Originally Posted by stanglx302
Now that I have my CPU and NB 100% stable, I want to start tweaking with the ram a little. I read somewhere that the ram controller on the X6 chips has a 1333 max. Is this true? So would there be any benefit at all to running it at 1600? It's is 1600 ram, but I just left it at 1333 for now.
AMD CPUs' benefit from both frequency and timings. However, finding the sweet spot for both is the key.
Since AMD's integrated memory controller (IMC) is only rated for 1333MHz anything above it is considered an overclock.
What you have to consider is what frequency (speed) you want to run at and at what timings.
For instance 1600MHz @ CAS9 is going to be noticeably slower than 1333 @ CAS7.
And then you throw CPU-NB in there as well. That is the direct link between the RAM and the CPU, so overclocking that as well affects your memory.
The first thing I'd do is see what the max frequency of the RAM I could get with auto timings. And testing with Memtest
just to see if the RAM itself could handle it. Who knows you might be able to run at 1800MHz.
When testing with Memtest be sure
to let it run at least
2 full passes. Some people suggest only tests 5 and 8, however, when testing my AM3 rig with DDR3, when I took my RAM too high I would get errors in other tests, and not once have I gotten an error in 5 or 8.
Also, be sure to jot down the timings that were auto selected for the given frequencies. That way later on it'll be a good reference point as to where to start when trying to put it all together.
After you've reached the RAM max stable frequency, go in to windows and run Prime95
blend test for at least an hour. If that's stable you can move on to the timings.
When doing your timings I typically adjust the CAS latency first. However, there is a formula to follow when adjusting the timings that I personally like to use: CL + tRCD + tRP = tRAS; tRAS + tRP = tRC.
There is some leeway with the timings, the formula isn't rigid. figure +/- 1 or 2 with tRCD and tRP, however I try to keep those three (CAS, tRCD, tRP) with 1 of each other. tRAS no more than +/- 4; tRC no more than +/- 8. However, if you want tighter timings the deviation will not be that much.
So no matter which timing you change, follow the formula through.
Also, your command rate should be 1T. I have found no noticeable improvement nor stability gain with DDR3 between 1T and 2T command rates.
At any rate, with this particular type of testing, you can go sort of quick. Run 1 full pass of memtest, then prime it for about 10 minutes. Then tighten another timing, repeat.
Memory testing with AM3 is semi tedious and time consuming. When I ran my AM3 rig, it took me 3 days to fine tune my rig to 1600MHz with 7-7-7-21-28-1T timings @ 2600MHz CPU-NB. Not bad for a C2 revision chip.