F3-16000CL9D-4GBTD @ 1600MHz on GA-P55-UD4P

[Bumrush]

Hey gang!

Just received my (2x) F3-16000CL9D-4GBTD and we passed an 8-hour memtest86+ run with no errors. So far, so good! Now I'd like to request some advice from the support staff regarding BIOS settings for this RAM.

I'm interested in limiting my BCLCK to 160-170 so that I can continue leveraging my Core i5-750's Turbo Boost and power saving features without increasing CPU voltages much, if at all. Currently, I run a BCLCK of 160 so that my
- 4 core speed is 3.2GHz (160 x 20)
- 3 core speed is 3.36GHz (160 x 21)
- 2 core speed is 3.68GHz (160 x 23)
- 1 core speed is 3.84GHz (160 x 24)
- 0 core (idle) speed is 1.44 (160 x 9)

Again, I'm very interested in keeping my BCLCK within the 160-170 range, so my RAM will consequently be clocked at 1600MHz-1700MHz. At the moment I'm running 7-7-7-20-72-1T with DRAM voltage @ 1.6 and things appear to be stable, but I'll be conducting some prime95 and OCCT tests today to confirm stability. What are the recommended latencies and voltages for this kit at 1600MHz-1700MHz?

Thanks!

[GSKILL TECH]

What you have it set at is good, you can try lowering timings from there to see how low you can go.

Thank you
GSKILL TECH

[Bumrush]

Follow-up

OK, so after some extensive testing I've determined that my lowest bootable settings at DDR3-1600MHz (BCLCK 160 x 10, DIMM voltage 1.65v, QPI voltage 1.21v) are 7-7-7-15-72-1T. While I would have been ecstatic to drop to CL 6, these are perfectly fine for my purposes.

Now, what strikes me as odd are the results of my benchmarking in Everest 5.30.1900. I ran a slew of memory benchmarks and I've found that, for the most part, the results are very similar regardless of certain settings. My testing methodology was to calculate the average of 5 trials of each benchmark at a variety of settings, focusing on tRAS from Auto (29) to 15 and tRFC from Auto (88) to 72.

For example, my lowest bootable settings (7-7-7-15-72-1T) yield the following results:
Read: 14361.4
Write: 11542.2
Copy: 16174.6
Latency: 48

With tRAS and tRFC set to Auto (7-7-7-29-88-1T):
Read: 14377.8
Write: 11551.8
Copy: 16444
Latency: 47.98

Lastly, with timings set manually using values that returned high average benchmark results (7-7-7-22-86-1T):
Read: 14491.8
Write: 11551
Copy: 16267.8
Latency: 47.98

(As you can see - and probably as should be expected - the Write and Latency benchmarks don't vary much at any setting, so those are a wash.)

I did expected to see a more dramatic variance with the Read and Copy results, especially at the lowest settings. However, the lowest settings actually returned the lowest average results, and the highest settings returned the highest average Copy but the middle average Read.

Do these results seem consistent with your findings? Are there any significant benefits to be gained by refining the settings further? I've never experimented with OCing RAM latencies so my expectations may be unrealistic, and I can move on if I've already reached the approximate peak of my RAM's performance. Any more advice/information would be much appreciated!

[GSKILL TECH]

Good work there. Yes, these findings are fairly normal. Generally, the lower the timings the better, but not always as the timings work together to run optimally. So if one number is too low, such as tRAS in this case, it can have a negative effect. tRAS should be tCL + tRCD + tRP. As your tests show, 22 ended up being the best because it is closest to 21.

7-7-7-21-1T should be your best configuration. Now to further enhance latency and bandwidth, you need to overclock the CPU, and these numbers should jump even more.

Enjoy!

Thank you
GSKILL TECH

[Bumrush]

Thanks so much for the prompt reply!

[Goldwater64]

I have the same board and memory but should I bump up the voltage slightly for 8 gigs? I'm having difficulty calibrating a stable voltage setting for 8 gigs.

[GSKILL TECH]

Not for memory, you will want to bump the QPI/VTT Voltage (memory controller).

Thank you
GSKILL TECH