A G.SKILL "True Believer:" Looking for insights per Command Rate=1

[BonzaiDuck]

I've used all sorts of memory kits -- OCZ, Crucial, Corsair. Started using G.SKILL kits during the DDR2 days, even while running some Corsair DDR2's. I dumped Crucial, but that's a story many here will have in common. I think -- during 2007/2008 -- Crucial had an "RMA" crisis . . .

So I have this two-stick kit (2x4GB=8GB) of G.SKILL Ripjaws "-GBRL" DDR3-1600's in my Z68 system.

Wrote to G.SKILL -- and got a great answer -- per setting the CMD=1 command-rate. "Yes, you probably can. No, it won't damage the memory if the modules aren't over-volted." "But . . . . CMD = 1 MAY lead to data corruption . . . "

I'd done this before with some Crucial Ballistix, but had to volt them to just under their warranty limit to make it work. By contrast, I left these RipJaws at their 1.51V setting, and set CMD=1. I picked up HCI Memtest(64) "Pro" for $14, and ran it for "1000% coverage" or ten iterations of "100% coverage." This is supposed to be totally "thorough" testing. No problem.

Assuming I re-run stress tests at my current over-clock settings with that command rate and validate that the system is still rock-stable -- can anyone think of ANY reason why this would cause data-corruption later, or make my system any less reliable or unstable?

Of course . . . if I ever buy a matching set of RipJaws to fill the two remaining slots, I'll have to drop back to CMD=2. I'm just wondering if there is some "risk" I haven't thought of.

 

[Shmee]

I would think it should be fine as long as no errors are found and testing is thorough. and since you didnt up the voltage, the ram should be fine. of course you never know, many other factors at play.

so far though, sounds pretty good! Id keep at it!

 

[BonzaiDuck]

I would think it should be fine as long as no errors are found and testing is thorough. and since you didnt up the voltage, the ram should be fine. of course you never know, many other factors at play.

so far though, sounds pretty good! Id keep at it!

I ran HCI MEmtest(64) for "1,000% coverage" -- what the HCI "notes" calls "thorough." Only need to redo a stress test in Windows. The performance gains for CMD=1 is supposed to be 5% to 6%. If it's stable and reliable, it's probably worth it . . . .

 

[StrangerGuy]

What? People still care about memory timings in this age, let alone a freaking SB system?

 

[Soulkeeper]

1T works flawless on my 2x4GB kit of mushkin 2133 with 1.5v
i'd assume it'll work with the gskill kit too

the only time i've noticed people setting these recent ddr3 kits to 2t is when they are overclocking very high with loose timings .. or maybe with 4 sticks in a system.

just set 1T and do your standard round of memtest86+ validating

"MAY lead to data corruption . . . " can be said about any memory setting ...

 

[=Wendy=]

Running G.Skill RIPJAWX (1600Mhz 2x 4GB) on an Asrock Z68 Extreme 4 with 2600K .
Voltage 1.5V
Speed 1600MHz
Timings 8.8.8.24 1T

No problems at all.

 

[Idontcare]

I don't know what it means but I run my 4x4GB GSkill kit at 1T (even though all four dimms are populated) and it passes HCI 1000% coverage.

 

[BonzaiDuck]

I don't know what it means but I run my 4x4GB GSkill kit at 1T (even though all four dimms are populated) and it passes HCI 1000% coverage.

I'm trying to remember which motherboard you're using. But it's sure good to know that you can do it -- in defiance of conventional wisdom.

What? People still care about memory timings in this age, let alone a freaking SB system?

If I didn't want to pick enthusiast-fly-s*** out of pepper, I'd buy a DELL. But the command-rate -- it's SUCH an EASY tweak to do!! I think I'd even seen where someone had tightened the stock 9,9,9,24 latencies for these RipJaws. I'm inclined to think a CAS of 7 instead of 9 would certainly require a voltage increase. And I'm not inclined to bother with it.

There's also the bank-cycle-time [or is it row cycle time?] -- if I recall correctly, it was "tRC." It, too, can be reduced to something like 1 or 2 greater than tRP + tRAS.

I'm a bit rusty on this since tweaking my DDR2 RAMs, and it pays to "refresh your memory" before making such tweaks.

You're abso-freakin-lutely correct, though. The payoff for fiddling with latency timings has almost disappeared. And there was an article by Anand on this site showing tests done with RAMs spec'd at > DDR3-1600 showing that it wasn't worth hardly anything to buy kits faster than that.

 

[Idontcare]

I'm trying to remember which motherboard you're using. But it's sure good to know that you can do it -- in defiance of conventional wisdom.

Its the MIVE-Z.

I've got all 4x4GB running at DDR3-1866 and 1.5V with 10-10-10-28-T1 timings.

I was surprised.

 

[Rifter]

Im running my 3 sticks of corsair XMS3 1600Mhz at 1T even though they are spec'd for 2T. I have run memtest and they are fine, and i have noticed zero signs of data corruption over the last 2 years i have had this system.

 

[BonzaiDuck]

[replying to both IDontCare and Rifterut] . . .

IDontCare:
I was pretty sure when making a comparison a couple months ago (and -- yeah -- I remember -- you got the Maximus IV Extreme-Z) that your flagship ASUS board and my -V-Pro board had the same power-design features/specs. Either 8-phase or 12-phase -- whichever. How that would affect this command-rate issue, I don't know.

I'd only guess that my board can do approximately what yours does in OC settings . . . . maybe not . . . . not sure . . .

So this is indeed "good intelligence" about the command rate issue and the G.SKILL modules.

Rifterut: Since you're obviously using a triple-channel kit (and even a different brand of DDR3) -- just as encouraging.

I'll probably spring for another double-module kit of the same and give it a try . . . . sometime, I suppose . . . .