9NDA3+ - 4 dimms - SS or DS ?

[Spycat78800]

Hi all,

I'm still trying to solve my problems overclocking my newcastle 3800+ (memory problems...explained in a former thread), and then i find a very strange "issue" on other nf3 ultra mobo like MSI (it's an example, don't take it for you).

It seems that this mobo doesn't support 4 Double sided dimms on DDR400 but only on DDR333. In such case, you can fix manually memory frequency to DDR400 and it works but for some reasons, bios (or cpu controller i dunno) consider memory as Overclocked so you can't overclocked it more very much without errors. In fact, it seems i have exactly same problems on my epox 9nda3+ (memory detected as ddr333 when you set 4 dimms but detected as ddr400 when you set only 2....) so ...

1 ) Does anyone using 4 dimms with epox 9nda3+ use single sided dimms or double sided dimms ?

2 ) Does anyone if there is a limit on newcastle memory controller (or on nf3) for a certain number of memory chips used on dimms ?

3 ) Does anymore know the address of a good computer exorcist ?

Thanks for all

[Timelord]

It's not the motherboard or NF3 chipset, but the memory controller in your Newcastle cpu limiting you to 333MHz with 4 Dimms.

I had the same with my "old" Venice 3500+ and 4 sticks of 512MB OCZ but since upgrading to an X2 4400+ the memory runs at 400MHz stock (and about 450Mhz overclocked) albeit at 2T rather than 1T timings - this is on a 9NDA3+. From all accounts I've read the 2T timing limit seems to be down to the BIOS (fingers crossed for a new BIOS to fix it soon).

[Lazgoat]

I've run 4 x 512MB TCCD modules up to about 250-255HTT (1:1) with a 3500+ Newcastle with very good stability at 2T.

[Spycat78800]

Lazgoat, your 4 dimms are single-sided or double-sided ?

[Lazgoat]

Originally Posted by Spycat78800

Lazgoat, your 4 dimms are single-sided or double-sided ?

I believe double sided.

[Spycat78800]

According to a reply from AMD about my memory problems with newcastle core, it seems that newcastle's memory controller isn't supposed to handle 4 double sided dimms at frequencies over DDR333 (and 2T...with 1T we cannot go over DDR200). AMD said also that this kind of problem have been solved with venice (for mono core) and x2 core...

As DDR400 works with 4 dimms on 9nda3+, i should consider as lucky ... ...gasp...

At least, i managed to overclock to 212 fsb (same voltage, same timings)....so it's not so sad either...

[Lazgoat]

I'm surprised you only managed 212HTT. Are the memory modules matched pairs? If they are two sets of matched pairs have you tried each pair individually?
Also try relaxing the memory timings, at speeds over 230 my four modules at 3,4,4,8 are good to about 270HTT. Separately one pair are good to 305HTT and the other to 270-275 at these timings and 2.8V.

[cadaveca]

Originally Posted by Timelord

From all accounts I've read the 2T timing limit seems to be down to the BIOS (fingers crossed for a new BIOS to fix it soon).

Not true. might as well uncross those fingers now. you need tccx ram, the right motherboard with low crosstalk on the mem interface, and ultra-high drive strength.

Bios is only part of the story. I've spent about 2 grand now trying to figure out this issue from user's perspective, and it really does seem to be due to the cpu and really not much else. Any reports of 1t working with 4x512mb might be just a "placebo effect"...i'm awaiting delivery of the board that is supposed to hlep this come to be reality to confirm, but boxes have to be made yet from what i hear(canada requires all merchandise sold in our country to bear both french and english on the box).

[Hozee]

And be gladly to be able to run 4 x 512 ( double sided) @ 2t. It is not always so obvious. Example: my 9npa+ sli has not wathever problem with 4, double sided sticks @ 2t....., but..., my other 9npa3j mobo d'ont support 4 sticks at all. With 2 sticks.... super overklocker, with four sticks..... worthless, even windows would'nt install normally ( all at stock speeds !), even after several times.
And for the record: on my 9npa+ sli it is a winchester 3500+ @ 2630 mhz running with 2 sticks @ 1T or 4 sticks @ 2T with the same latencies and mhz's while the 9npa3j runs with a venice 3000+ rev.E @ 2610 mhz with 2 sticks. 1T or 2T problems are mensioned to be related to the CPU controller but 4 memorysticks on the mobo seems to be an other issue.
So, i ame confused, is it the CPU memorycontroller or ... the mobo who is going wrong. In mi case not the 9npa+ sli ... but the 9npa3j ... so, it is the mobo in my case, ore not?...... Sorry Epox, but nevermind, Epox is still my bigest favorite. Every Mobo has its faults.

[Spycat78800]

In fact, i can just handle 210fsb not 212 (after some tests, i'm unstable at 212 even with upper voltage ->0.1v ). For my dimms, i chose them by serial number. So my 4 dimms are exactly the same (same serial#), and have been tested separatly without errors.

For timings relaxing, i just read an article about newcastle controller. I this article, we can read that on common bench(3dmark,...) and games (ut2004, farcry...) the results are the quite same between cpu @250 (1:1) with 3-3-3-8 timing and cpu@215 (1:1) with 2-2-2-5 (but it seems that it isn't the same results with venice and x2).

With these results, i'm quite disturbed about timing relaxing now ...Nevertheless, I will continue my own tests...