Command Rate for DDR2

[chevmaro]

I currently just built a new system. I have some crucial ballistix DDR2. Im just wondering why everyone is saying to run DDR2 in 2T vs. 1T. I know when I had my last computer with DDR it was always 1T. So are any of you guys running your DDR2 in 1T?

 

[MadScientist]

1T for 2 sticks, 2T for 4 sticks

 

[LOUISSSSS]

i'm using 2 sticks and CPU-Z says i'm running 2T. i'm using a P35 chipset and have heard that this chipset doesn't support 1T...

 

[JustaGeek]

Originally posted by: MadScientist
1T for 2 sticks, 2T for 4 sticks

I would agree with the above on the AMD platform.

There is a very little difference between 1T and 2T on the Intel platform, and 1T might cause instability, even with 2 sticks.

You might try running RAM at 1T, but I would generally recommend 2T for Intel, regardless of number of memory modules.

 

[BonzaiDuck]

JustaGeek --

I assume, for clarity, that you mean "chipset" when you say " . . . recommend 2T for Intel . . . ?"

 

[JustaGeek]

Originally posted by: BonzaiDuck
JustaGeek --

I assume, for clarity, that you mean "chipset" when you say " . . . recommend 2T for Intel . . . ?"

I meant "Intel platform", regardless of the chipset used.

My NVidia 650i chipset would default to 1T on the "first boot" with the 2x1GB G.Skill HZ memory, since the BIOS would read the EPP.

I promptly set it to 2T, since I did not want to experiment with 1T.

It doesn't happen with the 3GB of RAM, since the 2x512MB modules do not have EPP.

 

[BonzaiDuck]

I haven't had any disasters with 1T command-rate, but I want to be a sensible guy.

I found the VDIMM requirements for several configurations with 2T settings, then re-ran stress-testing to determine how much to kick up the voltage to be stable with 1T.

It seems to be about one notch among the BIOS VDIMM choices.

So I simply set the VDIMM at the higher value, so I can flip back and forth between 1T and 2T as my whim or need directs me.

 

[JustaGeek]

Stability has always been more important to me than a little gain in performance.

That's why I keep my "highly overclockable" RAM at 800MHz, and the command rate at 2T.

I let others test it to the "max". 😉

 

[BonzaiDuck]

Well, it really makes the memory do a lot of work, but the VDIMM for the 4:5 settings is either 2.05 or 2.125V (2.88 Ghz v 3.15 Ghz), with monitored value showing a bias against the "set" value of +0.35V at all settings. The VDIMM for 3.0 Ghz must be 2.15V for 1:1, {3,3,3,8,1T,tRC=11}. I can run them at lower power, and still choose the command rate without needing to change anything else.

The memory-cache bandwidth benches show Everest "Read" of fully 500 MB/s below the 4:5/3.15-Ghz setting.

If I loosen the latencies to 3,4,4,9 [1T or 2T], I can probably keep the voltage at the same level. Or I could test the voltage against looser timings and higher FSB, but sooner or later I would run up against a need to loosen more, make CMD=2T, or increase voltage.

But given certain benchmarks shown in the print and internet media for DDR2 and DDR3 modules, I'd be interested in the complete configuration details for anyone whose Everest Read result is >= 10,000 MB/s. If TRUE for DDR3, I haven't any need to go farther with this.

Knowing tht there has for a long, long time been things like "DMA/Direct_Memory_Access," VGA use of system RAM option, and other data paths independent of CPU, it may be that usage is another important determinant of "performance." I'll throw that idea into the fire . . . but we've already heard it.

 

[LOUISSSSS]

what? how are you guys doing 1t with a p35 chipset? whats it called in your bios?

 

[MadScientist]

Louisssss,
Some P35 MBs, i.e., the Asus P5K Deuxe, with a bios upgrade, have the Command Rate option of running at Auto, 1T, or 2T. If you have it on your Gigabyte MB it should be called something like Command Rate or Command per Clock.