Q6600 G0 SLACR OC Adventure

[taltamir]

So I just got my Q6600 a few hours ago, luckily it was a SLACR as I expected it to be (bought from the egg, OEM for 185$ with free shipping, others wrote they were getting SLACR).

Took some effort to disassemble my giant Noctua air cooler with its push-pull-pull (two 120mm fans on it, one 120mm on the case less then an inch from it) setup, but I got it done.

made sure it works (by booting into windows and checking the stepping in CPU-z)
Then to OC.

My E8400 was at 3.6ghz @1.27v
First I was trying to get a rough feeling of what this baby can do. At 400x9=3.6ghz it would only boot at 1.4v+, and it will not enter windows at 1.5v, I did not try above that.
At 378x9=3.4ghz it got into windows, but was not stable at 1.45 to 1.475.
At 367x9=3.3ghz it is running stable so far with 1.45v (the first setting I tried). the voltage actually seems to not droop as much as with 3.4ghz while on load (it now droops to 1.34 vs 1.32 before). And it is stable for 42 minutes so far @ OCCT v2.0.0a
with load temp of 56/56/60/60 (with the occasional 57/57/61/61 for a second).

60c load and 1.45v (or lower, after I am done tweaking) seems A-OK to me. So this is my new target OC. I don't think I will try to aim higher, even though I am sure I could, that extra 100mhz are not gonna make a big difference in speed, but are making a big difference in power consumption, heat, etc...

My ram is right now running at 752mhz, I will go and OC it as soon as I get to 1 hour stable on the CPU. It shouldn't be a problem though since the ram is rated for DDR2-1000 to begin with.

Then I am going to try dropping the cpu voltage slowly while maintaing stability.

Advice is welcome. I am trying to get a beefy, but not extreme OC with decent temps and voltage. raw performance is the most important, but performance per watt, thermals, and longevity is also important (not as much, but important enough for me not to not aim at 3.6ghz)

 

[sgrinavi]

Is it on that Gigabyte EP35-DS3R in your sig?

I had mine running in the 3.2 neighborhood, with my Q6600 G0, at just a hair above stock voltage. It will go 3.6 in xp at 1.4 volts, but only with the RAM sloppy.

 

[taltamir]

yea, it is with this board in my sig. I am now running my ram at 2.5x, so it is 367x2.5 = 917.5 mhz at 5-5-5-15 with default voltage (my board only allows +0.1, +0.2, etc to ram voltage, not setting a number).
But OCCT shows my voltage to be 1.9v on ram, not 1.8. It would not surprise me if this board overvolts the ram by default (considering that all the early DDR2 ram originally would not post at 1.8, only at 1.9 and above)

it lists my stock voltage at 1.345 or something, and when I set it to 1.45v @ 3.3ghz in mobo I am getting 1.39 idle 1.33 load in measuring software.

I haven't actually tried lowering the voltage yet in this speed. 3.4ghz was not stable at 1.475
but this is stable at 1.45 (1.39 idle 1.33 load) which is the first setting I tried on it. I will try lowering the voltage further. maybe 3.2ghz would give me much better voltage.

 

[Extelleron]

Unfortunately the newer Q6600's are generally not as good (or at least as consistantly good) as the Q6600's that we saw last year when you could expect 3.6GHz with reasonable voltage from nearly every G0. A lot of the 8xx batches seem to have 1.325V VID and not be very good oc'ers. I'm assuming yours is 1.325V VID based on the results..... you can check in CoreTemp.

If your apps are mostly multithreaded I suppose it is a good upgrade from an E8400. The Q6600's are just not clocking well enough for me though, since pretty much all I do is game an E8400 was better for me.

EDIT: So your real voltage is actually 1.32V under load though? That's a pretty big vdroop, so your chip isn't actually that bad then. Does that mobo have loadline calibration? My DS3P does..... just have it set @ Auto and vdroop is quite low. Usually loadline calibration isn't recommended but I think if vdroop is that bad it might be worth considering....

 

[taltamir]

yea. my BIOS shows the default voltage as 1.325v VID, but at 3.3ghz @1.45 on bios I am getting 1.33v load after droop, 1.39 idle.

What is load line calibration? there is no mention of it in the big OC guide in this forum.

 

[Extelleron]

Originally posted by: taltamir
yea. my BIOS shows the default voltage as 1.325v VID, but at 3.3ghz @1.45 on bios I am getting 1.33v load after droop, 1.39 idle.

What is load line calibration? there is no mention of it in the big OC guide in this forum.

Loadline calibration helps eliminate vdroop; so if you set 1.45V in BIOS with LLC you would be getting very close to 1.45V real voltage. Usually it isn't recommended because vdroop is actually part of Intel's specifications and is actually a good thing. But when you have 0.1V+ vdroop like you have, it can really get in the way of overclocking. Even though you have 1.45V set in BIOS, that is meaningless because under load when the voltage matters, you are only getting 1.33V. So without LLC what some oc'ers would just keep increasing BIOS voltage to compensate. So maybe they would set voltage to 1.6V and then real voltage under load may be 1.45V. But that is a dangerous thing to do and that is why some people will use LLC.

 

[John2777]

Hope you don't mind if I join in here and ask a few questions. I also just got a Q6600 SLACR from Newegg. I picked up a GA-EP35-DS3P Rev 2.1 on Ebay and thought I'd try a bit of overclocking. I'm a new to overclocking... in fact, this is really my first time... but I'm thinking I might be doing something wrong because my numbers look pretty high in comparison to some others I have read about.

My setup:
GA-EP35-DS3P
Patriot DDR2 1150
Xigmatek S1238 Cooler
Q6600 "SLACR"

I set my voltages in the Bios (F4c) to AUTO, then went for 9X400 with the divider set to 3 and PCI-E buss to 100. No problem booting into XP. I did some testing with an audio program which uses all 4 cores and ran CPU usage up to 95% on all 4 cores. But I have noticed that by system will not reboot into the same settings at these levels, but switches the CPU Clock setting to "Disable" and then goes into stock settings. Is that normal?

I can get it to reboot into the OC settings if I am about 3.2-3.4ghz, but that also depends on how hard I was driving the CPU before I rebooted. The harder I drove it the less likely it would reboot into the OC settings. Do I just need to let the CPU cool down more?

I also did a 8x475 for 3.8ghz, which kinda surprized me. It booted fine and was able to do some stress testing on the audio program. But I noticed that it didn't perform as well as 9x400 at 3.6ghz.

So I guess I am wondering if I might be doing something incorrectly, or is my CPU unusually good?

 

[taltamir]

normally using auto voltage increases for OC in bios is considered a very bad thing to do. I tried it with my board (EP35-DS3R v2.1) and it was pushing over 1.5volts on an E8400, which is way too high. (permanently damages the chip, kills it in under a month)

If it switches to disable it means that the OC is so unstable that it will not even boot (which prevents you from going into bios and changing it), to prevent the need to mess with jumpers it detects such a thing and disables your OC. Try again with lower OC or more volts.
Make sure you change the ram multiplier as well, and manually increase the CPU voltage. but don't go too high as it will damage your CPU. (over time it will kill it, but for quick durations it will just degrade it, lowering its max oc).

never, ever, reduce the cpu multiplier while overclocking. You want it as high as possible.

there is a sticky thread in here with a guide to overclocking, read it from begining to end. It is not too long.

 

[firewolfsm]

Dunno how 2.9GHz (415*7) with 1.1625V plays into all this, I haven't tried to OC any higher than this. Loads at 50C though, kinda hot considering how low the voltage is.

 

[Acanthus]

Just fair warning, the MOFSETs get extremely hot on the DS3R and can fail after a long time running an OCed quad. Tossing some sinks on those would be a very good idea.

RIP my DS3R :brokenheart:

 

[Acanthus]

Originally posted by: firewolfsm
Dunno how 2.9GHz (415*7) with 1.1625V plays into all this, I haven't tried to OC any higher than this. Loads at 50C though, kinda hot considering how low the voltage is.

Thats not hot at all for a quad on a freezer 7, the ACF7 is a midrange cooler.

50C loads on a Q6600 is sexy regardless of the cooler actually.

 

[John2777]

Originally posted by: taltamir
normally using auto voltage increases for OC in bios is considered a very bad thing to do. I tried it with my board (EP35-DS3R v2.1) and it was pushing over 1.5volts on an E8400, which is way too high. (permanently damages the chip, kills it in under a month)

Well, I definately don't want to kill anything! I would say that understanding the voltages and how to set them is the challening part of overclocking. I had read an article reviewing the GA-EP35-DS3P and the reviewer mentioned how nice the AUTO voltage feature was. I'll look to see what voltages are being applied and try to determine if my Q6600 is being degraded. I have to admit that the extra power the overclocking gives is quite alluring, but not at the expense of killing the CPU.

 

[Acanthus]

Originally posted by: John2777

Originally posted by: taltamir
normally using auto voltage increases for OC in bios is considered a very bad thing to do. I tried it with my board (EP35-DS3R v2.1) and it was pushing over 1.5volts on an E8400, which is way too high. (permanently damages the chip, kills it in under a month)

Well, I definately don't want to kill anything! I would say that understanding the voltages and how to set them is the challening part of overclocking. I had read an article reviewing the GA-EP35-DS3P and the reviewer mentioned how nice the AUTO voltage feature was. I'll look to see what voltages are being applied and try to determine if my Q6600 is being degraded. I have to admit that the extra power the overclocking gives is quite alluring, but not at the expense of killing the CPU.

65nm CPUs are more tolerant to voltage than 45nm. Your Q6600 should be able to handle up to 1.45v just fine if its a G0.

 

[OCGuy]

Taltamir, what multi-threaded apps do you run that would make you go from a E8400 to a quad? I did it for benchmark-whore purposes only.....do you encode or something else that actually uses 4 cores?

 

[dakels]

One strange problem I have on my Q6600 G0 is that my third core seems to be a lot more prone to failure then my other cores. Core 3 always fails first for me in Prime95. Is that common for people to see 1 core on a regular basis failing faster then others? It is stable at its current 3.0ghz OC. Thankfully most apps hit core 1 and 4 for some reason although it does seem a bit random at times.

 

[firewolfsm]

With my old athlon X2, I had one core that could only hit 3.1, the other could hit 3.4. I tried running two instances of prime, on the bad core I just disabled error checking so it would generate heat, and the other core was still stable. It can happen, and it sucks, a lot like one bad stick of ram.

 

[RussianSensation]

Download MemSet 3.4. Set Performance Level to 6 or 8. Although I selected board settings at Turbo, this was locked for me at 12. I set it to 8. WinRar skyrocketed from 1750 to 1900KB/sec for "free".

When finding top overclock, dont set Vcore to Auto. You wan't to incrementally raise Vcore to figure out what works and what doesn't. Also disable all power saving features for now like EIST, C1E, etc.

 

[John2777]

Well, my GA-EP35-DS3P is trying to kill by Q6600! When I had the voltage set to AUTO in the bios, it pushed the vCore to 1.55 at just 9x333. I makes me wonder what my board was running at when I ran it at 9x425! That would be about 3.8 ghz and pretty hot.

Well, I guess I'll just have to do things the old fashioned way... start at the bottom and work the vCore voltages up. It was fun to see the 'puter running at 3.8ghz... but not so fun to know it was probably frying the inards. Oh, well, being a novice was bliss, until I found out all the things I was doing wrong. My God! I'm dangerous with this motherboard!

 

[Acanthus]

Sounds like youve done pretty well, yeah that Auto setting is a killer. Kind of ironic that its when you set the voltage manually that it warns you about giving it too much voltage

 

[taltamir]

Originally posted by: John2777
Well, my GA-EP35-DS3P is trying to kill by Q6600! When I had the voltage set to AUTO in the bios, it pushed the vCore to 1.55 at just 9x333. I makes me wonder what my board was running at when I ran it at 9x425! That would be about 3.8 ghz and pretty hot.

Well, I guess I'll just have to do things the old fashioned way... start at the bottom and work the vCore voltages up. It was fun to see the 'puter running at 3.8ghz... but not so fun to know it was probably frying the inards. Oh, well, being a novice was bliss, until I found out all the things I was doing wrong. My God! I'm dangerous with this motherboard!

yea, same voltages here... it was rock solid and just worked, it will OC higher and higher and higher and kept on working, but it was giving very high voltages which would fry the chip in under a month.

Originally posted by: Ocguy31
Taltamir, what multi-threaded apps do you run that would make you go from a E8400 to a quad? I did it for benchmark-whore purposes only.....do you encode or something else that actually uses 4 cores?

For one thing, certain games... but I am also an extreme multi tasker. firefox allows you to save your open tabs when you exit. I have over 100 tabs in 10 seperate windows open at one time (I have gotten to 700 tabs before).
Opening and closing it is SOOOO much faster now. It is also EXTREMELY faster in 64bit mode (which I use).

Originally posted by: dakels
One strange problem I have on my Q6600 G0 is that my third core seems to be a lot more prone to failure then my other cores. Core 3 always fails first for me in Prime95. Is that common for people to see 1 core on a regular basis failing faster then others? It is stable at its current 3.0ghz OC. Thankfully most apps hit core 1 and 4 for some reason although it does seem a bit random at times.

Well, since each core is basically a seperate peice of hardware glued together, that is how it should work, the weakest one will always fail first. Also each core runs at a different temperature, always! We don't assemble thing atom by atom, so every item is different in the molecular level, and when you deal in nanometer wide items, then a few atoms difference makes a difference.

Originally posted by: Acanthus
Just fair warning, the MOFSETs get extremely hot on the DS3R and can fail after a long time running an OCed quad. Tossing some sinks on those would be a very good idea.

RIP my DS3R :brokenheart:

Thanks for the suggestion, I will look into it. Anyone knows a good one to buy? is arctic ceramique the best thermal glue to use for those?

 

[Acanthus]

Originally posted by: taltamir

Originally posted by: Acanthus
Just fair warning, the MOFSETs get extremely hot on the DS3R and can fail after a long time running an OCed quad. Tossing some sinks on those would be a very good idea.

RIP my DS3R :brokenheart:

Thanks for the suggestion, I will look into it. Anyone knows a good one to buy? is arctic ceramique the best thermal glue to use for those?

Ceramique is great, it's what i use for thermal epoxy.

 

[taltamir]

very interesting tidbit here... It would go for hours and hours of memtest and OCCT tests without a hitch... but if I opened firefox 64bit with a saved session of about 100 tabs. (in 10 seperate windows), WHILE OCCT was running a burn in of both ram and CPU at once... then it will blue screen. I raised the voltage a big, again, hours of individual tests, but FF WHILE doing a test (not by itself) got a blue screen... raising the voltage again stopped it though.

I wonder what is it that firefox taxes that caused it. You know what, I think I might know, 64bit FF is FPU starved, and 64bit calls for having 8 instead of 3 FPU in 32bit. FF64 gets a tremendous boost to speed from 64bit processing when handling massive amount of pages at once like that. I bet the OC tests just don't stress all the FPU at once...

Or maybe my guess is completely off, it is just interesting to see that tidbit.

Anyways, the system is rock solid at 3.3ghz, bios setting: 1.46250V, measured voltage 1.41 idle 1.36load with a +/- of 0.02v (it stays on the voltages listed above, but I measure a fluctuation of 0.02 every few minutes, lasts less then a second). and the room was getting hotter and hotter over time, it was "beating the AC" so to speak...

However the ram and the northbridge were scalding hot to the touch, the cpu was running 58-63c load (coolestcore-hottest core).

I decided to try aiming at improving me performance per watt and reducing the heat it generates.
So I lowered the FSB to 333. Lowered the ram multi from 2.5x to 2.4x, and lowered the voltage setting on the mobo from 1.46250V to 1.35V. I am now measuring 1.26v on load and 49-52c temp. I am going to see about lowering the voltage even further while maintaining the 3ghz mark. My mobo is reading the VID from the CPU as 1.325, and I am now at 1.35. It will be very awesome if I could actually go below that, and end up undervolting while over clocking...
Touching the ram and northbridge heatsink now shows me that they are hot, it is significantly less hot then before.

Also my room feels cooler now, I have it with a ceiling fan on and with the door open and with the AC on in the house. With the 3.3ghz setting the room was getting hotter and hotter. Now it is getting colder.

 

[taltamir]

OH MY GOD! I overclock to 3.0ghz AND I am undervolting at the same time! this processor is made of pure win!

 

[John2777]

I have a question on stability. When do you consider a OC setting stable and adequate? If Prime 95 runs for 15 minutes without errors, is that sufficient to delcare victory? Or do you have to continue doing this for several hours?

My next observation, and taking into acount that I am a newbie to this overclocking stuff, I am not having much luck with the methods outlined in the OC thread. I would get errors in Prime 95 on just about any setting when the BIOS has the STANDARD setting under performance.

So, I started doing things from a different perspective, and did the following:

Performance: Enabled EXTREME
EIST: Enabled
CE1; Enabled
9x375 (3375)/ divider 2.5 (950)
vCore+1.4

All other settings are pretty normal according to the guide. I had Core Temp and CPU Z open to watch whats going on and I ran Prime 95 for 20 minutes and all was Great!

So, I decided to move things up at bit. I left everything the same, except I changed the CPU settings to 9x390 (3510) and bumped up the vCore to 1.4125.

Everything looked good on Prime 95 for about 7 minutes, then one of the cores errored on a ROUNDING Error. Says that it returned 0.5 and was expecting 0.4 or less. Is that really a big error? I am letting the test run and the other 3 cores made it to level 10 and started over. Temps are pretty go, at 65 C on the most stressful moments of the testing, but doing well.

I noticed that the core that errored was basically the one running the coolest. Perhaps it didn't get enough voltage momentarily as it clearly is not accepting as much current at the others. So, perhaps the fix is to bump up the vCore temp another notch.

Anyway, this isn't according to the rules, but it's doing great.

P.S. Can you serf the net when prime is running inthe background?

 

[taltamir]

I ran 10 hours of memtest+, (I ended up reducing the oc on it after this stability test because ram and northbridge were SCALDING to the touch, now they are just warm during load, cool during idle). Make sure everything is on manual, otherwise the board might automatically tighten timing as you reduce speed which might introduce instability.

I ran OCCT v2 for hours and hours in mixed mode (maxes out CPU and RAM at once). I found out something very interesting, typically stability goes from: blue screen to occt reports CPU error and aborts the test to perfectly stable as you increase voltage. I noticed that even if it appears perfectly stable, if OCCT is testing the RAM and CPU AT THE SAME TIME (it can do it) and then I open a firefox3 64bit with a saved instance of 10 windows of 10 tabs each (100 pages) while OCCT is doing it (CPU only test isn't enough), I would blue screen, unless the voltage was high enough... Even if everything else INDICATES I am stable... Two ticks below the voltage I am running right now was not stable (with firefox test), even if one tick below is stable, I don't think I would use it as a "just in case" kind of deal. Running long hours stable on your benchmark program doesn't mean the computer is stable, it means that the computer is stable with the stress that the benchmark software is giving it. This is why I like to always be a little under my max achievable OC. gives a bit of headroom.

I noticed that windows vista welcome screen will have parts of it missing as if they were smeared with black. I am not too sure about THAT one but it appears it MIGHT to be OC related, and sensitive when everything else shows stability.

Basically there is a range of acceptable voltage, go too low and you are unstable, go too high and you are unstable. The min and max voltage both increase as you increase the clockspeed.
Disable the performance thing (automatic, maybe dynamic overclock, bad idea, it is just not good), disable EIST (allows software in windows to dynamically change those settings, bad idea), disable CE1 (reduced vcore and multiplier when in idle, reducing you performance by about 5% due to errors in the idle detect mechanism and reducing your OC stability). leave thermal protection 2 on (lowers clock and voltage if temp rises over a max temp, this will likely prevent your computer from physically and literally catching on fire, it will not however prevent you from damaging the chip).