Compatibility of GA-P35-DS4 with 2 pairs of TWIN2X4096-8500C5C

[slinger]

Hello,

 

I'm floating between deciding whether I have a hardware problem, or just a less than optimal compatibility situation with my Gigabyte motherboard and 2 new kits of TWIN2X4096-8500C5C. My modules happen to be v4.1.

 

I was experiencing months of BSODs, where I blamed my previous inexpensive (cheap?) memory. I had consistent failures in memtest86+ and Vista's Memory Diagnostic (it seemed that the Vista test found issues that memtest missed...64-bit related?). I was happy to pick up two 4GB sets of 'the good stuff'...giving me a 4-8GB upgrade and hopefully getting rid of my stability issues. Unfortunately, the errors became more frequent. I ended up RMA-ing my motherboard, since I didn't believe I could get bad memory yet again. I pretty aggressively test my memory before deploying the system and was lax only this last time since I had gone through 3 rounds of returns before giving up. The mobo came back NTF leading me to go through another round of testing to try to a> justify a 2nd return attempt on the mobo, or b> figure out if there is actually something wrong with the memory.

 

I started with one pair in dual-channel mode @1066 Mhz, with advertised 5-6-6-18 @2.1V (I saw references to 2.2V in this forum and tried this also w/o success) timings. Memtest failures, typically in tests 5 and 6. I then tried both chips individually in slot 1 (closest slot). Both tested perfectly. A little baffled that there wasn't a failure, I tried both chips in slot 3 (matching dual channel slot for #1). Both now had failures when tested individually. I tried both individually in slot 4 (in case slot #3 had an issue), I still have failures. The trend so far is 'close slot good', 'far slots bad'... with single chips.

 

Thinking that there may be a trace length issue on the PCB, I tried scaling back to DDR2-800. No luck at all with CAS4. 5-5-5-18 would post @800. It turned out that at this very relaxed timing, i could run cleanly overnight. I was able to creep up to 4-5-5-15 @800 with all four 2GB modules installed.

 

For another test, I had my first set of DDR2 1GB chips (same competitor as above) that are DDR2-800 4-4-4-12 rated. These modules ran perfectly at this speed in dual channel at the 'farther' slots... and to add insult to my situation, ran @1066 5-5-5-15 2.1V.

 

This suggests to me that the bus can take the speed (mobo is ok) ... *unless* four 2GB modules (opposed to two 1GB modules) puts great enough of a load capacitance to pollute the bus....except at 800 Mhz 4-5-5-15. I'm still not 100% convinced the memory itself is bad. It does test cleanly individually at rated speed ... so at least the memory cells are intact. Could this simply be a compatibility issue? I'm not overclocking this system, my goal is to get the performance I paid for, but I want system stability to be #1. 4-5-5-15 @800 Mhz is fine if that's all I can get ... while being stable.

 

I don't see alot of confirmed information on the website about the 'C5C' modules. No data sheets on the XMS product page, not listed on the XMS2 timing sheets in Ram Guy's sig links, contradicting information in forums... Discussions say they should be the same as the 5-5-5-15 rated 'C5' chips, but I've seen the 'C5C' quoted anywhere from 5-5-5-15 to 5-6-6-18. Also, what is the expected 800 Mhz performance for these chips? The SPD register is a CAS5 set, but I would expect these modules to be able to handle 4-4-4-12+. What chip manufacturer is v4.1? In general, documentation for this line seems lacking. Any additional information that can be provided would be appreciated.

 

Sorry for the dissertation ... Thanks,

 

Jason

[RAM GUY]

With 4 modules I would suggest setting the memory frequency at DDR667 or DDR800 and set the memory Voltage to 2.1 Volts and set the NB/MCH/SPP Voltage to +.2 Volts as well and test the system with http://www.memtest.org. In addition, with some MB's (Mostly ASUS) you have to disable legacy USB in the bios when running any memory test.

[slinger]

I have been able to successfully get all 4 modules testing cleanly (memtest booting from a USB stick) at DDR2-800 4-5-5-15 @2.1V (+0.05V MCH). I tried going as high as 2.2V based on feedback from other forum questions. The Gigabyte board seems to add 0.025V to the expected setting, so it actually had 2.225V.

 

If I can't get DDR2-1066, I guess I'm hoping to see DDR2-800 4-4-4-12 since this is what I was able to get with my previous PC2-6400 1GB modules. With the 2GB modules, it seems that specs tend to be less aggressive than 1GB modules. Do the denser modules put more load on the bus making the slower timings necessary? I guess I'm asking if I have realistic expectation from these denser modules.

 

Also, as a side note, I'll be happy to report that these Corsair modules with the relaxed 4-5-5-5-15 timings are reporting (in Memtest) a better bandwidth than my old modules (non-Corsair) with 4-4-4-12 timings. I think the Corsair 800 MHz bandwidths were pretty close against 1066 MHz timings also... It makes the less aggressive timings easier to take.

 

I'll also try the USB Legacy setting you suggest. I'm pretty sure this setting was enabled in my BIOS. Does this setting just pollute ram tests? If I turn it off for successful testing, can it be set on for normal use without causing stability issues?

 

Thanks for the quick support!!!

[RAM GUY]

I would suggest you lower the memory Voltage to no more than 2.1 Volts and increase the NB/MCH Voltage if needed. Other wise you will VOID the warranty on the modules its 2.1 Volts Max.

[slinger]

Not a problem. I didn't like the idea of running at 2.2V anyway.

 

I have the v4.1 modules. The 2GB C5C line doesn't seem to be in the 'DDR2 IC Inquiry'.

v4.x in the 1GB C5 area are PSC chips. Do you happen to know what chips are

in the v4.1 2GB modules? Updates to this area, the 'Qualification and Testing'

section, and adding the data sheet to the XMS product page would help

reduce confusion.

 

Thanks!

[RAM GUY]

The version number will be the same no matter if the "C" or just C5 the "C" at the end is only for the heat spreader. And I am sorry Data Sheets are no longer available.

[slinger]

No problem about the data sheets. It'd just be helpful.

I guess I was just surprised to see the 'newer' C5C product to have

less aggressive timings compared to the older C5 products with the

older heatsink design.

 

I was able to run the test you suggested with the 'Legacy USB' BIOS

setting disabled. I had similar results.

With all four modules: At 800 MHz (with RAM +0.3V and MCH +0.05V),

I can go as fast as 4-5-4-12 with flawless memtest runs.

At 1066 MHz, I get fewer errors as I relax the timings (with RAM +0.3V

and MCH +0.2V). For example, 5-6-6-18 had a ton of errors. 6-7-7-21

had *many* fewer. 6-8-8-24 had just a few, but still not perfect.

The slowest my mobo can time the CAS is 6 or else I'd try 7... but oh well.

 

Its sounding like this is accepted/expected for 4 modules, so my hardware

is doing the best it could do...

 

Thanks for all the help.

 

Jason

[RAM GUY]

Yes running 4 modules is more load on the MB and memory controller so the chipset will have an impact and the design of the MB but with 4 modules it is common to run the memory at a slower frequency.

[slinger]

Its been a few days now since I've reassembled the system.

So far the memory has been running solid with my new timings.

4x2GB TWIN2X/CM2X2048-8500C5C: DDR2-800, 4-5-4-12 @2.1V, +0.05V MCH

 

I guess my only other question for future reference, is if I ultimately

end up running DDR2-800 with 4 modules, could I have opted for a pair of

TWIN2X2048-6400C5C kits and had similar results? Are the PC2-8500 modules

binned for performance so I might have settled on looser timings if I had

purchased PC2-6400 kits? Just curious....

[RAM GUY]

No with 2 sets of any modules the next lower speed grade would be suggested so if you did that you may be limited to DDR667.