Gigabyte P965-P35 X38 overclocking and BIOS tweaking Guide

I have the Crucial Ballistix 2GB (2 x 1GB) 240-Pin DDR2 SDRAM DDR2 1000 (PC2 8000) Dual Channel Kit Desktop Memory, similar to this: http://www.newegg.com/Product/Product.aspx?Item=N82E16820148069 (I can't find the exact model on newegg).

It's rated at 2.2V and I have confirmed that it is micron D9 but I can't overclock it beyond the specs given at 2.2V. Should I increase the voltage even more to overclock it?
 
I have some inexpensive DDR2-800 memory that I'm using in an overclocked E4300 system. I was wondering whether it would be better to use the memory synchronized with the CPU or at a higher ratio to get a higher clock speed. I also tested the effects of memory timings. I didn't run a complete (or completely scientific) test, but here are my results, which convinced me that a higher ratio and clock speed -- even with relaxed timings -- are better than synchronizing the memory and CPU at a lower clock speed with tighter timings (at least at the speeds/timings that I tested).

Test: SuperPi/Mod 1.5xs, 2M digits
CPU: Intel C2D E4300 3.15GHz, 9x350MHz bus
Memory: G.SKILL F2-6400CL5D-2GBNQ (2 x 1GB, dual channel)
Motherboard: GIGABYTE GA-965P-DS3 (F10 BIOS)
Ratios are from CPU-Z (1:1 = Gigabyte 2.0 multiplier, 4:5 = Gigabyte 2.5 multiplier)
Each test repeated at least 3 times and averaged (results varied by as much as 0.5 seconds for each test).

Ratio, Timings, Time
1:1 (700MHz), 4-4-4-12, 47.1s
1:1 (700MHz), 5-5-5-15, 47.9s
4:5 (875MHz), 5-5-5-15, 45.6s
4:5 (875MHz), 5-6-6-17, 46.0


I know this test has some shortcomings, but the results demonstrate that, for this application, the benefit of speeding up the clock by 25% overcomes by a wide margin the overhead of a non-1:1 ratio, even when the timings are relaxed by 25%.

Holding the timings constant, the overall system speed improvement for this test at the 25% higher memory speed was about 4.8%. I used SuperPi, even though it is single-threaded, because it is a combined CPU/memory test, not just a memory bandwidth test, so it would better simulate the type of applications, such as video editing, that I run. I have no idea what this would translate into in terms of the performance of other types of applications, such as games. And this test doesn't necessarily tell us anything about the relative speed increase between 800Mhz and 1066MHz because that would have to run at an entirely different ratio (probably worse, with more wait states). I'd be really interested to hear what others find when they use different test programs or faster memory.
 
mrh said:
I have some inexpensive DDR2-800 memory that I'm using in an overclocked E4300 system. I was wondering whether it would be better to use the memory synchronized with the CPU or at a higher ratio to get a higher clock speed. I also tested the effects of memory timings. I didn't run a complete (or completely scientific) test, but here are my results, which convinced me that a higher ratio and clock speed -- even with relaxed timings -- are better than synchronizing the memory and CPU at a lower clock speed with tighter timings (at least at the speeds/timings that I tested).

Test: SuperPi/Mod 1.5xs, 2M digits
CPU: Intel C2D E4300 3.15GHz, 9x350MHz bus
Memory: G.SKILL F2-6400CL5D-2GBNQ (2 x 1GB, dual channel)
Motherboard: GIGABYTE GA-965P-DS3 (F10 BIOS)
Ratios are from CPU-Z (1:1 = Gigabyte 2.0 multiplier, 4:5 = Gigabyte 2.5 multiplier)
Each test repeated at least 3 times and averaged (results varied by as much as 0.5 seconds for each test).

Ratio, Timings, Time
1:1 (700MHz), 4-4-4-12, 47.1s
1:1 (700MHz), 5-5-5-15, 47.9s
4:5 (875MHz), 5-5-5-15, 45.6s
4:5 (875MHz), 5-6-6-17, 46.0


I know this test has some shortcomings, but the results demonstrate that, for this application, the benefit of speeding up the clock by 25% overcomes by a wide margin the overhead of a non-1:1 ratio, even when the timings are relaxed by 25%.

Holding the timings constant, the overall system speed improvement for this test at the 25% higher memory speed was about 4.8%. I used SuperPi, even though it is single-threaded, because it is a combined CPU/memory test, not just a memory bandwidth test, so it would better simulate the type of applications, such as video editing, that I run. I have no idea what this would translate into in terms of the performance of other types of applications, such as games. And this test doesn't necessarily tell us anything about the relative speed increase between 800Mhz and 1066MHz because that would have to run at an entirely different ratio (probably worse, with more wait states). I'd be really interested to hear what others find when they use different test programs or faster memory.
Click to expand...

yep, cpu MHz are more powerfull than memory MHz. With C2D's the best stragety for advanced users is to max out the cpu speed anyway you can (loose timings) and let the memory fall where it can. And people with 1066 or other high speed memory can see performance gains using a mulitplier other than 1:1 to raise the memory speed closer to rated specs. I emphasize the 1:1 for the stability when starting off, and if I ever get the advanced part of the quide written I will suggest doing just what you did. You can read all the guides in the world but they are just guides, not rules and methodical testing is the way to determine what works best with your machine.
 
ok i have tried everything to get this board to do over 400fsb , an its just not doing it for me ...

i have a E6420 CPU @ 3.0Ghz orthos stable , im using

THIS Ram - http://www.newegg.com/Product/Product.asp?Item=N82E16820220144 (patriot eXtreme performance ddr2 800mhz ram (4-4-4-12)

heres a screen shot of cpu - z , the important stuff to know


I have tried F11/F12 & currently using the F10 bios , my ram isnt even running at spec so i don't think its my memory ...

what happens when i set it for 400FSB ( i bump the volts up to say 1.5 (yea its high but i figure hey it should boot it than i can mess with it from there) , save the bios than it just goes into a cycle - - power on , off , on , off , on an it goes on until i flip the psu switch an clear cmos ...

anyways heres my config

-Intel Core 2 Duo E6420
-Artic Freezer 7 PRO (56FULL LOAD @ 3.0Ghz)
-Gigabyte DS3 V 3.3
-Patriot Extreme Memory ddr2 800mhz as in link above
-OCZ 700watt psu
-EVGA 8800GTS SC
 
RedLancerOz said:
ok i have tried everything to get this board to do over 400fsb , an its just not doing it for me ...

i have a E6420 CPU @ 3.0Ghz orthos stable , im using

THIS Ram - http://www.newegg.com/Product/Product.asp?Item=N82E16820220144 (patriot eXtreme performance ddr2 800mhz ram (4-4-4-12)

heres a screen shot of cpu - z , the important stuff to know


I have tried F11/F12 & currently using the F10 bios , my ram isnt even running at spec so i don't think its my memory ...

what happens when i set it for 400FSB ( i bump the volts up to say 1.5 (yea its high but i figure hey it should boot it than i can mess with it from there) , save the bios than it just goes into a cycle - - power on , off , on , off , on an it goes on until i flip the psu switch an clear cmos ...

anyways heres my config

-Intel Core 2 Duo E6420
-Artic Freezer 7 PRO (56FULL LOAD @ 3.0Ghz)
-Gigabyte DS3 V 3.3
-Patriot Extreme Memory ddr2 800mhz as in link above
-OCZ 700watt psu
-EVGA 8800GTS SC
Click to expand...

My system with E4300 absolutely cannot reach 400 Mhz either even when I reduce the multiplier so instead, I just set it to 356 Mhz x9 = 3.304 Ghz.
 
I am having a bit of trouble. I can't boot with an fsb of 425 or higher on my e6420. On a 7 multiplier i can do 475fsb (1.55v), so i know it is not my ram holding me back. I tried incrasing the voltage, but it just doesn't work, i also loosened the timings. Is there anything i can do?

My specs are:
DS3 rev 3.3 F12 Bios
e6420 cooled by an Thermalright Ultra-120 eXtreme
G.Skill 2*1gb HZ which proved stable at 950mhz (haven't tried anymore)

Thanks for the help, or any suggestions u have!

ps i idle at around 32, and load under about 47, with 1.56, should i try higher, and is anything safe as long as it doesn't get too hot?

EDIT: I loosened the timings to 6-6-6-16 and it booted at 426, is that bad? should i rather go with 4-4-4-12 and 424 fsb?
with 6-6-6-16 and 426 it had pretty bad stability, so my good stable right now is 420*8, with 4-4-4-12 timings, what can i do?

it is looking giid at 424*8= 3392 at 1.5v and 4-4-4-12 timings, is there anyting u guys can think of to get it to 3400??? anything i should change
 
well i managed to pull off 400fsb with 7x400 (2.8Ghz) , memory isnt the culprit must be the chip :( , will even do higher so im guessin my chips max is 3ghz
 
Do I really need to add a fan to my northbridge even though the FSB is at 356 Mhz right now? It does feel a little hot though.
 
i put a 80mm fan sitting ontop of my video card blowing towards the northbridge , that thing gets SCORCHING hot ... without it , don't know why most board manufacturers dont put active cooling there :mad:
 
I've noticed that if I plug in my external HDD, my system has trouble starting up when overclocked.

Without the HDD, my E4300 is stable at 3.204 Ghz at 1.4V. When I connect the HDD before starting up the system. It cycles on and off a few times and reverts back to the original 1.8 Ghz. Does anyone know why it does this?
 
i would put a fan on it, it only costs about $3 and u can just hot glue it on, it is easy and better safe then sorry
 
petern said:
I've noticed that if I plug in my external HDD, my system has trouble starting up when overclocked.

Without the HDD, my E4300 is stable at 3.204 Ghz at 1.4V. When I connect the HDD before starting up the system. It cycles on and off a few times and reverts back to the original 1.8 Ghz. Does anyone know why it does this?
Click to expand...

No idea why, but you are not the first person to have issues, the board seems not to like USB drives. I dont use one so I have no way to test etc.
 
and is anything safe as long as it doesn't get too hot?
Click to expand...

no, over voltage stress can cause transistor junctions to just "short out" its an extremely complicated and technical topic. Vcore absolute max is 1.55 volts. You are killing your CPUs reliably and service life. Its your stuff do what you want.
 
petern said:
I have the Crucial Ballistix 2GB (2 x 1GB) 240-Pin DDR2 SDRAM DDR2 1000 (PC2 8000) Dual Channel Kit Desktop Memory, similar to this: http://www.newegg.com/Product/Product.aspx?Item=N82E16820148069 (I can't find the exact model on newegg).

It's rated at 2.2V and I have confirmed that it is micron D9 but I can't overclock it beyond the specs given at 2.2V. Should I increase the voltage even more to overclock it?
Click to expand...

I would not unless I put a fan on it, OCZ and Corsair sell coolers made for memory. Dont beleive me, go to memory formum and look at the posts of burnt up micron D9.
 
Does anyone, by any chance, know if either the F11 or F12 BIOS updates contain fixes for the Intel C2D bugs mentioned on Slashdot and on http://marc.info/?l=openbsd-misc&m=118296441702631 ?

They both came out around the time, or after, the bugs were announced, but the descriptions on Gigabyte's web site don't mention it. Perhaps one of you has knowledge beyond what is there?

Microsoft, by the way, has released an update: http://support.microsoft.com/?kbid=936357
 
mrh said:
Does anyone, by any chance, know if either the F11 or F12 BIOS updates contain fixes for the Intel C2D bugs mentioned on Slashdot and on http://marc.info/?l=openbsd-misc&m=118296441702631 ?

They both came out around the time, or after, the bugs were announced, but the descriptions on Gigabytes web site don't mention it. Perhaps one of you has knowledge beyond what is there?

Microsoft, by the way, has released an update: http://support.microsoft.com/?kbid=936357
Click to expand...

that /marc.info link info is the opinion of (IMO) a complete idiot, Intel have been releasing chip errata (bugs) since the 8088 processor and they all contain bugs and I have seldom seen such an overblown analysis and FUD (worthy of the Inq) in a public post. Pure crap.

Your machine has not burst into flames, right ? , Software runs fine , right ? If it was so freaking horrible would you have heard about it before this , right ? .

To address your question directly, I am not sure, but I think the latest bios mentioning an i/o chip update does because it makes no sense otherwise. Obviously this is nothing but a guess. A major issue to me, you noticed it also, and more important than the subject one, is the crappy lack of information we are fed over bios update contents. I have gone to Gigabytes support site and complained, I suggest you do too, put in both a complaint to tech support and non-tech support (marketing) saying that the lack of details on what is changed in new bios releases makes you unlikely to purchase their product again. Every Gigabyte owner needs to do that.

We don't know, just download the MS update and install it if it will make you feel better.

I have not had any issues but I run a gen 1 CPU.

Here is more:
http://www.hardforum.com/showthread.php?t=1202416
 
Now that I have a better idea of what my memory can handle, I've started raising the FSB for an overclock.

FSB: 373
Memory multiplier: 2
vcore: 1.35
Timings: 4-4-3-8
Overclock is 2.975 GHz

I tried raising the FSB up to 390, as I am aware of the strap. The board wouldn't boot with those settings. Now, at 373, I'm very stable. Temperatures are very good, with 24/28 at idle, and 35/38 after two hours of gaming (or 10 minutes of Orthos, whichever comes first!). Temps in Orthos don't rise after the first 5 minutes. I've a tuniq tower, and haven't bothered to increase the rpms past the minimum.

At the 390, the board just sat there. No noise. Nothing. Ooops! Well, I powered off, cleared the CMOS, and rebooted. This time, I didn't have to restore the BIOS from the harddrive, so I was able to load my default profile--tweaked--but no oc'd.

If I've been reading the different threads correctly, I'd probably need to relax my timings a bit, though perhaps I ought to raise vcore by a small amount, too.

My goal is to get above a 3.0GHz oc.

--ceolstan
 
ceolstan said:
Now that I have a better idea of what my memory can handle, I've started raising the FSB for an overclock.

FSB: 373
Memory multiplier: 2
vcore: 1.35
Timings: 4-4-3-8
Overclock is 2.975 GHz

I tried raising the FSB up to 390, as I am aware of the strap. The board wouldn't boot with those settings. Now, at 373, I'm very stable. Temperatures are very good, with 24/28 at idle, and 35/38 after two hours of gaming (or 10 minutes of Orthos, whichever comes first!). Temps in Orthos don't rise after the first 5 minutes. I've a tuniq tower, and haven't bothered to increase the rpms past the minimum.

At the 390, the board just sat there. No noise. Nothing. Ooops! Well, I powered off, cleared the CMOS, and rebooted. This time, I didn't have to restore the BIOS from the harddrive, so I was able to load my default profile--tweaked--but no oc'd.

If I've been reading the different threads correctly, I'd probably need to relax my timings a bit, though perhaps I ought to raise vcore by a small amount, too.

My goal is to get above a 3.0GHz oc.

--ceolstan
Click to expand...

bah you deserve some fun
Vcore 1.375
set timings to 4 4 4 12
set FSB to 410 (to get past that strap)
cpu mulit to 8x

see if it will boot, if not same but 5 5 5 15 (set Vdimm memory voltage to upper range of manuf specs, you can test and see if you can back off later)
 
Actually, I've been having a blast!

I relaxed the timings to 4-4-4-12, and bumped vcore to 1.375. I set the fsb to 410. The cpu multi has been 8. vdimm is at 2.2 as per SuperTalent's specs for overclocking RAM. I booted into windows no problem, but Orthos started beeping within 15 seconds. Temps at idle were 25/29, and didn't seem to budge very high. However, the system was clearly unstable, lol!

I relaxed timings to 5-5-5-15, rebooted just fine. Orthos ran for a good 20 seconds before beeping at me. I bumped up the vcore to 1.4, and ran 30 seconds, but I could see the difference in temperatures just from raising the vcore. I'm not sure that vcore is the issue here. I will try bumping up a bit higher, as it seems to be the case that the strapping issue can be somewhat variable according to the board. If that doesn't work, I'll see what happens if I back down. From my experience with the memory testing, it seems as if a few MHz goes a long way toward determining whether or not something is stable.

I could also play around with the multiplier, too. :)

--ceolstan

Edit:

I dropped the FSB by 3 while keeping the relaxed timings. Orthos ran without a hitch for 10 minutes. Temperatures were never above 40. RAM was definitely warm, but I could keep my fingers on it for a good 10+ seconds, probably longer, but certainly not something that'd give me 3rd degree burns. I kept the voltage at 1.37. With vdroop, it was at 1.35. Tomorrow, I'll see what happens when I tighten the timings a bit. I'll also see what happens if I raise the FSB to 413.

It was a thrill to see what looked like the potential for a stable oc around 3.2GHz. I know that's a mild overclock to a lot of folks here, but wow! That was fun to see!
 
Is anyone getting halfway decent ram clocks on their rev 1 c1 DS3/S3? I cannot break 1066mhz cas 5 in dual channel and 1100mhz cas 5 single channel on my Rev 1 S3. I've tried every bios version, timings voltage up to 2.7V ect. and no matter what I cannot break the wall. I've got 2x512mb PC8000 Ballistix that have done 1350mhz STABLE on a buddies P35 board. I've got the vmch hardmod and the vdimm hardmod on my board, vmch up to 1.74V doesn't help nor high voltages. The sticks will do 1060mhz 4-4-4-8 w/ 2.3V on this board and 1000mhz 3-4-3-10 w/ 2.81V (yes I know, single stick no more than 15 min w/ 80mm fan blowing negative ambient air over it.) I'm not sure if I've missed something really simple or the board just won't clock any higher (which I just can't let myself believe, there's got to be something!)

Also, I can't break 470mhz FSB on this board regardless of voltage or multiplier. I've got an extremely good L628B E6400 that will do 3400mhz w/ 1.22V(actual voltage w/ multimeter.)

The vdrop (not vdroop) is horrendous on this board. Set 1.35V in bios 3.4ghz, speedfan reads 1.32V (NEVER TRUST SOFTWARE FOR VOLTAGES PLEASE, IT'S NEVER CORRECT), actual voltage measured with multimeter reads around 1.26V idle and 1.24V under load.

When trying to set vcore above 1.59V, the board won't boot. Yes I understand that more vcore is bad for the chip, but 1.59V reads around 1.46V idle and 1.43V load. All power saving options are always disabled.

I've been probing around trying to find the vFSB mod for this board but I can't even find the read point. The closest I've found was a readpoint that read out a voltage of 1.20V, which didn't change when I increased the vFSB, which leads me to believe it's the secondary SB voltage.

Thanks guys!

P.S Incredible article Bill, I'm sure it's been a lifesaver for 100's of people! Taught me a thing or two too!
 
:) Just wanted to thank Bill Parrish for the excellent guide and advice. Finally completed my first build (for my daughter) with the S3 board with the slight OC you recommended and it runs great. Very much appreciated by both of us.
Will now be waiting for Bill's Penryn MB guide this Fall/Winter when I'm ready to build for myself ;) :D
 
The vdrop (not vdroop) is horrendous on this board. Set 1.35V in bios 3.4ghz, speedfan reads 1.32V (NEVER TRUST SOFTWARE FOR VOLTAGES PLEASE, IT'S NEVER CORRECT), actual voltage measured with multimeter reads around 1.26V idle and 1.24V under load.
Click to expand...

Where do you measure this? I want to actually measure mine myself instead of relying on hardware, but just don't know where to do this.
 
The proper way to do it is to look at the socket pinout and find Vcc and Vss and backtrack using continuity tests to a point just outside the socket but being absolutely sure the point is after the inductors in the VRD. I am not sure how the measurement mentioned in the quote was made but I find it suspect.

You want the Datasheet titled Intel® Core™2 Extreme Processor X6800Δ and Intel® Core™2 Duo Desktop Processor E6000 and E4000Δ Sequences

found here:
http://www.intel.com/design/processor/datashts/313278.htm

In section 3.1.8 you will find a picture of how the coordinate system for the signals is laid out. In section 4 will be a pictorial of what signal is where in relation to the socket.



Start picking a Vcc and/or Vss place your meter on a resistance scale (and with one lead on the CPU socket pin) find it on the backside with a continuity check (zero resistance) and start looking at where traces go and hope to find it on the topside. Or you can just have your board out of the case standing on its edge and use the pads you have identified on the bottom side. If you short something while power is on, well, you know.

It might a tedious search, you must use Vss for ground, a common ground on the board will not do.

Now why did I go through all that ? Because unless you do it yourself or someone who "says" this "place" is the point to measure had better detail exactly how they found the point or its BS.


Another interesting point would be the accuracy of the meter used, even a DVM has usually a 1/2 digit error in the last place so a meter that reads to .xx on a say 12 volt scale, for instance 1.35V the "5" would be very suspect. I cant remember how to translate that into a percent,
 
First of all, thanks to Bill - I'm chugging along happily at 400MHz FSB and 3.2GHz on my E6400.

I noted your remarks on Gigabyte's reported CPU temp being low. I believe this is a BIOS problem that has been fixed in F11 (for DQ6). Whereas before, the temperature reported by ET5 was always 5-10C cooler than Intel TAT or CoreTemp, the Gigabyte temperature reported by ET5 is now 5-10C HOTTER than Intel TAT or CoreTemp.

The reason why I believe this is a BIOS problem (now fixed in F11 for DQ6) is that I had been using Intel QST for fan control (not ET5) and noticed that my fan was running consistently faster with F11 (I have a loud CPU fan). The fan speed was also jumping around less. Installing ET5 verified this.

I don't know what was wrong with the old BIOS, but I also remember seeing the Gigabyte reported temperature jumping around a lot. It would read 45 one second and then 53 the next second and then back to 45. This is also fixed and the temperatures fluctuate normally - about 2-3C range of fluctuation.

So for those who like the custom CPU fan control of ET5, there is no longer a reason not to use it!

P.S.
board: 965P-DQ6 v3.3
firmware: F11
ET5: May 2007 dated download
 
I guess I have an E6600 CPU that just doesn't OC as well as others. Using stock HSF, I was able to get a stable overclock at 9x366Mhz but it required me putting a 50mm fan on the MCH and overvolting the CPU to 1.475v. At that voltage, the CPU, while stable, ran hotter than I like - my little room was even noticeably hotter! I went back to 9x333Mhz at 1.375v and am happy there.
 
BillParrish said:
The proper way to do it is to look at the socket pinout and find Vcc and Vss and backtrack using continuity tests to a point just outside the socket but being absolutely sure the point is after the inductors in the VRD. I am not sure how the measurement mentioned in the quote was made but I find it suspect.

You want the Datasheet titled Intel® Core™2 Extreme Processor X6800Δ and Intel® Core™2 Duo Desktop Processor E6000 and E4000Δ Sequences

found here:
http://www.intel.com/design/processor/datashts/313278.htm

In section 3.1.8 you will find a picture of how the coordinate system for the signals is laid out. In section 4 will be a pictorial of what signal is where in relation to the socket.



Start picking a Vcc and/or Vss place your meter on a resistance scale (and with one lead on the CPU socket pin) find it on the backside with a continuity check (zero resistance) and start looking at where traces go and hope to find it on the topside. Or you can just have your board out of the case standing on its edge and use the pads you have identified on the bottom side. If you short something while power is on, well, you know.

It might a tedious search, you must use Vss for ground, a common ground on the board will not do.

Now why did I go through all that ? Because unless you do it yourself or someone who "says" this "place" is the point to measure had better detail exactly how they found the point or its BS.


Another interesting point would be the accuracy of the meter used, even a DVM has usually a 1/2 digit error in the last place so a meter that reads to .xx on a say 12 volt scale, for instance 1.35V the "5" would be very suspect. I cant remember how to translate that into a percent,
Click to expand...

I've triple checked with 3 different multimeters and all of them read the same voltage. There's atleast 10 different vcore readpoints around the cpu socket that all read within .01V of each other. I'll throw up a diagram showing markings of the vcore readpoints in a few, working on fixing my gf's old pc (she killed my 3rd and final laptop.) Also, if the last digit on the dmm has an error margin, it would be the 10^-3 (.001) decimal place. Most multimeters I've used have 2v/20v/200v/500v DCV adjustments.
 
NHutch said:
I've triple checked with 3 different multimeters and all of them read the same voltage. There's atleast 10 different vcore readpoints around the cpu socket that all read within .01V of each other. I'll throw up a diagram showing markings of the vcore readpoints in a few, working on fixing my gf's old pc (she killed my 3rd and final laptop.) Also, if the last digit on the dmm has an error margin, it would be the 10 to the negative 3rd power (.001) decimal place.
Click to expand...

I knew that $12 sperry DVM was a mistake to buy.

I read your post again and I am have no answer for you 470mhz FSB
on a $130 board ain't bad. The 500+MHz reports are typically suicide runs with one stick of "golden" memory etc on a cherry picked board. My DQ6 with extensive rework of the Heatpipe and extra cooling for the chipset is iffy over 475, will not load windows over 482ish, and the fastest I ever got it to boot was 492 and I have tried everything but hardware mods and phase change. I does appear your board has an appreciable Vdrop, but nothing you cannot adjust for as the bios will let you set crazy bios voltages. I wonder if the VRD is saturated, it just cant put out any more current. Anyway, its all about the MCH in my opinion, even cooling it with fan etc, an internal hot spot due to it being so heavily OCed and game over.

One trick is to use a can of compressed air and try supercooling the different chips, dont forget the PLL chip, and see what happens, it will help identify a weakness.
 
BillParrish said:
I knew that $12 sperry DVM was a mistake to buy.

I read your post again and I am have no answer for you 470mhz FSB
on a $130 board ain't bad. The 500+MHz reports are typically suicide runs with one stick of "golden" memory etc on a cherry picked board. My DQ6 with extensive rework of the Heatpipe and extra cooling for the chipset is iffy over 475, will not load windows over 482ish, and the fastest I ever got it to boot was 492 and I have tried everything but hardware mods and phase change. I does appear your board has an appreciable Vdrop, but nothing you cannot adjust for as the bios will let you set crazy bios voltages. I wonder if the VRD is saturated, it just cant put out any more current. Anyway, its all about the MCH in my opinion, even cooling it with fan etc, an internal hot spot due to it being so heavily OCed and game over.

One trick is to use a can of compressed air and try supercooling the different chips, dont forget the PLL chip, and see what happens, it will help identify a weakness.
Click to expand...

Yea I've done some work with my northbridge, cleaned to perfection (goo gone, qtips + 91% iso alcohol, Arcticlean 2 Thermal surface purifier, AS5) also done to the southbridge, have 4 fans laying on top of the board (1 over memory, nb, 2 around the cpu socket.)

470mhz FSB has been my absolute max I can post with, boots into windows fine but vaguely stable. Another 1mhz and the board will not post. Anything over 1066mhz on the ram in dual channel will fail during post on the memory testing stage. The voltages aren't all that crazy on this board, sure you can do 1.6V, 1.8V and 2.0V but there's nothing inbetween. Your DQ6 has MUCH better voltage regulation too. I've tried supercooling my board in the winter but I didn't have the ram to break 465mhz FSB (my previous best on normal ambients was 440mhz with the poor clocking ram.) Opened the window in february on a really cold night (it was around -10F!) Room temperature was hovering around 0F and somehow my hoses didn't completely freeze solid. They felt like they were but I could still see the water flowing.

Is your nb on water? I've been thinking about putting mine but I need to buy a nb block with the proper mounting, I don't think my extra block (apogee GT) will mount properly!

How cold is compressed air? That's something that really interests me.

Jeez I just went out to smoke a butt before finishing this, was pacing back and forth the whole time and all the sudden when I turned around a skunk walked right by me! Not even a foot away from me just walked right by like I wasn't even there! This would be one of the worst nights to get sprayed by a skunk, girlfriend is sleeping at her moms tonight and I don't have my car or any tomato juice!
 
I was not aware of the DS3 voltage adjustment limitation, the manual does not go into details, I missed that.

No I didnt WC the MCH/Northbirdge becasue the heatpipe also cools the SB and the voltage regulator MOSfets I would have had to lay out a fair amount of cash to recool it all the way I felt it should be done so I did basically what you did and replaced the crappy push pins with "real" hardware and added some fans.
http://mysite.verizon.net/ressdxka/sitebuildercontent/sitebuilderpictures/computer003.jpg

As you can see I went with an el-cheepo WC setup to save money and time, it actually works pretty well but I figured also that putting my video card in the loop was probally all the extra it could handle. Damn now you got me thinking, I didnt know the WC was going to work as well as it did and I probally should consider it, maybe with the X38.

I havent used it in a while so I am not sure, pretty darn cold, below freezing, its mainly due to the physics of it changing pressure as it comes out of the can. They even sell a product, or used to, called "freeze it" or "Freeze off" for electronics technicians, it used to be the fast way to fix TVs etc back in the day people actually did that kind of thing. Take a can and start spraying parts and 1/2 the time you would hit something and the unit would start working, saved tons of time trouble shooting circuit by circuit. If you have an air compressor with a dryer on it, it should work too, not quite as well but good enough.

Yikes, I too live in the "wilds", our problem is hungry bears, I think I would almost rather be eaten than sprayed, never had it happen but I have smelled "strong" skunk before and man that is one awfull smell.

LOL quck google seems to indicate my memory of brand names was way off, hair spray and wart remover, I love being senile, every day is like brand new.

Apparently any canned air will do and it all in how you hold the can.

http://en.wikipedia.org/wiki/Canned_air
 
Update:

I bumped vcore to 1.45 per UncleWebb's advice, and got Orthos to run stable.

I then tried 9 x 390 = 3.510 and was able to get into Windows. Orthos ran for about 15 minutes or so, but produced a Delta Temp of 11C and as I went to get a drink, I heard the PC rebooting. Think it crossed the heat threshhold?

Should I lap my CPU and heatsink? That temp seems way too high for a Big Typhoon VX?

I've had no luck in any 8x multiplier, not even 8x400. So I am curious what else I can try to get about 3.33GHz?

Thanks guys!
 
JStudrawa said:
Update:

I bumped vcore to 1.45 per UncleWebb's advice, and got Orthos to run stable.

I then tried 9 x 390 = 3.510 and was able to get into Windows. Orthos ran for about 15 minutes or so, but produced a Delta Temp of 11C and as I went to get a drink, I heard the PC rebooting. Think it crossed the heat threshhold?

Should I lap my CPU and heatsink? That temp seems way too high for a Big Typhoon VX?

I've had no luck in any 8x multiplier, not even 8x400. So I am curious what else I can try to get about 3.33GHz?

Thanks guys!
Click to expand...

Lapping the cpu and heatsink will drop your temps a couple of deg C at least. I dont know if it will help you max OC much you are pushing that chip pretty hard. I will help stabilize an OC that is iffy, probally.

If you have not reworked or enhanced the cooling on the MCH/northbridge you are not going to be able to do high overclocks. The Memory Controller Hub becomes overclocked just like the cpu when you set high FSB values. Without a fan on the heatsink it gets hot as fire and will be unstable. If you rework your main cpu heatsink, think about removing the MCH heatsink, redoing the thermal compound under it, replacing the hardware with #4-40 screw nuts and lock washers. Be sure to use a nylon washer under the head of the screw on the backside of the board and dont tighten too tight, smartly snng will do, just compress the split lockwasher or you might crack the core. A 40 or 50mm fan can be hot melt glued or sheet metal screws can be forced down in the fins to hold a fan on. Or replace the existing MCH heatsink with an aftermarket one the HR-05 is popular. You have reached the OCing area (greater than 3.2GHz) where "tweaks" like the above are required to get much higher and be stable. You are in an area of diminishing returns and should consider if another 100MHz of CPU speed will be worth all the work required. Of course its "fun" to us crazies. But you need to do something with the MCH.

http://www.hardforum.com/showthread.php?t=1194043&highlight=DS3+NB
 
BillParrish said:
Lapping the cpu and heatsink will drop your temps a couple of deg C at least. I dont know if it will help you max OC much you are pushing that chip pretty hard. I will help stabilize an OC that is iffy, probally.

If you have not reworked or enhanced the cooling on the MCH/northbridge you are not going to be able to do high overclocks. The Memory Controller Hub becomes overclocked just like the cpu when you set high FSB values. Without a fan on the heatsink it gets hot as fire and will be unstable. If you rework your main cpu heatsink, think about removing the MCH heatsink, redoing the thermal compound under it, replacing the hardware with #4-40 screw nuts and lock washers. Be sure to use a nylon washer under the head of the screw on the backside of the board and dont tighten too tight, smartly snng will do, just compress the split lockwasher or you might crack the core. A 40 or 50mm fan can be hot melt glued or sheet metal screws can be forced down in the fins to hold a fan on. Or replace the existing MCH heatsink with an aftermarket one the HR-05 is popular. You have reached the OCing area (greater than 3.2GHz) where "tweaks" like the above are required to get much higher and be stable. You are in an area of diminishing returns and should consider if another 100MHz of CPU speed will be worth all the work required. Of course its "fun" to us crazies. But you need to do something with the MCH.

http://www.hardforum.com/showthread.php?t=1194043&highlight=DS3+NB
Click to expand...

I've got a fan on the MCH, but I'll look into redoing it's seating and thermal paste. I also bumped voltage there to +0.2 and on memory to +0.1.

yeah, I figure that I am nearing the limit of my setup, but am very very happy with where she is now anyway. And you're right, it's just fun at this point to see how much more I can coax out of her :)
 
I think the whole MCH cooler needs to go on anything other than the DQ6. The stock coolers are aluminum, which as we all know stinks. Adding some weight would help bring temps down too (that's why air cpu coolers are working pretty good now, the best ones weigh around 2LBS!!!!)
 
Got her to run 9x380 = 3.42GHz now. But not Orthos stable, she reboots after 15 minutes. But I have no issues running games at all, or anything else.

Heading to Frys tomorrow to get a better MCH cooler I think. Is the Zalman blue thing better than a 50mm fan on the stock MCH heatsink?
 
JStudrawa said:
Got her to run 9x380 = 3.42GHz now. But not Orthos stable, she reboots after 15 minutes. But I have no issues running games at all, or anything else.

Heading to Frys tomorrow to get a better MCH cooler I think. Is the Zalman blue thing better than a 50mm fan on the stock MCH heatsink?
Click to expand...

I dont know, maybe someone who has tried one will check in, If the Zalman will take a fan it might be worth a try, but if its just one of those fan shaped ( as in an old hand fan, shit no under the age of 30 has probally ever even seen what I am trying to talk about :D ) , well anyway if it wil take a fan or you can get a fan blowing right on it it would probally be ok. The HR-05 seems to be the winner, with a fan, if you have room to install it.

since you are going for every little bit, there is some new diamond filled (really, tiny industral diamonds) HS compound that is susposed to be better than AS5. /shrug
 
BillParrish said:
I dont know, maybe someone who has tried one will check in, If the Zalman will take a fan it might be worth a try, but if its just one of those fan shaped ( as in an old hand fan, shit no under the age of 30 has probally ever even seen what I am trying to talk about :D ) , well anyway if it wil take a fan or you can get a fan blowing right on it it would probally be ok. The HR-05 seems to be the winner, with a fan, if you have room to install it.

since you are going for every little bit, there is some new diamond filled (really, tiny industral diamonds) HS compound that is susposed to be better than AS5. /shrug
Click to expand...

I haven't heard of that yet. I always thought Silver was the best thermal conducter of heat. I have a feeling that stuff could do more harm than good, instead of filling the microscopic gaps it CREATES more of them... Diamonds are the absolute hardest thing in the world...
 
You must log in or register to reply here.
Back
Top