Can't get my Q6600 Higher than 3.00GHz

D

Daemas

Gawd
Joined
Apr 11, 2009
Messages
791
Title says it all. I can't for the life of me get my Q6600 above 3.00GHz IBTv2.4 Maximum stable.

Components:
GIGABYTE GA-EP45T-UD3LR Rev1.0 with latest (F5) BIOS
Intel Q6600 B3 VID 1.3125
GeIL 4GB (2 x 2GB) 240-Pin DDR3 SDRAM DDR3 1333
Enermax Galaxy Evo 1250w
Prolimatech Megahalems Rev. B with 2x Yate Loons Mediums in Push/Pull.
Termalright HR-05/IFX + Panaflow 80mm fan

Temps are 23c-24c idle and 56c-57c load @ 23.2c ambient


here are my settings:
Code: 
CPU BClock and Multi: 333x9
MCH Frequency Latch: 333MHz
System Memory Multiplier: 4.00B 

MEMORY (All these settings are the JEDEC specs)

Cas Latency Time: 7
tRCD: 7
tRP: 7
tRAS: 24
tRRD: 4
tWTR: 5
tWR: 10
tRFC: 60
tRTP: 5
CMD: 2

VOLTAGES

CPU Vcore: 1.35625v
CPU Termination 1.200v
CPU PLL: 1.550v
CPU Reference 0.805v

MCH Core: 1.100v
MCH Reference: 0.760v
ICH I/O: 1.550v
ICH Core: 1.100v

DRAM Voltage: 1.500v
DRAM Termination: 0.750v

It won't even do 3.06GHz let alone 3.1, 3.2, or 3.3 :/

It's obviously not the power supply. The RAM goes up to at least 1522MHz if I loosen the timings (passes 12 passes of memtest86+). I think the motherboard just sucks. It can't do 400BCLK, but I really don't need it to. I'd be really happy with a 3.2GHz overclock which is 9x355 but I can't seem to do anything to make it work.

I've tried upping the voltage way past what is needed (1.45v) as well as bumping the CPU Termination to 1.4v and the PLL to 1.750v (which I got from an anandtech article on OCing a Q6600) and the MCH Core to 1.4v. Everything I do is met with a BSOD as soon as IBT ramps up to 100% all cores.

Is it the motherboard? Can the chip just not do it? fucked up northbridge?
Any help would be much appreciated.
 
Try dropping the multiplier and raising the FSB. If you can raise the FSB, than you've probably hit the chips' max. If you can't raise the FSB, than you've hit the boards' max FSB.
 
B3s can sometimes be a pain to OC and 3.0 seems to be a common wall for that stepping. I've never been a big fan of touching PLL voltages and 1.750 seems dangerously high in my mind unless you plan on replacing the chip in a short period of time.

Could your motherboard be holding you back, maybe, the L suffix boards from gigabyte are usually the limited versions of the higher end boards like the UD3R. The board doesn't have the elaborate heat pipe cooling system for the MCH and would certainly have an effect on heat. However, I think the boards were still rated to support 400mhz .

I'm on a EP45-DS3L, which is actually a half-generation older than the UD3 series. I have had zero luck getting it to work at 400FSB. Even though my CPU can do 9x355 (3.2 ghz) on stock volts (1.2875v) I think the chip is able, I'm simply limited by the board and the CPU cooler (AC Freezer 7 pro the older version).
 
My G0 has no problems hitting 3.4 at 1.25V. B3s in general clock nowhere near as well as the G0s.
 
mwarps said:
My G0 has no problems hitting 3.4 at 1.25V. B3s in general clock nowhere near as well as the G0s.
Click to expand...

Pretty much sums it up

Dont expect a great OC with the B3 stepping. The G0 was great for overclocking, but the B3 consumes more power due to a larger manufacturing process (I think), so hitting a higher clock on lower volts is not going to be possible.
 
shiarua said:
Pretty much sums it up

Dont expect a great OC with the B3 stepping. The G0 was great for overclocking, but the B3 consumes more power due to a larger manufacturing process (I think), so hitting a higher clock on lower volts is not going to be possible.
Click to expand...

regardless of steppings or revisions, all Q6600s use the same manufacturing process. I really don't even care about pumping the volts, my megahalems can handle it. I just want the speed.

Update: 375x8 (3000MHz) is stable (albeit I had to up my voltage one more notch from 1.35625 to 1.36250) 410x7 is also stable. I guess it’s time to go get a Q9550 for $170 at microcenter. How much you think I could get for the Q6600 full retail? Like $85-$100?
 
Daemas said:
regardless of steppings or revisions, all Q6600s use the same manufacturing process. I really don't even care about pumping the volts, my megahalems can handle it. I just want the speed.
Click to expand...

These are horribly ignorant things to say in regards to overclocking. The worst thing you can do to a chip is overvolt it. If you have the money to blow on a new CPU (You're looking at the Q9550, good choice), then by all means, continue to volt the hell out of the B3. It won't last very long at all.

All Q6600s were produced at 65nm, yes, but that's not a manufacturing process. That's a process node. There is a big difference between the heat and OC performance of the two steppings and ignoring it is just foolish.

If you want the speed, get a better CPU. Buying a Q9550 will be QUITE an upgrade from the B3. 45nm and 12MB cache ... and it will clock to the sky comparatively. Good choice.
 
If you are willing to go out and get another CPU that quickly, throw more voltage at it.

Have you tried Load Line Calibration on?

All I can tell you is my B3 hits 3.6GHz on a UD3P @ 1.5V with LLC off. With it on, probably 1.45V

These days, I normally run 3.4GHz @ 1.344V to prolong CPU life though.
 
Last edited:
mwarps said:
These are horribly ignorant things to say in regards to overclocking. The worst thing you can do to a chip is overvolt it. If you have the money to blow on a new CPU (You're looking at the Q9550, good choice), then by all means, continue to volt the hell out of the B3. It won't last very long at all.

All Q6600s were produced at 65nm, yes, but that's not a manufacturing process. That's a process node. There is a big difference between the heat and OC performance of the two steppings and ignoring it is just foolish.

If you want the speed, get a better CPU. Buying a Q9550 will be QUITE an upgrade from the B3. 45nm and 12MB cache ... and it will clock to the sky comparatively. Good choice.
Click to expand...

I don't normally try to OC with voltages like that. I was just making sure that it wasn't a problem of the voltage being too low. Right now everything is at stock voltages except the Vcore.

Spare-Flair said:
If you are willing to go out and get another CPU that quickly, throw more voltage at it.

Have you tried Load Line Calibration on?

All I can tell you is my B3 hits 3.6GHz on a UD3P @ 1.5V with LLC off. With it on, probably 1.45V

These days, I normally run 3.4GHz @ 1.344V to prolong CPU life though.
Click to expand...

Yes, LLC is on.
 
Grab a Q9550 and sell your Q6600. The exchange will probably run you $70 after selling your CPU, and you can OC to 4Ghz if you play around enough.
 
Lets try a rational planned approach.

Make a memtest96+ boot CD (download the iso and burn it) to ensure you dont trash your OS.
Set the machine to boot from CD and use the boot disk.

Set everything to stock.
Set memory mulitplier to 2.0 to underclock your memory
Give the MCH the second available increase in voltage from the menu.
Give the memory the first available increase in voltage from the menu, do not exceed 1.6V.

Reduce cpu mulitplier to 8x and starting at 266MHz fsb increase FSB in 10MHz jumps until it will not boot. Do not jump more than 10MHz at a time.

take note of what strap you use as FSB increases and try different ones.

Boot see if it passes POST (install a speaker if you dont hear the BEEP), and let it run the first 3 tests (only takes a min or two)

When it will not boot or goes into the reset cycles, back of 50MHz of FSB and try windows and record your temps at idle and with prime95 OCCT etc. (stop using IBT for the time being).

Post back how far you got.
 
BillParrish said:
Lets try a rational planned approach.

Make a memtest96+ boot CD (download the iso and burn it) to ensure you dont trash your OS.
Set the machine to boot from CD and use the boot disk.

Set everything to stock.
Set memory mulitplier to 2.0 to underclock your memory
Give the MCH the second available increase in voltage from the menu.
Give the memory the first available increase in voltage from the menu, do not exceed 1.6V.

Reduce cpu mulitplier to 8x and starting at 266MHz fsb increase FSB in 10MHz jumps until it will not boot. Do not jump more than 10MHz at a time.

take note of what strap you use as FSB increases and try different ones.

Boot see if it passes POST (install a speaker if you dont hear the BEEP), and let it run the first 3 tests (only takes a min or two)

When it will not boot or goes into the reset cycles, back of 50MHz of FSB and try windows and record your temps at idle and with prime95 OCCT etc. (stop using IBT for the time being).

Post back how far you got.
Click to expand...

Code: 
Multi*BCLK       Latch*mem multi = ddr speed     MCH Core      DRAM Voltage
8x266                 266x3.00A=800              1.180v         1.520v
8x276                          =828
8x286                          =858
8x296                          =888
8x306                 333x2.40B=734
8x316                          =758
8x326                          =782
8x336                          =806
8x346                          =830
8x356                          =854
8x366                          =878
8x376                          =902
8x386                          =926
8x396                 400x2.00D=792
8x406 POST FAIL                =812

8x356 333x2.40B=734 27.3Amb 27-31Idle (probably 27-29 if I gave it a few more minutes 20minutes Prime95 Blend avg45-48 high of 50-51 PASS )

EDIT: I only use this computer for gaming/internet so I guess I really don't need it 100% super stable. I know IBT is really grueling for a system but I was only using it because 6min fail of IBT is much faster than 14hr fail of prime95. So whatever program you think I should use to mimic my stress needs I'll use.

EDIT2: I left it running while I was at the gym. Still going after 1.5hr. 24c ambient 43c-46c load

EDIT3: 23.7 ambient 40c-43c load
 
Last edited:
wow good job, little late to discuss now, I will take a look and get back to you tomorrow (Sunday) .

Quick glance says board should do 375 FSB (3.3GHz) which is promising. I will get back to you. If you want to mess with it tonight leave it on the 2.0 D strap (memory will be way underclocked but we will get to that later) and run same test with 9x cpu mulitplier and see what happens.

We are charcterizing the board to get a feel for where trouble will start. Once we know that we can play with VTT and FSB voltage and maybe a extra bit more of MCH voltage, one thing at a time and see what helps or not. The we attack getting memory up to speed.

Edit. Sun AM.

Temps are excellent so no worries there. We can pretty much eliminate those as concerns.

Now set the Vcore to 1.375V

When you get the time, do go ahead and see what happens with the 9x, expect it to crap out earlier but stay on the 2B strap and you can start a little higher in FSB to save time. Another thing we are doing by the smallish jumps in FSB is allowing the board to reset timings of things not found in bios settings screen. Once you find the FSB limit on 9x back off 10MHz and use it/test it for a couple of hours (memory is still way underclocked, dont worry about it yet) and if all is well, then give the MCH another notch or two and see if you can get more out of it. You have an excellent cooling solution on the MCH but I still like to keep it under or near Intels absolute max datasheet value of 1.21V (1.25v is ok IMO) which is a big reduction from the P35 chipset MCH max of 1.38V. A lot of people are used to the much higher MCH values and (again IMO...) raise P45 MCH voltage way too high which I think actually hurts the OC. This is something you can come back to later if you wish, the present goal is to get you to about 3.2 stable as at least with my machine once my Q6600 hits 3.2 ish I dont see a lot of difference in the "feel" of the machine, but my Q6600 is not a great OCer either.

I am keeping in mind that you have the "L" version, as mentioned in a post above this version lacks the cooling and "heftyness" of the MOSFETs of the more expensive versions. A quad OCed can pull a lot of current and Vdroop. Once we get to where you want to be, ensuring some active airflow on those components around the cpu area would be a good idea as the 3rd party heatsinks typically eliminate the "blowby" of the downwards facing stock heatsink that normally provides air movement over those parts. No worries now but wanted to mention it before I forgot.

After we see what 1.375Vcore and 9x will do and then 9X with a bit more MCH (or not depending on what works best) we will take a look at memory speed and VTT/FSB voltages.
 
Last edited:
Code: 
Multi*BCLK       Latch*mem multi = ddr speed     MCH Core      DRAM Voltage
9x266                 266x3.00A=800              1.180v         1.520v
9x276                          =828
9x286                          =858
9x296                          =888
9x306                 333x2.40B=734
9x316                          =758
9x326                          =782
9x336                          =806
9x346                          =830
9x356                          =854
9x366                          =878 POST fail 1st time, but worked the 2nd time
9x376 POST FAIL                =902              1.800v
9x376 POST FAIL                =902              1.220v
9x376 POST FAIL                =902              1.240v
9x376 POST FAIL                =902              1.260v

9x366 Windows loading fail     =878              1.800v
9x366 Windows loading fail     =878              1.220v
9x366 Windows loading fail     =878              1.240v
9x366 Windows loading fail     =878              1.260v

9x356 Windows loading fail(124)=854              1.800v
9x356 Windows loading fail(124)=854              1.220v
9x356 Windows loading fail(124)=854              1.240v
9x356 Windows loading fail(124)=854              1.260v

9x346 Windows loading fail(124)=830              1.800v
9x346 Prime95 BSOD (15-20min)  =830              1.220v W2 rounding error ~3min
9x346 Prime95 BSOD (10-20min)  =830              1.240v W2 rounding error ~1min
9x346 Prime95 BSOD (10-20min)  =830              1.260v W2 rounding error ~5min

Currently at 9x341 Prime95 Blend for 45minutes everything still going strong.


I do have some questions about temp monitoring though.

Coretemp has my cores at 51c-53c, Speedfan has them at 58c, and HWMonitor has them at 69c-70c. Which ones should I be going off of?

Also, with speed fan and HWMonitor I can monitor two other temperatures on my motherboard. The first is always between 32c-36c(36c right now at load) but the second almost mirrors my core temperatures (58c right now in both Speedfan and HWMonitor). Which one is my northbridge? What's the safe temp range for it?

If the second is my NB and that's too high, it might be because the megahalems is fucking huge and they both barely fit(should have measured before I ordered). They are currently touching and the NB heatsink is kind of off kilter but I am getting the same temps on that field as I did with the stock HS. So I'm wondering if I'm gonna have to go find a dremel or angle grinder or if that temp field is just the mobo's "CPU temp" sensor.


thanks in advance.
 
Wauw - great thread - compliments all!!!

Don't want to hijack this thread just some personal observations:

1/ I have a q6600-g0 and I can OC to 3.3Ghz - stable for 30 mins then the network just stops working (without any clear sign why!) - I have already increased voltages on north- and south bridge but does not work
2/ On 3.1 GHz network seems stable but LightRoom hangs when doing complex image manipulations
3/ Only if I reduce to 3.0Ghz (1333) and set all voltages to stock will network and lightroom continue to work
4/ same question here on difference between CoreTemp and SpeedFan - there are huge differences in core temp monitored between these two tools (CT ~55C vs SF~48C) - I thought these tools just read some data from the MoBo? And then in SF there is this mysterious 'SYSTEM' which keeps steady at 92C!!! anyone knows what this is (my system has Noctua CPU cooler and 6 Noctua 12+14cm fans)

I am not an expert in overclocking so any advise is much appreciated.

Q6600G0 - XFX 790i Ultra Sli - Silverstone Strider 1500W - 8GB OCZ Fatal1ty at 1.7V 9-9-9-20
 
not sure why the PCIe lock won't work: the 2xLAN is integrated on the MoBo - on the PCIe I only have a SB and a GPU and both work fine. moreover in the BIOS settings I have stayed away from these settings. How can I check this?
 
csaba73 said:
not sure why the PCIe lock won't work: the 2xLAN is integrated on the MoBo - on the PCIe I only have a SB and a GPU and both work fine. moreover in the BIOS settings I have stayed away from these settings. How can I check this?
Click to expand...

In some boards, if you set the PCIe frequency to 101 or something like that it will lock. I've never used one of those Nforce boards, so I'm not even sure if that's the issue, but since parts of the motherboards work at different frequencies, it seems like it might be an issue. Either that or maybe the memory is slightly flakey when OC'd depending on the settings used.
 
hmm, in Italy it retails at 200 Euro (283 USD) - not sure if it's not better to go I7 immediately
 
csaba73 said:
hmm, in Italy it retails at 200 Euro (283 USD) - not sure if it's not better to go I7 immediately
Click to expand...

If you have the cash, isn't a bad idea. You can sell your old stuff and get some money for that too.
 
Gentlemen - read the rules on thread hijacking, my fault for not enforcing them as well as I should and I don't want to have to start. Lets keep it related to the OPs topic please - at least to Q6600's - thanks. I am old and get confused easily :p

--------------------------------------------------------------------------------------------------------

Daemas - You excellent note taking and documentation should be a tutorial on how to do it. Outstanding job and really helps me a lot. Hang on to that data as I find every now and then I come back to my machine and play and having the baseline information you are developing now absolutely invaluable. Answers about temps first.

At around 100C your CPU puts out an "I am hot" signal to the motherboard and will begin throttling if C1E is enabled in the bios. This number is factory set for each individual CPU and we have no way of knowing if it is 98,99,100,101 etc.. so software reading the internal Digital Temp Sensors (DTS) should be using the target 100C value. This 100C for the Q6600 is called TjMax. There is another higher temp Intel has not disclosed likely 110-115C where the chip will just shut down to save itself. The temp monitoring software reads a register on the CPU which provides "number of deg C until TjMax". Think of it as a countdown timer. This is called "delta T". Both Realtemp and Coretemp have options to display delta T instead of normal temp. That number should match within a deg. C regardless of software as it is being read from the CPU. Here is how you could look at it yourself. Here is a link to a post that links back to the very first classic posts on reading core temps of C2D. It was quite the hot topic with C2D's first came out. http://hardforum.com/showpost.php?p=1031827510&postcount=31

What is happening is that it is likely the programs are using different values for TjMax as that is something the programmer has to assume based on CPU and until recently the proper value was in question. It actually still is somewhat debated but I go by the meager info Intel has released at IDF 2009 in San Fransisco. TjMax for a 6600 is 100C. (Its all relative anyway, the temp you read is determined by math in the program. TjMax - DTS reading = reported temp. so you can see there that differing values of TjMax will report different temps. As the DTS reading is delta T and you can read/show delta T with most of those programs to know how much thermal headroom you have, knowing you have a value of say, 20, for delta T, you know you have 20C of thermal headroom/more to go before the CPU will send out the PROCHOT spin up the fans I am getting hot signal. Frankly that is more meaningful than my CPU runs at 65C at load because using the delta T (distance to TjMax) takes out of the picture any software inconsistencies. Slight differences of 1C or so between programs is due to the "window" of time the DTS values are read and averaged over as well as roundoff differences. When speedfan, Coretemp and realtemp all have the same TjMax value you will find they track one another almost perfectly. I like Realtemp but after a month or two making sure all is well, I stop constantly monitoring and just check it occasionally to tell me if all is well or I need to clean the crap out of my rad fins.


The first is always between 32c-36c(36c right now at load) but the second almost mirrors my core temperatures (58c right now in both Speedfan and HWMonitor). Which one is my northbridge? What's the safe temp range for it?
Click to expand...

The first is a "board" temp from a thermistor mounted near the end of the 1st PCI slot and is an overall system temp. About the only use for this is an indicator that case airflow is restricted by dirty fan grills etc. It should change very slowly with load and after long periods of gaming/load tend to run a bit higher. You could also use it to access changes made to case cooling to see if the changes help or not but in general it is a kind of useless temp once you get an idea of what is normal for it. The second is from a thermistor mounted under the CPU socket and if EIST and/or C1E are enabled is used as the CPU temp for fan control and throttling for power saving features. You will notice it tracks CPU load but in general responds slower and likely is a few deg C cooler (usually but not always) than the internal DTS temps reported by speedfan and realtemp. Use the core temps once you figure out the TjMax setting is correct.

The Northbridge/MCH uses the same strategy as the CPU for thermal control/trip. It will send out an I am too hot (tied to the CPU's "I am too hot" ) signal so either chip can trip the motherboard to spin the fans at 100%. And it has a thermal shutdown temp to save itself. What it does not have is the reporting register and on P45 Northbridge/MCH there is no way to get "actual" temps. The finger method or gluing a thermistor of a temp monitor onto it is the only way. The MCH/Northbridge just does not report its internal temp to the outside world but it does track it inside the chip.

Getting that chipset cooler firmly down and seated onto the MCH should be a priority. The MCH is a little computer on its own and cooling it is a primary key in getting high FSB out of the system. Touch the other heatsink a bit does not concern me. Making sure the base of the HR-1 (or whatever its called) is properly sitting on the MCH/Northbridge and making good contact is critical. If you can remove/bend some fin from the HR-1 (don't cut the heat pipe :eek: ) and maybe "persuade" the heat pipe to bend just a bit more, careful ! Do what you need to do to get it seated. Some touching - meh, not a big deal.

OK looks like about 340 (I like round numbers) x 9 for 3.06 GHz is about where we are going to end up for now. This is just a baseline to get a basic OC established for day to day and something you can work on improving as time and inclination permits.

HOLY Crap I do ramble on. 3 cups of coffee worth. Break for now. Gotta go. Will be back to comment on OC later today/tonight (Monday), I hope.

Looks like MCH voltage is no help. Go with 9 x 340 and pick memory mulitplier just below and just above rated memory speed and check stability/use. Hopefully we can use the one that OCs your memory a bit. If the over one fails testing try giving you memory .1 ish volts more. The manuf page for your memory does not indicate a range of voltage so I would not push it, .05 ish (.025 more to start would even be better but not sure what options/values you have available, try first increase, then maybe second if second value is less than or about .1 v. I would not go over 1.6V without learning more about how your memory responds to additional voltage. This is something you may be more up on than me. Its your stuff :D

If you must, back off FSB by a couple of MHz to get the memory speed close.

Oh damn, forgot, if you have the setting "memory performance enhance" or something to that effect, set it to normal or fast (see if on normal you can get higher FSB if you want to play some). Just make sure it is not on the highest/best setting - that will hamper OCing too much.
 
csaba73 said:
Wauw - great thread - compliments all!!!

Don't want to hijack this thread just some personal observations:

1/ I have a q6600-g0 and I can OC to 3.3Ghz - stable for 30 mins then the network just stops working (without any clear sign why!) - I have already increased voltages on north- and south bridge but does not work
2/ On 3.1 GHz network seems stable but LightRoom hangs when doing complex image manipulations
3/ Only if I reduce to 3.0Ghz (1333) and set all voltages to stock will network and lightroom continue to work
4/ same question here on difference between CoreTemp and SpeedFan - there are huge differences in core temp monitored between these two tools (CT ~55C vs SF~48C) - I thought these tools just read some data from the MoBo? And then in SF there is this mysterious 'SYSTEM' which keeps steady at 92C!!! anyone knows what this is (my system has Noctua CPU cooler and 6 Noctua 12+14cm fans)

I am not an expert in overclocking so any advise is much appreciated.

Q6600G0 - XFX 790i Ultra Sli - Silverstone Strider 1500W - 8GB OCZ Fatal1ty at 1.7V 9-9-9-20
Click to expand...

1-3 no clue - sorry I dont mess with Nvidia chipsets and all I can recommend is a rational plan and testing routine to establish a stable OC. Google "Nvidia OC guide" Nvidia puts one out on their website and it might have a tip you missed.

4- see post above. as far as the system temp. what all these programs do is scan the sensor buss and attempt to guess what is connected to what and guess at what it is monitoring, It is normal for speedfan to find unconnected senosrs and report insane temps. However if that temp starts out low on cold boot and runs up to 92C most likely it is your MOSFET temps and it is not out of line as mosfets (used in the cpu voltage regulator/power suppy on the board) are made to run at 120C or more. Check your manual/bios and see if the bios is reporting a temp that also runs about that value. If so its got to be a board thermistor near the mosfets. You can rename sensors in speedfan somewhere in the settings or not display it at all. So 2 possible things.
1) unconnected sensor and speedfan reporting "insane" value - disable.
2) onboard sensor, probably thermistor near mosfets, check bios reported temps to attempt to figure out what it is called and where it is.
 
BillParrish said:
snip
Click to expand...

I'm at work right now but when I get home I'll try bending the NB HSF like 5 degrees away from the CPU HSF which should allow me to tighten both of the screws all the way. If that doesn't work, can/should I run with the cpu fan turned 90 degrees (blowing hot air up (or down)? I use an NZXT Beta Evo case, which means I could have the one fan on the top back fan slot pushing/pulling and only have one attached to the HSF which would cut down on the turbulence. Do you think this solution would cool as effectively as my current solution?

66372496.gif
this look like it will work?

also, how far do you think I can bend my NB HSF?

Also, i currently have my 4850 OCed as high as it can go in CCC, do you think that this could possibly cause any of the problems that I'm currently seeing? When I had it at 3.00 stable, it was fine and never crashed, so I think the OC of the VGA isn't causing anything but I wanted to make sure.
 
I love those cases, I have quite a bit of experience with the Antecs with the big fan in the top. The best way I have found for cooling is close to what you are thinking of. Yes rotate the CPU heatsink 90 deg so that it is blowing up into the top fan, heat rises so that will almost instantly shoot the warm air out of the case. Now my little twist on the deal. What I found, and you should try it both ways to see what works best for you, is to flip the rear fan around so it blows into the case and onto the cpu heatsink. This has resulted in a couple of deg lower temps when the main HS CPU fan is not running at high speed. Apparently the extra cool air hitting the side of the heatsink helps and it is not strong enough to disrupt the upward airflow. But check and see how it does with your setup. It should be pretty easy to swap around if it does not work as expected. But the upwards pointing CPU fan/heatsink is for sure the way to go and if that will allow your chipset cooler to seat properly - major bonus. If you have to remove one of the fans I think it would be the "pull" upper fan and leave what would become the bottom "push" fan. With that big top fan and if the heatsink ends up fairly close (you could fabricate a duct if you felt industrious) I do not think your temps will suffer a bit and temps are not your main issue anyway. Getting that MCH heatsink to sit down has a better chance of helping with an OC espically when we try and get the memory tweaked up than a couple of deg C on the cpu that had good temps but is resisting a high OC.

If 5 deg of tilt clearance is all you need you should be able to do that, gentle firm steady pressure while supporting the base. However I must admit I have never done it and am basing my suggestion on bending various kinds of copper tubing over the years for AC and plumbing work and I have never bent a heatpipe tube. But 5 deg is not much. If uncomfortable with it, perhaps you can bend a couple of fins on the HR-1. That should not impact cooling too much as you have a fan on it.

All just my thoughts, you are there and have the stuff in hand and I can tell by your posts your a smart guy, go with your gut and dont let me talk you into something you are not comfortable with.

will be busy tonight, so likely not post again today, still thinking and want to see what happens when you pick a memory mulitpler close to your memory specs, we still have an ace in the hole, the FSB voltage. It usually helps but want to hold off on it until we get the memory in the ballpark and the machine proven stable at 3.0 or above with the memory somewhere near rated speed.

(ouch - despite our current effort it appears we are not much better off than when started. )

I dont think the gpu is causing any issue unless you use something like 3Dmark for stablity testing, crashes in games is when I would start looking at that. OCCT, prime95 etc. are just math programs that stress the cpu, video card never even goes into 3D mode.
 
BillParrish said:
snip
Click to expand...

I'm about to try to remount the cpu HSF 90 and fix the NSF heatsink. One thing that I found when I did your testing is that I was able to run at ~2.93GHz stable at stock voltage, but to get 3.00GHz (what I was running before) I was running it like +.04375v which is like 6 or 7 steps. So something is FUBARing my quest. Hopefully the NSF heatsink thing fixes it.

well, than again, that 2.93 @ stock wasn't IBT stable though :/
 
Last edited:
She wasn't too happy, but I gave her some liquor and some lube and it all fit in.:D

img0007vo.jpg


img0006ujv.jpg


img0005zp.jpg


it's like 2mm from the NB HSF to the CPU HSF

EDIT: I''m currently priming @ 3.06 1380DDR3 overnight. I failed 3.1 8XXDDR3 last night.
 
Last edited:
Now that the heatsink for the mch is ok, run Vcore at 1.375 or 1.3825 and see where you get with a 9X multiplier, couple of bumps of MCH voltage and the 2D strap.
 
345*9 3.1GHz 1.3825v Vcore 1.220v MCH 2.40B strap (lowest I can go using the 333MHz latch)

been priming for a good 2 hours so far. gonna leave it over night. It failed after about 45min with 1.80v MCH so I bumped it up to 1.220v. If it's still going in the morning I'm gonna bump it up to 3.2GHz.
 
Quick question, what is the max vcore that you've ever thrown at it? Since I have a similar Gigabyte board with basically the same bios options and the same CPU stepping as you (Q6600 B3), I have to note that my settings at current are:

379x9 = 3.4GHz
C1E enabled, LLC enabled
CPU Vcore: 1.375v (reads 1.344 in CPU-Z)
CPU Termination 1.300v
MCH Core: 1.200v

I have not touched anything other settings.

Idle Temperature: 35°C (fan is only 800 RPM for noise reduction)
Load Temperature goes up to 60°C

But I have experimentally bumped Vcore up to 1.60+V before to pull 3.8GHz but that was frightening and I quickly stopped doing it.

I think you are compulsively obsessing too much over a stress test though. Your overclock does not need to pass IBT or Prime95 to be stable for what you need. When I goto 3.6GHz, Prime will calculate a wrong value after about an hour. Set your overclock to what you want and then use your computer normally. If it stays stable for hours & days just doing what you normally do and in the games that you usually play and doesn't crash - why bother?
 
Last edited:
I tend to agree but its a debate that will generate multiple page posts. IBT is really tough and its my opinion people misuse it for long term stress testing when it is a thermal analysis tool that to my thinking should be run for a hour to determine your temps and thermals are good. Then switch to prime and the other math based tests and top it off with looping 3D mark or something that stresses both the cpu and gpu/video card if you are a gamer. But stability is in the eye of the beholder. I am of the camp that is the machine will do what you want it to without ever having an issue, its stable. Others certainly disagree. Myself I like OCCT as it is easy and has temp monitoring, logging etc. and I let it run a couple of hours (or an evening while I watch TV) and call it. Then use the heck out of the machine, as my sig says, I can make my machine go faster but in its current configuration I have never had a BSOD, game lock or anything else happen that I can attribute to OCing. Absolutely hate that and so I never run the machine flat out in normal use. That said, in testing like we are doing I did want to find the max mainly for my own edification and would let it run all day stress testing. When it fails I just back off 5MHz FSB and try again later. Over a peroid of weeks I came to know the machine and found the spot I liked as the trade off in speed vs stressing the entire system vs my personal defination of stablitly. I guess the key idea is that there is no hurry, my machine is over 2 years old and I still occasionally "press the gas" on a rainly day to see if I can get more out of it playing with settings combinations.

Of great interest with Gigabyte boards is the Memory Performane Enhance option that on some boards affects the MCH latency. tRD. It can have a huge impact on overall system performance once you get the cpu tuned up and characterized. That is why I am only worring bout the CPU at the moment. Once the OP gets the cooling sorted out, a good idea of what the cpu will do the plan is to tweak the memory with the idea that if necessary minor reduction in FSB that allow Memory Performance Enhace to get off the "Normal" setting may result in better overall system performance (or not).

been priming for a good 2 hours so far. gonna leave it over night. It failed after about 45min with 1.80v MCH so I bumped it up to 1.220v. If it's still going in the morning I'm gonna bump it up to 3.2GHz.
Click to expand...

Nice, a little progress, frankly 3.2 is a good target with that board. I know lots get much higher numbers but as I mentioned earlier thats a nice speed for performance/day to day use. Of course play and see how high you can get it and if temps stay in line you can certainly play with voltages around 1.4V area but with that 'L" version of the board I am concerned about stressing the cpu voltage regulation circuitry. If there is any way you can rig a fan (another reason I like flipping the rear fan to blow in) to blow air onto the board under that huge heatsink around the cpu area (look for the little flat black squares - 2 groups of 3 behind the cpu and a line of 6 above the cpu). Air on those guys would help long term realibity and perhaps stability and with OCing and long stress tests a lot of current is being pulled through those. They are made to run hot but with 4 cores flat out, if you intend to keep the board a while, a spare fan blowing in there is cheap insurance. Anything to get even a little air directly hitting them will do.

Carry on, keep your notes concerning where you can get with increased Vcore, I suspect you will find a point where increased Vcore does little or no good and finding that voltage is a great thing to know if for no other reason than to set it at/near its max useful value and then leave it alone eliminating one setting to adjust out the the many you have to deal with and allowing you to play with Vtt and or Vfsb (I have to go check your manual for what is available) without the complications of changing Vcore too. You are getting to the point where you want to start locking down (not adjusting) stuff, mainly Vcore and Vmch, so we can see what affect the other settings have. Vmch is still something we will change as we get into the buss voltages but only a little so finding the same "up to here helps but more does not" voltage is a good thing as you should set it to its max "this helps" voltage and then make only very minor changes.



Intel data sheet absolute max for a Q6600 is 1.55 volts and if you consider that per Intel spec (what the board manuf actully meet is another question) the CPU voltage regulation circuity is allowed a 50mV, .050V overshoot during load transitions, somewhere around 1.50V should be the max Vcore if you care about Intel specs.

Me personally with my cpu I have found that from stock voltage (mine is 1.25 I seem to recall, it was not one of the better lower Vid ones - bad luck of the draw) to 1.382 I get improved OCing and then there is a flat spot from about 1.39 to 1.45-1.4625 where the increased voltage does little or no good. Around 1.475 more voltage starts to help a little bit and I bailed out testing around 1.5125 becasue the minor speed increase over what the cpu was doing at 1.375-1.382, to me, was not worth the worry and temps jumped up quite a bit. I just did not see it as worth it for day to day operation to stress the chip. But every board/cpu is different.

Anywho, I am just talking, none of this is written in stone, and there is no hurry. Enjoy your machine and I am glad it is looking better for OCing. I be around.
 
Last edited:
I've got my Q6600 at 1.3825V at 3.5GHz (9 x 389MHz).

Works great for me at 23-27C idle and 52-54C using Large FFT's in P95. Blend isn't the higher heat - so that isn't accurate to me. I also run a ten cycle IntelBurnTest v2.40 and that passes too (but can get me to 62C).

I'll try to get around to a screenshot and info later.
 
ecomputer_ca said:
I've got my Q6600 at 1.3825V at 3.5GHz (9 x 389MHz).

Works great for me at 23-27C idle and 52-54C using Large FFT's in P95. Blend isn't the higher heat - so that isn't accurate to me. I also run a ten cycle IntelBurnTest v2.40 and that passes too (but can get me to 62C).

I'll try to get around to a screenshot and info later.
Click to expand...

Is it a G0 or a B3 stepping though? That would be more important to the OP has he (and I) have the B3 which is historically the worse overclocker.
 
Sounds like that CPU is just a dud. Until my "golden" E6400 I was convinced that most people exaggerated their overclocks and especially whether they were stable or not. Not all CPU's are created equal, even with more voltage, lol.
 
well, I was able to hit 3.1 and 3.2

3100MHz is with 1.3875v Vcore, default all other voltages.

3200MHz is with 1.45v Vcore, 1.34 CPU Termination which raises the CPU ref to 0.893v and the MCH ref to 0.848. MCH Voltage is 1.300v. It's been Prime Blend stable for 30minutes now, hopefully it lasts overnight.

Are there any ways to try to lower the VCore needed? Does raising CPU Termination, CPU PLL, or MCH Voltage have that potential ability? (I guess if this passes prime overnight I could start trying to drop my mch and stuff, but I'd like to save myself some time.)


I was also screwing around with my mom's email machine with a E6300 and a p5wDH Deluxe and hit 3.73GHz at 1.4v so I may put that chip in my board and see if the board really does suck or if it is the chip.

EDIT: BTW, what's the max i should be cooking this chip? Coretemp has a Tj. max of 85. Is that about right? I know that 1.5v is pretty much the limit, but idk about temps as it tends to get hot as hell in my room when I'm running both the computer, plasma and ps3 with the door closed (like 90 degrees ambient lol)
 
Last edited:
Max temp would be 85C of course and this is where the chip will start to slow itself down to preserve itself, etc. Crashes might occur above this too.

Personally, I like to keep everything under 60C. I also have a very high ambient temp but I counter that with an open case and a Tuniq Tower 120 with a 110cfm fan. It's rare for it to go much above 50 normally but my cooling is above average for air.
 
Look for the "Distance to TjMax" to be at least 10C-20C under full load, that number will count down toward zero and at zero the CPU will start to throttle and send out the "I am hot" interrupt.

I am sitting just surfing the web and running a virus scan in a cool room with a Distance to TjMax of 68 on core 1 and 72 on core 3. This makes a TjMax of 85 unlikely as 85-68= 17C core temp which is ambient or less (aprox 63 F) and my WCing is not that good. I have mine set at 95C, I often say the value should be 100C. Its all relative, its Distance to Tj that is the only thing you can know for sure.


Truth is Intel has never fully released to the public what TjMax (actually it would be the manuf target Tj as each CPU is slightly different and set individually at the factory) is. At IDF 2009 in San Francisco they gave the info for the 45nm chips. Nothing was said about the 65nm chips/6600. There is a slide floating around with Intel's name on it with the 65nm info but it was pulled and not presented. Unfortunately the Intel PDF of the presentation is now gone from Intel's site.

Distance to TjMax is your thermal "headroom" and an absolute indisputable indication of if the die is close to being stressed thermally, any other number is for edification only.
 
Daemas said:
well, I was able to hit 3.1 and 3.2

3100MHz is with 1.3875v Vcore, default all other voltages.

3200MHz is with 1.45v Vcore, 1.34 CPU Termination which raises the CPU ref to 0.893v and the MCH ref to 0.848. MCH Voltage is 1.300v. It's been Prime Blend stable for 30minutes now, hopefully it lasts overnight.

Are there any ways to try to lower the VCore needed? Does raising CPU Termination, CPU PLL, or MCH Voltage have that potential ability? (I guess if this passes prime overnight I could start trying to drop my mch and stuff, but I'd like to save myself some time.)


I was also screwing around with my mom's email machine with a E6300 and a p5wDH Deluxe and hit 3.73GHz at 1.4v so I may put that chip in my board and see if the board really does suck or if it is the chip.

EDIT: BTW, what's the max i should be cooking this chip? Coretemp has a Tj. max of 85. Is that about right? I know that 1.5v is pretty much the limit, but idk about temps as it tends to get hot as hell in my room when I'm running both the computer, plasma and ps3 with the door closed (like 90 degrees ambient lol)
Click to expand...

Unfortunately Vcore is the primary voltage affecting CPU overclock. You can test reducing it a step and see if you can maintain the same OC but in general you are adjusting the other voltages to "catch up" to the CPU and its increased voltage/speed.

Remember you have 4 cpus in a 2 cpu bag. In this board (sig) my E6300 would do 475FSB easy, with the quad I can just barely make it go 385FSB regardless of anything I try. But give it a go, it would be interesting to see what happened when you swap. Do keep in mind one of the cardinal rules of male survival, "It never pays to piss off mom."
 
Last edited:
You must log in or register to reply here.
Back
Top