Memory Timing question for overclocking...

[CdnAtWork]

Hello all,

First off, here's my setup info:

Codegen 6066-CA case
ASUS A7V600 400FSB, SATA, Gbit LAN 6.1, Audio, 8xUSB, With Bios Rev 1006, VIA_Hyperion 4IN1_V451v, Vcore@1.7
AMD Athlon XP 2500+ Barton Core 333MHz 512KB L2 Cache (overclocked to 3200+)
Vantec Aeroflow Heatsink with Arctic Silver 5 compound (keeps me around 47 idle and 52 full load while overclocked)
2 x Infineon 256MB PC3200 400MHz DDR Memory
ATI Radeon 9200 128 DDR with wxp-w2k-catalyst-7-991-040224m-013831c Drivers
80GB Maxtor HD 8mb
LG CD/RW 24x10x40
OKIA 420w ATX P4 Power Supply

Current bios memory setup with available options in brackets:

Ram set to 400FSB
Config: Auto - By SPD
CAS Latency: 3 (2.5,2,1.5)
RAS to CAS Delay: 3 (5,4,2)
RAS Precharge Delay: 4 (5,3,2)
Active Precharge Delay: 8 (9,7,6)

Info on ram, incase this helps:
GPM400X64C3/256/HR
HYB25D256800BT-5

I'm totally lost when it comes to changing my memory timings. With having the cpu overclocked, should I manually change my timings or leave them at stock? I just want to make sure I have things optimised properly. I currently have the occasional reboot for no apparent reason and my temps are a bit higher than i'd like but i can live with them.

So any advice fellas? As usual, any input is appreciated.
Thanks!

 

[LostInSpace927]

Most people I have communicated with try for 2-2-2-5 or 2-2-2-6. The tighter the better(in most cases) If you feel you have reached a wall try loosening them up a bit. But if you set memory to manual and change the timings and it doesn't boot. My suggestion would be to add some voltage to the vdimm.

 

[CdnAtWork]

thanks for the advice. I tried going to 2-2-2-6 but i couldn't boot, even with upping the ram voltage to my max of 1.85,
so I've got it set back to manual again. Would there really be a noticible difference in performace with changing the timings? or is it more for stability?

Thank again.

 

[LostInSpace927]

You would really only see a performance boost in benchies. Just nonaseconds in real world terms. Its just better stablity. But if you already OC'd to 3200+ and nice load temps, then you've prolly reached the stabilty wall.

 

[Super56K]

max ram voltage of 1.85? Seems to me that you changed your cpu vcore and not your ram. Try looking again and setting your ram to 2.7 or 2.8v But honestly like LostinSpace927 said your not going to see any big difference from it other than benchmark scores.

 

[CdnAtWork]

your right superk, meant to put in 2.85. Guess i'll keep it at stock. I figured out why my temps were going so high. QFan Control was enabled and it was screwing things up. i disabled it and now i'm keeping steady at 44C

Thanks folks

 

[Soulkeeper]

you can run the memory 1:1 with fast timings 2-2-2-6 (via chipsets have great memory controllers)
or run it with it's spd timings at a ratio to make the memory 400mhz
you can also up the voltage on it if you have to

i like your setup
if you can post a few benchmarks
you could prob stomp half of nforce2 systems atleast

 

[0roo0roo]

lowest until stable. really doesnt matter much. tomshardware.com has a piece on memory timing. difference is so small, its not worth your time almost. esp if you value stability.

 

[CdnAtWork]

well, i just shut down all running processes and ran 3DMark03 and got a whopping 709...Perhaps someone can take a peak at the results below and give some advice on why i crapped out so bad, aside from the poor vidcard...

Result Details Here

 

[Algere]

Originally posted by: CdnAtWork
well, i just shut down all running processes and ran 3DMark03 and got a whopping 709...Perhaps someone can take a peak at the results below and give some advice on why i crapped out so bad, aside from the poor vidcard...

Result Details Here

3DMark03 is mostly video card dependant. So your not gonna score that well unless you test it with DirectX 9 hardware.

 

[Fricardo]

Yeah, Algere is right, the 3D marks are basically graphics card tests. Hence being "3D" Mark. A slow video accelerator will bottleneck your system on this type of application no matter how good your other components are.

 

[CdnAtWork]

makes sense...
So what should i use to benchmark it?