Jump to content
Corsair Community

Unstable performance from RMA replacement Dominator 8500's


Recommended Posts

Hey y'all,

 

Yesterday I received a pair of Twin2X2048-8500C5D's from TSXpress in the Netherlands, RMA'd as a result of Post ID #371204

 

The replacements are marked 2.10v, ver 6.1. My original sticks were rated 2.20 volts, but as they only ever needed 2.075, I didn't see the slightly lower rating on the new ones cramping my style to any degree. However...

 

I installed the new Dominators and just booted straight into Windows on the same settings my "spare" memory had used - 1:1 memory bus and 1.925 volts.

 

Then, I restarted and DELed into the BIOS to set it back up for the faster RAM. I loaded up exactly the same configuration the system was running (with complete stability) immediately prior to the first set failing.

 

This was 375MHz FSB, 1:1.5 memory bus (so DDR1125, slightly overclocked), 5/5/5/15/36, 2.075vDIMM.

 

Except now, with the replacement memory, the system wouldn't even POST, the debug meter again showing C1. I reset the CMOS and set it back up at stock 333MHz FSB (so DDR1000 memory bus). Booted into Windows and ran stable for the few minutes I had it up.

 

Then I tried raising the FSB to 356MHz - so pushing the memory to its rated DDR1066. The system booted into Windows, but locked up a few seconds after the desktop appeared. I reset, and pushed the vDIMM up to the rated 2.100 volts to see if that would help.

 

Restarted, and the system locked up again a few minutes into Windows - while idling, too. I've since had to drop the memory bus down to 1:1.20 to get stable operation for anything more than a few minutes.

 

The system has been running completely stably with a pair of CM2X1024-6400's (on the 1:1 DRAM multiplier) at 375 FSB while the Dominators were being replaced so I know the other parts of the system are still okay at this speed. It seems, at this point, like these new sticks just aren't a patch on the old ones.

 

I could have lived with the new sticks maybe not having quite so much headroom over their rated speed as the old ones, but as things stand, these new ones won't even run stable at close to DDR1066, let alone above it, in a context where I know the rest of the system is capable of doing so, and was, with the older pair of modules.

 

I've been a bit busy work-wise to stop and do a lot of detailed stability testing today, so in the meantime does anyone have any suggestions? Thanks in advance. :)

Link to comment
Share on other sites

Thanks for the quick reply RAM GUY, please see inline:

 

Do you have the same issues if you leave the CPU at default settings and then configure the following settings manually:

 

Memory Frequency: 1066MHz

Memory Voltage: 2.1v

TCL: 5

TRCD: 5

TRP: 5

TRAS: 15

Command Rate: 2T

 

I'd already set those timings for my old sticks, and can confirm this is what I'm currently running on the new pair. For the time being I've set all the other memory timing parameters in the BIOS to Auto, as I was again having difficulty getting the system to POST if I used some of the timings for my old sticks.

 

However, with regard to your comments about leaving the CPU at default settings and setting the memory frequency to 1066, this is impossible on these motherboards given the selection of FSB to memory bus multipliers/dividers available. Either you can have the CPU at default, or the memory at default, but not both at the same time.

 

If the processor is set at default (333MHz FSB) speeds, the DDR frequency on the nearest (1:1.5) FSB to DRAM multiplier is only 1000MHz. Conversely, in order to obtain a 1066MHz memory frequency, an FSB of 355MHz must be selected.

 

Test with http://www.memtest.org and let us know if you are getting errors!

 

I've run some Memtest sweeps this evening, with the following results:

 

With the CPU at default speed (333MHz FSB), and, DRAM at 1000MHz with all other timings and voltage as recommended by you above, Memtest runs completely without any errors.

 

If I then switch the FSB to 355MHz, thereby giving 1066MHz memory frequency, Memtest starts to throw off errors are early as test #3.

 

If I keep the FSB at 355MHz, but drop the memory multiplier to 1:1.25, the test completes successfully, thereby indicating that the RAM is directly causing the errors in the previous run.

 

Just to see what would happen, I put the memory bus multiplier back up to 1:1.50 and dropped the vDIMM to 2.075. On these settings Memtest crashed and caused a spontaneous reboot within a couple of seconds of starting.

 

Finally, again out of idle curiosity I kept the speed the same but tried upping the vDIMM to 2.125. With this higher voltage I was able to complete a successful Memtest run at 1066MHz memory frequency. I’ve now dropped the voltage back to 2.100 and am running at 340MHz FSB (1020MHz memory) in order to try and ensure stability.

 

So, the conclusion of this testing so far is that these new sticks will not run stably at their rated 1066MHz without needing more than their rated voltage. Conversely they will only run stable somewhere below their rated performance level if the recommended voltage is adhered to).

 

While I can accept that my old pair of sticks might have been unusually good, it appears on current evidence that the replacement set aren’t quite in the ballpark. Any further advice or testing tips appreciated.:):

Link to comment
Share on other sites

I see where you're coming from with the BIOS, but if that was at fault wouldn't that also imply that the system shouldn't have been stable on the old sticks either?

 

Again it would ran absolutely rock steady (clear through repeated Memtest passes, and any task I threw at it) with the old RAM at 1125MHz with the timings above and just 2.075 volts. In fact if I recall right it'd pass at 2.050 volts - the extra click was just to add a bit of headroom.

 

Anyway, if the board would perform like that on the old sticks, it tends, to my mind, to suggest there's nothing hugely wrong with the hardware or the BIOS.

 

Is there anything significantly "less good" about the more recent rev 6.x 8500's when compared with the earlier ones? If anything I'd have hoped that yields and stability might have improved over time - although again it's possible that my old pair might have been on the "better than average" side of normal.

Link to comment
Share on other sites

  • Corsair Employees

I do see where you are coming from but the modules we sent to you were tested by Tech support before they were sent I believe and the fact they will run at a lower setting would suggest they just react different with these modules. Can you test them in another make of MB? Maybe a friend can help you just to be sure?

I don't have a problem testing the modules for you personally if you want to try that. Just call in and ask for me 800-205-7657.

Link to comment
Share on other sites

Thanks, I have another system I can build the memory (and if necessary processor) into, but it's over in my office, and I obviously need a bit of downtime on both to swap things over. I'll report back once I've done that.

 

I hear what you're saying about Abit - it's a shame they're pulling out of the board business but to be honest their BIOS development has been inconsistent for a couple of years. I'm an embedded systems designer/programmer by trade so it's not unknown for me to do some low-level BIOS tweaking myself from time to time. For the sake of this testing with the new RAM though, I've been using a stock Abit build.

Link to comment
Share on other sites

Archived

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

×
×
  • Create New...