• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

My X79 RAM adventures

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

yoadknux

Member
Joined
May 6, 2016
I seem to open topics about memory about once a week, so I think that from now on I'll just write stuff here... It's probably not most informative post, more like a diary.

I have played around a lot with RAM... Probably changed more sticks than any other component in my system. I have a 4930k @ 4.4GHz, Aorus 1080ti, Crucial MX500, Asus P9X79 LE, and some Corsair 80+ 750W PSU (I think RM750i)... And I have done A LOT of ram mixing, and some overclocking. Here are my findings:

1. I have mixed 2x8 G-Skill Ares (CL9, 1600MHz) with 2x8 G-Skill RipjawsX (CL9, 1600MHz). Memory was Memtest and Prime95 stable at default timings (CL9 1600). Max stable overclock was CL13 2133MHz at 1.65V voltage. CL12 2133MHz also booted but applications were crashing. At some point I was getting a BSOD when booting the PC here and there, it was probably due to OS corruption because of the overclock. RipjawsX managed to do CL11 2133MHz stable but not together with the Ares.

2. I have mixed 2x4 Corsair Vengeance (CL11, 2133MHz) with 2x4 G-Skill RipjawsX (CL9, 2133MHz). Memory was Memtest and Prime95 stable at default timings (CL11 2133). Max stable overclock was 11-13-13-36 2400MHz at 1.65V voltage. I am still surprised that those sticks mixed so well together. Also sticks are dual rank which from what I understand do not have the same overclock capabilities but give better performance.

3. I have mixed 2x4 Corsair Vengeance with 2x4 G-Skill RipjawsX with 2x8 G-Skill Ares and 2x8 G-Skill RipjawsX. It was a funny attempt that actually managed to boot and all 48GB was detected, but Memtest showed errors regardless of timings so I gave it up. Also memtest took a looooong time with that much memory and I was running out of patience.


4. I have bought a 2x4 ADATA XPG (CL10, 2133MHz) set. It did not work at all, even in dual channel couldn't detect more than one stick :) I suspect faulty set or maybe designed to work with different memory controller.

5. I have bought a 4x8 Corsair Vengeance (CL10, 1866MHz) set. They were not working at all on my motherboard after many attempts. The memory identified by the BIOS and OS varied a lot from 8GB, 12GB, 16GB, 20GB, 24GB, 28GB, 32GB.... regardless of timings. I have tried several times with various timings, with and without XMP, high VCCSA, high VTT, high DDR voltage, nothing. I managed to use 32GB at some point but Memtest gave errors. I have searched around the internet and several users have reported issues with this kit when using X79.

6. I have bought a 4x4 G-Skill TridentX (CL10, 2400MHz) set. The memory was easily detectable and I have manually inserted the timings and they worked well. There are two things that have disappointed me about this set, the first is that the memory is single rank and not dual rank. The second is that I was still unable to fulfill the overclock potential of this kit: CL9 2400MHz - boot but almost corrupted the OS. CL11 2666MHz - no boot. Maybe CL9 2133? I'll try eventually. Woomack has once stated that X79 doesn't do much better than CL10 2400MHz regardless of brand - I'll test that further but I think he's right. With BCLK I was able to get 2500MHz but didn't test stability.

Just a few things I learned along the way:
1. If you care about your data, don't overclock your memory. Bad memory OC can corrupt the OS: I have personally seen random BSODs (even on stable timings because of previous failed OC attempts) and failed OS boots.
2. VCCSA and VTT, at least from what I've seen, don't really help you achieve better memory overclock. I have increased it from 1.1V to 1.15V/1.2V/1.25V/1.3V but other than increasing CPU temperatures I did not notice anything special. TridentX is CL10 2400MHz stable even with both VTT and VCCSA at 1.1V.
3. X79 heat sinks feel super HOT, like touching the backplate of my 1080ti (cannot keep fingers on them for more than a few seconds).

I will of course try to improve the overclock on TridentX, and later mix with the Corsair and Ripjaws set to see what happens. Single rank and dual rank together, here we goooooo
 
Last edited:
One important thing that I forgot to mention configuration #2 (mixed ram) has outperformed configuration #6. This may raise some eyebrows: TridentX is one of the best DDR3 memory kits. But like I mentioned in my post, it is single rank, there are chips on only one side of the stick. The mixed configuration used only dual rank sets. We can compare the results I obtained from Userbenchmark:


TridentX 10-12-12-31 2400 (single rank)
MC Read 49.7
MC Write 38.1
MC Mixed 45.7
127% 44.5 GB/s

SC Read 20.4
SC Write 16.5
SC Mixed 15.5
50% 17.5 GB/s

Latency 52.1

Mixed Corsair + RipjawsX 11-13-13-36 2400 (Dual rank)
MC Read 54.2
MC Write 40.4
MC Mixed 52.6
140% 49.1 GB/s

SC Read 19.9
SC Write 16.5
SC Mixed 17.1
51% 17.8 GB/s

Latency 53.4
75% 53.4 ns

The tradeoff here is about 10% higher bandwidth vs 3% latency.
 
Back