F3-12800CL8D-8GBXM x2 with ASUS Maximus Extreme IV I7 2600K

[Hogbert_1999]

Fairly new to overclocking and needing some help if possible

System:

Asus Maximus Extreme IV
I7 2600K
16GB of Memory F3-12800CL8D-8GBXM (x2)\
Corsair Force GT SSD

I have the CPU O'clocked 12hrs stable on prime 95 @ 4.4Ghz with following Bios Settings
LLC 25%
VCORE Offset Mode +0.025
VRM Fixed Frequency 350
Phase Control Optimised
VCCSA Auto
VCCSIO Auto
Memory Frequency 1600
DRAM Voltage 1.55V
Timings 8-8-8-24-2T

The above is stress testing fine and passes 5 passes of Memtest 86+

I then tried to OC the Memory some more as follows:

Tried up to 1.65V DRAM Voltage for the following
1600mhz 7-7-7-21 2T
1600mhz 7-7-7-24 2T
1600mhz 7-8-7-21 2T
1600mhz 7-8-7-24 2T

1866mhz 9-9-9-25 2T

All had errors show up in Memtest 86+ :-(

The only tweak I can get to work is as follows:

1600Mhz 8-8-8-24 1T and this goes through memtest with zero errors 3 passes so far.

So please can you assist by advising how I can get more from these modules and whether the tweaks may help with the CPU wall I have also hit (44 mutiplier) for a stable setup?

I know my Vcore (1.406V full load in CPUZ) is high for a 4.4ghz setup (66 degrees on air) however I can not get it lower and if I go up in voltage and multiplier it becomes unstable.

I have kind of come to the conclsuion it may be the VCCSIO that needs a manual imput instead of auto, however I am not experienced enough to know what and how to go about it.

All help appreciated :-)

Jody

[emissary42]

I guess you also tried 1600 7-8-8-24?

Not all ICs like high voltages, so it might be worth a try to cut Vdimm back a little bit. For 1866 you could start at 10-11-10 and if thats stable go for 9-11-10, then 9-10-10 and 9-10-9 with voltage at 1.55 to 1.60V.

[Hogbert_1999]

Thanks so much for the comments that gives me something to try and I will:-)

I am still confused by the whole vccio voltage and whether I should not have it on auto?

Jody

[emissary42]

For DDR3-1866 you don't need to raise system voltages, that comes way down the road (at 2400+).

[Hogbert_1999]

I have set the VCCSA manually to 1.0V and the VCCIO to 1.15 based on what real temp was telling me i had from auto values when testing with Prime 95 on a stable setup.

and with 1.55v i was able to run a short test with 9-10-9-25 2T will need to try more when time permits, however quickly tried 9-9-9-25 2T and that failed at test 4 with these settings.

Is my manually set voltages OK for overclocking thsi stuff to 1866 or the next frequency up?

[emissary42]

I don't think you need those voltages above normal, for the memory clock at least*. If you want to try for DDR3-2000 and/or DDR3-2133, then it could be necessary. However with your 2600K a memory overclock of DDR3-2133 is not guaranteed regardless of memory. Ivy bridge i5/i7 do DDR3-2400+ easy, but most sandy bridge processors including the i7 edge out around DDR3-2100/2200.

I don't know what ICs are used on your kit, but if you had to relax timings beyond CL11-11-11-31 2T, then DDR3-2000 is probably not worth it. Compare performance between your 1600 8-8-8-24 and 1866 9-10-9-25(27) 2T settings via typical memory benchmarks and then chose the faster one. You could also try to reduce command rate from 2T to 1T for both settings, should give them a small performance boost if it is stable.

*However raising VCCIO does help with cpu overclocking at some point.

[Hogbert_1999]

Thanks for the info it is invaluable to me right now.

As I am working and have a young family it will take me some time to work through the above scenarios.. When I have I will report back :-)

[emissary42]

No worries, take your time.

[Hogbert_1999]

Ok I did some testing last night and was able to successfully tweak and achieve the following with Memtest stability..

no voltage adjustment needed from my std setup of 1.55DRAM VCCSA1.000v and VCCIO 1.150V

1600 8-8-8-24 1T
1833 10-10-10-25-1T

However if I bring the 1600 or the 1833 first timing down one I can not get it through memtest test#6 even at 1.6v DRam as well as increasing VCCSA and VCCIO :-(
so to clarify I tried

1600 7-8-8-24-2T Failed Memtest
1833 9-11-10-25 2T Failed Memtest

So I am now wondering am I better testing 2000mhz and timings or trying to reduce the other timing values @ 1600 and 1833 knowing that I can not reduce the first timing on either...?

Thanks for all help and advice.

PS is there a way you can get memtest to just test test#6 to save some time initially playing with values?

[Hogbert_1999]

Ok further update I watched this tutorial video

http://www.youtube.com/watch?v=6kNWM...r7t_gQ&index=1
and it help me speed up my testing and I have the following results!

1.750v

1833 9-9-8-28 2T and it boots and runs MaxxMem witht he following reult :-)




I just then tried to run it in memtest and again in test 6 it failed (small numbers but it did.

What would i be best adding or removing to stabilise it in memtest?

[emissary42]

Does it also fail in MT#6 when you lower your CPU overclock?

Also i would not go as far as 1.75V on Vdimm, because it is rarely worth it.

[Hogbert_1999]

Sorry Typo

VDIMM 1.575 ........

Have not tried it on lower CPU clock will try it and see what happens?

Also tried another config which seems to be still failing but much better I found that the timings in my last post would not allow Prime 95 workers to start!!!!!

SO i have now

10-9-8- 28-1T

System boots
Prime will work on full stress test (well it starts now, not had time to run it for hours yet).
Benchmark runs

Voltage tried Vdimm 1.575v failed test#6 on Memtest
Trying Vdimm 1.600v now on memtest???

I did ensure my CPU O'C was fully stable at 44Ghz before starting Memory O'C.

[emissary42]

When you raise your memory clock, your cpu overclock might need more voltage. That is quite common and would actually make sense here, since you seem to be getting errors in #6 regardless of timings or voltages.

[Hogbert_1999]

Ok i have a final clock after much fiddling about..

1866 10 - 10 - 8 - 24 - 1T

Vcore still at 1.4v
VCSSA 1.00v
VCCIO 1.10v

I am happy with that at the moment it is definately faster than the XM profile

thanks for your help it was truly good guidance during the testing.

Jody