Why cannot I squeeze my Q6600?

[Narynan]

I am in need of help, and could really only think of one place to go, so please bear with me guys.

Right now I have just got done putting together a new machine, and am looking for "a little more" out of my Q6600. Before I get TO much into what is going on, I will give you a little bit of an idea as what is in my machine.

I got for a mobo asus p5e.
Q6600 /// g0 stepping
zalman 9500 cooler
Corsair XMS2 2x2 = 4gb
Zalman 850w psu.
Sapphire 3870
WD 320G hdd.
Antec Nine Hundred case

thats about all we need to know I think, the rest is pretty basic.

So as it is right now, I can get a 9x333 = 3.0 ghz overclock easy, I can drop my voltage in my bios to 1.26 and get 10 hours of p95 steady.

However, it seems if I go to something a little more aggressive, I get system reboots and freeze ups, ie. 8x400. 9x360.

I dont know if its a voltage issue, as I have tired the 8x400 overclock at 1.3v and get crashes ( i do however get about 45min - 120min) , and i know its not a heat issue, because before it crashes, coretemp records a max core on core 1 of 60c. I dont know if I want to move my voltages much more above 1.3 - 1.33 because of the heat concerns. Not because I am afraid of the wear and tear it MIGHT put on the processor, because I just dont like seeming anything above 65c. Right now I am not playing with timings, or anything other than voltage and mult/fbs speeds.

Reguardless of my paranoia, what recommendations do you think should be the highest core temps? 70 is probally warmer than I would like. As it is, I can get that 9x333 with the 1.26 and have a top temp of 50c give or take 1 in either direction. What I really want is that magical 3.2ghz, if only because that was the speed of my old p4.... stock.

What do I do guys? I get more out of this processor?

 

[taltamir]

Don't lower your multiplier, you want it as high as possible for any OC.
I think its your voltage, it needs to be higher... but dont go over 1.4
actually why not just set it AT 1.4, and then lower it bit by bit (1.39, 1.38) until you find the point of instability.

Are you sure your ram is not erroring? you did remember to change the ratio on it, right?

 

[Marty502]

That's what I was thinking. Aren't you supposed to drop your memory ratio as a first measure when the OC doesn't go any further? At least that's what I'd do, instead of cranking voltages, which should be the last thing to try to squeeze some extra Mhz in my opinion.

 

[taltamir]

btw, did you read through this entire guide?
http://forums.anandtech.com/me...=2057083&enterthread=y

 

[isaaclim]

I had the same problem as you with my q6600. I had to push my voltage up to 1.468v and with vdrop it show as 1.42v in windows. That was a higher than expected voltage for me since I read all these people achieving 3.2ghz at lower voltages. But my temperatures are really good due to some awesome airflow.

 

[Narynan]

yeah, have the Ram ratio set to auto. So I would be inclined to that that it's taking care of itself....

Alright when I get home I will try looking at 1.4 and then going from there. It just SEEMS like a rather large increase from what I am using right now for that extra 200mhz, especially working heat into the problem. But temps are not really my worry. I can handle those. I will take it off auto and move it to the DDR 720 bench. because my fsb multi will be 360 when I try my new OC.

and i did read the whole guide, top to bottom. However, some of it was so far removed from my board, or so much more complicated than what I was looking for. I am not looking to tweek my ram timings or anything else.

And to Isaaclim.... what speed are you running on yours with that kinda OC? because thats almost .2V higher than what I can get nice and cool at the 3.0 mark.

Also, is running it at 8x400 not worth it so I can run the 1:1 ration with my ram? I thought that would be something to look forward to, but if its not, then I am surely not going to cry over it. And if it's not, why would be nice. Just so i know. I like knowing.

 

[isaaclim]

my rig is at 3.2ghz
ram is at 960mhz
8x400 for my multi and fsb.

my story is very similar to yours. For 3.0ghz, I think I had to run 1.36v for a stable OC. But when I looked for 3.2ghz, 1.46v was the only voltage that gave me stability. it sux, but then again my temperatures are so good I'm not worried. Plus its within the 1.5v limit the CPU is rated at.

 

[PlasmaBomb]

If your board supports the newer intel 400fsb CPUs it shouldn't be holding you back. Check what the voltages are for the ram and chipset, you may need to up them. Don't leave the ram on auto.

 

[TaylorTech]

Your XMS RAM should be able to handle some decently high voltages. Try MemTest in your BIOS [or download if it doesn't include it] and see if that isn't the error. Make sure your divider is set accordingly. I've heard it's great RAM but can be stubborn when it comes to overclocking.

 

[Narynan]

Well right now, its set to auto, but I'll check it out and see what its able to do. It looks like it is "normal" at 1.9V.

I dont understand what you mean when you say "set the divider accordingly".

At this point, I am keeping it as basic as possible for the overclock. I will happily do the voltage at the standard 1.9V, instead of auto. Or do I need to give it more? Never really toyed around with RAM voltages before.


Another question. Which is the better overclock?

9x355=3.2ish
8x400=3.2

Is the 1:1 RAM / FSB ratio better than the multipler?

 

[sgrinavi]

Originally posted by: Narynan
Well right now, its set to auto, but I'll check it out and see what its able to do. It looks like it is "normal" at 1.9V.

I dont understand what you mean when you say "set the divider accordingly".

At this point, I am keeping it as basic as possible for the overclock. I will happily do the voltage at the standard 1.9V, instead of auto. Or do I need to give it more? Never really toyed around with RAM voltages before.


Another question. Which is the better overclock?

9x355=3.2ish
8x400=3.2

Is the 1:1 RAM / FSB ratio better than the multipler?

9x355

 

[Drsignguy]

Originally posted by: sgrinavi

Originally posted by: Narynan
Well right now, its set to auto, but I'll check it out and see what its able to do. It looks like it is "normal" at 1.9V.

I dont understand what you mean when you say "set the divider accordingly".

At this point, I am keeping it as basic as possible for the overclock. I will happily do the voltage at the standard 1.9V, instead of auto. Or do I need to give it more? Never really toyed around with RAM voltages before.


Another question. Which is the better overclock?

9x355=3.2ish
8x400=3.2

Is the 1:1 RAM / FSB ratio better than the multipler?

9x355

9x355 = 3195 your( Corsair XMS2 PC2 6400 2x2 = 4gb ) Ram @ 1:1 = 710Mhz
8x400 = 3200 your( Corsair XMS2 PC2 6400 2x2 = 4gb ) Ram @ 1:1 = 800mhz

You have a bit more headroom with your ram @ 355 rather than a higher FSB so you can attempt to get a higher OC.

 

[taltamir]

Originally posted by: Narynan
yeah, have the Ram ratio set to auto. So I would be inclined to that that it's taking care of itself....

wrong assumption. auto just means it sets it based on the type of chip you have, not based on the FSB you set. a CPU saying 1333mhz FSB uses 2.4 ram multi (E8400), a CPU saying 1066mhz FSB uses 3 ram multi (Q6600).

It means it is ocing the hell out of your ram.

Set it manually, it will show you the end result of the ram in mhz, i suggest you set it for 2.00 multiplier at first while finding a stable CPU OC, then see abour raising your ram multi, possible with raising the ram voltage and loosening timing to get a stable ram OC at something higher.

For example, on my E8400 the default ram multi is 2.4 because its base is 333mhz (333x2.4 = 800mhz, aka DDR2-800)
When I switched to Q6600 today the default was 3, because 266x3 = DDR2-800.

I set it to 2, tweaked with my CPU speed, finally settled for a modest OC of 3.3ghz (I didnt want to push the CPU voltage too high... performance/watt, lifespan, heat, etc are also important to me).
@3.3ghz and 9x CPU multi I am at 367mhz base,
367x2 = 734, modest underclock in which I benched CPU stability. After I was satisfied I went and OCed the ram by changing its multi to 2.5, so now it is 367x2.5 = 917.5mhz @ 5-5-5-15 with min voltage of 1.9v (it is supposed to be 1.8, but this mobo is trying to be more compatible because the first gen of DDR2-800 all required at least 1.9v to post).

At 3.2ghz on a Q6600 you would be at 356x9
your ram then is trying to run at 356 x 3 = 1068
Now they SELL ram that is marked for 1066 operation... but only at high voltages like 2.1 or 2.2 volts on the ram.


PS. Every OC should start with a memtest run to see if the ram is stable, if the ram is not stable then there is no point even trying to go into windows or testing the CPU (as it will error). Make a COMPLETE successful run (about an hour) using http://www.memtest.org/
Alternatively, if you can underclock the ram then you are fine testing the CPU first.

 

[Narynan]

Excellent thats some of the best information I have gotten so far, and would really probally explain alot.

However, I have no I dea where to find this RAM multi you are talking about. Im not at my home comp right now, so I cant really look at my bios, but I will check it out when I get home. Nothing that I saw on my Bios steps out and grabs me as to where this RAM multi would be.

Could anyone help me out with that.

 

[OCGuy]

Originally posted by: Narynan
Excellent thats some of the best information I have gotten so far, and would really probally explain alot.

However, I have no I dea where to find this RAM multi you are talking about. Im not at my home comp right now, so I cant really look at my bios, but I will check it out when I get home. Nothing that I saw on my Bios steps out and grabs me as to where this RAM multi would be.

Could anyone help me out with that.

On my P5KC, you dont actually select a divider outright, such as 1:1, 1:2. You just go to RAM Frequency, and it will give you a list of speed options, that the divider is already calculated in to.

As far as people saying dont go over 1.4v, they are referring to 45nm, of which the Q6600 is not one. 65nm handle higher. Good luck!

 

[Narynan]

Alright yeah, I have seen that option. Opened it up and closed it just as fast because I while I under stood the relation to the numbers, I didnt understand the effect it would have on the rest of the system.

Having looked around at a large number of sites and getting my feed back from here, there is a LOT more than just upping the FSB to get a good stable overclock.

A question that is stemming up from all of this in my Ram voltage, if I plan on staying with a 2.4 RAM multi at 350fsb 9x multi. Should I be leaving the voltage at the documented 1.9V for this RAM? I cant seem to find anything other than 1.9V listed and I have heard that it is kinda picky ram.

And on this board and with this chip, should I have Loadline Cali on or off?

 

[taltamir]

these gigabyte boards have a specific voltage number for the CPU only, the rest you choose "normal" or +0.#
With ram the default is 1.9, because first generation DDR2-800 all required at least that much to post (they were overclocking DDR2-667 modules and overvolted, actually they still are... only DDR2-1000 or -1066 can run at the JEDEC standard for DDR-800).

Anyways. the multi for ram is on the same screen as the CPU, I don't remember what exactly it is called at the moment. In many motherboards it is a ratio, like 1:2 and 5:6...
In gigabyte boards it is a multiplier...
I have the following options:
2.0, 2.4 (default for E8400) 2.5, 3.0 (default for Q6600), and above 3... not ideal choices, it is obviously the standard ratios just with a more easily understood multiplier instead of ratio. Honestly I like it better, it is very easy to see what you are setting your ram at.

EDIT: I looked it up, gigabyte calls it "system memory multiplier" in the bios, same window as the other overclocking stuff.

Also, I got it stable at 3.3ghz, but the room was getting hotter even with the AC running, doors open, and ceiling fan on, the north bridge and the ram was scalding to the touch, and the voltage was rather high (aka, performance per watt was not good), I lowered it to 3.0ghz and I lowered the voltage a LOT. the temperatures dropped 11c at load (63 to 52c), the room is actually getting colder and colder now, and the north bridge and ram are much MUCH colder to the touch now (not cool, but much cooler).

So I saved all the data on the stable 3.3ghz oc i got, I will keep it in mind if I feel like it at some point, but for general usage I am sticking with 3.0ghz. I'll actually try to undervolt while OCing it to 3ghz soon. (if you count vdroop i already am)

 

[Narynan]

Taltamir. Thats the same experience I was getting. Sence starting out this page I have gotten a lot of useful help, and have hit 3.2, but nothing stable as of yet.

I think for the summer I am more than ok with 3.0G on the board. However, I after all the discussion and everything I have learned I am going to force a bunch of voltages and set thing manually. It just seems like a better idea. I'm not looking for the fastest OC.

I'm going to do some work on this when I get home in 2 hours and see how everything goes. I'll keep everyone up to date.

PS if you have a similar mobo / cpu / ram setup and have a 3.0 - 3.4 ghz OC, please let me know what your settings are at.

 

[taltamir]

the only difference between my mobo and yours is that I have the raid edition southbridge from intel... so i got 2 more SATA plugs and better raid support.
The northbridge and other stuff are the same... both are revision 2.1 which added the EPU (energy chip thingie) and are both very very similar.

Check my signature:
gaming: Q6600 OC to 333fsbx9 for 3.0ghz, undervolted 1.26v idle /1.23v load (VID 1.325v, mobo setting 1.3125v), 4GB DDR2 ram @ 800mhz 5-5-5-15 1.87v-1.9v, eVGA GTX260 SC

here is my OC log:

3.3ghz @ FSB: 367, memory: 2.5x, CPUv: 1.46250v
even one level of CPUv below it is unstable.
northbridge and ram scalding to touch. room temperature rising constantly.
60c load.

3.0ghz @ FSB: 333, memory: 2.4x, CPUv: 1.3125V
two CPUv levels below it is unstable, one below was not tested.
50c load. northbridge and ram are hot, but not too hot at load. cool to the touch at idle

If you noticed that the CPUv i set it too is 1.3125V (two ticks BELOW the default on my chip), with vdroop it comes out to 1.26v idle and 1.23v load

people tell me early G0 units had a VID of 1.2 and supposedly could OC much higher and still remain cool and stable and at lower voltage, but it seems intel stopped giving us such a free meal

 

[Jessica69]

Originally posted by: taltamir

people tell me early G0 units had a VID of 1.2 and supposedly could OC much higher and still remain cool and stable and at lower voltage, but it seems intel stopped giving us such a free meal

That's very true, taltamir. Head over to XS and look around at all the OC'ing resullts on Q6600's. The earlier G0 stepping cpus allhad lower VIDs compared to later spins of them....I have one, while not the earliest, that has a VID of 1.275V and I managed to get a 4.0GHz run out of it, Prime 95 stable for an hour, at 1.55V....but it's also water cooled.

But my experience is not unique. The lower the VID on the cpu, the higher the OC that can be achieved with lower voltage applied. Seems the later production units sometimes even have trouble hitting 3.6GHz with any amount of voltage run through it.....while I run mine 24/7 at 3.6GHZ on 1.34V.

Just wonder what happened to the production of these Q6600's that made them less and less able to achieve the OC's that the earlier ones could.....

 

[taltamir]

my guess is, whatever intel did to improve the quality in G0 stepping, they undid it. There is more then enough headroom in it without it to be perfectly stable in any speed they sale, and it reaching 4ghz makes it compete too strongly with future parts. They are not big fans of OCing after all..

 

[Drsignguy]

Originally posted by: taltamir
my guess is, whatever intel did to improve the quality in G0 stepping, they undid it. There is more then enough headroom in it without it to be perfectly stable in any speed they sale, and it reaching 4ghz makes it compete too strongly with future parts. They are not big fans of OCing after all..

Still, you undervolted to 1.3125v from your chips VID of 1.325v, while over clocking, you must be doing something right. And I am a fan of OC'ing.

 

[Narynan]

Hey everyone just a quick up. I changed alot of settings at once ( I know its not the ideal way to do it, but by this point and having slaved over it, I can say that I got a good idea where things were going to be to get this done right) and booted up. Started up P95, coretemp, and CPU-z. And let it run over night. The good news is that I am 9x360 for P95 stable for the last 8 hours. I will post a change log when I get back from the wedding I am going to this weekend. I will have my voltages ready Sunday night. However the best part of this whole news, is that I (so far) have hit a highest temp at any given time, of 54C on any of the given core. This is all in a room without AC. Heres hoping its stable the whole time, and gold when I come back.

 

[Idontcare]

Originally posted by: Jessica69
Just wonder what happened to the production of these Q6600's that made them less and less able to achieve the OC's that the earlier ones could.....

Originally posted by: taltamir
my guess is, whatever intel did to improve the quality in G0 stepping, they undid it. There is more then enough headroom in it without it to be perfectly stable in any speed they sale, and it reaching 4ghz makes it compete too strongly with future parts. They are not big fans of OCing after all..

More than likely what people are noticing is that the Q6600 parts are coming out of different fabs and not all the fabs are producing chips with the same electrical qualities. So as Intel switches production from 65nm to 45nm in their premium fabs, those fabs ship less of the higher quality 65nm chips thus the market is having more and more of the Q6600's coming from the remaining 65nm fabs which seemingly have less electrical quality (in terms of margin to overclocking).

Some people around here may be old enough to remember the Celeron 300A's that overclocked to 450 and higher...they were packaged in Malaysia and Costa Rica and while you could get chips from either location that would hit top clocks most people recognized the highest liklihood of getting a random chip from the market and hitting those top clocks came with the chips from Malaysia. (dubbed "Malays")

So it only stands to reason that as Intel shifts production there will be a shift in the general overclocking distribution of Q6600's. Also they may be adding late qualifying fabs that haven't ran much 65nm product as they shift their leading edge fabs to more and more 45nm capacity.

 

[taltamir]

mmm. makes sense idontcare. I just assumed intel would use the same techniques in all of them, but its not like all intel employees are telepathically linked, and I doubt they put too much effort into their second tier fabs, and various pieces of equipment are probably different, and so on.

Originally posted by: Drsignguy

Originally posted by: taltamir
my guess is, whatever intel did to improve the quality in G0 stepping, they undid it. There is more then enough headroom in it without it to be perfectly stable in any speed they sale, and it reaching 4ghz makes it compete too strongly with future parts. They are not big fans of OCing after all..

Still, you undervolted to 1.3125v from your chips VID of 1.325v, while over clocking, you must be doing something right. And I am a fan of OC'ing.

Yea, it's awesome . I opted for that instead of a slightly higher clock with an overvolt (that got higher QUICKLY)