Possibly bad VS1GBKIT400C3

[gothick]

Hi there!

 

I recently bought a matched pair (VS1GBKIT400C3) of Corsair ValueSelect modules. I found that my machine crashed a couple of times, so I ran memtest86 across them and it found faults. I sent the RAM back to the supplier, and they tested it, said they had found it faulty too, and sent me a new replacement part.

 

As soon as I got this part, I ran memtest on it, just in case, and this set's coming up as faulty, too. So, I'm starting to wonder whether I'm very unlucky, or whether there's something else going on. Can you help?

 

The setup is a little unusual, as I'm slowly upgrading an old PC, and the RAM is the first part I went for. That said, this means that it's being underclocked for DDR400. The memory is in an ECS K7S5A motherboard (tested and OK with the above module according to the Corsair product page), running on a 133 FSB with an Athlon 1200 (i.e. CPU 133, DRAM 133, CPU multiplier 9x.)

 

Basically, I'm quite surprised to get two DDR400 modules in a row that can't cope on a 133MHz bus, so I thought it was worth getting some help here if I could before I blindly send these modules back!

 

Here are the details:

 

Motherboard: ECS/EliteGroup K7S5A (SiS 735 chipset). The matched pair are the only modules installed. There's an old nVidia GeForce II graphics card in an AGP slot, and an Ensoniq/Soundblaster PCI 128 in a PCI slot. One DVD, one CD, one hard disk. 300W ATX PSU. That's pretty much it.

 

BIOS Settings (everything's pretty much at the "most reliable" settings at the moment):

 

DRAM Timing Configuration: Normal (default)

SDR/DDR CAS Latency: SPD (default)

SRD/DDR RAS Active Time: 7T (upping this from default 6T didn't change anything)

SRD/DDR RAS Precharge Time: 4T (default)

Auto Detect DIMM/PCI Clk: Enabled (disabling it didn't change anything.)

CLK GEN Spread Spectrum: Disabled (default)

Dram Driver Slew Rating: Normal (default)

CPU Frequency: 133MHz

DRAM Frequency: 133MHz

 

The old memory was SDRAM (the K7S5A supports both SDRAM and DDR) and worked for years without problems, but I've never had DDR in this motherboard before.

 

Today I left memtest-86 v3.2 running for about 9 hours; I came back to find 69 errors in Test 3, 18 errors in test 5 and 342 errors in test6. The tail end of the error log looked like this:

 

Test Pass Failing Address Good Bad Err-bits

6 6 00023ed5a80 ffffdfff ffffffff 00002000

6 6 00025ed5a80 ffffdfff ffffffff 00002000

6 6 00029ed5a80 ffffdfff ffffffff 00002000

6 6 0002ded5a80 ffffdfff ffffffff 00002000

6 6 0002fed5a80 ffffdfff ffffffff 00002000

6 6 00033ed5a80 ffffdfff ffffffff 00002000

6 6 00037ed5a80 ffffdfff ffffffff 00002000

6 6 00039ed5a80 ffffdfff ffffffff 00002000

6 6 0003bed5a80 ffffdfff ffffffff 00002000

6 6 0003ded5a80 ffffdfff ffffffff 00002000

 

I re-ran test 6, and the beginning of the errors it found look like this, a similar pattern:

 

6 1 0003c0b1a80 ffffdfff ffffffff 00002000

6 1 000320b1a80 ffffdfff ffffffff 00002000

6 1 0002c0b1a80 ffffdfff ffffffff 00002000

6 1 000240b1a80 ffffdfff ffffffff 00002000

6 1 000220b1a80 ffffdfff ffffffff 00002000

6 1 000200b1a80 ffffdfff ffffffff 00002000

6 1 0001e0b1a80 ffffdfff ffffffff 00002000

6 1 0001c0b1a80 ffffdfff ffffffff 00002000

6 1 000180b1a80 ffffdfff ffffffff 00002000

6 1 000140b1a80 ffffdfff ffffffff 00002000

 

...continued all the way to 000020b1a80 (from 960 to 32Mb, basically)

 

Unfortunately, I don't have anyone who can lend me any known-good modules to try out, and I don't know anyone who has time at the moment to check out the memory in their machine.

 

Does any of the above point to any particular cause, do you think?

 

Thanks very much,

 

Matt

[RAM GUY]

Please make sure that you have the latest bios and then load setup defaults and set the following settings and test the modules one at a time with http://www.memtest.org.

CPU Freq: 133 MHz

Memory Frequency: 100%/1-1

Dim Voltage to 2.7 Volts

Resulting Frequency: 133MHz

SDRAM CAS Latency: 2.5T

SDRAM RAS to CAS Delay (tRCD): 3T

SDRAM Row Precharge (tRP): 3T

SDRAM Active to Precharge Delay (tRAS): 7T

[gothick]

Hi again.

 

I couldn't find a RAS to CAS delay setting anywhere (the only memory-related settings I could find are mentioned in my first post), nor could I find a Dim voltage setting -- there are no voltage settings of any kind from what I can see. The DIMM sockets themselves actually "2.5V" engraved on the plastic, if that's any help...

 

Having said that, I did everything else you said, and left one single stick in today to be tested using memtest86+ (rather than memtest-86) while I was at work (11 hours.) I came back to find 1256 errors, all from test 5. No errors from any other tests. The last chunk of errors on the screen were all with a "Good" pattern of FFFFFFFF, with errors in fairly random-looking portions of the lower bits, e.g. "Err-Bits" of 00000001, 00002000, 00040000, etc.

 

I re-ran test 5 and got errors within about 3 minutes, similar to the previous results. I then swapped the DIMM with the one I'd taken out, re-ran Test 5, and got very similar results -- 44 errors in first 3 minutes.

 

I've played around with the other low-level timing settings a bit, and they either make no difference or make things worse.

 

However, if this is any help, once I changed the FSB speed to 100MHz (DDR200), down from 133, I couldn't quickly reproduce the error (I left test 5 running for about half an hour just now with no problems.)

 

Any ideas?

 

Thanks,

 

Matt

[RAM GUY]

Matt,

I would suggest testing the modules in another system if you can, but I have no problem replacing the modules if you want to try that. But what you have posted would suggest some other problem.

Please follow the link in my signature “I think I have a bad part!” and we will be happy to replace them or it!

[gothick]

Hi,

 

Thanks for the help so far! I thought the results looked suspicious myself, combined with the fact that I've never had any DDR in this motherboard before, so I've just ordered a new motherboard (Gigabyte) to replace this old one. That should arrive Tuesday, so I'll give the memory a try in the new board next week.

 

If it's still showing as faulty, I'll send it back to my supplier (as I'm in the UK, so posting things to California doesn't sound like a good option!)

 

Thanks again,

 

Matt

[RAM GUY]

No problem, please let me know how you make out!

[gothick]

No problem, please let me know how you make out!

 

Hi,

 

Interim update: Got a brand new Gigabyte motherboard today (7VT5600P-RZ.) Plugged in the processor, the power supply, a cd-rom drive, a graphics card and monitor, and the RAM. Nothing else. I ran test 5 of memtest86+ first of all -- and it failed near the end with about 50 errors, just like it did on the old ECS board.

 

Can you think of any reason why one of those components would be affecting the RAM in the same way in a different motherboard? If not, I'm pretty convinced now that I've just had a couple of bad sets of memory... I've applied for an RMA number from the supplier, anyway!

 

Cheers,

 

Matt

[RAM GUY]

Matt,

Please make sure that you have the latest bios and then load setup/optimized defaults and set the Dim Over Voltage control to +.2 Volts and test the modules one at a time with http://www.memtest.org.

[gothick]

Matt,

Please make sure that you have the latest bios and then load setup/optimized defaults and set the Dim Over Voltage control to +.2 Volts and test the modules one at a time with http://www.memtest.org.

 

I'll do that this evening so that I can memtest tomorrow.

 

Here's an interesting change, though: my new processor arrived yesterday, and I fitted it into the new Gigabyte board to test it. I didn't have time to set up a proper test as you advised, but I left the memory testing (both sticks) all day today at the optimised default values of the motherboard, and came back to find only two errors in ten hours. Both from test 5, in the 7th pass (at 333.1mb and 349.1mb addresses.)

 

This is far, far fewer and less repeatable errors than I was getting when I had the old Athlon 1.2 in there -- and the board is now running at 200MHz (400MHz FSB for an Athlon XP 3200+...) I find this rather odd, but at least the error count is going in the right direction!

 

Incidentally, can you tell me what "Bank Interleave: 4 bank" means? This is one of the optimised default settings on the memory page of the BIOS. I thought it would be 2 bank, what with their being two stick of memory...

 

Will report back tomorrow.

 

Thanks for all the help so far,

 

Matt

[RAM GUY]

Taken from Rojakpot.com

 

 

SDRAM Bank Interleave

 

Common Options : 2-Bank, 4-Bank, Disabled

 

Quick Review

 

This BIOS feature enables you to set the interleave mode of the SDRAM interface.

 

Interleaving allows banks of SDRAM to alternate their refresh and access cycles. One bank will undergo its refresh cycle while another is being accessed. This improves memory performance by masking the refresh cycles of each memory bank. A close examination will reveal that since the refresh cycles of all the memory banks are staggered, this produces a kind of pipelining effect.

 

However, bank interleaving only works if the addresses requested consecutively are not in the same bank. If they are in the same memory bank, then the data transactions behave as if the banks were not interleaved. The processor will have to wait until the first data transaction clears and that memory bank refreshes before it can send another address to that bank.

 

Each SDRAM module is internally divided into either two or four banks of memory. Double-banked SDRAM modules generally use 16Mbit SDRAM chips and are usually 32MB or smaller in size. Quad-banked SDRAM modules, on the other hand, usually use higher density (64Mbit-256Mbit) SDRAM chips. All SDRAM modules of at least 64MB in size are quad-banked in nature.

 

If you are using a single double-banked SDRAM module, set this feature to 2-Bank. This is the only option available for the single double-banked SDRAM module.

 

If you are using at least two double-banked SDRAM modules, you can use the 4-Bank option as well as the 2-Bank option. Of course, it is recommended that you select 4-Bank for better interleaving performance.

 

If you are using quad-banked SDRAM modules, you can use either interleave options. Of course, it is recommended that you select 4-Bank for better interleaving performance.

 

Because a 4-bank interleave always allows for better interleaving performance, it is highly recommended that you select the 4-Bank option if your system supports it. Use the 2-Bank option only if you are using a single double-banked SDRAM module.

 

Please note that Award (now part of Phoenix Technologies) recommends that SDRAM bank interleaving be disabled if 16Mbit SDRAM modules are used. This is because early 16Mbit SDRAM modules have stability problems with bank interleaving. The good news is all current SDRAM modules support bank interleaving.

[gothick]

*Snipped bank interleave explanation*

 

Wow! Thanks for that; at least I know what's going on there now (and that it's perfectly fine.)

 

I've now had time to do a very thorough test on one of the sticks, and a reasonable test on the second one. Having loaded optimised defaults, and bumped the DRAM voltage up to 2.7V as you instructed, I can't get any errors out of memtest86+. I left the first bank testing yesterday while I was at work, and got no errors in 36 passes of memtest. The second bank I left to run for a while this morning, and got through 6 passes without a single error. I'll set it up for a longer test tomorrow, just to make sure.

 

Assuming that the second stick will test out okay in a longer test, I this means that they're working individually, but are having problems when they're together. I ran a second brief test with them both together the other morning and hit two errors in pass four before half an hour was up -- also with optimised default BIOS setup and 2.7V selected.

 

What do you think this means? I'm afraid I'm a bit clueless when it comes to hardware; I'm a software engineer!

 

Thanks for all the help so far!

 

Matt

[Wired]

May need 2.8 volts for both sticks. Try them in single channel mode and then in dual channel mode. This will help to determine if the memory controller and/or a slot is having a problem.

 

Also, generally 3 passes is sufficient.

[gothick]

May need 2.8 volts for both sticks. Try them in single channel mode and then in dual channel mode. This will help to determine if the memory controller and/or a slot is having a problem.

Okay, I'll bump up to 2.8V and try both sticks now.

 

Also, generally 3 passes is sufficient.

Don't worry, I'm no test fetishist :):, it's just that I normally leave the machine testing when I'm out at work, so eight to ten hours of testing is the minimum!

 

Also, previously I did one test with both of the sticks where I got only two errors in ten hours of testing... Is there an "acceptable minimum" level of errors? I mean, if I'd only seen one in the course of a day, I'd have put it down to a cosmic ray, but to get two spaced out over some time seemed to indicate something being a little broken. But I'm no expert, which is why I'm here!

[Wired]

Well there are 2 general trains of thought when it comes to extended testing of memory:

 

Some say you have to test for 20 passes because some memory errors only come up when memory is taxed for that long, or that certain conditions have to exist for errors to occur.

 

Others say that when you test for 20 passes that inevitably you will get errors.

 

As for the testing, test one stick in each slot, just to make sure the slots are ok. However, your mobo may not support single stick booting from each slot.

 

If all the slots are known to be ok, then when you test in single channel and dual channel mode, you know that a bad slot won't skew the outcome.

[gothick]

Well, this is interesting. If you like that sort of thing ;):

 

So far: I've bumped the voltage up to 2.8V, and tested with the memory (both sticks) in DIMM1 and DIMM2 sockets.

 

1) 2.8V, with a bank interleave of "4 bank": 4 errors within 20 minutes, in test 5.

2) 2.8V, with a bank interleave of "disabled", 2 errors within 20 minutes, in test 5.

 

Interestingly, this was far worse than I'd seen before with both sticks in. So, for an experiment, I left the bank interleave "disabled", and re-ran the last test with the DIMM voltage set to "Auto" (although the BIOS unfortunately doesn't tell me what voltage it's actually using.) In this case, I got through 3 passes of memtest without a single error.

 

So:

 

3) Auto Voltage (2.5?), with a bank interleave of "disabled", 0 errors in 3 passes.

 

I'm about to do the same test again, with bank interleave set back to "4 bank", but with voltage at "auto", and then manually set to 2.5V, to see whether it's mainly the voltage that seems to be affecting the problem, rather than the interleave.

 

Questions that are forming in my mind so far:

 

a) I thought that increasing the voltage would make the RAM more stable? (Could this be a heat-related problem? Would bumping the voltage up increase the temperature of the RAM?)

 

b) Will each individual stick of memory work at 2.8V when tested alone?

 

c) What's the motherboard using as the voltage when set to "Auto"?

 

I'm going to try a few more tests, including trying each slot individually, as you said. I don't think I've tested a stick singly in the "DIMM2" socket yet, so if the motherboard will let me, that seems to be the next sensible test, as I'm using DIMM1 and DIMM2 for the dual tests.

 

I'll report back in a while!

 

Cheers,

 

Matt

[RAM GUY]

Matt we can try replacing your modules if you like, but what you have posted so far might suggest a power problem. Can you test the modules in another system? If you want to try replacing your modules please follow the link in my signature “I think I have a bad part!” and we will be happy to replace them or it!

[gothick]

Matt we can try replacing your modules if you like, but what you have posted so far might suggest a power problem. Can you test the modules in another system? If you want to try replacing your modules please follow the link in my signature “I think I have a bad part!” and we will be happy to replace them or it!

 

I was thinking myself there might be some odd power problem going on. But now I've tried to investigate that and I'm getting such odd results I don't know what to think.

 

I was, like I posted, easily getting errors, especially at higher DIMM voltages.

I unplugged a few things, to give myself a nice bare-minimum system, and ran memtest again. Without a SoundBlaster PCI 128, the hard drive, the floppy and without anything plugged into the USB ports, I got through a pass OK at 2.8V.

 

So I started plugging things back in, one by one. The first thing I tried was the Soundblaster. With it plugged back in, I got two errors in the first pass of memtest.

 

So I plugged everything else in except the soundblaster, and got through three passes fine at 2.8V.

 

Just to confirm, I plugged the SoundBlaster back in. And got no errors in two passes!

 

This morning, I set it off for a "soak" test, and came back home at about pass 14. To find *two* errors on the screen, both from pass 0.

 

Now I'm just tempted to give up on computers and go back to pen and paper

:sigh!:

 

I'll see if I can borrow someone else's system for a while, but I don't know anyone who's using a 400MHz FSB, so I probably can't try an exact comparison...

 

Thanks for bearing with me!

[RAM GUY]

No problem, thats what we are here for.

And as I said, I have no problem replacing your modules if you want to try that.

[gothick]

Well, I returned the memory again, and got another replacement. Which failed tests very quickly running at 2.5V, and then got through about one and a half cycles of memtest before failing once in test 5 at 2.8V.

 

So. Any ideas, guys? Does this ValueSelect memory just not work in this motherboard for some reason, or is there something odder going on, do you think?

 

Thanks,

 

Matt

[RAM GUY]

Matt errors in test 5 of memtest may suggest some other problem. Can you tell me the make and model of MB you have along with the CPU speed and it’s FSB as well? In addition, please tell me the bios settings you have set for both CPU and memory and any performance settings that you may have set?

[gothick]

Matt errors in test 5 of memtest may suggest some other problem. Can you tell me the make and model of MB you have along with the CPU speed and it’s FSB as well? In addition, please tell me the bios settings you have set for both CPU and memory and any performance settings that you may have set?

 

I have a Gigabyte 7VT600P-RZ(-C) with an Athlon XP 3200+, running at 400MHz FSB. As suggested, I'm running with the "load optimized defaults" settings with the DIMM voltage bumped up to 2.8V. Pretty much everything is set to auto-detect; the FSB is definitely 400; I can pull the exact details of the memory settings on next boot, which will probably be tomorrow.

 

Cheers,

 

Matt

[RAM GUY]

Have you tested the modules one at a time, and if you have more than 2 slots please try them in slots 1-3 and see if that helps?

[gothick]

Have you tested the modules one at a time, and if you have more than 2 slots please try them in slots 1-3 and see if that helps?

 

These modules, like the last ones, test fine individually. I've tried both of them individually in slots one and two and they're fine, but they consistently fail (just now in tests 2 and 5) when run together. I've not tried them in combination in anything other than slots 1 and 2 yet (although I did with the last pair and that didn't help.)

 

I can eliminate one more thing, I think: the behaviour is unchanged with my shiny new 500W power supply, which I fitted yesterday. So really, it's got to be the motherboard or the memory, or the combination of the two, unless the graphics card is having some bizarre effect. (Heck, maybe in the end I'll find out the _case_ is causing the problem ;): )

 

I'll try them in slots 1 and 3 and slots 2 and 3 next -- any other suggestions in the meantime? In the end, I'm planning on building another little machine, so I guess I could try a different motherboard at some point, too, but I probably can't afford it this month. And I do quite like the Gigabyte board; it suits my setup well and wasn't vastly expensive...

 

Thanks,

 

Matt

[RAM GUY]

According to the Chipset white papers the modules should be placed in slots 1-3 and can you tell me the exact bios settings you have set for both CPU and memory?

[gothick]

According to the Chipset white papers the modules should be placed in slots 1-3 and can you tell me the exact bios settings you have set for both CPU and memory?

 

Ah -- I'll try them in slots 1 and 3 then. I swear, the motherboard manual doesn't mention this; I did look, twice, and I've just double-checked!

 

I'll see if I can see all the exact settings I'm using -- but the BIOS isn't telling me what values it's using for some things when it's using the "Auto" setting. Basically, I'm using "optimised defaults" with 2.8V DIMM voltage set manually.

 

Thanks,

 

Matt