GA-Z68MX-UD2H-B3 New EFI

[jambek2003]

Is it my system or do anybody also experience boot-problems and stability-problems with UEFI U1c?! Settings which worked for BIOS F13 with DSDT don't work anymore. Sometimes Hackintosh even fail to POST. Had to set 16GB Corsair from 1600Mhz BIOS F13 to 1333Mhz in UEFI U1c. After one night, after maybe week working without problems, system crashed and from then on stability problems. Already resetted Mobo for like 4 times cause it just wouldn't even POST. DVID was @ -0.190V with BIOS now @-.0050V and even not stable with same max Multi's (4.2ghz by setting multi to 43,43,43,42 and RevoGirl's SSDT-gen-script. Hack just crashed just before writing this with Max multi to 37. I don't get it!

Do you guys experience stability problems? My setup z68mx with U1c, i2500K, Sapphire HD5770, 4x4GB Corsair Vengeance LP. I'm thinking about going back to BIOS and DSDT. Any thoughts?!

EDIT: In Memory-section I enabled XMP-profile upped Vmem to 1.46Volts (Is Low voltage RAM should work [email protected]) and Upped Fan-speeds from Normal to Manual 1.75..something.... Ran CPUTest to stress Cores and also Unigine Heaven Demo for stressing GPU. both ran parallel for 1 hour without any problems. So no stability issues. Restarted en halted system and also seems to be okay.. knock on wood!

 

[genzai]

EDIT: In Memory-section I enabled XMP-profile upped Vmem to 1.46Volts (Is Low voltage RAM should work [email protected]) and Upped Fan-speeds from Normal to Manual 1.75..something.... Ran CPUTest to stress Cores and also Unigine Heaven Demo for stressing GPU. both ran parallel for 1 hour without any problems. So no stability issues. Restarted en halted system and also seems to be okay.. knock on wood!

I have no stability issues with my 4.2ghz OC and 1600mhz memory. So hopefully your tweaks will get you stable. memory and CPU all perform a bit differently and some can be affected by firmware changes.
g\

 

[valkeriefire]

@jambleck,

Many people report overclocking on the UEFI bios is more difficult than the legacy bios. I happen to disagree, UEFI bios is different because we no longer have Load Line Calibration (which I am a fan of), BUT I have had good success with the new bios. Set your Vcore to "Normal", and then try DVIC of +0.20 (under load this should result in a Vcore of about 1.45v (which is a little high but certainly still safe if you have a good aftermarket cooler, don't try it on the stock cooler). Test with Prime95 blend test for 1 hour. If it doesn't freeze or crash you are probably good. You should be able to OC up to at least 4.6ghz with that voltage (I can do 4.8 with it). If your target speed is 4.2ghz, then slowly reduce the voltage by 0.02v until instability returns, then go back to the lowest stable voltage. You should be able to run around +0.10 for 4.2ghz. My 2600k runs at +0.16v @ 47/46/46/46. My 2500k runs at +0.2v for 45/44/44/44 (I think I can run lower volts but have not had time to test it).

 

[jambek2003]

@valkeriefire I don't want to throw so much juice on my CPU. I try as low as possible.

i2500k was running Vcore normal+DVID -0.190V@4200Mhz with bios F13 so that's even undervolted! I don't think I enabled LLC in bios F13.

My hack has frozen a few times since last post. Also memory had errors at [email protected] which was completely stable before at 1600Mhz. Upped Vmem to 1.46V and after 2.16Hours at generated 2 errors in Memtest86+ at 1600Mhz. What's going on I thought maybe airflow so vacuumed the machine! Now upped DVID to 0.00Volts Multi's at 43/43/43//42 and currently running CPUTest for 20 about minutes. Hopefully it helps.... cross fingers.

 

[valkeriefire]

@ Jambek,

I would be very surprised if you could successfully undervolt your CPU and sustain an overclock under load (and based on your crashes I'd say you are not). Take a look at these overclocking thread...

http://www.overclock.net/t/1100100/info-intel-2500k-2600k-overclocking-tips

http://www.overclock.net/intel-general/910467-ultimate-sandy-bridge-oc-guide-p67a.html

Based on these threads, which I've read a lot of, and my own trial and error, I can honestly say only a golden chip is going to run at the voltages you are trying without crashing.

What kind of cooling are you using? Even a $20 aftermarket cooler like a Coolermaster 212+ should be more than enough to cool at voltages up to 1.4v. Also everyone is afraid of voltage, but it is paper tiger. As long as you keep your Sandy Bridge CPU under 1.5v (intel rates them up to 1.53), then you will be buying a new CPU long before you burn out your current one.

Also what memory are you running? Most 1600 RAM runs at 1.5v. Some, like the Samsung 30nm chips run at 1.35v, but if yours is 1.5v or 1.65v and you are trying to undervolt it, then it will cause stabilty issues. Best to only deal with one variable at a time. Don't overclock your RAM or undervolt it until you have stabilized your CPU, or else you won't know which is causing the problems.



Goodluck.

 

[jambek2003]

Okay had a quick check at these guides maybe playing with C3-state and C6 would be helpfull. But for now; Hack is running allready 20H:57M stable at 4200Mhz, verified with MSRDumper, in CPUTest with testtype "all" on all 4 cores --> DVID is increased to 0.00Volts. Looks promising. By the way my Cooler is a Thermalright True 120MM Black lapped and the i2500k also lapped with 2500grid sandpaper ;-) Memory is Corsair Vengeance LP according to webshop it was Low Voltage @1.35Volts.

Corsair.com --> Vengeance® Low Profile — 16GB Dual Channel DDR3 Memory Kit (CML16GX3M4A1600C9);

Choose from 1.5V and 1.35V Models

1.5V Vengeance Low Profile DDR3 memory is designed to meet the memory specification of the 2nd Generation Intel® Core™ processor family. The Special Edition Arctic White Vengeance Low Profile memory modules operate at the ultra-low voltage of 1.35V, for ultra-quiet systems and other low-voltage applications.

But in Tech specs it stated SPD 1.5Volts??

I'm also going to check if I can test with custom FFT-setting in CPUTest. According to guide should be set to 1344 and 1792 FFT. Keep you guys posted maybe helpfull for other OC'ed Hackintoshes. EDIT: No can do with CPUTest but also Prime95 is available for OSX; ftp://mersenne.org/gimps/p95v277.MacOSX.zip so testing now FFT between 1344 and 1792! Going to run this for tonight.

 

[valkeriefire]

Sound like you've got it figured out. Running the 1344 and 1792 FFTs will test your system very quickly.

Goodluck.

 

[jambek2003]

Sound like you've got it figured out. Running the 1344 and 1792 FFTs will test your system very quickly.

Goodluck.

Hah yes the Hack has been running several days with several of tests. okay looks like I had to up the DVID to 0.00Volts, trying later on to lower this step by step, ran for like 13H Prime 95 stable and also 15hours of CPUTest. Outcomming of Memory testing is interesting. I was running test with Memtest86+ v5.0b1, it's way faster then previous versions; it's multi-core. So a lot more of runs in the same time BUT; I got 2 errors within 18hours,in Test 5 in pass 11(like 10.5hours), every time the same errors at same location ;-( Upping Vdimm no luck, upping VTT to 1.08V same errors BUT within 9 hours, upping VTT to 1.10V, 127 errors in 7 hours. Okay looks like putting more juice on the Mem-controller does increase errors. I was like "why always test 5?!" Did some googling and found there is a bug in the Beta version which could trigger errors at test 5. WTF?! Testing with v4.20, way slower, but Vdimm @1.5V and VTT to 1.02V no errors after 15 passes/36H:30M. This is stable okay! I lowered the Vdimm to 1.40V and only did test 5; 156 passes in 9hours. going to run full test with Memtest86+ v4.20 this weekend and see if still no erros occur.

Now going to have some real live working on my Hack and cross fingers it will not be unstable. Long story short; it looks like EFI is stable but upping the DVID( offset voltage) was essential for my system. By the way I didn't disable C3- and C6-state, I left it on enable.

 

[pavlinbl4]

Hello Have you USB 3.0 working ?

 

[randycatanach]

Have you tried the universal USB 3.0 kext in Multibeast?a