Low Clock Spi32M Eficiency

[badatgames18]

cheers guys

hey bad what's your bclk for 6m57 ?
btw, the JNL was 1.82v, 2 of my sticks won't take much more vdimm without errors.
but higher bin like 1866 Blue might do CL5

Zen: your 25/26 RTL explains it very nice!
i only noticed latency but actually your read was better too.

my aida tfaw was same as bios set 16 (4 x tRRD)
For bug did you mean that Aida doesn't show tWTR, tRRD, tRTP < 4, but can set < 4 in bios ?
think aida is correct with min = 4.
i didn't see perfomance increase < 4 so i set all @ 4

pretty high.. 109.4 iirc in os it's .01 higher i think. i am bad at efficiency lol.. it's on my xp hard drive, i'll post when i'm on it again

[ZenEffect]

cheers guys

hey bad what's your bclk for 6m57 ?
btw, the JNL was 1.82v, 2 of my sticks won't take much more vdimm without errors.
but higher bin like 1866 Blue might do CL5

Zen: your 25/26 RTL explains it very nice!
i only noticed latency but actually your read was better too.

my aida tfaw was same as bios set 16 (4 x tRRD)
For bug did you mean that Aida doesn't show tWTR, tRRD, tRTP < 4, but can set < 4 in bios ?
think aida is correct with min = 4.
i didn't see perfomance increase < 4 so i set all @ 4

4 act can be set to lower than 4x. i will post up diff between 16 and 13 as there is a gain

*edit*

TFAW 13


TFAW 16


wish someone else would run 107.1, or re-bench with a better time. i really really really dont want to buy dice to run 32m on 1333 divider.

[Morphling]

This is getting realy interesting

[ZenEffect]

This is getting realy interesting

yeap, i hope people are picking up on the tweaks and are re-benching. its always great to see efficiency progression in threads like this

on a side note:

107.2 bclk gets me into the 7:00:8XX range, so scaling is moving along just as expected

@ cheapseats

read gets better with tighter tfaw, when you are too tight, it drops as does the 32m efficiency.

I agree that dividing the scores would be better (Like 105.X range , 106.X range ... etc) , but first we have to compare Gigabyte & Asus & MSI boards . i'm not sure if Gigabyte gives correct clocks in CPU-Z as it reads 0.2Mhz lower than Bios settings. Easytune reads the same clocks as BIOS .

Stelios, please take a look at easy tune one more time. The behavior I am seeing is the bclk slider reports the same as what is set in bios, however the .2 loss is apparent when you click the cpu tab. I cant confirm for the boards you have, but thats what Im seeing with P67A-UD7 B3 and Z68XP-UD4. I suspect it has something to do with some skew options that arent visible in bios.

[cheapseats]

hmm Zen, can you test tFAW 16 performance again but on XP ?
if tRRD 4 = min, then 4*tRRD 4 = min 16 clock window..
you can't change 'four activate window' into '3.25 activate window'

some new runs with your 25/26 RTL

note still old aida version (it's showing around +150Mb/s read, +1Gb/s copy, but latency same as newer version)

107.3 6m59.797



108.1 6m56.844




these 2 also 25/26 with tRAS 13 (was 14), tRFC 34 (was 36), tRC 18 (was 25)

107.2 6m59.906



108.1 6m56.687

[ZenEffect]

hmm Zen, can you test tFAW 16 performance again but on XP ?
if tRRD 4 = min, then 4*tRRD 4 = min 16 clock window..
you can't change 'four activate window' into '3.25 activate window'

some new runs with your 25/26 RTL

note still old aida version (it's showing around +150Mb/s read, +1Gb/s copy, but latency same as newer version)

107.3 6m59.797

[/IMG][/URL]

Nice runs! Good to see RTL is working out for you

I will re-bench it tonight... but I im 99.999% sure there was a gain when I tested previously. Normally I'd agree with you, but trrd x 4 = 4 I think aida is bugged.

will try to run off 107.2 XP, but the board is being weird.... 107.2 takes some patience and fanagalling. should net me 7:00 time.
I see why now gtx2 are truly worth the money. trc stops at 28 for me, 31 is optimal. trfc stops at 35, 36 is slow 37 is optimal. tras 14 is as low as it goes. im only giving 1.65v... i dont want to fry hypers at low clocks... though they are worthless to me at high clocks so i may throw that into the continuous testing/tuning

*edit*

TFAW13


TFAW16


so mabye it doesnt make sense, im going off of what ive tested. tfaw below 16 = gain. my 32m is very consistant +-.02 every time. sorry, dont have aida on the xp hd to show timings but i assure you only tfaw was changed. i can install it to show tfaw change and re-run with pretty much the same results if you really really really want me to.

[l0ud_sil3nc3]

no hypers here just working on the same multi as I can't even boot on the 1:5 divider unless I am at 6-7-6

[Counter CS]

I'm glad i dicided to open this thread, learning some tips and tweaks

Keep on the good work.

Research is the key for WR on SPi32M

[ZenEffect]

I'm glad i dicided to open this thread, learning some tips and tweaks

Keep on the good work.

Research is the key for WR on SPi32M

see sig (not there yet, but some day... im fairly confident in my 2d ability)

lets see some re-benching! Counter CS, you should be at my times, all the info is given in this thread pretty much, dont let the cas4 fool you

anybody care to share some os tweaks? i will if you will

@ loud

we are using 1:5 ala 1333 strap. i will switch to gskill blue on giga right now to see where things are at.

*edit*

now i wonder if tras12 is doable on my "good" sticks. bsod at windows boot 1.5v



*edit edit*

not going to be as fast as hypers. subs are higher, rtl is 27/28

[l0ud_sil3nc3]

@ zen

what volts on for the 1:5 strap and those timings on bbse.

[ZenEffect]

@ zen

what volts on for the 1:5 strap and those timings on bbse.

1.65v