Jump to content
Corsair Community

DK8N and TWINX2048RE-3200


signals42

Recommended Posts

Well, I seem to have a problem... I just bought a new system based on the Iwill DK8N motherboard. (2x Opteron 250, if it matters) I also picked up 4 GB of Registered ECC DDR400 memory in the form of 2x Corsair TWINX2048RE-3200. I have 2 modules per CPU in DIMM0 and DIMM1 as the MB instructons specify.

 

The problem is: when I boot the machine, it correctly identifies "4096 MB Physical Memory" in the startup message. Then it's POSTs and counts/tests the memory and it only counts up to "3232 MB." I've lost 864 MB of ram.

 

When I boot into Linux (multiple distros and kernels) I've only got 3232M available to me. I do not see the problem in Windows, as XP can only identify 2048 MB.

 

I've talked to Iwill support about this, and their answer (so far) is that no Corsair memory is on the supported memory list. Kingston and Smart modules work fine.

 

Since I couldn't find a retailer who carried the appropriate modules from their "approved" list, I went to Corsair's web site and looked up the DK8N:

 

http://compatible.corsairmemory.com/memorysearch.aspx?modelid=395

 

As you can see, the TWINX2048RE-3200 modules are on Corsair's DK8N list.

 

Anyway, I'm stuck and a little P.O.ed that I spent all of that money on 4 GB of memory only to have 864 MB of it disappear.

 

Does anybody have any suggestions about troubleshooting/fixing this problem? Let me know if I need to provide more information.

 

Thanks,

-Kevin

Link to comment
Share on other sites

  • Corsair Employees

Kevin,

First thing, I think they are just passing the buck here and not trying to help you! This issue you have described will happen no matter what module you have installed. First with Win2K or newer you will only have access to 2.0 Gig of ram for applications and that may be why it's showing 2.0 Gig, second with Win2003 you can change this in the Boot.ini to have full access. With http://www.memtest.org I would see how much memory is showing it should show at least 3232M if not the whole 4 Gig. But if not that would be due too the way the bios reserves memory for PCI devices, and they may be able to adjust this with bios, but this is the default limits set in the reference bios for this chipset! So I would test the system with http://www.memtest.org and let’s make sure there is not a bad module and go from there.

Link to comment
Share on other sites

Well, I set the BIOS back to optimal defaults, then put in the timings you sent me in e-mail. I was unable to adjust the voltage or frequencies as you suggest because the BIOS doesn't let me. I ran memtest86+ for an entire pass and this is what I found out:

 

First, memtest only sees 3231M of memory. (With an AGP aperture of 128M. Less with a bigger aperture, more with a smaller one.)

 

Second, it thinks the RAM settings are "600 Mhz (DDR120) / CAS : 3-3-3-8 / Dual Channel (128 bits)" and it says the chipset is "nVidia nForce3-S (ECC : Disabled)" both of which are wrong. It's DDR400 and ECC is enabled in the BIOS. I've tried using the ©onfiguration menu to change it to ECC Mode "On" but that doesn't change it. It still says "ECC : Disabled."

 

Then there are the errors. 45 of them in the pass. Tests 2, 4, and 7 report errors at the following addresses:

 

000000001f4 - Bad: 000002ff

000000001f0 - Bad: 00000000

000000001fc - Bad: a0008050

000000001f8 - Bad: 7ac3a010

 

(These addresses came up once in test 2 and several times in tests 4 and 7, but the "bad" is always the same?!?)

 

I don't have time tonight to disassemble the thing to check individual memory modules, but will tackle it tomorrow. (The Thermalright XP-90 HSFs on the CPUs have the memory blocked in. It's no small task to swap the memory in this thing. The worst part is that I have to put the HSFs back on to test each stick. :(: )

 

-Kevin

Link to comment
Share on other sites

Well, I had a little time and tried physically switching the memory around. Now I'm even more confused:

 

With 2GB (2 modules) in DIMM0 and DIMM1 on CPU0 and no memory in CPU1 the machine POSTs and counts all 2048 MB of memory. Yay!

 

With 1GB (1 module) in either CPU's DIMM0 or 2GB (2 modules) one in each CPU's DIMM0 slot, the machine won't even POST. I just get a black screen.

 

I haven't had a chance to try both modules in CPU1, or any modules in DIMM2 and DIMM3 yet. I guess I'll try that tonight.

 

Now, the strange part... With 2GB on CPU0 and none on CPU1, I ran memtest. I got the exact same errors in the exact same locations with the exact same "Bad" values as I did with 4 GB?!? Could this be some IO device mapped into memory or something?

 

Also, memtest reports the memory even more strangely with 2GB on CPU0 than it did with 4GB. Now it reports the memory as: 600 Mhz (DDR300) CAS 0-0-0 single channel.

 

Since both modules are on the same CPU in DIMM0 and DIMM1, it should report dual channel, and it's still DDR400 memory. How the CAS timings are set to 0-0-0 is a mystery too.

 

So, I'm not sure what's up. I'll have more time to switch memory around tonight, but I don't know what to try next. Are the memtest errors real memory errors, or just something expected that memtst doesn't know about? Is the fact that it won't POST with only 1 module installed something to worry about? Does the fact that it counts all 2GB of memory when there's only 2GB of memory mean anything?

 

I spent too much money on this thing for this... :mad:

 

-Kevin

Link to comment
Share on other sites

Why does my system display less than 4GB even though I have installed 4 GB?

 

AMD chipset architecture requires memory above 3.5GB to be reserved for PCI devices. You will typically see 3.6-3.8GB available.

 

This is from a Tyan FAQ (I'm using an Iwill, but I'd bet this holds true for any AMD chipset.)

 

Ok, so I guess I can see what you meant about the PCI 2.1 spec reserving memory up there. I just read the PCI 2.1 spec and a ton of other technical info. I'm assuming that with the normal PCI mapping, and a 256MB graphics adapter with a 512MB aperture I could loose a ton of memory.

 

I'm not at the machine at the moment (I have to pretend to work :o: ) but I'm wondering if I can't do an 'lspci -vv' and get the memory ranges that the PCI devices have mapped and figure out if I am seeing the rest of the memory or not? Sounds feasible to me.

 

If it is indeed a HUGE PCI address map that is masking the physical memory, I guess I'll just have to live with it.

 

-Kevin

Link to comment
Share on other sites

  • Corsair Employees
Well I would try the other 2 modules and see if you get the same failure, if so that would suggest some other issue, most likely what you have suggested, due to memory mapping. I would check with the MB maker and ask if they have seen this issue and or if they have new bios you can try! But if you did try both sets one set at a time and get the same errors, I would suspect some other issue!
Link to comment
Share on other sites

Well I would try the other 2 modules and see if you get the same failure, if so that would suggest some other issue, most likely what you have suggested, due to memory mapping. I would check with the MB maker and ask if they have seen this issue and or if they have new bios you can try! But if you did try both sets one set at a time and get the same errors, I would suspect some other issue!

 

Well, while I haven't tried 2 modules on CPU1 only, I've tried both sets of 2 modules on CPU0 and get the same errors at the same addresses with both sets of memory. I suppose if it happens on CPU1, it's probably some sort of memory map, but if not, is it possible I have a bad Opteron (memory controller)?

 

I will see what I can get from Iwill, but they have not publicly released a BIOS newer than what I'm running.

 

Thanks for all of the assistance.

 

-Kevin

Link to comment
Share on other sites

Just thought I'd post the reason I got CAS 0-0-0 from Memtest86+. Apparently it doesn't support multiple CPUs and just queries one of them for the memory parameters. It picked the CPU that had no memory, so it got bad memory parameters.

 

I switched the memory to the other CPU and Memtest looked normal again. Of course it still identifies the RAM as 600 Mhz DDR120, but, whatever...

 

Oh, and I got the exact same errors and "bad" values at those 4 memory addresses on the other CPU, so I am assuming it's something being used as memory mapped IO that Memtest86+ doesn't know about.

 

Thanks again!

 

-Kevin

Link to comment
Share on other sites

  • Corsair Employees
No Problem, I have spoke to I-Will and they are taking a stance of if they have not tested it will not work basically, so I have asked that we try and get them our modules so they can validate them. If I get any information I will let you know ASAP, as I think you have sent us an email as well!
Link to comment
Share on other sites

  • Corsair Employees

Kevin,

I have the DK8X running with 2 sets of this same module at my home, and I do not have a problem with the memory other than basic limitations which I think you understand! In addition, it will not run with an ATI 9800 Pro and some other driver issues that should be resolved with the next release from AMD. And I am using SUSE 9.1 64 Bit Linux and WinXP SP2 so I am confident it's not the memory at fault.

Link to comment
Share on other sites

Archived

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

×
×
  • Create New...