Weak memory controler on X2 3800

[GuniGuGu]

Hi, i'm not sure about the rest of you guys, but what i''ve noticed, is that my X2 3800 has a pretty weak memory controller, espically with 2x1gig modules.

I can say this, because I've been able to test the memory out on very low multi and have it come up flawless easily at 233 LL (1t CL2)

However, if I try the same, with the CPU going at 2.65ghz, it'll hang in p95...

So with some testing, I found 2 settings that give me 2 choices...

I can either run the memory at:

220Mhz@ 2.5-3-2-2 -2T
or
132.5MHZ@ 2-4-2-2 -1T

I've also noticed that it doesn't respond to stable at increases to the DRam voltage... However it can post higher (perhaps i need more testing), but the above are the 2 extremes i've found between high bandwith and LL at the stock 2.6 voltage.

I guess i should benchmark em, to find the difference, but what do you guys think I should do? and also, how have you guys gone as far as 2x1gig modules in terms of memory settings.

 

[Shimmishim]

what do you mean by 233 LL ?

 

[Duvie]

Week?? Weak!!!


I think this is hit or miss and can very likely have to do with motherboard shortcomings as well as specific bios and timing compatability with certain ram sticks...

I had noticed with my winchester is that I could run at 2.48ghz 9x274 at 1:1 cas 2.5-4-4-7, yet as I reached 290's I had problem running a 166 divider at cas 2.5 with ddr speeds lower then before....I think it has to do with stress and where you start to reach the chips limit...

 

[GuniGuGu]

Originally posted by: Shimmishim
what do you mean by 233 LL ?

233 low latency (2,3,2,2)

 

[Budman]

www.spelling.com

 

[GuniGuGu]

Originally posted by: Duvie
Week?? Weak!!!


I think this is hit or miss and can very likely have to do with motherboard shortcomings as well as specific bios and timing compatability with certain ram sticks...

I had noticed with my winchester is that I could run at 2.48ghz 9x274 at 1:1 cas 2.5-4-4-7, yet as I reached 290's I had problem running a 166 divider at cas 2.5 with ddr speeds lower then before....I think it has to do with stress and where you start to reach the chips limit...

Well that was pretty much my point... i was trying to get at.. when you start to stress the chip, it's memory controller starts to stutter and give problems... I.E. Even if your memory can do 300 at 2-2-2-2 1t 1:1 or whatever, if the chip is near it's limit, it will not be possible.

And I also think that 2x1gig modules are prolly harder and more taxing on the cpu then 2x512 like most run.

and I do assume you were talking about 1t command rate.

 

[GuniGuGu]

lol...

thanks for that very helpful and informative reply

hehe.. until i read your post i didn't understand what duvie meant when he said:
"Week?? Weak!!!"

lol.. i completly overlooked it, even though he pointed it out as well.. i thought he meant like "Weak weak??!!!" (being sarcastic and suggesting that they were very potent and strong)

I think i'm just very tired

Getting back to topic though...

 

[coomar]

that 4 in 2-4-2-2 is way too tight, its usually cas (the first timing) + the other timings which would 6, its board specific about which it likes

 

[Shimmishim]

there are no 1 gig modules that are capable of 2-2-2-3 or even 2-2-2-6

that's like 1 gig bh-5 modules and i don't think they exist..

be sure you set the timings correctly cuz i hav a feeling your timings are way off...

and also could be the mem controller on the chip as you said... but i'd like to think its your timings first

 

[GuniGuGu]

Originally posted by: coomar
that 4 in 2-4-2-2 is way too tight, its usually cas (the first timing) + the other timings which would 6, its board specific about which it likes

That was a type and meant to be 3.

I have no idea what you meant by, "he other timings which would 6, its board specific about which it likes" are you implying the the other timings are all up to the boardboard and the ram and cpu have no effect on them?

 

[GuniGuGu]

Originally posted by: Shimmishim
there are no 1 gig modules that are capable of 2-2-2-3 or even 2-2-2-6

that's like 1 gig bh-5 modules and i don't think they exist..

be sure you set the timings correctly cuz i hav a feeling your timings are way off...

and also could be the mem controller on the chip as you said... but i'd like to think its your timings first

Do you know what we're talking about here?

My whole point, was that I have tested the ram ok at (233LL) at lower CPU speeds without any problems. But as CPU speed increases the ram must run slower in order to accomidate for the faster CPU.

Also, I never said it was 2,2,2,6, it was 2,3,2,2.

 

[Markfw]

Yes, exactly ! As speed increases for the same memory chips, timings must be loosened, thats the way it works.

 

[imported_Kensai]

2-3-2-2 is... Idiotic to say the least..
Hell, I could have an Intel P-D system outperform your clunker if you're going to use 2-3-2-2 for Ram..
2-3-2-2 isn't tight.. 2-2-2-5 is generally known as "tight".

I'm sure at those speeds, it's not the memory controller's fault but your ram.

 

[Hyperlite]

Originally posted by: Kensai
2-3-2-2 is... Idiotic to say the least..
Hell, I could have an Intel P-D system outperform your clunker if you're going to use 2-3-2-2 for Ram..
2-3-2-2 isn't tight.. 2-2-2-5 is generally known as "tight".

I'm sure at those speeds, it's not the memory controller's fault but your ram.

heh yeah, there is no way you will be able to run stable at those timings... like kensai said, 2-2-2-5 would be pretty amazing, but 2-3-2-2 is just plain unreasonable.

 

[GuniGuGu]

Yeah, I screwed up..

Meant to read:
220Mhz@ 2.5-3-2-5 -2T
or
132.5MHZ@ 2-3-2-5 -1T

 

[imported_Kensai]

Originally posted by: GuniGuGu
Yeah, I screwed up..

Meant to read:
220Mhz@ 2.5-3-2-5 -2T
or
132.5MHZ@ 2-3-2-5 -1T

Loosen your timings to overclock more...
X-X-X-5 is hard to overclock with... -7 and -8 outperform the -5 on A64 systems anyways.

Maybe something like 2.5-3-3-8 @ 220-240+ MHz @ 1T

 

[JEDIYoda]

Have you tried to Loosen your timings to overclock more...
X-X-X-5 is hard to overclock with... -7 and -8 outperform the -5 on A64 systems anyways.

Maybe something like 1.5-2-2-7 @ 320-340+ MHz @ 2T

 

[ssvegeta1010]

Originally posted by: JEDIYoda
Maybe something like 1.5-2-2-7 @ 320-340+ MHz @ 2T

WTF?

I hope you mean effective speeds.

 

[imported_Kensai]

Originally posted by: JEDIYoda
Have you tried to Loosen your timings to overclock more...
X-X-X-5 is hard to overclock with... -7 and -8 outperform the -5 on A64 systems anyways.

Maybe something like 1.5-2-2-7 @ 320-340+ MHz @ 2T

Oh wow. I want 1.5-2-2-7 @ 320-340+ MHz @ 2T...

:roll:

 

[GuniGuGu]

OK well thanks for the advice guys...

I just wanted to make a point, that the "quick and dirty" overclocking method, really doesn't apply here, as the highest cpu speed, will not even come close to allowing you to run your memory at it's highest possible speed, because the on-die memory controller really started to weaken under pressure situations..

So i'm not sure what it's like for you single core guys, but as i've learned the hardway Overclocking the X2, best way, is to lower all your ram timings and speed, and ht, start from your base cpu speed/voltage and increase it in steps. I.E. Overclock the FSB by 10mhz, prime95 it, for at least 8hrs, if success try another 10mhz, if fail, increase voltage and test until you have the fastest most stable speed.

Once you have the fastest cpu speed possible then increase your memory, try for something realistic, and p95 test, if ok then you can tighten timings/increase freq gotta keep p95 testing it, untill p95 starts to fail within 8hrs (at least). I don't think memtest is very effective because the cpu needs to be stressed as well..


This o'c process takes a while, but i have a full time job, and i need sleep and i think it'll produce the most stable results.