Memory upgrade

[davmax]

After recapping my motherboard (Epox 8RDA+) and the power supply I increased memory to 1Gb PC 3200. Occasionally I would get odd things happening, various error messages and sometimes a flickering video and rarely beeping in the system loud speaker requiring a system restart.

I ran Memtest86 and found that with the two 512Mb memory sticks inserted memory errors would occur with the block move test. No problems with each running on it's own. Lowering the Vdimm made it worse. By making Vdimm 2.9V (maximum) both run without error. the problem occurred if both sticks were on the same memory channel or one in each memory channel. I have the dual memory channel NF2 chipset.

My current conclusion in that the Samsung chip based generic sticks become marginal at the normal operating voltage.

Has anybody got an alternate view?

I have ordered a quality brand matched pair of memory sticks.

[tylerlough]

Re: Memory upgrade

did you try changing the memory timings to a higher number. dual channel sometimes is picky, they should be the same stick for best results. hope this helps

[Rainbow]

Re: Memory upgrade

I've read somewhere that nForce 2 chipsets don't produce valid timings according to DDR400 spec.

[davmax]

Re: Memory upgrade

Thanks guys.

Re Timings. changed CAS from 2.5 to 3.0. Problem still there.

The point to be aware of is that this problem consistently occurred in the top 5% of memory near the 1Gb value and is made better or worse by shifting Vdimm. This to me points to a bad thing happening at a particular block move situation in a particular area of addressing or memory. Does the fact that it is sensitive to voltage have any relation to timing? Perhaps? Maybe the block test spans the two modules making a mismatch critical!!!!

[WNG]

Re: Memory upgrade

late to the party, but just wanted to add, I think you're on to something. Running an Abit board with the same NF2 chipset. Memtest86 showed the same error as you for dual channel operation with non-matched brand 256 MB PC-2700 DIMMs, one being Samsung. System running a Thoroughbred XP overclocked at 333MHz FSB.

The Samsung chipped DIMM seems very picky. The block move error is very rare on my system. The slower FSB may be why.

[trodas]

Re: Memory upgrade

davmex, I got a good results from such situations by throughly cleaning by syntetic alcohol the ram dimms contact's as well, as the sockets. Usualy I apply vacuum cleaner on the sockets (with very small end extension piece suction to maximize the power) and then throughly clean with alcohol the dimm's contacts and make sure I inserting the dimm alreay wet from the alcohol, so it clean a bit the contacts too with alcohol.
Harder way is to apply special contacs cleaning spray on the sockets...

The point is, that they become weak and dirty over times, so... you got to fix that. I use 3 mainboards with 1G DDR 3200 modules and all double channel and all nF2. I must say that double channle is nowhere near picky, it can run with two entierly different dimms and so on...

Truth is, that is much harder convince nF2 to run at higher FSB with 2x512 ... But I still managed 250MHz FSB stable, so... so clean up the contacts

PS. BigPope has a link to pretty cool modded bios and if you used GC Samxons for the recap, then you should be able finally get a 250Mhz FSB from your Epox 8RDA+ easily Yea, as unbeliable, as it sounds. But I know you just want standard clock now, but stable

[Topcat]

Re: Memory upgrade

BigPope has a link to pretty cool modded bios and if you used GC Samxons for the recap, then you should be able finally get a 250Mhz FSB from your Epox 8RDA+ easily Yea, as unbeliable, as it sounds. But I know you just want standard clock now, but stable

If the CPU can hack it... Overclocking success and failure rides more on the CPU than it does the motherboard. Good caps help, but they aren't the key to a successful overclock. Pannies, Rubies, Sanyos, and any other good quality cap will yield the same results.

[gg1978]

Re: Memory upgrade

I always thought the original 8RDA+ was limited to 166FSB stock, with a later revision allowing 200FSB operation.. I remember when a lot of people fragged their BIOS chips by running at 180FSB+. Anyway, 2.9 sounds high for VDIMM.. I would see if you could RMA the sticks. Any DDR DIMM that needs more than 2.7V at "stock" type speeds like PC3200 is probably flaky..

[bgavin]

Re: Memory upgrade

Davmax,

Memtest86 tests 5,8 are the most troublesome with Athlon processors on particular mother boards. For example, these tests fail 100% on ABit KG7 boards. I have a good number of KG7 in service, and none have memory errors in normal operation.

I just finished two nForce2 boards at stocking clocking (333 DDR) using Crucial branded PC2700. Memtest86 runs all tests without errors for extended repetitions. The Crucial sticks came from Newegg, and also have Samsung on them. This is notably different than memory purchased directly from Crucial (which is more expensive).

[Fizzycapola]

Re: Memory upgrade

Hi Davmax,

I just thought I'd share I have also had problems since replacing my Teapo's at RAM slots with Rubycon ZL's. Pretty much same as your problem.

[gdement]

Re: Memory upgrade

I've sometimes been able to stabilize a board with memory errors by upsizing/adding capacitors to Vmem and Vtt. But the effect is pretty subtle, so if it's badly unstable it probably wouldn't help much.

[tazwegion]

Re: Memory upgrade

For what it's worth... I've experienced similar issues with other generic types of DDR modules, however I found by lowering the CAS latency to 3.0 (despite them claiming CL2.5) the issue was resolved, in my situation they both ran fine @ CL2.5 when being used individually


It's often better to pay the extra and get some decent low latency modules IMHO, I've never had any issues with Corsair XMS or Geil Ultra-X series

[davmax]

Re: Memory upgrade

Since I started this thread the problem of failure with two memory modules was solved by installing higher quality memory. GEIL two matched modules for dual channel operation. These are currently running at 2.5 4 4 8. These appear in my signature system info.

Apparently the original problem was simply modules that were not well matched for dual channel.

Thanks guys for your inputs over time.

[tazwegion]

Re: Memory upgrade

Yeah it can be a real PITA to get mismatched or poorly matched memory modules to play nicely together (without resorting to looser settings)