Memory timing curiosity

[WW2Planes1]

First, my system specs:

AMD Athlon64 3000 NewCastle Skt754 (CG Stepping, 2GHz base freq)
2x512MB Crucial/Micron Single Sided (rated 3-3-3-8 timings) link
Asus K8N-E Deluxe rev 1.03 bios 1006 (mem voltage limited to 2.7V)
PNY GeForce4 Ti4200 64MB 8x
SATA hard drives

Second, some background:
As a prelude to possibly doing some slight overclocking, I decided to see what memory timings my memory could handle at stock speed. Originally, I would have liked to see some 2-2-2 timings (even though it is value ram), but in the end I could only get 2.5-2-2. However, as the title says, I found that I could get the 2-2-2 timings if I ran my processor with a 4x multiplier, instead of the 10x default. Everything else was running at full speed.

800MHz HT Bus (200*4)
DDR400 speed (w/ 2-2-2 timings)
but CPU at 800MHZ (4x mult)

if I tried upping the multiplier to 10x, I would get a System Failed due to CPU Overclocking on POST.

currently, I'm running stable with
800MHz HT Bus (200*4)
DDR400 speed (w/ 2.5-2-2 timings)
CPU at 2000MHZ (10x mult)

so, why does the CPU speed make the difference?


EDIT:
Vcore is stock
Vdimm is 2.6V (i don't remember what the rated voltage is... either 2.5 or 2.6).
in truth, i wasn't expecting the memory to run 2-2-2, i just found it weird that it would be dependent on the CPU frequency
power supply is a 450W Antec SurePower
I've got 4 Hard drives in the system now, and 2 optical drives. I don't think i'm that close to overloading the power supply.
And i didn't try 2-2-2 at a 2T command rate. I figured that 2.5-2-2 at 1T is better than 2-2-2 at 2T
/EDIT

 

[imported_Kensai]

Maybe your system just can't take it?

 

[jpeyton]

The problems with an on-die memory controller.

 

[WW2Planes1]

Originally posted by: jpeyton
The problems with an on-die memory controller.

On-die memory controler and my memory, you mean? because the chip can handle 2-2-2 timings with other memory...

 

[yliu]

No i don't think it's the memory controller. I think there are two possibilities here:

1. Although your system boots with your memory at 2-2-2 and CPU at 800mhz, it's still not really stable, and running it at 200x10 just exposed the problem sooner.

2. Tighter timings cause memory to draw more power; it's possible that your PSU can't deliver clean power for both the memory at tight timings and your CPU at 2ghz.

This is assuming that you never changed the voltage of vcore or vdimm.

 

[Erssa]

I wouldn't say it's because of the PSU. I haven't seen any reviews where that memory could have done 2-2-2, so I find this pretty strange too.

 

[Amaroque]

Originally posted by: yliu
No i don't think it's the memory controller. I think there are two possibilities here:

1. Although your system boots with your memory at 2-2-2 and CPU at 800mhz, it's still not really stable, and running it at 200x10 just exposed the problem sooner.

2. Tighter timings cause memory to draw more power; it's possible that your PSU can't deliver clean power for both the memory at tight timings and your CPU at 2ghz.

This is assuming that you never changed the voltage of vcore or vdimm.

I'd have to agree here. I had an old PII that would run at 133 MHz with PC 100. No errors in any memory tests at all. When I dropped a much faster 133 FSB slot 1 CPU in there, my RAM was no longer stable at 133 MHz with any timings. I also had an oversized (for the time) PSU in there, so I took the PSU out of the equation. The new CPU was meant for 133 MHz and was not OC'd.

This is what I was left with... The RAM was always marginal at 133 MHz. By putting a faster CPU in, I was pushing more data through the RAM (or pushing the RAM harder) at any given time. Thus exposing any weakness that had not been prevalent at the time, before a faster CPU was put in.

 

[WW2Planes1]

Originally posted by: Amaroque

Originally posted by: yliu
No i don't think it's the memory controller. I think there are two possibilities here:

1. Although your system boots with your memory at 2-2-2 and CPU at 800mhz, it's still not really stable, and running it at 200x10 just exposed the problem sooner.

2. Tighter timings cause memory to draw more power; it's possible that your PSU can't deliver clean power for both the memory at tight timings and your CPU at 2ghz.

This is assuming that you never changed the voltage of vcore or vdimm.

I'd have to agree here. I had an old PII that would run at 133 MHz with PC 100. No errors in any memory tests at all. When I dropped a much faster 133 FSB slot 1 CPU in there, my RAM was no longer stable at 133 MHz with any timings. I also had an oversized (for the time) PSU in there, so I took the PSU out of the equation. The new CPU was meant for 133 MHz and was not OC'd.

This is what I was left with... The RAM was always marginal at 133 MHz. By putting a faster CPU in, I was pushing more data through the RAM (or pushing the RAM harder) at any given time. Thus exposing any weakness that had not been prevalent at the time, before a faster CPU was put in.

That makes a lot of sense (and after thinking about it, that's pretty much what I guessed the problem would be). I didn't do any stress testing of the RAM at 800MHz, so it probably was marginal and the faster CPU speed just pushed it over the edge. Weird though, that I couldn't even get past POST at 2GHz, but could get all the way into windows at 800MHz (on the CPU)

also, added PSU/Voltage info to OP