Dolphin, the GameCube and Wii emulator - Forums

Try to get one that's 1.5v or lower if possible. Sandy bridge doesn't like high memory voltages and the low voltage + decent frequency/latency implies it's good (and therefore highly overclockable) memory.

(01-29-2011, 03:43 AM)NaturalViolence Wrote: [ -> ]Try to get one that's 1.5v or lower if possible. Sandy bridge doesn't like high memory voltages and the low voltage + decent frequency/latency implies it's good (and therefore highly overclockable) memory.

I'm looking at these 1.5v 4GB kits from G.SKILL made for Sandy Bridge:
1600 8-8-8-24 $65
1600 7-8-7-24 $70
1866 9-9-9-24 $75
1600 6-8-6-24 $80
If you have any advice or know of any other kits I should look at, let me know. I see some other G.SKILL RAM with the "Made for P67/H67" labeling, but running at higher than 1.5v. Of course, they really just up the voltage and change the timings a little and sell them for more. How dumb.

So, I didn't jump on the Newegg sale but it's only $10 or so.

Go for the $65 kit.

(01-29-2011, 06:22 AM)NaturalViolence Wrote: [ -> ]Go for the $65 kit.

Because the price or is there any other reason it is better? The 1866 technically has better actual latency (cas/frequency*2000 = ns latency is the formula I found) and significantly more bandwidth.

1866 also requires that the memory controller on the cpu be overclocked leading to more heat and voltage requirement.

Quote:Because the price or is there any other reason it is better? The 1866 technically has better actual latency (cas/frequency*2000 = ns latency is the formula I found) and significantly more bandwidth.

I wouldn't call that significant. They're all so so close to each other that you might as well grab the cheap $65 kit.

Quote:1866 also requires that the memory controller on the cpu be overclocked leading to more heat and voltage requirement.

^That as well.

Oh, I didn't know that the memory controller needed more voltage or anything. I thought it was just changing the memory multiplier? (unless that's what you meant)

Also, if the difference between 1333 and 1600 is significant with EFB to RAM, wouldn't the difference between 1600 and 1866 be significant as well?(though just a little less) or at 1600 is the bottleneck essentially gone?

I just don't want to be playing X game and have EFB to RAM being a bit too slow because I didn't pay the extra $10, heheh.

If they both had the same latency ddr3 1866 MHz ram would be about 15% faster than ddr3 1600 MHz ram. However you are comparing ddr3 ram running at 1866 MHz 9-9-9-24 (higher frequency and latency) against ddr3 1600 MHz 8-8-8-24 (lower frequency and lower latency). So their is little to no difference between them. The second set would only be about 5% faster in random read/write and probably the same speed in sequential read/write.

And yes the memory is controlled by a memory controller. Since it needs to be running in sync with the memory if you raise the memory frequency (clock rate) you are also raising the memory controller frequency. And higher frequencies require higher voltages. So the memory controller will be running at a higher frequency and voltage.

Wikipedia: "Because modern DRAM modules' CAS latencies are specified in clock ticks instead of time, when comparing latencies at different clock speeds, latencies must be translated into actual times to make a fair comparison; a higher numerical CAS latency may still be a shorter real-time latency if the clock is faster. " Hence the formula I noted earlier (cas/frequency*2000). 1866 CAS 9 actually has less latency than 1600 CAS 8. But its still close so it doesn't make much of a difference either way. At least I learned a lot right? heheh

(01-29-2011, 05:05 PM)iluvdolphin Wrote: [ -> ]Wikipedia: "Because modern DRAM modules' CAS latencies are specified in clock ticks instead of time, when comparing latencies at different clock speeds, latencies must be translated into actual times to make a fair comparison; a higher numerical CAS latency may still be a shorter real-time latency if the clock is faster. " Hence the formula I noted earlier (cas/frequency*2000). 1866 CAS 9 actually has less latency than 1600 CAS 8. But its still close so it doesn't make much of a difference either way. At least I learned a lot right? heheh

FYI I just ran benchmarks on 1600MHz 8-8-8-24 vs 1866 9-9-9-24 and the only notable difference was in the read test. If my memory serves me correctly the 1866 read @~ 21,000MB/s where now it reads @ 19,626MB/s. I'll switch back to 1866 and run benchmarks to get a more accurate comparison.

What is nice about the 1600MHz is I shaved about 6-8 degrees Celsius off of each core. I think I'll keep it here regardless of what the benchmarks show. It saves me some $$ since I was considering water cooling.