My overclocking project.. the goal 3.6 Ghz

[Wolfhound]

Ok guys, I had been expending a lot of time (since monday) to overclock my q6600 since I got my new Asus ENGTX260 (216 cores) and new shiny 4Gigs OCZ Reaper PC-8500 kit (nice memory and fast), right now I had reached 9x356 (3200) at 1.35 Vcore (CPUZ says it is at 1.2526V) and memory is at 1068Mhz (have read in a lot of reviews that this memory can reach at least 1100Mhz), for now is stable, right now is running prime95 since I woke up at 5:00 AM, maximum temps are between 48ºC to 54ºC on load (prime 95 maximum heat test) and idle temps are between 27 to 32ºC, cooling is a thermalright ultra extreme with a Papst 1600rpm fan and Zalman STG1 goop (when I get home I´ll post pics of my rig). I had removed the heatsinks of north and southbridge and changed the goop with AS5 and adding a 40x40mm fan on northbridge (I had upped the voltage 0.1V on both for stability), my goal is to reach 3600Mhz without overvolting too much, below you´ll see a screenshot of my last stable result.

Do you have some advice or comment, please share it! ;)

http://img103.imageshack.us/img103/4706/3200fp3.jpg (http://imageshack.us)

[Toss3]

Just try putting your FSB to 400Mhz and lower your memory timings and start testing and tuning in different voltages. I couldn't get my q6600 to boot at 3,6Ghz unless I put the Vcore to 1,5V. :o

You could also try finding a guide on how to fix the vdroop on your motherboard(helped me alot with my P5K-E). It can usually be fixed by a simple pencil mod. :)

[Wolfhound]

Just try putting your FSB to 400Mhz and lower your memory timings and start testing and tuning in different voltages. I couldn't get my q6600 to boot at 3,6Ghz unless I put the Vcore to 1,5V. :o

You could also try finding a guide on how to fix the vdroop on your motherboard(helped me alot with my P5K-E). It can usually be fixed by a simple pencil mod. :)

Thanks Toss3 will try :thumbsup:

[walterman]

Adding a fan to the north bridge is a very good idea. It will allow you to raise the chipset voltage, to tight the memory timings & raise that FSB. Once you finish raising the FSB, try to lower the tRD. This will lower your memory latency, and raise the memory bandwidth.

[Bman212121]

Good luck, that's about all I can add. I can get mine booted at 3.6 but it takes about 1.5V just like Toss3's does.

The one advantage you have is your VID is much better than mine. I have a 1.3250 VID on my Q6600. For me anything above 3400 seems to start taking a lot of voltage to just get it booting. I haven't been able to get it stable at even 3.5 yet.


1.35 Vcore (CPUZ says it is at 1.2526V)

That's what's going to kill you trying to overclock. From what I vaguely remember seeing posted a few of the gigabytes boards have really bad vdrop/droop. With my q6600 set at 1.5V in the bios I only drop about .02V into windows, and maybe another .01V loaded. (Mine is pencil modded for vdroop.) I do remember looking for the pencil mod on that board for CaptN. Apparently it doesn't work from what I've seen, so it's best just to leave it be and see what you can get without modding it.

Also I would suggest just leaving your ram underclocked while you work in your cpu clock. Doing everything separately makes it a lot easier to get each part stable.

OT: I also found a little bit more info about vdroop modding. They suggest using a multimeter and testing the resistance on your mod. They prefer to leave just a tiny bit of vdroop so they aren't going too heavy with the graphite. http://www.ocforums.com/showthread.php?t=558347

I'm still trying to figure out the whole idea behind vdroop... I think it has something to do with varying voltage instead of varying amperage to keep the same amount of current. That being said it would seem like vdroop is actually bad because the less voltage the more amperage needed for the same amount of total power. Too many amps is how you would hurt the wiring, but on the same token lowering the voltage would also lower the total amount of power coming in provided the amps are constant. :banghead:
I just don't know what to make of it. This guys post makes a little sense though. http://www.hardwarecanucks.com/forum/overclocking-tweaking-benchmarking/10516-vdroop-2.html

[Wolfhound]

Adding a fan to the north bridge is a very good idea. It will allow you to raise the chipset voltage, to tight the memory timings & raise that FSB. Once you finish raising the FSB, try to lower the tRD. This will lower your memory latency, and raise the memory bandwidth.

I have added a 40x40x10mm papst fan @ 6300rpm to the northbridge now temps are 32ºC with +0.1V added

[Wolfhound]

Good luck, that's about all I can add. I can get mine booted at 3.6 but it takes about 1.5V just like Toss3's does.

The one advantage you have is your VID is much better than mine. I have a 1.3250 VID on my Q6600. For me anything above 3400 seems to start taking a lot of voltage to just get it booting. I haven't been able to get it stable at even 3.5 yet.



That's what's going to kill you trying to overclock. From what I vaguely remember seeing posted a few of the gigabytes boards have really bad vdrop/droop. With my q6600 set at 1.5V in the bios I only drop about .02V into windows, and maybe another .01V loaded. (Mine is pencil modded for vdroop.) I do remember looking for the pencil mod on that board for CaptN. Apparently it doesn't work from what I've seen, so it's best just to leave it be and see what you can get without modding it.

Also I would suggest just leaving your ram underclocked while you work in your cpu clock. Doing everything separately makes it a lot easier to get each part stable.

OT: I also found a little bit more info about vdroop modding. They suggest using a multimeter and testing the resistance on your mod. They prefer to leave just a tiny bit of vdroop so they aren't going too heavy with the graphite. http://www.ocforums.com/showthread.php?t=558347

I'm still trying to figure out the whole idea behind vdroop... I think it has something to do with varying voltage instead of varying amperage to keep the same amount of current. That being said it would seem like vdroop is actually bad because the less voltage the more amperage needed for the same amount of total power. Too many amps is how you would hurt the wiring, but on the same token lowering the voltage would also lower the total amount of power coming in provided the amps are constant. :banghead:
I just don't know what to make of it. This guys post makes a little sense though. http://www.hardwarecanucks.com/forum/overclocking-tweaking-benchmarking/10516-vdroop-2.html

Thanks for the info I´ll check it, yesterday I tried to overclock to 9x400 Vcore 1.4 Memory at 1:1 but the board doesn´t post regarding I upper the vcore to 1.5V that is the maximum vcore for quad core

[methimpikehoses]

Mine didn't like 3.6 unless the voltages were high also.

[Wolfhound]

Mine didn't like 3.6 unless the voltages were high also.

How much Juice you have to put on it?

[Bman212121]

Thanks for the info I´ll check it, yesterday I tried to overclock to 9x400 Vcore 1.4 Memory at 1:1 but the board doesn´t post regarding I upper the vcore to 1.5V that is the maximum vcore for quad core

Have you made sure that your board can run a 1600fsb? Another common thing to try is dropping the cpu multiplier way down, to like 7x, then upping the fsb to see how high it can go. You might need to change something related to the fsb to get it to post at 1600. On my 780i I need 1.4V into the North Bridge (SPP) to push it up there.

[methimpikehoses]

How much Juice you have to put on it?

Over 1.5V (400x9=3.6ghz), and I didn't kept it there long because it was too hot... idled at about 45C.

I currently have it set at 1.47V which reads 1.44V in everest. It's much cooler than 1.5V for some reason, at 440x8=3.52ghz. Still, it idles at 36C, which isn't great, but good enough.

[bob saget]

Over 1.5V (400x9=3.6ghz), and I didn't kept it there long because it was too hot... idled at about 45C.

I currently have it set at 1.47V which reads 1.44V in everest. It's much cooler than 1.5V for some reason, at 440x8=3.52ghz. Still, it idles at 36C, which isn't great, but good enough.

change sig :bleh:

swi rpl

Keep going wolf, i am cheering for ya :D

[Wolfhound]

change sig :bleh:

swi rpl

Keep going wolf, i am cheering for ya :D

Thank you :thumbsup: I´ll try today to push it a little more and cross fingers tht it boots, I have been searching th net for vdroop mods for P35-DS3R but found some bad histories about fried mobos and cpus so I think I´ll pass on it and try to find the correct Vcore to work OK

[$n][pErMan]

So apperently my P5B is not the only mobo that suffers from Vdroop... I thought they fixed that on newer boards. :(

[Misfire]

use intel burn test instead of prime95. It will push your cpu a lot harder but a 10 minute IBT = 40+ hours of prime95. Its the same program that intel uses to test their chips. Intitially, Prime95 reported no errors with my overclock when i ran it for 7+ hours, but running IBT for 5 minutes detected unstability, so I up'd my voltage by a little bit and am now IBT stable.

This should save you lots of hours.

http://www.xtremesystems.org/forums/showthread.php?t=197835


Benefits of using Linpack:
1. More accurate than Prime95 Small FFTs/Blend.
2. Takes less time to tell if your CPU/RAM is unstable than Prime95 (usually
something like 8 minutes Linpack vs 40 hours under Prime95).
3. Use the same stress-testing engine that Intel uses to test their products
before they are packed and put on shelves for sale.



Linpack by Intel(R) is an extremely stressful program that will
put even the most powerful X86/X64 CPU in the world at its knees. Load temp
under Linpack will be up to 22*C higher than the competing software Prime95.
This program will make usage of Linpack easier and more practical.

Keep in mind, use this program at your own risk. By using this program, you
agree that neither I nor Intel shall be responsible for including, but not
limited to: burned up CPU, fried motherboard, spontaneous room temperature
increase, hair loss, or mental stress.

[Bman212121]

Over 1.5V (400x9=3.6ghz), and I didn't kept it there long because it was too hot... idled at about 45C.

I currently have it set at 1.47V which reads 1.44V in everest. It's much cooler than 1.5V for some reason, at 440x8=3.52ghz. Still, it idles at 36C, which isn't great, but good enough.

Nice. I can't seem to find a sweet spot for this 780i board at 1600mhz+. It seems to have a lot of fsb holes and even though I can pass a prime95 session, it doesn't start up properly upon reboot. I just did some more testing and I think I have some solid numbers for what I wanted to run.

380 fsb (1520mhz effective)
cpu 3420mhz (380 x 9)
Not sure what the BIOS voltage is yet, as I still have it set to auto. The 780i seems to be pretty good about figuring out the voltages it needs to run the cpu, so I let it figure it out for me and then just manually set it. I'm showing 1.424 in windows using CPU-Z, so I just need to match it. I'm guessing somewhere around 1.46 in bios.
EDIT: CPU Core 1.45V set in bios.
FSB 1.4V
SPP 1.4V (Northbridge)
MCP 1.5V (Stock, southbridge)
SPP -> MCP 1.2V (Stock)

I have to go check what the rest is... :lol:

EDIT:
RAM 1.850V
Running underclocked at the moment. 1:1 fsb:dram ratio. 760mhz effective. I'll probably try another divider that is close to 1000mhz and see if it helps or hurts performance.

GTLVREF Lane 0: 00mv
GTLVREF Lane 1: 110mv
GTLVREF Lane 2: 00mv
GTLVREF Lane 3: 00mv

use intel burn test instead of prime95. It will push your cpu a lot harder but a 10 minute IBT = 40+ hours of prime95. Its the same program that intel uses to test their chips. Intitially, Prime95 reported no errors with my overclock when i ran it for 7+ hours, but running IBT for 5 minutes detected unstability, so I up'd my voltage by a little bit and am now IBT stable.

This should save you lots of hours.

http://www.xtremesystems.org/forums/showthread.php?t=197835

+1. It does load the CPU much better than P95 does. For now I've been using a combo of both. For testing different speeds I just run a quick P95 of like 5 - 10 minutes, until I get to where I want to test longer. Then I've been doing about an hour of P95, followed by 5 runs of linpack. Surprisingly using a blend on P95 I hit about 63C max, and 69C max after 5 runs on linpack. I know I've hit 71C before using higher voltage. P95 usually runs lower than that too if I don't let it go as long.

I've also been trying to make sure that it stays stable when running SLI, so I've been loading up crysis DX10 with 8xAA/16x AF VH settings and looping the CPU and GPU tests a few times. I'm pretty happy with these clocks and as long as everything stays running okay I think it's a pretty good 24/7 setup.


@wolf: Just keep going at it little by little. You will hopefully get there. I've been trying on and off for a while to get stable at around 3.4 and I think I have good numbers now. Voltage is a little lower than what I've tried with before, so it helps keep the heat down. I just checked what I would need to set it in bios, and I'm now running 1.45V in bios at 3420mhz and can pass an hour of P95 and 5 tests of linpack. What this means is that I probably don't need 1.5V to hit 3.6ghz, it's just that the boards fsb doesn't seem to want to stay stable at 1600mhz. If your board is happy with 1600 fsb, you might be able to reach it with less voltage.

[Wolfhound]

use intel burn test instead of prime95. It will push your cpu a lot harder but a 10 minute IBT = 40+ hours of prime95. Its the same program that intel uses to test their chips. Intitially, Prime95 reported no errors with my overclock when i ran it for 7+ hours, but running IBT for 5 minutes detected unstability, so I up'd my voltage by a little bit and am now IBT stable.

This should save you lots of hours.

http://www.xtremesystems.org/forums/showthread.php?t=197835

Nice find, I get it to 3600Mhz but is very hot for my taste (cores reached 65ºC, vcore on bios at 1,4250V)
Had tried this Intel burn test at 3600Mhz it gave me errors after 10 minutes, prime 95 had been running for 10 hours without error, nice find :thumbsup:

http://img168.imageshack.us/img168/5089/overclock3600zk7.jpg (http://imageshack.us)

[Wolfhound]

Nice. I can't seem to find a sweet spot for this 780i board at 1600mhz+. It seems to have a lot of fsb holes and even though I can pass a prime95 session, it doesn't start up properly upon reboot. I just did some more testing and I think I have some solid numbers for what I wanted to run.

380 fsb (1520mhz effective)
cpu 3420mhz (380 x 9)
Not sure what the BIOS voltage is yet, as I still have it set to auto. The 780i seems to be pretty good about figuring out the voltages it needs to run the cpu, so I let it figure it out for me and then just manually set it. I'm showing 1.424 in windows using CPU-Z, so I just need to match it. I'm guessing somewhere around 1.46 in bios.
EDIT: CPU Core 1.45V set in bios.
FSB 1.4V
SPP 1.4V (Northbridge)
MCP 1.5V (Stock, southbridge)
SPP -> MCP 1.2V (Stock)

I have to go check what the rest is... :lol:

EDIT:
RAM 1.850V
Running underclocked at the moment. 1:1 fsb:dram ratio. 760mhz effective. I'll probably try another divider that is close to 1000mhz and see if it helps or hurts performance.

GTLVREF Lane 0: 00mv
GTLVREF Lane 1: 110mv
GTLVREF Lane 2: 00mv
GTLVREF Lane 3: 00mv



+1. It does load the CPU much better than P95 does. For now I've been using a combo of both. For testing different speeds I just run a quick P95 of like 5 - 10 minutes, until I get to where I want to test longer. Then I've been doing about an hour of P95, followed by 5 runs of linpack. Surprisingly using a blend on P95 I hit about 63C max, and 69C max after 5 runs on linpack. I know I've hit 71C before using higher voltage. P95 usually runs lower than that too if I don't let it go as long.

I've also been trying to make sure that it stays stable when running SLI, so I've been loading up crysis DX10 with 8xAA/16x AF VH settings and looping the CPU and GPU tests a few times. I'm pretty happy with these clocks and as long as everything stays running okay I think it's a pretty good 24/7 setup.


@wolf: Just keep going at it little by little. You will hopefully get there. I've been trying on and off for a while to get stable at around 3.4 and I think I have good numbers now. Voltage is a little lower than what I've tried with before, so it helps keep the heat down. I just checked what I would need to set it in bios, and I'm now running 1.45V in bios at 3420mhz and can pass an hour of P95 and 5 tests of linpack. What this means is that I probably don't need 1.5V to hit 3.6ghz, it's just that the boards fsb doesn't seem to want to stay stable at 1600mhz. If your board is happy with 1600 fsb, you might be able to reach it with less voltage.

Thanks for the tips, I found stability putting the bios settings on auto (by SPD)

[methimpikehoses]

Nice work, Wolf! That's a low voltage for 3.6. :cool:

[Wolfhound]

Nice work, Wolf! That's a low voltage for 3.6. :cool:

Yep is a low voltage but i had to upper the FSB Voltage and the MCH voltage to +0.3V on the mobo

Now I´m trying to reduce it a little and checking stability, but I think 65-70ºC is too much for a 24/7 overclock, I need more fans on the case or a more powerful fan on thermalright Ultra 120 Extreme, now I have a Papst @ 1600rpm