DDR3 X79 - Better overclocker: G-Skill CL7 1600 or Corsair CL11 2133?

[yoadknux]

I'm looking to achieve 2400MHz CL11 OR 2133MHz CL9 on quad channel on my X79. I do this by doing voodo ram mixing. I have the following kits:

G-Skill RipjawsX, XMP: 9-11-11-1.65V 2133MHz (2x4)
Corsair Vengeance, XMP: 11-11-11-1.5V 2133MHz (2x4)
AData XPG, XMP: 10-11-11-(something) 2133MHz (2x4)

Right now I'm mixing the RipjawsX with Vengeance. My memory is at 10-11-11-32 1.65V 2133MHz, and VCCSA is 1.1V. I have decided that for fun I want to be able to achieve better memory speeds on quad channel. With the current set I can do something like 12-13-13-32 1.67V but VCCSA is 1.2V, don't want to go over that. I thought that maybe the Corsair sticks are holding me back so I bought an AData kit but I was unable to get it to work completely (also those memories are single rank, the rest of the kits are dual rank, not sure if they play along well).

Now I'm going to get G-Skill Sniper (2x4) that if I remember correctly has an XMP of 7-8-7-24-1.6V 1600MHz single rank. I am buying everything used so it's not a big deal to me and I like the experimenting. From your experience, and if you try to predict, which one should be the better overclocker, the CL7 1600 or CL11 2133?

Thanks!

 

[Woomack]

Motherboards usually take timings from first modules so you can try better or worse sticks in first slots to check which one will work better. It's matter of sub-timings rather than main and some are not even listed.
If you look for good RAM then check TridentX 1866 8-9-9, 2133 9-11-11, 2400 10-12-12 or 2600 10-12-12. Sniper series were not so good. RipjawsX depends on luck, some were good but most were Hynix based. More relaxed timings in G.Skill are saying it's Hynix too.
Good Hynix MFR can make 2400 10-12-12 or 2600 11-13-13 but I would look for Samsung IC for the best results as in quad channel you won't run much above 2400.

Corsair RAM is really random. 1.5V kits were from older series so are probably Hynix based and won't run at tight timings. ADATA looks better but also hard to say how it will really work.

There are generally no good kits at low frequency which are using 4GB modules. Most late DDR3 on good and more expensive IC were higher frequency. There are exceptions like TridentX but these kits were already more expensive. Early DDR3 were good also in lower frequency but most good IC are in 2GB modules like Elpida or PSC.

 

[yoadknux]

Motherboards usually take timings from first modules so you can try better or worse sticks in first slots to check which one will work better. It's matter of sub-timings rather than main and some are not even listed.
If you look for good RAM then check TridentX 1866 8-9-9, 2133 9-11-11, 2400 10-12-12 or 2600 10-12-12. Sniper series were not so good. RipjawsX depends on luck, some were good but most were Hynix based. More relaxed timings in G.Skill are saying it's Hynix too.
Good Hynix MFR can make 2400 10-12-12 or 2600 11-13-13 but I would look for Samsung IC for the best results as in quad channel you won't run much above 2400.

Corsair RAM is really random. 1.5V kits were from older series so are probably Hynix based and won't run at tight timings. ADATA looks better but also hard to say how it will really work.

There are generally no good kits at low frequency which are using 4GB modules. Most late DDR3 on good and more expensive IC were higher frequency. There are exceptions like TridentX but these kits were already more expensive. Early DDR3 were good also in lower frequency but most good IC are in 2GB modules like Elpida or PSC.

Yes! It's Woomack!

So I'm guessing the Sniper isn't a great overclocker. Eh, no big deal. I'll try it out anyway. Any point in bumping VCCSA for 2400 above 1.1V?

 

[Woomack]

I have good memory but short so I don't really remember how high I was setting SA on X79. Can check some of my threads and maybe you will find the answer -> http://www.overclockers.com/forums/showthread.php/732290-Woomack-s-memory-test-list

 

[yoadknux]

I have good memory but short so I don't really remember how high I was setting SA on X79. Can check some of my threads and maybe you will find the answer -> http://www.overclockers.com/forums/showthread.php/732290-Woomack-s-memory-test-list

So the memory master has short memory.... hahaha

I've set stick voltage to 1.66V and VCCSA to 1.15V (reported as about 1.18V in bios), and sticks to 11-13-13-36-2T 2400MHz. It booted and went into windows. Then I ran Prime95 Blend for an hour. Temps were a bit higher (I guess because higher VCCSA?) but I'm fine with max 85c on prime95. It passed so "inital" stability is good, of course I will run MemTest as well and play games etc since it's the most important factor for stability.

One odd thing though, maybe you can explain it, but when I checked RealTemp after a while it reported my BCLK was about 103 (which resulted in my cores overclocked to almost 4.6GHz instead of 4.4), although it is set to 100 at bios. Is this normal behavior?

Thanks!

 

[EarthDog]

Normal. After a certain memory speed, it raises the bclk to help reaxh those speeds.

 

[wingman99]

I had some RipjawsX DDR3 1600 F3-12800CL8D-8GBXM speed good memory to overclock.

 

[yoadknux]

Normal. After a certain memory speed, it raises the bclk to help reaxh those speeds.

Huh, interesting. I have the following questions then:

1. I don't see the BLCK constantly overclocked, most of the time it's at 100. When should I expect this BCLK increase to occur?
2. Should I consider downclocking my CPU in this case (since the BLCK causes the cores to be at about 4.6 which is a rarity for 4930k)?
3. What about overclocking the BLCK myself to the value that was shown in Realtemp? Is BLCK overclocking bad?

Also, at those voltages, should I consider adding some fans around the RAM or is RAM cooling just a gimmick? the case is pretty well ventilated with a side 25mm intake and top 25mm outtake but some sticks may not be receiving good airflow (On my board for 4x4 quad channel 2 sticks can be placed at an open area but the other 2 sticks must be placed behind the CPU radiator which means they are kinda cramped in a tight place).

 

[EarthDog]

1. Not sure the threshold...3000 mhz? 3200? Not sure. You can find out by testing it to see....

Im surprised to see it so slow. Perhaps i am not correct....(see testing above to find out)
2. That isnt our call. If voltages and temps are in order though....
3. Its not bad, its just limited and whwn adjusting bclk, raises the speed over everything.

 

[yoadknux]

So... I have managed to achieve CL11 2400MHz on my 4x4 mixed RAM. great.

But I got bored. So I bought a used corsair vengeance 4x8 (32GB total) 1866MHz CL10 kit for about 150$. Just some odd thoughts about this kit:

1. It took me about an hour to get them to work, I am not sure why. I think my MB starts hating all those RAM changes. It detected all the sticks but the detected memory capacity itself (in both the windows and bios settings) has changed wildly, sometimes it was 8GB, sometimes 12GB, 20GB, 24GB, 16GB, 28GB, finally after a lot of setting changes (mostly back-and-forths) I managed to obtain 32GB. I am not sure why this happens, it's very weird behavior.
2. I talked about the XMP, well the voltages on the XMP of this kit is 1.5V for RAM (which is about average, the higher clocked kits usually require 1.65V), and, wait for it... 1.3 VCCSA? wtf? Why so high? Do higher memory kits require more VCCSA? Anyway after finally making the MB recognize the entire 32GB kit I tweaked the timings to match the XMP (10-11-11-27 1866) and got it to work with 1.1V for the VCCSA. I left Memtest running but it takes a long time with 32GB so I'll post the results later.

Is there a way to explain what I'm seeing (like why the MB doesn't always detect 32GB but some odd fractions instead, why VCCSA is so high, etc)? I think if I plan on keeping VCCSA high I will need to cool the heatsinks on the MB itself, which are pretty hot to the touch. It's a p9x79 LE and the processor itself is water-cooled with Corsair H80i in push-pull.