Please, some tuning assistance

R

Renee

n00b
Joined
Jan 14, 2007
Messages
6
Hi,
I was wondering if I could get some tuning help. I’ve been in computing for about thirty five years and have a lot of software and hardware experience. At the same time I don’t pretend to know all the details of this. I have a brand new very complex system and it hangs or crashes about once a day. The system will run complete superpi without errors. It will pass one pass of the Vista extreme memory tests which is quite a battery. I let it run two passes over night and it hung on the second pass so the system is still not stable.

OS: Vista Ultimate 32 RTM

My hardware configuration looks like this:

CPU: Core-2 Duo Extreme x6800 – currently not overclocked – Aircooled Temp 44 degrees C at idle according to MBM5 Voltages are Asus Defaults -

Cooling: CoolerMaster Hyper 48 - (KHC-L91-U2)

Memory: 2 Gigabytes Kingston HyperX KHX8500D2K2/2g run at 1066 Mhz 2.35 volts
The board has set the latencies to be higher than the ratings, I have done very little tuning

Mother board: Asus P5WDG2 WS Professional Bios Firmware Rev 604
CPU Spread Spectrum Turned off

PSU: Antec Neo HE 550 Watts

Video Card EVGA 7950 GT OK

SCSI Adapter – Adaptec 39320 in PCIX slot run at 133 MHZ (not using raid)
Disks (2) Seagate Scsi Cheetah 15K RPM 140 GB disks in Antec coolers

Audio – Soundblaster Live PCI128
DVDs (2) Plextor 755-SA (Sata)

Turning CPU Spread spectrum off made a huge difference in stability. When it was on, superpi had rounding errors in the first iteration. Turnng it off superpi will complete out to 32 million places routinely without error. Superpi in no way saturates the two cores when it runs.

CPU-Z Report

CPU-Z version 1.35.
CPU(s)
Number of CPUs 2

CPU#1 APIC ID = 0
Name Intel Core 2 Extreme X6800
Code Name Conroe
Specification Intel(R) Core(TM)2 CPU X6800 @ 2.93GHz
Family / Model / Stepping 6 F 6
Extended Family / Model 6 F
Package LGA 775
Core Stepping B2
Technology 65 nm
Supported Instructions Sets MMX, SSE, SSE2, SSE3, SSE4, EM64T
CPU Clock Speed 1602.7 MHz
Clock multiplier x 6.0
Front Side Bus Frequency 267.1 MHz
Bus Speed 1068.5 MHz
L1 Data Cache 32 KBytes, 8-way set associative, 64 Bytes line size
L1 Instruction Cache 32 KBytes, 8-way set associative, 64 Bytes line size
L2 Cache 4096 KBytes, 16-way set associative, 64 Bytes line size
L2 Speed 1602.7 MHz (Full)
L2 Location On Chip
L2 Data Prefetch Logic yes
L2 Bus Width 256 bits

CPU#2 APIC ID = 1
Name Intel Core 2 Extreme X6800
Code Name Conroe
Specification Intel(R) Core(TM)2 CPU X6800 @ 2.93GHz
Family / Model / Stepping 6 F 6
Extended Family / Model 6 F
Package LGA 775
Core Stepping B2
Technology 65 nm
Supported Instructions Sets MMX, SSE, SSE2, SSE3, SSE4, EM64T
CPU Clock Speed 1602.7 MHz
Clock multiplier x 6.0
Front Side Bus Frequency 267.1 MHz
Bus Speed 1068.5 MHz
L1 Data Cache 32 KBytes, 8-way set associative, 64 Bytes line size
L1 Instruction Cache 32 KBytes, 8-way set associative, 64 Bytes line size
L2 Cache 4096 KBytes, 16-way set associative, 64 Bytes line size
L2 Speed 1602.7 MHz (Full)
L2 Location On Chip
L2 Data Prefetch Logic yes
L2 Bus Width 256 bits



Mainboard and chipset
Motherboard manufacturer ASUSTeK Computer INC.
Motherboard model P5WDG2 WS PRO, Rev 1.xx
BIOS vendor American Megatrends Inc.
BIOS revision 0604
BIOS release date 11/30/2006
Chipset Intel i975X rev. C0
Southbridge Intel 82801GB (ICH7) rev. 01
Sensor chip Winbond C1

Graphic Interface PCI-Express
PCI-E Link Width x16
PCI-E Max Link Width x16



Memory
DRAM Type DDR2-SDRAM
DRAM Size 2048 MBytes
DRAM Frequency 534.2 MHz
FSB:DRAM 1:2
CAS# Latency 5.0 clocks
RAS# to CAS# 6 clocks
RAS# Precharge 6 clocks
Cycle Time (TRAS) 18 clocks
Bank Cycle Time (TRC) 21 clocks
# of memory modules 2
Module 0 Kingston DDR2 PC6400 - 1024 MBytes
Module 1 Kingston DDR2 PC6400 - 1024 MBytes



Software
Windows version Microsoft Windows Vista (6.0) Ultimate Edition (Build 6000)
DirectX version 10.0


*** The cores were at idling when this snapshot was running so the clocks are the idle clock speeds.


I’m not interested in any extreme overclocking. My goal would be to run the cpu overclocked at about 10% optimized for stability and durability. I’m not a gamer. I’d really appreciate some help with the stability and tuning issues.

Thank you.
 
If your running everything at stock speeds I would be inclined to think that it may be an issue with Vista. Drivers and whatnot are still not optimized for the new os. See if your system is logging the error and do a search on google for it.
 
The reason I don't think that's true is that the Vista memory tests do a reboot and load very little and run in VGA mode. I'm running stock drivers and the video drivers are the latest from Nvidia. I have two friends running Vista Ultimate RTM and for them it's very stable. The hangs appears to be random.

But I will take your advice and trace through the event logs. The thing is they are going to be full of flukes. The system really was flaky before I turned spread spectrum off of the CPU last night. I was amazed at how reliably, unreliable superpi was. Turning spread spectrum off made a day and night difference and somehow, I think thats a large clue.
 
I just received the following results from prime95

Intel(R) Core(TM)2 CPU X6800 @ 2.93GHz
CPU speed: 2937.90 MHz
CPU features: RDTSC, CMOV, Prefetch, MMX, SSE, SSE2
L1 cache size: 32 KB
L2 cache size: unknown
L1 cache line size: 64 bytes
L2 cache line size: unknown
Prime95 32-bit version 24.14, RdtscTiming=1
Best time for 512K FFT length: 8.652 ms.
Best time for 640K FFT length: 11.731 ms.
Best time for 768K FFT length: 14.426 ms.
Best time for 896K FFT length: 17.254 ms.
Best time for 1024K FFT length: 19.100 ms.
Best time for 1280K FFT length: 24.303 ms.
Best time for 1536K FFT length: 29.755 ms.
Best time for 1792K FFT length: 35.431 ms.
Best time for 2048K FFT length: 39.467 ms.
Best time for 2560K FFT length: 51.989 ms.
Best time for 3072K FFT length: 63.560 ms.
Best time for 3584K FFT length: 76.582 ms.
Best time for 4096K FFT length: 85.846 ms.
Best time for 58 bit trial factors: 3.789 ms.
Best time for 59 bit trial factors: 3.804 ms.
Best time for 60 bit trial factors: 3.813 ms.
Best time for 61 bit trial factors: 3.822 ms.
Best time for 62 bit trial factors: 6.073 ms.
Best time for 63 bit trial factors: 6.103 ms.
Best time for 64 bit trial factors: 5.830 ms.
Best time for 65 bit trial factors: 5.825 ms.
Best time for 66 bit trial factors: 5.820 ms.
Best time for 67 bit trial factors: 5.775 ms.
FATAL ERROR: Rounding was 0.5, expected less than 0.4
Hardware failure detected, consult stress.txt file.
 
I just backed my memory speed down to 889 MHZ. Prime95 stress/torture ran without error until I stopped it after 8 iterations. It's quite the test and I don't understand exactly what occurred. I watched physical memory and virtual memory be absolutely consumed.

All performance monitors sugested that there was plenty of available cpu headroom left but if so, task scheduling was really poor and the system didn't act as if there was the available cpu that all monitors suggested.

I'm confused about several statements made by kingston. This memory is rated and tested at 1066 mhz according to them but the memory is overclocked at 1066 Mhz. According to Kingston:

"The memory timings (CAS latency, etc) is lower (faster) than is normal for this type of memory. Low latency, as it is called, is a form of overclocking. And there is no DDR2 memory out today rated for 1066Mhz yet. The DRAM in this memory is actually overclocked 1066Mhz. But it has been tested for 1066Mhz. So you may want to adjust the latency timings to ones that will run stable in your system. According to ASUS, the highest speed memory this board is rated for is 800Mhz. "

I always seen Asus say the this board supports 1066 MhZ memory. Kingston says it doesn't. I really paid a lot for this memory and I'm running it what I believe is below rated speeds. Is what I am seeing to be expected? Is the 975x chipset really supposed to run at 800 MHZ or is there something else about the board.
 
Run Orthos Blend. If you get the same rounding error, keep working with your RAM timings.
Orthos is Prime95 for dual core CPU's.
 
Using CPU #1
Beginning a continuous self-test to check your computer.
Press Stop to end this test.
Test 1, 4000 Lucas-Lehmer iterations of M19922945 using 1024K FFT length.
Test 2, 4000 Lucas-Lehmer iterations of M19922943 using 1024K FFT length.
Test 3, 4000 Lucas-Lehmer iterations of M19374367 using 1024K FFT length.
Test 4, 4000 Lucas-Lehmer iterations of M19174369 using 1024K FFT length.
Test 5, 4000 Lucas-Lehmer iterations of M18874369 using 1024K FFT length.
Test 6, 4000 Lucas-Lehmer iterations of M18874367 using 1024K FFT length.
Test 7, 4000 Lucas-Lehmer iterations of M18474367 using 1024K FFT length.
Test 8, 4000 Lucas-Lehmer iterations of M18274367 using 1024K FFT length.
Self-test 1024K passed!
Test 1, 800000 Lucas-Lehmer iterations of M172031 using 8K FFT length.
Test 2, 800000 Lucas-Lehmer iterations of M163839 using 8K FFT length.
Test 3, 800000 Lucas-Lehmer iterations of M159745 using 8K FFT length.
Test 4, 800000 Lucas-Lehmer iterations of M157695 using 8K FFT length.
Test 5, 800000 Lucas-Lehmer iterations of M155649 using 8K FFT length.
Test 6, 800000 Lucas-Lehmer iterations of M153599 using 8K FFT length.
Test 7, 800000 Lucas-Lehmer iterations of M147455 using 8K FFT length.
Test 8, 800000 Lucas-Lehmer iterations of M143361 using 8K FFT length.
Test 9, 800000 Lucas-Lehmer iterations of M141311 using 8K FFT length.
Self-test 8K passed!
Test 1, 560000 Lucas-Lehmer iterations of M212991 using 10K FFT length.
Test 2, 560000 Lucas-Lehmer iterations of M210415 using 10K FFT length.
Test 3, 560000 Lucas-Lehmer iterations of M208897 using 10K FFT length.
Test 4, 560000 Lucas-Lehmer iterations of M204799 using 10K FFT length.
Test 5, 560000 Lucas-Lehmer iterations of M200705 using 10K FFT length.
Test 6, 560000 Lucas-Lehmer iterations of M196607 using 10K FFT length.
Test 7, 560000 Lucas-Lehmer iterations of M194561 using 10K FFT length.
Test 8, 560000 Lucas-Lehmer iterations of M188415 using 10K FFT length.
Test 9, 560000 Lucas-Lehmer iterations of M186369 using 10K FFT length.
Test 10, 560000 Lucas-Lehmer iterations of M184319 using 10K FFT length.
Self-test 10K passed!
Test 1, 4500 Lucas-Lehmer iterations of M17432577 using 896K FFT length.
Test 2, 4500 Lucas-Lehmer iterations of M17432575 using 896K FFT length.
Test 3, 4500 Lucas-Lehmer iterations of M17115073 using 896K FFT length.
Test 4, 4500 Lucas-Lehmer iterations of M16815071 using 896K FFT length.
Test 5, 4500 Lucas-Lehmer iterations of M16515073 using 896K FFT length.
Test 6, 4500 Lucas-Lehmer iterations of M16515071 using 896K FFT length.
Test 7, 4500 Lucas-Lehmer iterations of M16297569 using 896K FFT length.
Test 8, 4500 Lucas-Lehmer iterations of M15997567 using 896K FFT length.
Test 9, 4500 Lucas-Lehmer iterations of M15597569 using 896K FFT length.
Self-test 896K passed!
Test 1, 5300 Lucas-Lehmer iterations of M14942209 using 768K FFT length.
Test 2, 5300 Lucas-Lehmer iterations of M14942207 using 768K FFT length.
Test 3, 5300 Lucas-Lehmer iterations of M14155777 using 768K FFT length.


I'm runing Ortho Blends and both processors really are super saturated and it runs at a low priority which allows me to do other things.

With memory set to 889 Mhz this appears to be as stable as it can possibly be. MBM5 reports a core temperature of 63 - 68 degrees C under full sustained load/stress.

Operation could not be more stable. These were results after 66 minutes. I recognize that longer testing is needed to fully validate the system but it seems to be working fine. I'm continuing to run ortho without interrupting it.

Where does one begin in memory tuning?
 
You must log in or register to reply here.
Back
Top