Guide To Changing Memory Timings using Memset and Subsequent Performance Increase

As there have been a few questions about which memory someone should get; the fast one or the one with better timings, I've decided to put together this little info session/guide. Before we begin, I want to state that this is only an outline as to what performance change you can expect by changing your memory settings and that in no way am I responsible for any damage that you may cause to your hardware. I would also like to state that in my testing, I had many lockups caused by lowering settings too much. However, I have not experienced any hardware failure or data corruption. Now that the legal stuff is out of the way, onwards to performance!

Required software: Memset - used to change memory timings
Sisoft Sandra - for performance testing of memory latencies and memory bandwidth
CPUz - verification of changed memory settings (**OPTIONAL**)

In order to change your memory timings, you need to be running an Intel chipset. If when you start the program, you get missing entries, uncheck the box at the bottom and try again. If it fails, look around on the web for fixes since I don't know much more.

The changing of timings is incredibly simple, but time consuming. Try each setting one at a time, lowering it by a single count until it freezes (you will need to do a hard shutdown). Once this happens, change the setting to one higher than when it crashed, then run a latency test for stability. Do this until it doesn't freeze running the test.

I spent the better part of last night and this morning testing each individual memory setting to see how each setting affected both the memory bandwidth and memory latencies. Now, I'm not going to provide info as to how each individual setting affected performance as many settings changed performance within what I would believe is the margin of error. What I will do however is use a before and after system to show the change in performance. Included is a zip file (attached to post) with the detailed information of each of the three runs.

First, we begin with default settings of raw data

http://img41.imageshack.us/img41/7696/memsetdefaultsettings.jpg
http://img33.imageshack.us/img33/7210/defaultbandwidth.jpg
http://img43.imageshack.us/img43/8291/defaultlinearlatency.jpg
http://img42.imageshack.us/img42/277/defaultrandomlatency.jpg

Now, post tweaking the results are as follows

http://img40.imageshack.us/img40/9945/memsetoptimalsettings.jpg
http://img39.imageshack.us/img39/3254/optimalbandwidth.jpg
http://img38.imageshack.us/img38/8871/optimallinearlatency.jpg
http://img37.imageshack.us/img37/5804/optimalrandomlatency.jpg

As we can see, memory settings have only a slight effect on both bandwidth and latency.

Now we'll take a look at some "real world" tests to see how much of an impact this increased bandwidth and decreased latency makes. First up is Super Pi.

http://img36.imageshack.us/img36/1199/superpidefault.jpg
http://img33.imageshack.us/img33/6074/superpioptimal.jpg

Next up is wprime

http://img43.imageshack.us/img43/4898/wprimedefault.jpg
http://img42.imageshack.us/img42/835/wprimeoptimal.jpg

Last up is Passmark Performance Test 7.0 (only best results for both tests shown)

http://img39.imageshack.us/img39/144/defaultmemory.jpg
http://img38.imageshack.us/img38/7710/optimalmemory.jpg

So, from this testing, we can see that changing your memory timings does have an impact on your system's performance, albeit how much is up to the memory (yours may or may not be able to run the same timings as mine). This relates to a few percentage points of increased speed above stock.

I hope this info session has been useful, if you have any questions or comments post them below.

The files here have been named by the Memset timings with the test performed appended (ie. my default settings for the bandwidth test are named 5-5-5-15-7.8us3-35-11-13-11-5-7-5-6-6 bandwidth 1/2/3)
http://rapidshare.com/files/236849360/memory_performance.zip

 

Hey, thanks for guide. Will try soon. BTW what is your laptop, CPU etc...?

 

It's a Toshiba A200-TH7 with the RAM swapped with some OCZ sticks and the CPU upgraded to a T8300. I'm at work right now but once I get home (done in 2 1/4 hours + 1 hour drive) I'll do a CPUz/Sisoft output.

 

SiSoftware Sandra

System
Model : TOSHIBA Satellite A200
Workgroup : WORKGROUP
Host Name : Matt-PC
User : Matt

Processor
Model : Intel(R) Core(TM)2 Duo CPU T8300 @ 2.40GHz
Speed : 2.6GHz
Cores per Processor : 2 Unit(s)
Threads per Core : 1 Unit(s)
Type : Mobile, Dual-Core
Integrated Data Cache : 2x 32kB, Synchronous, Write-Thru, 8-way, 64 byte line size
L2 On-board Cache : 3MB, ECC, Synchronous, ATC, 12-way, 64 byte line size, 2 threads sharing

System
Platform Compliance : Intel Centrino Duo Mobile Technology
Mainboard : TOSHIBA ISKAA
BIOS : TOSHIBA V2.60 (08/28/2008)
Bus(es) : ISA PCI PCIe IMB PCMCIA CardBus USB FireWire/1394 i2c/SMBus
Multi-Processor (MP) Support : No
Multi-Processor Advanced PIC (APIC) : Yes
Total Memory : 2GB DDR2 SO-DIMM

Chipset
Model : Toshiba Mobile PM965/GM965/GL960 Express Processor to DRAM Controller
Front Side Bus Speed : 4x 200MHz (800MHz)
Total Memory : 2GB DDR2 SO-DIMM
Memory Bus Speed : 2x 333MHz (666MHz)

Memory Module(s)
Memory Module : OCZ OCZ26671024VSO 1GB DDR2 SO-DIMM PC2-5300U DDR2-666 (5-5-5-15 3-20-5-3)
Memory Module : OCZ OCZ26671024VSO 1GB DDR2 SO-DIMM PC2-5300U DDR2-666 (5-5-5-15 3-20-5-3)

Video System
Adapter : ATI Mobility Radeon HD 2600 (120 SM4.0, 256MB DDR2 Integrated Graphics)

Storage Devices
WDC WD3200BEVT-22ZCT0 (320GB, SATA300, 2.5", 5400rpm, NCQ, 8MB Cache) : 298GB (C (D
TSSTcorp CD/DVDW TS-L632D (ATA33, DVD+-RW, CD-RW, 2MB Cache) : N/A (E

Logical Storage Devices
Windows (C : 122GB (NTFS) @ WDC WD3200BEVT-22ZCT0 (320GB, SATA300, 2.5", 5400rpm, NCQ, 8MB Cache)
Shared (D : 122GB (NTFS) @ WDC WD3200BEVT-22ZCT0 (320GB, SATA300, 2.5", 5400rpm, NCQ, 8MB Cache)
Optical Drive (E : N/A @ TSSTcorp CD/DVDW TS-L632D (ATA33, DVD+-RW, CD-RW, 2MB Cache)

Peripherals
LPC Hub Controller 1 : Toshiba ICH8M-E (ICH8) LPC Interface Controller
Audio Device : Toshiba 82801H (ICH8) HD Audio Controller
Audio Codec : Realtek Semiconductor 0268h
Audio Codec : Lucent/Agere Systems (Was: AT&T MicroElectronics) HDA softmodem
Audio Device : ATI (AMD) HD Audio Codec
Serial Port(s) : 3
Disk Controller : Toshiba 82801H (ICH8) Ultra ATA Storage Controllers
Disk Controller : Toshiba Mobile SATA AHCI Controller
Disk Controller : Intel Flash Cache Logic Chip
USB Controller 1 : Toshiba 82801H (ICH8) USB UHCI #4
USB Controller 2 : Toshiba 82801H (ICH8) USB UHCI #5
USB Controller 3 : Toshiba 82801H (ICH8) USB2 EHCI #2
USB Controller 4 : Toshiba 82801H (ICH8) USB UHCI #1
USB Controller 5 : Toshiba 82801H (ICH8) USB UHCI #2
USB Controller 6 : Toshiba 82801H (ICH8) USB UHCI #3
USB Controller 7 : Toshiba 82801H (ICH8) USB2 EHCI #1
FireWire/1394 Controller 1 : Toshiba PCIxx12 OHCI IEEE 1394 Host Controller
CardBus/PCMCIA Controller 1 : TI PCIxx12 Cardbus Controller
SMBus/i2c Controller 1 : Intel 801xx/63xx SMBus

Printers and Faxes
Printer : Microsoft XPS Document Writer (600x600, Colour)
Printer : HP Officejet 5600 series (1200x1200, USB, Colour)
Fax : Microsoft Shared Fax Driver (200x200)

Network Services
Network Adapter : Realtek RTL8101E Family PCI-E Fast Ethernet NIC (NDIS 6.0)
Network Adapter : Bluetooth Personal Area Network
Wireless Adapter 1 : Intel(R) Wireless WiFi Link 4965AGN (802.11g (ERP), AES-CCMP, 54Mbps)

Power Management
Mains (AC) Line Status : On-Line

Operating System
Windows System : Microsoft Windows 7 Ultimate 6.01.7100
Platform Compliance : x86

Windows Experience Index
Current System : 4.7