Asus P5K Pro + E8400 + GSkill F2-8500CL5-2GBPK * 2

[PsySc0rpi0n]

Hi guys...

I am trying to test the top max overclock i can get with my kit of 40GB.

I'm with this settings now

FSB - 475Mhz
CPU Multi - 8.5
CPU Freq - 4037Mhz
Mem - DDR2-1141
FSB Strap to NB - 333Mhz

Dram timings
CL - 5
RAS# to CAS# Delay - 6 ->7 -> 8
RAS# PRE Time - 6 ->7 -> 8
RAS# ACT Time - 17 ->19 -> 21
RAS# to RAS# Delay - 6 -> 7 -> 8
REF Cycle Time - 70
WRITE Recovery Time - 8
Command Rate - 2N
.
.
.
All the rest is auto



CPU PLL Voltage - 1.52v
FSB termination Voltage - 1.22v
DRAM Voltage - 2.10v
NB Voltage - 1.45v
SB voltage - Auto
Clock Over-Charging Voltage - Auto
NB GTL Voltage Reference - 0.61x

Even with this settings, memtest86+ v4.20 is reporting errors in test #5.

Can anyone give a hand on getting this stable???

I'm at latest BIOS in my board...


Edit;

Changed values to the red ones...
I've been increasing values in red but i'm getting errors at memtest86+...

[PsySc0rpi0n]

Can't make'em stable at memtest86+ v4.20...

Can anyone give a help?

[Tradesman]

Is this 2 single sticks, or a 2x2GB set or 2 2x2GB sets (the model number is incomplete)

[PsySc0rpi0n]

It's a set of 2 * 2 Gb... F2-8500CL5D-4GBPK... Sorry... The title is corrected now...

[GSKILL TECH]

Try 5-6-6-20 2.20V

Thank you
GSKILL TECH

[PsySc0rpi0n]

Ok i will try when i get home today.

Let me know something

Is it true that the 4th value in memory timmings is the result of ading the first three, theorically??? I mean 5-5-5-15 => 5+5+5 = 15...

So, i've been trying things based on this assumption...

So i've not tried 5-6-6-20 but i've tried 5-6-6-17.

Well, i have to be honest. I have never been able to fuuly understand how timings work. I mean that in what, you people, base to change the 2nd, 3rd and 4th values?

I cansee people with settings like 5-7-6-10 or 5-5-8-12 (just random examples) and so but i can't figure outo how do you people get to these values, because if i change values like this i use to see around the net e never get the pretended results...

For instance, is it possible to have the 2nd and 3rd values lower the the 1st?

[Tradesman]

The old rule of thumb was 1+2+3 = 4 (5-5-5-15 4-4-4-12) and to a degree was set by DRAM standards laid down by JEDEC to try and make DRAM interoperable across platforms and chipsets. , but when OCing became more mainstream and people began wanting higher performance, the manufacturers and people began experimenting with the timings to push the envelope for better performance. In general the lower the numbers the better the performance, but at times you can come up with really odd combos, using one set of sticks you might have better performance at 8-8-8-24 while with another you might have a higher performance sweetspot at 8-9-8-24 or even 9-8-8-23 (theoretically). Basically it comes down to getting the system stable and from there (especially in your free time) one can experiment with the timings, try different things and benchmark as you go to find higher performance and throughput, often, it's not something you 'see' but will show up in benchmarks, OC enthusiast in particular take this route to get the 'ultimate' performance

[PsySc0rpi0n]

So, i've been trying to stabilize by kit at 1141Mhz with that setting from first post...

In memtest86+ v4.20 i get errors no matter if i have 5-5-5-15 or 5-6-6-17 or 5-7-7-19 or 5-8-8-21...

I can assume for previous tests with my other kit (a PC5300 Micron D9GMH) that if i increase CAS from 5 to 6 i will loose a lot of performance and to match the performance from CAS 5 i have to increase Mhz and if i increase Mhz i will get errors in memtset86+ v4.20 again.

So, this looks like a circle.

Today i will try 5-6-6-20 as GSKILL TECH suggested.
And about other timmings, should i leave them in auto or do you recommend to change any of them beyond 5-6-6-20?

[Tradesman]

I'd leave at auto to start, if stable could try changing CR from 2T to 1T for a bit of a boost

[PsySc0rpi0n]

Ok...

I asked because i've seen around the net that the tRFC in large amounts of memory should be raised to increase stability and this concept applies to to the tWR too...

I don't know if it makes any difference in 4Gb (2*2Gb sticks) too...

[Tradesman]

I was referring to the DRAM base timings and yes it's true, as the amount of DRAM goes up tRFC generally needs to go up also and other advanced timings can vary. Generally the BIOS will make those adjustments for you based upon the base timings that are loaded, and what it pulls from the SPD in conjuction with it's own tables of reference (the BIOS programming itself)

[GSKILL TECH]

tWR does not change, 8 is max for DDR2-1066+. 6 for DDR2-800.

Thank you
GSKILL TECH

[PsySc0rpi0n]

Ok Gskill Tech...

I'm testing it now at 5-6-6-20, 2.10v and tRFC at 50...

All rest is AUTO..


Edit;

Test #5 at MesmTest86+ v4.20 is the most stressing, right? I think i'm going to test my mems only at test #5 and when they are ok here i'll perform a total test...

Edit 2;

Errors were found with 5-6-6-20... Going to test with 5-6-6-22... What you say?

[PsySc0rpi0n]

Well, 5-6-6-6-20 didn't work and 5-6-6-22 didn't work too, neither 5-6-6-25 nor 5-6-6-27 nor 5-6-6-28...

What else you suggest?



Uploaded with ImageShack.us

[PsySc0rpi0n]

Well, after several tries i'm about to give up...

All the next timming settings came up with errors...

5-6-6-25 (already tried before)
5-6-7-25
5-6-8-25
5-6-9-25

5-7-6-25
5-7-7-25
5-7-8-25
5-7-9-25

So, what should i try next?



Edit;

Well, I've tried some more timming settings:

5-8-6-25
5-8-7-25
5-8-8-25
5-8-9-25

None of those worked out...

Please, help...