SOLVED Looking to tighten the timings on my G.SKILL F3-17000CL9Q-16GBZH.

[Ajay57]

What can i say this is AMAZING guy's, with my Overclock @ 42 so far i have dropped yes dropped, from 1.40v to 1.350v in the Vcore in my Bios and it just keeps passing P95 blend. Without any problems at all, its got to be the extra stability due to the Ram here, that was set up!!




Regards,

AJ.

 

[Ajay57]

I just begin to wonder what this means for my higher overclocks if i can lower my Vcore by so much!! I still think that they is more to come, and possible to get as low as 1.300v for a OC of 42!!!

Its worth a try at least to see what the outcome would be, and to test with P95 Blend each time!!

The Amazing news is that i have gone from 1.400v in the Bios Vcore to now 1.300v and its 100% tested by P95 Blend and runs Stable, with no problems. This is with my 4.2 OC which i set up my new ram with and Overclocked to 2400Mhz.




AJ.

 

[Woomack]

I think you can make at least 4.4-4.5GHz at 1.3V. Your old 4930K was running @ 4.7 1.5V or something near so when you got much better 4960X then why are you keeping it at so high voltage at 4.2GHz ? Simply set higher clock and check if it's passing HyperPi 32M , if yes then run something more.

My 4930K is running like that:
4.2GHz ~1.2V
4.5GHz ~1.3V
4.7GHz ~1.4V

Maybe 0.02V more for full stability but I don't really care to run it in prime95 or any other test for longer. It passed 1h XTU @4.7GHz ~1.41V as I remember but that's all.

101 or 124 blue screen errors are caused by too low or too high voltage and also overheating. When you set 1.5V+ then check how high is max temp. Some IB-E users are already using sub 0 cooling for voltages above 1.5V.

 

[Johan45]

Now you got it Ajay!

 

[Ajay57]

What can i say, i am blown away with this so far, maybe i can do 1.200v @ 4.2Ghz!!

It seems like a total different PC altogether, all because you have the Correct Ram and its dialed in and working properly with your CPU. The extra stability is just mind blowing, and i have never seen anything like this before.

So i would say to any OCF Members if you want extra stability and possible lower Vcore you just got to try this!!!

Thanks Woomack for your time and Help and also Johan45 for your encouragement as well, WAY TO GO OCF!!

Well still here and guess what still going down with the Vcore and passing Blend @ P95.



Respectfully,

AJ.

 

[Ajay57]

Sorry everyone for the double post, but just reached 1.200v in the Vcore fully tested with blend P95!! Yep i am really happy the way this is going, so fingers crossed this could mean a lot! When i start to Overclock the other way towards 4.8 etc or even dare i say 4.9!!

Regards,

AJ.

 

[Woomack]

you can go up with cpu multi ... when you see errors then add 0.03V and try again, when it passes then raise cpu ratio ...

 

[Ajay57]

Well the news is not to bad, but a little disappointing as well, up to 4.7 its great the Vcore is 1.425v only. It passes P95 Blend with no problems at all, on the other hand to reach 4.8 its still the same as before.

The Vcore is still 1.500v to 1.510v that is to get passed Cinebench and other benckmarks, it seems to me that after 4.7 it drops off a cliff!!

I have not tried using the 125mhz Strap yet to see if that would help to reduce some of the extra Vcore to be able to run things. I have tested it out on some of the 2D Benchmarks there seems a big improvement to me.

So what to do next and where to go is the next question here???

Respectfully,

AJ.

 

[Woomack]

Straps are not helping to reduce vcore. You will faster need ( or not ) higher VCCIO or other voltages that will slightly raise CPU temp.

Your 4960X is scalling almost the same as my 4930K so 4.7GHz is pretty easy but for higher clocks are bigger steps.

Now all depends what you want to achieve as you checked memory and CPU. You can try higher straps and compare performance but I doubt it will help you much.

 

[Ajay57]

What i was looking and hoping for was a Vcore of 1.475 or 1.485 something in that area, with the LLC bumping into 1.520v!! What it is and just need to accept it, the whole thing works but. I do not like very much the higher CPU voltage, that's why i was looking for another way. To try to achieve the same result, what if we lowered the RAM SPEED to 2133Mhz!!

I was thinking about this last night, what if we lowered the RAM SPEED to 2133Mhz and then tweaked it for max performance?? Would that help to stabilize the CPU and drop the amount of Vcore being applied to it??

What do you think about that Woomack, have you tried to do this before and what was the results??

 

[Woomack]

I saw no difference in max CPU clock if I set 1600 or 2500+ memory clock. This platform can make more and you have probably best board to make high memory clocks on X79. My MSI ( RIP ) couldn't run stable at more than ~2500 but as I said no matter how high I set memory clock, cpu clock wasn't any better.
Most my 5.2GHz+ results on 4930k are with memory @2520.

 

[Ajay57]

Well if that's the way it is then the thread is SOLVED my friend, again many thanks for your time and help!! Its been great to work together, i will try this out on some Benchmarks, and see what happens.

And yes its nice to be running @ 42 @ 1.200v for my everyday Web Chat!!

Respectfully,

AJ.

 

[Ajay57]

Well today i had some time spare so i decided to have another go at OCing my RAM some more after reading some excellent reviews by Woomack!

Been comparing his timings to mine and decided to give it the green for GO!

So here is my results and i am happy to say its made things better than before as you can see!! What i am showing you is the BEFORE and AFTER the tweaking of the Main and SUB timings. Its quite a lengthy job as you can only move one at a time and in some cases only one digit as well!!

My Mem Tweak It score at the beginning was = 47875, now its = 47913.






Respectfully,

AJ.

 

[Woomack]

As I said before , compare these results in AIDA64. Especially write test will show you if everything is alright. Maxxmem Preview = single thread test.

I also see you stabilized your rig and you are getting bored That's actually good as you still want to set some more and it's better than spend money on new hardware that won't be much better ( if not worse ).

Save your current settings in a profile in BIOS and try something like:
2666 11-13-13-35 1N 1.65V , all other settings auto , tRFC 340. If it won't boot then try the same but 1.70V.
If you make it work then compare results in AIDA64 and HyperPi 32M. Don't waste time on longer stability tests before you set higher performance than 2400 CL10.

I would check something on my board but well, 2 power phases burned down and I sent it to RMA so no X79 for next couple of days ( if not weeks ).

 

[Ajay57]

Sorry to hear that Woomack buddy, yeah just tweaking and slowly adjusting things no rush lol! But looking at the results a good step in the right direction that is and yes i will try what you posted. When i have some more spare time, going to be really busy this afternoon.

Thanks for looking in!!

AJ.

 

[Woomack]

Ye, I killed 2x mobo, cpu and a flash drive in 2 days. Gigabyte Z87X-OC already came back from RMA, support didn't say what they fixed but I see freshly soldered both BIOS chips and one other chip next to them. 4770K CPU was showing only 00 code on ASRock board and a flash drive ( which was pretty nice SLC ) is only blinking.

I'm telling you to try 2666 clock as you already set something good at 2400 and it won't be much better so to keep improving try something at higher clock instead of tightening timings. When you make it stable then you can try tighter timings. With some luck maybe it will run at CL10. One of my Patriot 2x8GB kits which is also on Hynix is running at 2600 10-12-12-32 1.75V ... my brother got it so knowing him it's running at auto 1600 right now

 

[Ajay57]

Last time i tried 2666 it crashed and did not like it at all, but last night i up dated to the new Bios 0507!! Which to me seems better that the Factory Stock one that came with the Mobo.

But like i said, i have to go now as its getting late here lol!!

See you later,

AJ.