All seems OK for now, running LinX long 20 runs without problems.
Thankx Thrax
@Leo : That's why i had fear to go over 1.65 too, but many people talk on the forum about a rule that says : if you are within .5v between VTT voltage and VDIMM , you're ok. I've seen a lot of people running at 1.8 VDIMM with increased VTT voltage, but in any case i've seen also people saying its dangerous to do so lol
0
LeonardoWake up and smell the glaciersEagle River, AlaskaIcrontian
edited September 2009
Assumptions can get one in trouble! So much for assuming on my part. Yeah, Thrax, I understand now. Thank you. I'm still quite happy with Socket 775/DDR2. I don't yet have practical experience with the newer stuff.
You seem to have a golden i7 in your hands. Matched with your oc skills, that is one of the most powerful PCs available over there. Congratulations! I am envious.
Yeah this PC is screaming for work processing , i almost feel bad each time the screen saver comes up :P
In less than 6 month i'll change my video card too (9600 GT atm) and in a year or less i'll put in a SSD drive (Raptor 150GB atm) and then it will be perfect
Bclock frequency = 181
CPU multi = 20
QPI multi = 36 (for QPI Link Speed of 6.51Ghz)
Uncore multi = 17
Memory multi = 8
Memory voltage = 1.64
CPU voltage = 1.20 in bios (~1.15 in windows, drop at 1.13 on full load)
CPU load temps = ~60C on RealTemp
TURBO = ON
HT = ON
C1E = ON
The weirdest thing that i didnt get an answer for tho, is why is i am able to get DDR-1448-7-7-7-21 STABLE , but cant boot up from cold state at that setting >.<
If i boot at 8-8-8 then go into bios and change it to 7-7-7 it will work for that time, then i close the computer, wait 3 secondes, boot again, and it wont boot and it will come back to last secure settings to boot.
Just got an idea, maybe i could use Easy tune to change the settings once in windows, and load from that profile at each boot. Once power is in and after the bios screen i can do 7-7-7-21 , it was memtest 45mins stable. Gotta test it in windows tho cause i didnt last time.
Alright, I'd like to test memory overclocking first. Set the following parameters:
Turbo mode and EIST: DISABLED
Bclk: 200MHz
Core voltage: 1.2v (Default 1.2v)
DRAM Voltage: 1.65v (Default varies)
Uncore voltage: 1.35v (Default 1.2v)
CPU PLL Voltage: 1.88v (default 1.8v)
CPU Multiplier: 12x (2400MHz)
QPI Link Speed: 36x (3.6GHz - or - 7.2GT/s)
Uncore Multiplier: 16x (3200MHz northbridge)
Memory Multiplier: 8x (DDR3-1600)
Loosen the memory timings as tight as they will go, and start testing with memtest. If it fails, reduce the uncore and memory multipliers by one step and try again.
A few notes: These settings should get DDR3-1600 capable memory running at DDR3-1600 speeds (200*8 = 1600). If this fails, your memory may simply be unable to rise to that challenge. Secondly, many Core i7s top out at 3.8GHz, even with the C0 stepping.
Wow, after some testing with the settings you suggested i've found something that should be very near the max i could achieve 45 mins memtest-stable (a bit more than one complete pass).
DDR-1600 8-8-8-19 !!!
When it booted i was sure it wont pass memtest, but it did lol.
I first tried 8-8-8-24, after 20% memtest i tried 8-8-8-21, after 20% i went 19 and when it booted i tried any setting to reach 7-7-7, it wont boot at 7-7-7 even with 7-7-7-28.
So, my RAM is capable of 1600 at 8-8-8 which is unexpected (for me) , don't know if i have reasons to be happy, but i am atm lol
Badass! You can leave the timings loose, because the Core i7 doesn't give two shits about memory latency. All the Ci7 cares about is bandwidth.
Next step, let's get that CPU clockspeed up. Boost the vcore to 1.375v and increase the CPU multiplier to 19x... That will give you 3.8GHz. If your temperatures are reasonable under load, try increasing the CPU multiplier to 20x. That will give you 4GHz.
Didnt know i7 didnt care that much about timing , its good to know
Why boost the vcore that much ? I was getting 3.6 (3.8 with turbo ON) with 1.2v
You plan on keeping turbo OFF and overclock as high as possible at max voltage ? (for me, 1.375 is effectively the maximum i would go while saying its safe from what i read) Or you just want me to try that to see if i can reach that than reduce voltage to try to find the minimum voltage windows-stable ?
3.8 was stable and temp were at 61C underload so i tried 4GHZ :
With VCORE of 1.375v , 4GHZ is very stable : 12 long runs of LinX, didnt go to 20 complete runs [its what i do with 'final test'] because i think i'll try to decrease voltage after. Temps were OK i think with that much voltage, 79C maximum on one of the core. It will go down if i manage to get stable at lower VCORE voltage.
So what now, try to decrease voltage ?
BTW, just for curiosity i tried to decrease the Uncore voltage of 1.35v (its shown in red in my bios screen) for 1.34 wich is not red, but i didnt even make it to windows background (freeze upon 'loading' screen). Get it back to 1.35v and it booted like a charm, seems like this voltage will remain that high to keep my OC stable :P
Still can't believe DDR-1600 is stable now with 8-8-8-19 tho
Go ahead and back the CPU voltage down to see if you can get it stable at 4GHz with a lower voltage, otherwise you're done. All the voltages I've had you use today are within Intel's safety specifications, including the vcore!
Comments
Thankx Thrax
@Leo : That's why i had fear to go over 1.65 too, but many people talk on the forum about a rule that says : if you are within .5v between VTT voltage and VDIMM , you're ok. I've seen a lot of people running at 1.8 VDIMM with increased VTT voltage, but in any case i've seen also people saying its dangerous to do so lol
In less than 6 month i'll change my video card too (9600 GT atm) and in a year or less i'll put in a SSD drive (Raptor 150GB atm) and then it will be perfect
Bclock frequency
CPU multi
QPI multi
Uncore multi
Memory multi
Memory voltage
CPU voltage
CPU load temps
CPU multi = 20
QPI multi = 36 (for QPI Link Speed of 6.51Ghz)
Uncore multi = 17
Memory multi = 8
Memory voltage = 1.64
CPU voltage = 1.20 in bios (~1.15 in windows, drop at 1.13 on full load)
CPU load temps = ~60C on RealTemp
TURBO = ON
HT = ON
C1E = ON
The weirdest thing that i didnt get an answer for tho, is why is i am able to get DDR-1448-7-7-7-21 STABLE , but cant boot up from cold state at that setting >.<
If i boot at 8-8-8 then go into bios and change it to 7-7-7 it will work for that time, then i close the computer, wait 3 secondes, boot again, and it wont boot and it will come back to last secure settings to boot.
As for overclocking, I won't have much advice until I know the QPI multiplier for sure.
There's a F8 Beta available tho.
I'll post my QPI as soon as i am at home.
Thankx for the help
QPI multi = 36 (for QPI Link Speed of 6.51Ghz)
QPI LINK SPEED :
Auto
X36
X44
X48
SlowMode
(I was wrong before in and older post when i was speaking about 36,74,96 lol)
I got the calculator from icrontic and found it weird too that the multi was not 18,20,24...
But its working the same way, at 36 i got 6.51Ghz, i think if i go X44 it will go up to 7.8 or something like that.
Auto is selecting 36 by default.
Complete article :
http://www.overclock3d.net/reviews.php?/cpu_mainboard/gigabyte_ex58-ud5_x58_motherboard/4
Alright, I'd like to test memory overclocking first. Set the following parameters:
Turbo mode and EIST: DISABLED
Bclk: 200MHz
Core voltage: 1.2v (Default 1.2v)
DRAM Voltage: 1.65v (Default varies)
Uncore voltage: 1.35v (Default 1.2v)
CPU PLL Voltage: 1.88v (default 1.8v)
CPU Multiplier: 12x (2400MHz)
QPI Link Speed: 36x (3.6GHz - or - 7.2GT/s)
Uncore Multiplier: 16x (3200MHz northbridge)
Memory Multiplier: 8x (DDR3-1600)
Loosen the memory timings as tight as they will go, and start testing with memtest. If it fails, reduce the uncore and memory multipliers by one step and try again.
A few notes: These settings should get DDR3-1600 capable memory running at DDR3-1600 speeds (200*8 = 1600). If this fails, your memory may simply be unable to rise to that challenge. Secondly, many Core i7s top out at 3.8GHz, even with the C0 stepping.
Thankx
DDR-1600 8-8-8-19 !!!
When it booted i was sure it wont pass memtest, but it did lol.
I first tried 8-8-8-24, after 20% memtest i tried 8-8-8-21, after 20% i went 19 and when it booted i tried any setting to reach 7-7-7, it wont boot at 7-7-7 even with 7-7-7-28.
So, my RAM is capable of 1600 at 8-8-8 which is unexpected (for me) , don't know if i have reasons to be happy, but i am atm lol
Next step ? :P
i'll post a screenshot in 1 min.
Next step, let's get that CPU clockspeed up. Boost the vcore to 1.375v and increase the CPU multiplier to 19x... That will give you 3.8GHz. If your temperatures are reasonable under load, try increasing the CPU multiplier to 20x. That will give you 4GHz.
Why boost the vcore that much ? I was getting 3.6 (3.8 with turbo ON) with 1.2v
You plan on keeping turbo OFF and overclock as high as possible at max voltage ? (for me, 1.375 is effectively the maximum i would go while saying its safe from what i read) Or you just want me to try that to see if i can reach that than reduce voltage to try to find the minimum voltage windows-stable ?
With VCORE of 1.375v , 4GHZ is very stable : 12 long runs of LinX, didnt go to 20 complete runs [its what i do with 'final test'] because i think i'll try to decrease voltage after. Temps were OK i think with that much voltage, 79C maximum on one of the core. It will go down if i manage to get stable at lower VCORE voltage.
So what now, try to decrease voltage ?
BTW, just for curiosity i tried to decrease the Uncore voltage of 1.35v (its shown in red in my bios screen) for 1.34 wich is not red, but i didnt even make it to windows background (freeze upon 'loading' screen). Get it back to 1.35v and it booted like a charm, seems like this voltage will remain that high to keep my OC stable :P
Still can't believe DDR-1600 is stable now with 8-8-8-19 tho
Grats on 4GHz.
If you ever come to Montreal, let me know and i'll pay you a beer or two
Thankx to all the others who did their best to help me too
I'll let you know my final vcore and temp as soon as i find it