SLi_dog

heh heh, oh

So prices are about the same, few.

JAFFA

After what eva said about TRD I decided I better get to the bottom of what is going on with this board, so did a few tests at 3-3-3-8 this evening:-

All tests done with the following settings:- (VCore 1.45v, VDIMM 2.45v VMCH 1.75v, VFSB 1.4v, ICH 1.2v)
3-3-3-8 TRD 5



3-3-3-8 TRD 6



3-3-3-8 TRD 7 (tests from other day)



3-3-3-8 TRD 8



Summarised Results


As you can see the actual difference in clock speeds is not that big, a couple of MHz here and there. The impact of TRD is most noticeable in the 32m times where the tightest TRD of 5 beats down the Slackest TRD of 8 by 6 seconds despite the CPU being clocked 14 MHz slower. It's also surprising to see that the difference in times between TRD 7 and 8 is smaller than one would expect.

To be truthful I didn't think the board would even be able to run TRD 6 let alone TRD 5, it certainly struggles with TRD once the memory speeds are up into the 500 MHz range with slack alpha timings. I'm gonna keep testing and see how TRD affects the clocks acheived with different timings and all that jazz.


Take it easy fellas,

Maff

eva2000

nice as i suspected tRD = 5 is fastest for 32M

JAFFA

Moved on to 4-3-3-8

Unfortunately with 4-3-3-8 timings the system would hardlock when set to TRD 5 so I started at TRD 6 and booted in a little higher at 9 x 360 4:5 which automatically made some of the secondary timings a tad looser.

All tests done with the following settings:- (VCore 1.45v, VDIMM 2.45v VMCH 1.75v, VFSB 1.4v, ICH 1.2v)


4-3-3-8 TRD 6



4-3-3-8 TRD 7 (32m test from other day)



4-3-3-8 TRD 8



Summarised Results 4-3-3-8


The results are pretty much in line with the 3-3-3-8 results in the previous post, just a shame that TRD 5 won't hold. Kinda wierd actually being as the only major difference was CAS 4 which you'd think would make it easier.

alpha0ne

Thanks JAFFA for posting your findings, always useful and interesting (well to me anyway )

SLi_dog

That is some insanely nice RAM there Maff. 3-3-3-8 tRD 5 @463Mhz

How low can you drop the tRAS value? Can you get it down to one for 32M? I'd be interested to see the time difference

Try setting tRFC to 20 for SuperPi using memset too. (carefull you don't corrupt you OS though)

alpha0ne

I've been corrupted from a very early age and it aint hurt me none

JAFFA

Fellas, things are getting a bit wierd, I'm fairly certain that this board is just running out of steam and is struggling to hit the high memory speeds. I was getting bored waiting for 32m tests to finish so I'm just going with 1m tests for now, but on the whole 32m stable is about 12MHz lower than 1m stable.

All tests done with the following settings:- (VCore 1.45v, VDIMM 2.45v VMCH 1.75v, VFSB 1.4v, ICH 1.2v)

547 MHz 4-4-3-8 TRD 6 *I did actually get to 550 4-4-3-8 TRD 6, but saved the wrongs screenie*


555 MHz 4-4-3-8 TRD 7


556 MHz 4-4-3-8 TRD 8





550 MHz 4-4-4-8 TRD 6


552 MHz 4-4-4-8 TRD 7


552 4-4-4-8 TRD 8


Summarised Results


Well there you have it I can clock higher with 4-4-3-8 than with 4-4-4-8... who'd've thunk it!?! I did try higher with 5-4-4-8 and 5-5-4-8, but slackening the timings hurt more than it helped and only gave me literally a couple of MHz more RAM speed. The FSB isn't maxxed being as it will run 32m at 460MHZ, the RAM is running better with tighter timings so I can only assume that this board is not liking memory speeds much above 550MHz. I've not seen any evidence of anyone with these P5WDG2 WS Pro's having better luck. Unless I'm missing something fundamental to getting high clocks I reckon it's a 975x thing.

Thanks for reading,

Maff