Just picked up some DDR3 on craigslist. Anyone know much about this stuff? - Page 2

It's not possible to even run different pairs at different voltages and speeds (at least not on this motherboard). The motherboard runs all 4 slots at the same voltage and speed. The "ratings" on the memory have nothing to do with what the motherboard runs them at in this case except if the settings are left on "auto," in which case it will default to 1333-9-9-9 in this case, the worst performing is defaulted to in order to ensure stability and all slots are run at that speed. Nor is it possible to segregate one pair of memory from normal system memory and treat it as a separate heap. The system RAM is recognized as one "chunk" from the moment the BIOS posts.

Obviously my older 1600-8-8-8 kit is going to be able to run any setting that the generic PNY set will handle, so when tinkering with settings, it is the PNY set that I am finding the limits of, and the OCZ that is along for the ride. I understand that a small part of the stability issues can be created by the fact that the memory controller is working with all 4 slots loaded. Matched sticks are only significantly important for use in dual channel. Obviously, I'm using the OCZ as one "dual channel set" and the PNY as a separate "dual channel set."

"The following conditions do not need to be met:

Same brand
Same timing specifications
Same DDR speed
Memory channel speed is decided by the slowest DIMM module populated in the system."
- http://www.intel.com/support/motherboards/desktop/sb/cs-011965.htm#flex

---

Hello Tjj226,

I just tried the following:

Reference clock: 240MHZ
CPU multiplier: 13 = 3.12GHZ
NB Multiplier: 10 = 2400MHZ NB
~1.2V NB, ~1.4V CPU, ~1.66V RAM
RAM: 1333-9-9-9-24-36-1T (reference clock results in 1600 speed),

System is running fine (for now) but memtester came up with failures in a stress test. Will keep tinkering

Best regards,
Eric
Edited by mdocod - 10/20/12 at 5:52pm

Hi Tjj,

The multiplier was set for 1333, but the resulting speed was 1600 due to the reference clock.

I did a number of test cycles and found that for whatever reason, this RAM is not stable at any settings over ~1.6V. When I try to run 1.65V, it will produce errors regardless of speeds/timings. The good news is that I am getting promising results with lower speeds and tighter timings as suggested. In the past, I never took the time to really optimize this machine, so I am working on that now. Shooting for better NB clocks, a mild CPU OC, etc etc.

---

Hi Schmuckley,

We're not expecting great numbers here, I'm just tinkering with cheap RAM. I am well aware of what sort of numbers modern low voltage kits are able to run. The kit I picked up off of craigslist was about half the price of a sammy kit I can't justify fancy RAM for an Athlon II. I can justify the tinkering time on the cheap RAM because it's fun and I have the time!

---

UPDATE:

Was running:

Reference clock: 266MHZ
CPU: 1.4V, X12 = 3.2GHZ
NB: 1.275V. X10 = 2660MHZ
HT: X7 = 1862MHZ
RAM: 1418-7-8-7-20-24-4@1.61V

Bummer... booted but getting errors. Going to back off to 250/2500 on the reference/NB here as I don't like the idea of running the NB voltage so high.

---

Now running:
Reference clock: 250MHZ
CPU: 1.4V, X12.5 = 3.125GHZ
NB: 1.225V, X10 = 2500MHZ
HT: X8 = 2000MHZ
RAM: 1333-7-8-7-20-24-4@1.56V

Booted, no errors yet.

got ~11,000MB/sec in a quick benchmark.

I have tried a handful of speeds and timings, and it doesn't matter if I'm running 1066 on tight timings, or 1600 on loose, the bandwidth cap is about the same either way. I will probably stick with these latest settings as they seem to be stable and fast.

Eric
Edited by mdocod - 10/20/12 at 10:24pm

Hi Redwoodz!

The load on the power supply for the RAM, and the IMC changes when all 4 sticks are in place like you were saying, so anything I find out while running a set individually may not apply when all 4 sticks are loaded. I'd rather use the tinker time on the intended configuration. I don't see that as pissing into the wind personally. Though I appreciate your thoughts on the matter.

I run linux, no CPUz for me

The dump from dmidecode is not nearly as revealing as CPUz. I have a love/hate relationship with linux. This is one of those times for the hate.

These settings are working fine it seems:
Reference clock: 250MHZ
CPU: 1.4V, X12.5 = 3.125GHZ
NB: 1.225V, X10 = 2500MHZ
HT: X8 = 2000MHZ
RAM: 1333-7-8-7-20-24-4@1.56V

---

I am beginning to wonder though, if I CAN actually run different timings on the different pairs. Seems like the option is there in the BIOS, I just don't think it's a good idea to do that.
Edited by mdocod - 10/21/12 at 1:02am