 360-370MHz range (core frequency).

 

So, come the weekend I'll first explore raising vMCH to see if that permits a successful boot at 2.2v vDIMM. If not, I'll press on to the other voltage tests and single module/slot tests. Whatever my results, I'll report them back here once collected.

 

Thanks,

 

Daryl

[RAM GUY]

Raising the MCH voltage was going to be my next suggestion if multiple modules were not able to run at their spec. And just so you know, we have no problem replacing the modules for you at anytime, we would just like to be confident that the problem lies in the modules so that a replacement resolves the issue. Let us know how the testing goes and if it looks like the memory is causing the issues then lets get them replaced!

[Ambress]

OK, the results are in and, in brief, the memory appears at fault. The details:

 

Memory Type: Two pair TWIN2X2048-8500C5 XMS8505 v1.3

 

vDIMM: Voltage was measured at 0.05v increments over the BIOS-supported range from 1.80-2.4v, using DDRAM2 modules that would support a given voltage if the currently installed module would not. All voltages measured at 0.03-0.04v above the BIOS setting, with the same readings observed whether vMCH was set to 1.50v (min) or 1.65v (max). So the motherboard appears to be providing a good vDIMM.

 

DIMM Slots: With vDIMM at 2.1v (spec voltage indicated on module) and testing each module individually in each of the 4 DIMM slots, a successful cold boot POST was not observed for any module. When subsequent testing identified 1 module that allowed a succesful POST with vDIMM at 2.2v, the module was tested in all slots with a successful cold boot POST for each.

 

DDRAM2 Test: With BIOS at default settings (266MHz Core clock rate, 2.66GHz CPU Freq, DDR2 Frequency overlooked) but manually setting memory voltage to Corsair test spec 2.2v and CAS timing to 5-5-5-15-5, the PC was powered off for 30 seconds or more, then powered on for a "cold boot". Modules were tested both at vMCH = 1.50v and vMCH = 1.65v with identical results. Goal was to observe whether system would POST for a given memory voltage.

 

• Below 2.2v vDIMM, no single module in any DIMM channel would permit a successful POST
  
• Module s/n 1126431 - Successful POST at 2.2v or higher
  
• Module s/n 1126432 - POST failed at 2.2v. Successful POST at 2.25v or higher.
  
• Module s/n 903896 - POST failed at 2.2-2.25v. Successful POST at 2.3v.
  
• Module s/n 093897 - POST failed 2.2v. Successful POST at 2.25v or higher.
  
• As previously reported, with all modules installed, a successful cold boot POST occurs if vDIMM is 2.3v. Resetting vDIMM to 2.2v, a warm reboot allows successful POST and normal system operation. Cold boot fails POST.
  

 

So, it appears an RMA on all 4 modules will be needed. My first two experiences with RMAs were a little frustrating, given what I considered a slow response to simply get the RMA issued. If I remember correctly, I think it took at least full 3 weeks in both cases to get the new modules. Will referencing this thread expedite approval? My saving grace may well be the new memory I ordered, which I may keep as a hedge against more failures of the replacement modules...I do not have a very high confidence in these Corsair modules based upon my experience, as I'm sure you can appreciate.

 

Thank you,

 

Daryl

[RAM GUY]

Please contact our customer service at 888-222-4346 and dial "0" or (510) 657-8747 and they should be able to help you speed up the RMA process, just let them know you have done an RMA in the past and the modules which were returned to you have failed. Or you can fill out our the On Line RMA Request Form and post your case # in this thread, and I'll approve it ASAP.

[Ambress]

Thanks...I went ahead and just used the online RMA form, as that was convenient to do where I'm currently at, with no nearby phone. But, I goofed somehow it seems, as I saw no Case # reported when I submitted the information. I don't recall if the Case # might be presented via a pop-up? I do notice a pop-up blocker is active on the PC I'm using.

 

If I receive an e-mail ahead of any providing the actual RMA information but with a Case #, I'll come back and add that information here.

 

I appreciate your help!

 

Daryl

[RAM GUY]

You should receive an email shortly which will have a case # associated with it.