A little help needed with new build.

[suspekt]

First try the setting that is in color and that's in the layout Sharp made.. then if that doesn't work, try the things I wrote below, one step at the time

[en0tlaid]

and by steps do u mean each hard return as section ?

[suspekt]

Yes, the steps now have a * infront of them

 

• This is one step
  

[en0tlaid]

should i be running the same tests on each core ?

 

when i run prime it seems when i run the blend test on both cpus it seems to crash first on the prime that i start first.

 

im running large on one core and small on the other core and its already into test 5

and im using

settings from post 23

and

Raise tRWT from 2 to 3

Raise tWR from 2 to 3

[en0tlaid]

self test core 0 = 5 sets of 800000 Lucas-lehmer iterations using 8 k lenght =passed

self test core 1 = 5 sets of 4000 Lucas-lehmer iterations using 1024k lenght =passed

[en0tlaid]

core 0 still running second cycle of 560000 Lucas-lehmer iterations using 10k lenght core 1 crashed durring second cycle of

4500 Lucas-lehmer iterations using 896k lenght

[en0tlaid]

moving on to next bios setting

[suspekt]

ok

 

I think it is strange that it isn't stable.. I mean those timings provided in this thread look OK and you have qualified each stick at the time and also both pairs and now you are running all 4 of them...

 

And you have a good PSU also, quality brand and model, 700W. I must ask, you did connect all power plugs into the motherboard as Sharp suggested, right?

 

How about cooling, how is your processor cooled? Is the cooler firmly seated on the CPU?

 

Maybe all that's needed is a BIOS update.. either 704-2 BTA or the new 406-BT as wevsspot suggested.

 

But please continue with the test and even if all those settings fail I guess we can raise them more before we try to update the BIOS.. I guess it is just one or two settings that will turn the system from non-stable to completly stable... but at this point it is hard to say. (And I don't even have 4 sticks/4GB so I can test myself, just relying on theory here)

 

EDIT: In prime you can run either small FTU's/blend or what's the other one called? in place? (my computer is down so I'm not on my machine now so I can't check). At this point it doesn't really matter, just run the same test on both cores.

[en0tlaid]

16 reboots and 16 tests of prime later...... still get rounding errors everytime

 

and yes all power connectors are pluged in

 

processor is cooled with stock hsf. case is cooled with 5 led fans at 3000RPM

and yes its seated correctly and with Artic Silver

 

case temp is aprox 34 C

and processor temps have never gone over 40C (i have a cool drive 4 temprature monitor)

and vid card has never gone over 45C

[suspekt]

strange, really strange

 

have you applied the x2 patch for Windows? (don't think it has anything to do with it.. but just to rule that one out)

 

well you could try to use the following (all at the same time)

 

FSB BUS Frequency 195MHz

CPU/FSB Frequency Ratio 12

 

Cas latency (tCL) 3

tRCD 4

tRAS 10

tRC 16

tRFC 20

DRAM Bank Interleave Disabled

 

Max Async Latency 10ns

Read Preamble Time 8ns

 

After this if you are not stable I think you might reconsider upgrading the BIOS to a newer version and see if that helps.

[Praz]

I'm sure you're probably getting tied of trying all these settings but if you want to give it one more shot here are some more. Try these with just 2 sticks in the orange slots and if unsuccessful try the yellow slots. If you have to test with the yellow slots try with CPC enabled and disabled. Test these settings with your soundcard removed.

 

FSB Bus Frequency ......................... 200

LTD/FSB Frequency Ratio.................... AUTO

LTD Bus Transfer Width .................... AUTO

CPU/FSB Frequency Ratio ................... X 12.0

 

PCI eXpress Frequency ..................... 100Mhz

 

CPU VID StartUp Value...................... Auto

CPU VID Control............................ Auto

CPU VID Special Control ................... Auto

LDT Voltage Control ....................... 1.30 V

Chip Set Voltage Control .................. 1.60 V

DRAM Voltage Control ...................... 2.70 V

DRAM+ .03V If it's not 3.2 V............... Disable

 

DRAM Configuration

 

DRAM Frequency Set......................... 200 (DRAM/FSB:1/01)

Command Per Clock.......................... Enable

CAS Latency Control........................ 3.0

RAS# to CAS# delay ........................ 04

Min RAS# active time........................08

Row precharge time ........................ 04

Row Cycle Time ............................ Auto

Row Refresh cyc Time ...................... Auto

Row to Row delay .......................... Auto

Write Recovery Time........................ Auto

Write to Read Delay ....................... Auto

Read to Write Delay ....................... Auto

Refresh Period ........................... Auto

Write CAS Latency ......................... Auto

DRAM Bank Interleave ...................... Enabled

 

DQS Skew Control........................... Auto

DQS Skew Value ............................ 0

DRAM Drive Strength ....................... Auto

DRAM Data Drive Strength................... Auto

Max ASync Latency ......................... Auto

DRAM Response time ........................ Normal

Read Preamble Time......................... Auto

IdleCycle Limit ........................... Auto

Dynamic Counter ........................... Disable

R/W Queue Bypass .......................... 16X

Bypass Max ................................ 07X

32 Byte Granularity ....................... Disable (4 Bursts

[suspekt]

aah now I understand what can be the problem! why I havn't thought of this before...

 

Corsair have said that when they test their single PC4000 sticks they can do 250MHz in single channel, but if they test the same module in dual channel (two matching sticks), they might just be stable at ~245MHz or something, cause of the slight skew that occurs at the time of data transfer. They must then use better quality PC4000 modules if they are going to sell them as a "matching pair".

 

If your sticks are certified for dual channel, 2x1GB on both pairs, they may run 200MHz perfectly when they operate on their own, but if you run dual pairs, total 4x1GB they might be stable at something like 195MHz or even 190MHz. (Taking into account the slight skew and also the slight "overhead" while using dual channel mode)

 

So putting your memory on a divider might be the solution to the problem. Like the 9/10 divider or something.

 

Maybe I can find a link so you all can read about it.