My first overclock, seems to work, advice please

Associate
Joined
30 May 2007
Posts
96
Hi all. I've followed various guides as to how to overclock and I THINK I've done OK. I'm a little confused as to whether my memory settings are correct though (got a bit confused with memory multiplier :confused: ).

My computer spec is
GeIL 4GB (2x2GB) PC6400C5 800MHz Black Dragon DDR2
Intel Core 2 Duo E6600 LGA775 Conroe 2.40Ghz
Gigabyte GA_P35_DS4 Motherboard
EVGA GeForce 8800 GTX KO ACS3 Silent 768MB
Thermaltake VA9003SWA Kandalf Aluminium Full Tower
Tagan TG700-U25 700W ATX2.0 Dual Engine silent sli
Samsung Spinpoint HD501LJ 500GB SATA
Auzentech X-Meridian 7.1 Soundcard
Optiarc NEC AD7170S Serial ATA DVD
Dell Ultrasharp 2407WFP 24” Monitor

and Geil's website says the following about my RAM

4GB 800MHz CAS5 (2GBX2-SKU#GB24GB6400C5DC)
64Mx8 or 128Mx8 DDR2 FBGA Chips
240pin, Non-ECC, Un-buffered DIMM
8 Layers Custom Black Dragon PCB with LED
Retail Package
Working Voltage:1.9V - 2.4V
Lifetime Warranty

I've overclocked the CPU to 3Ghz and I've manually set the RAM timings to 4-4-4-12 (instead of the recommended, 5-5-5-15)

Everything appears to be running fine, have run Orthos for 30 mins (blend) and with 100% CPU usage in both cores, Everest is showing:

CPU 53C (didn't see it go above 55C)
Core1 53C
Core2 54C
Motherboard 49C

As I'm typing this (system almost idle) Everest is showing

CPU 30C
Core1 29C
Core2 31C
Motherboard 46C

It's the memory speed/voltage I'm not sure about. Also I've read that I need to set my PCI Frequency to 33.33 and cant see any way of doing so? Here's my BIOS settings. Do they look OK?

MB Intelligent Tweaker (M.I.T)
Robust Graphics Booster [AUTO] (options are AUTO, Fast, Turbo)
CPU Clock Ratio [9x]
CPU Host Clock Control [Enabled]
CPU Host Frequency Mhz [333]
PCI Express Frequency Mhz [100]
C.I.A.2 [Disabled]
Performance enhance [Standard]

System Memory Multiplier (SPD) [2.00]
(That gives a memory frequency of 800 667, Originally the multiplier was set to AUTO and gave a memory freq of 800 1000)
High speed DRAM DLL setting [Option 1]
DRAM Timing Selectable (SPD) [Manual]
CAS Latancey Time 6 [4]
DRAM RAS# to CAS# Delay 6 [4]
DRAM RAS# Precharge 6 [4]
Precharge Delay (tRAS) 18 [12]
ACT to ACT Delay (tRRD) 3 [AUTO]
Rank write to read delay 4 [AUTO]
Write to Precharge Delay 6 [AUTO]
Refresh to ACT Delay 52 [0]
Read to precharge Delay 3 [Auto]
trd 7 [AUTO]
trd Phase Adjustment 0 [AUTO]

System Voltage Control [MANUAL]
DDR2 OverVoltage Control [NORMAL]
PCI-E OverVoltage Control [NORMAL]
FSB OverVoltage Control [NORMAL]
(G) MCH OverVoltage Control [NORMAL]
CPU Voltage Control [NORMAL]

ADVANCED BIOS SETTINGS
HDD S.M.A.R.T Capability [Disabled]
CPU Hyperthreading [Enabled]
Limit CPUID Max to 3 [DISABLED]
No Execute memory Protect [DISABLED]
CPU Enhanced Halt (C1E) [DISABLED]
CPU Thermal Monitor 2 (TM2) [DISABLED]
CPU EIST Function [DISABLED]
Virtulization Technonogy [DISABLED]


PC HEALTH STATUS
CPU Warning Temp 70C/158F
CPU Fan Fail Warning [Enabled]
System FAN1 Fail Warning [Enabled]
Power Fan1 Warning [Enabled]
Smart Fan Control [DISABLED]

Many thanks

Phil
 
Last edited:
Re: memory / voltage.

Memory multiplier of 2.0 is good. The Gigabyte BIOS is a bit confusing, but 2. means 1:1 divider, which I personally find optimum. You can check the resulting divider in CPU-Z under the memory tab.

You'll find the actual rated memory of the Geil on the holo on the stick itself. I think that board defaults to 1.8v (please check), so you will need to manually add the difference in voltage in BIOS. Eg my Geil is rated at 2.1v, so I have put +.3 in BIOS.

Hope this helps.
 
Many thanks. I'll keep it at 3mhz for a while, if it remains stable, I'll try a little more.

Had a look on one of my sticks and it says

2GB PC2-6400
DDR2-800 CL=5-5-5-15
GB24GB6400C5DC
Voltage 1.8v

So I presume my voltage is correct as it is? (MB defaults to 1.8v)

CPU-z Memory says
Type DDR2
Channels# Dual
Size 4096MBytes

Frequency 333.0Mhz
FSB: DRAM 1:1
CAS# Latency (tCL) 4.0 clocks
Ras# to CAS# Delay (tRCD) 4 clocks
RAS# Precharge (tRP) 4 clocks
Cycle Tine (tRAS) 12 clocks

Command Rate (CR) 2T

CPU=Z SPD for both slot 1 and slot 3 says
Module Size 2048MBytes
Max Bandwidth PC2-6400 (400Mhz)

Timings
Frequency: 333 Mhz
CAS# Latency: 4.0
RAS# to CAS#: 5
RAS# Precharge: 5
tRAS: 15
tRC: 20
Voltage 1.8V

Frequency: 400 Mhz
CAS# Latency: 5.0
RAS# to CAS#: 6
RAS# Precharge: 6
tRAS: 18
tRC: 24
Voltage 1.8V
 
Last edited:
It looks to me as if you have plenty of headroom for a bigger clock in your own time.

Load temps are good (50s), excellent if that's with a stock cooler - you don't say. You haven't even touched your voltages yet, and you have yet to loosen your RAM timings to their rated levels.

Get going mate - what's stopping you?!
 
Yep it's the stock cooler. Mind you my case has a 250mm intake fan on the side, a 120mm exhaust and a 90mm exhaust on the back, a 90mm exhaust on the top and a 120mm intake on the front. That probably helps quite a lot.

Going to have to look up info on memory and loosing their timings. I presume that having it set at 4-4-4-12 is better than 5-5-5-15?

Many thanks

Phil
 
Going to have to look up info on memory and loosing their timings. I presume that having it set at 4-4-4-12 is better than 5-5-5-15?

Many thanks

Phil

It is. But the higher you clock, at some point you will need to loosen the timing (increase the bandwidth). Personally, I don't care to faff about with it manually, so I always set the timings to auto, and let the mobo decide.

edit... Actually, I may decide to faff about with it, now I started thinking about it :)
 
OK. I've just spent a merry hour benchmarking. I have a similar setup to you in as much as I am running an FSB of 333Mhz at 3Ghz and using PC6400 DDR2 RAM. So I thought, what the heck, lets benchmark the system with RAM timings at the auto (5-5-5-15), or manually changing them back to 4-4-4-12 (ie faffing about).

Here are the results:

Sandra 12 - Memory bandwidth - 5-5-5-15

memorytest55515ju7.jpg


Sandra 12 - Memory bandwidth - 4-4-4-12

memorytest44412jj4.jpg


Sandra 12 - CPU - 5-5-5-15

cputest55515gh5.jpg


Sandra 12 - CPU - 4-4-4-12

cputest44412ee4.jpg


See next post for the rest.
 
benchmark continued...

Sandra 12 - Cache and Memory - 5-5-5-15

cachememory55515bd9.jpg


Sandra 12 - Cache and Memory - 4-4-4-12

cachememory44412ls8.jpg


SuperPi 1.5 mod - 1Mb run - 5-5-5-15:

superpi55515vs8.jpg


SuperPi 1.5 mod - 1Mb run - 4-4-4-12:

superpi44412cj6.jpg


So by my reckoning, that's between a 1% and 2% performance gain. Not bad for a bit of faffing around. That said, coming from the school of you don't get something for nothing, I am not sure what price is being paid - anyone help?
 
Last edited:
Back
Top Bottom