Jump to content
Corsair Community

Two sets of TW3X4G1333C9DHX failing memtest86+ with GA-MA790XT-UD4P


Troubled

Recommended Posts

I have recently tried getting a twin 4G bundle (2x2G) of Corsair RAM (TW3X4G1333C9DHX) going in a system with a Gigabyte GA-MA790XT-UD4P motherboard.

 

After some instability I ran it through the latest memtest86+, and experimented with settings. One of the RAM sticks regularly failed- at around the same addresses (which screams RAM fault)- and the other ran flawlessly for over 16 hours, so I returned the RAM kit, which was tested by the store (and confirmed bad by them too), and was refunded.

 

All settings on auto defaults, and the timings were detected correctly. The motherboard was flashed to the latest BIOS as well. No overclocking was done- I'm actually not interested in overclocking this system.

 

Oh- and I tried increasingly conservative settings, bumped the speed down and such- but all it did was delay the onset of problems.

 

Anyway, I purchased another twin bundle (TW3X4G1333C9DHX again), this time from a different store. And the new set failed too! Whilst I haven't had enough time yet to isolate the specific fault, it seems to appear when both sticks are in, and at a fairly consistent address.

 

My new system has basically been sitting collecting dust because I've only every had bad RAM. I've bought Corsair RAM for a few years now, and as you can imagine, after getting what appears to be two bad sets of RAM in a row, my confidence is the brand is somewhat shattered at this point.

 

I'm happy to give things a try, and I'm hoping someone there has an idea of some BIOS settings that might make this RAM work. With marginal RAM, am I best off experimenting with the voltage, or timings? What works best with the TW3X4G1333C9DHX? What should I try? The GA-MA790XT-UD4P has a plethora of options, and it takes hours per test. Or should I just return this second set as well? Are there known problems with this combination or particular RAM set?

Link to comment
Share on other sites

  • Corsair Employees
Please make sure that you have the latest bios and load setup defaults and then set the memory Voltage to 1.7 Volts then set the timings manually to the tested settings for the specific module you have, and test the modules one at a time with http://www.memtest.org to be sure one is not failing. 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 some MB's (Mostly ASUS) you have to disable legacy USB in the bios when running any memory test.
Link to comment
Share on other sites

Already have and was using the latest BIOS and the latest memtest86+ (as of one week ago). Have tried with both 1.6V and 1.7V. The sticks fail when used together.

 

Tried default settings overnight with one module, and it worked AOK for about 8 hours. I usually see failure within 20 minutes so one of the sticks *might* be fine. Haven't run the other module on its own yet.

 

Tried in slots 3 & 4 rather than 1 & 2, failed at exactly the same memory address as when in slots 1 & 2.

 

Timings were set to auto on these tests, but they were the expected 9-9-9-24.

 

Tried again, this time with 1.7V (as suggested) in slots 3 & 4, manually set the settings that were detected by SPD anyway, and it failed again, exact same location.

 

When you say "set the timings manually to the tested settings", do you mean the settings here:

 

http://forum.corsair.com/forums/showthread.php?t=54052

 

This suggests 1.5V for the RAM, and suggests setting specific voltages for the CPU. I don't have that motherboard and don't know which CPU was used in that test. Setting the CPU voltage, for example, doesn't seem wise. Should I be looking somewhere else for the "tested settings"?

 

"Legacy USB Storage Detect" was already disabled.

 

NB: I have no secondary DDR3 motherboard to test with.

 

For now, I'm going to try the other stick in isolation (at all auto settings ie. 1.6V to match the conditions on the other stick), but it'll take some hours before I have results. If it fails, my guess is one faulty module. But if it passes, then the two are failing together at a consistent address, and I'm at a complete loss as to what to do next.

 

If I should be trying something else, please let me know.

Link to comment
Share on other sites

Looks like two defective sets in a row.

 

The first set could be tracked to the one module rather reliably.

 

The second set is odd. They both seem to run fine, when run individually, for hours. But run them together, and they fail at the same memory address. Confine the search to about 400MB around that address, and you get an error almost every ten seconds after the first few minutes. Put them in different slots- errors at the same address. Same err-bits. Always. Swap the modules with each other, and the error is at a second address. Basically, the errors follow the module order but are independent of slot. It seems one module is defective, but only when used with the other. Weird. Wish I had more RAM to diagnose it further, but what I've seen thus far seems to be enough to confirm the fault so I'll leave the rest to the pros.

 

There are some other oddities, but the gist is that the TW3X4G1333C9DHX pair isn't working on a GA-MA790XT-UD4P that otherwise is fine. I'll be returning it to the store on that basis, I imagine it'll make its way back to Corsair indirectly in time.

 

Oh, and: Error behaviour was different for each of the two sets, so let's not hear "must be the mobo". Besides- the first set was confirmed busted independently in a different mobo.

 

Anyway- two bad RAM sets in a row out of retail. That's a bit poor. Not sure if anyone reading has access to the Quality Control process at Corsair, but it might be worth a look to see why things are slipping through.

 

Let's see how I go with set #3 (store wants to replace rather than refund, so I'm stuck trying again).

 

I have the RAM for another 36 hours if anyone has any suggestions.

Link to comment
Share on other sites

Unfortunately, as per my previous post, I only had the modules for 36 hours. I no longer have the modules. They have been returned as per my previous post. Unfortunately it means I can no longer boot the PC to check the settings, nor look at the label to read settings back.

 

I honestly cannot see how repeated failures at the same one memory address, with the same error bits each time, with no other errors, independent of slot choice but dependent on module ordering (ie. failed address shift), consistent across multiple boots and reorderings, could possibly be *anything* but bad RAM (or marginal RAM that only works on some motherboards). I am not an expert, of course. Perhaps you could explain how it could be something apart from a RAM fault, maybe by way of a single example that would present these exact same symptoms?

Link to comment
Share on other sites

Well, this is rapidly moving into the realm of the ridiculous.

 

I've received my replacement RAM. I'm on my *third* set now. Packet says TW3X4G1333C9DHX, labels on each say CM3X2048-1333C9DHX. I think that means they're not a tested pair. SPD for the modules say they should run at 1066MHz rather than 1333MHz, with timings 7-7-7-? (I forgot to note down the last timing, but I'm guessing 20). So if I go for auto settings, I get 1066MHz instead of the 1333MHz I paid for.

 

Either someone's trying to pass off 1066MHz RAM as 1333MHz RAM, or the SPD wasn't programmed correctly for the module. Great. Neither fill me with confidence.

 

I noticed that the "fins" on one of the modules are pushed together a bit- I'm guessing it's been handled poorly by someone- squeezed to insert into a slot rather than being pushed in properly.

 

Since there's not a lot more I can do given the hour, I'm going to run memtest86+ overnight with manual settings set to 9-9-9-24. Let's see how it handles that. Ironically, it's the most stable set of Corsair DDR3 RAM I've had thus far- it's been running for an hour, no errors yet. Almost two full passes of memtest86+ at the time of this writing.

 

Are the incorrect SPD settings something that happens occasionally? Or is it likely that somewhere down the line somebody put the wrong sticker on- either by accident, or is someone trying to cheat me by putting new stickers on cheaper RAM?

Link to comment
Share on other sites

Thanks for letting me know. As the other two sets had the SPD set to 1333MHz and 9-9-9-24, seeing 1066MHz and 7-7-7-20 was a shock. I was concerned that someone in the supply chain might have slipped me lower-spec RAM and slapped a sticker on it, hoping nobody would notice. If it's normal, it's not a problem.

 

I took a guess at some settings yesterday evening (9-9-9-24, but had to guess the rest), and memtest86+ has been running, fault-free for 14 hours (15 full passes). Whilst I can't be sure that I've got the settings right, it would appear that I've got a stable configuration to work from.

 

Once in the BIOS, I hit Ctrl-F1 for the extended setting options. I began by loading optimised defaults on each of the two pages, then made a number of changes.

 

Initial settings:

 

Set Memory Clock: Manual (was Auto)

Memory Clock: x6.66 (1333MHz) (was x5.33)

DCTs Mode: Unganged (unchanged)

DDR3 Timing Items: Manual (was Auto)

 

Timings. For these settings, format is: current (SPD Auto). All of them started on the auto setting value, so most were changed. Note that RAM is in slots 1&2, slots 3&4 are empty.

 

CAS # latency: 9T (7T 7T)

RAS to CAS R/W Delay: 9T (7T 9T)

Row Precharge Time: 9T (7T 9T)

Minimum RAS Active Time: 24T (20T 25T)

1T/2T Command Timing: 2T

TwTr Command Delay: 5T (4T 5T)

Trfc0 for DIMM1 110ns (110ns 110ns)

Trfc2 for DIMM2 90ns

Trfc1 for DIMM3 110ns (110ns 110ns)

Trfc3 for DIMM4 90ns

Write recovery time: 10T (8T 10T)

Precharge Time: 5T (4T 5T)

Row Cycle Time: 34T (27T 34T)

 

RAS to RAS Delay was either 4T or 5T, I'm not certain. I've set it to 5T for now, but the previous test may have been done at 4T.

 

Checking the RAM modules individually suggests that they have the same SPD settings (they both report 7-7-7-20 on their own).

 

All of the following were unchanged:

 

Bank Interleaving: Enabled

Channel interleave: Enabled

DQS Training Control: Skip DQS

CKE Power Down Mode: Disabled

Memclock tri-stating: Disabled

Bank Swizzle Mode: Enabled

 

I've had to guess a few setings- mostly went with what "Auto" suggested unless I had the settings otherwise. Are these settings correct? If not, what should be changed?

Link to comment
Share on other sites

Thanks. Do the other settings I've listed also look okay? Listed here for convenience, full list in my previous post:

 

TwTr Command Delay: 5T (4T 5T)

Trfc0 for DIMM1 110ns (110ns 110ns)

Trfc2 for DIMM2 90ns

Trfc1 for DIMM3 110ns (110ns 110ns)

Trfc3 for DIMM4 90ns

Write recovery time: 10T (8T 10T)

Precharge Time: 5T (4T 5T)

Row Cycle Time: 34T (27T 34T)

 

RAS to RAS Delay was either 4T or 5T, I'm not certain. I've set it to 5T for now, but the previous test may have been done at 4T.

 

(voltage was at 1.6V, I believe, but I'll check next reboot)

Link to comment
Share on other sites

Archived

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

×
×
  • Create New...