Sign in with
Sign up | Sign in
Your question

Q6600 OC w/ P35-DS3L, sporadic Prime95 results

Last response: in Overclocking
Share
March 19, 2008 2:57:57 PM


Specs:
COOLER MASTER Centurion 5 CAC
Antec earthwatts EA500
GIGABYTE GA-P35-DS3L
Intel Core 2 Quad Q6600
ARCTIC COOLING Freezer 7 Pro
Crucial Ballistix 2GB (2 x 1GB) 240-Pin DDR2 SDRAM DDR2 800
BFG Tech GeForce 8800GTS (G92) 512MB
Windows Vista Home Basic 32 bit (OEM)

I'm using CPUZ, HWMonitor, Coretemp, and Prime95 latest version, Small FFT's
Multiplier stays at 9x, memory is at 2.0 for 1:1 ratio.


Memory set (auto): 5-5-5-18

3.4ghz@1.35v = crash after ~min
3.4ghz@1.375v= TM2 activated. TM2 is a setting in the bios that changes the multiplier from 9x to 6x when the cpu gets too hot. It seems to kick in when my core temps >65C. I haven't shut it off because I want to keep them below 65C anyway.
3.3ghz@1.3v = PRIME95 error < 1 min

Memory set (manual) 4-4-4-12

3.3ghz@1.325v = PRIME95 1 error after 2hr 26min, cores: 63, 61, 59, 63
3.3ghz@1.33125v = PRIME95 1 error after 2hr 48min, cores: 62, 62, 60, 58
3.3ghz@1.35v = TM2 kicked in. 2 cores reached 65C before the MOBO switched to 6x from 9x.
3.3ghz@1.3375= PRIME95, 2 cores failed with in 6 min
3.25ghz@1.3v=system crash while running PRIME95 after 1hr 45min. No core got hotter than 62C

Note the results in bold. These make no sense (at least to me).

HWmonitor and core temp both give the same numbers always. Are these numbers correct or do they have to be calibrated? The OC guide stickied here doesn't mention having to calibrate these programs.

What about Northbridge temps?
March 19, 2008 4:17:16 PM

Just to be safe, follow Comp's guide and calibrate your temps. You never know just how far off you are until you level the variables. His guide shows you exactly what temp you should see at specific settings.

It's totally worth it to take the time to do.
March 19, 2008 8:57:47 PM

The ones in bold, are they reproducible? If not, then it's totally normal.

One time I ran Blend test for 8hours fine, then the next time I ran it, it failed within 5mins.
Related resources
March 19, 2008 9:08:34 PM

Evilonigiri said:
The ones in bold, are they reproducible? If not, then it's totally normal.

One time I ran Blend test for 8hours fine, then the next time I ran it, it failed within 5mins.


Each of these results occured only once.

But if a processor can run the program for 8hrs (we'd usually assume this is stable), and then fail in 5 mins....doesn't this mean that it is NOT stable?

How then do we determine that our OC is stable?

Does anyone know how this works? I'm guessing that the processor has a certain probability to fail a calculation based upon the stress conditions. Running the test longer increases the likihood of observing a crash event (more calculations). We assume that if we run the processor for enough hours and don't observe a crash event, that the probability of the event is sufficiently low and thus the processor is stable. If this is true there could be "lottery" events (an extremely unlikely event occurs) and an otherwise stable processor crashes.
March 19, 2008 9:22:06 PM

Yes, if it fails at any given time, it's unstable.

Just by pure luck, the ram I had held on for 8hours, making me think it was stable. However I crashed a lot and had issues so I ran it again, only to have it fail.

It is ideal to run Prime95 for 24hours, and some people had it fail on the 23rd hour. However, this is only necessary if you want your system 24/7 stable. 8hours of small FFT and 12hours of blend test is recommended for normal use.
March 20, 2008 2:06:09 PM

I ran small FFT for 7hr and 40 mins last night (I shut it down this morning, it never crashed). I noticed that the program tests "8k", '12K", "16k" ...."64k." It takes ~6 hrs to test all the "k"s. I'm guessing this is why the guide suggests running it for six hours?

Anyway, I'm happy that I've got my first "stable" OC.
3.2ghz@1.3375v. Cores max temp 65, 65, 62, 60. I'm going to try decreasing the voltage some more as well. What vcore is would you expect w/ my setup?



March 20, 2008 2:21:46 PM

You will have to find that out for yourself. It's all trial and error. While some may run 3.2 at near or even below stock voltage, some will take a few bumps to get stable.
a b à CPUs
a b K Overclocking
March 20, 2008 4:12:29 PM

billiardicus, in my ongoing thermal testing for the Temp Guide, I have noted an interesting phenomenon whereby some Q6600's appear to be inherantly unstable at 3.3Ghz. Call it a "speed bump", but I suspect it is related to memory strapping or perhaps memory controller (Northbridge) issues. Regardless of FSB settings, Vcore and / or Vdimm and timing combinations, I found that these particular processor samples were impossible to stabilize from 3300 to 3383Mhz. Settings were found which were 100% stable at 3250 and 3400Mhz.

As I read your posts, I noted that your testing seems to skip around a bit. It is preferable to first choose a frequency, then incrementally ramp up Vcore until stability is established and settings are documented after the successful completion of at least 12 hours Small FFT's. It's a long and tedious process, but it's the only way to nail down the highest stable overclock within the constraints of safe temperatures.

Regarding temperatures, Hardware Monitor reports CPU and Core temperatures, while Core Temp will not report CPU temperature, which is the only temperature Intel specifies. Further, Hardware Monitor, Core Temp and Everest may over-estimate Core temperatures, because they all "guess" at the unknown value of "Tjunction Max" which Intel does not support or publicly document on Core 2 desktop processors.

To deal with this problem more effectively, there's a new Sheriff in town; Real Temp 2.24. The following link is to the Real Temp thread, http://www.xtremesystems.org/forums/showthread.php?t=17... which has exploded to 20 pages. It's a bit of a read, but lends fresh perspective and understanding to the perpetual Tjunction Max / Core temperature debate. The link to download Real Temp is on the first post, which also explains how the program works, and what problems it fixes.

Real Temp contends that all the popular temperature utilities such as Core Temp report excessively high values on many processors, so depending upon which variant you're running, Real Temp will typically report Cores from 5c to 10c cooler than other utilities. I strongly agree with Real Temp, however, I think it could be perfected if an additional variable was considered.

Only SpeedFan http://www.almico.com/speedfan.php can be "calibrated", which is precisely why I feature SpeedFan exclusively in the Core 2 Quad and Duo Temperature Guide http://www.tomshardware.com/forum/221745-29-core-quad-t... as the monitoring utility of choice. My calibration methodology is based upon the use of known values, so if you follow the Guide to the letter, then you'll have accurate SpeedFan temperatures, which will report Core temperatures right in the middle between Real Temp and Core Temp.

From the Guide:

Section 6: Scale

Scale 2: Quad
Q9x50: Tcase Max 71c, Stepping C1
Q9300: Tcase Max 71c, Stepping M1
Q6x00: Tcase Max 71c, Stepping G0

-Tcase/Tjunction-
--70--/--75--75--75--75-- Hot
--65--/--70--70--70--70-- Warm
--60--/--65--65--65--65-- Safe
--25--/--30--30--30--30-- Cool

Hope this helps,

Comp :sol: 
March 20, 2008 7:37:09 PM

Thanks Comp. I've read your guide several times and found it helpful.

I will check out Real Temp, but right now the TM2 feature of my motherboard changes the multiplier to 6x whenever one of the cores exceeds 65c. I really don't know how this function works, but it says it does this when the CPU gets too hot. I don't know it's criteria for determining "too hot" but I'm reluctant to disable the feature. But as long as TM2 is limiting my temps, it doesn't really matter how I measure them. My vcore is thus limited to ~1.3375 unless I improve the HSF.

Your comment that the Q6600 is inherently unstable at 3.3ghz seems to make sense. When I was playing around with 3.4ghz I was having less trouble.
a b à CPUs
a b K Overclocking
March 20, 2008 7:49:29 PM

If I'm not mistaken, the TM2 feature allows the temperature threshold to be adjusted, so it should be possible to raise your Vcore and OC ceiling.
March 20, 2008 8:41:06 PM

Hum...googled this and didn't find anything, but a lot of guys are disabling the function. According to your guide, I should be good up to 70C....

I've come to realize OCing is habit forming. I don't need the extra power/speed, I just WANT it. 3.2ghz just doesn't sound nearly as exciting as 3.4 or 3.5 or (whistfully) 3.6 :) 

I've already put in an order for the new Xigmatek HDT-S1283 based upon FrostyTech's review. It's cheap (~$35 from Mwave), and I really want to compare it to the Arctic Freezer Pro.
March 22, 2008 12:40:39 AM

move you TM2 temp to 75.as its the safest temp for overclocking.just to let you know when i OC i have got the core upto 85C.plus another thing is that if your CPU does get TOO HOT.the whole system will just crash .so really you cant really burn your CPu even when you max it.but what is your VID(you can get from Core Temp)?
March 22, 2008 12:59:20 AM

VID=1.25

How do you change the TM2 temp? Only immediate options are enable/disable.
March 22, 2008 3:46:21 AM

disable it!thats what i do.you know when you OC your CPU.if you see the temp go too high you can stop the test manually.the TM2 temp should be 100C set by Intel integrated into the CPU itself.well the system will shut down itself when everything is too hot.

i got my high volt Q6600 vid=1.3V and i have tried 3.4G stable at 1.475V with Scythe Infinity.according to that i would estimate you should achieve 3.4 with 1.41V remember to add about 0.1V to your VTT\FSB and your northbridge.As i would say dont let the cpu core go above 75C.65C is a VERY SAFE MARGIN.nothing will go bad under 100C.

Now i have a lapped(only grit 600) Artic Cooling Freezer Pro 7.Stable at 3G(375x8) with 1.33V.NB/VTT 1.4V full load would be at 69C.I wouldnt push it any further in my high standard volt.but yours should achieve higher easily:) 

REMEMBER!!!YOU CANNOT FRY CPU THESE DAYS THE MOTHERBOARD AND CPU ITS SO ADVANCE AND SELF PROTECTIVE!even you disable the protect mechanism there is another protection further up the temperature range.no one can disable that.and its not available to users.
March 22, 2008 3:58:02 AM

Yeah, I know its there as well. I had my water block on wrong, got to like 100+ c 10 times, and after 10 seconds, it turns off, even though I have every option disabled that can be disabled.

--Lupi
March 22, 2008 4:00:54 AM

so as lupi said!be wild with your voltage!!!:p  should be fun

lupi where you from?
March 22, 2008 4:04:05 AM

I am in seattle! Yee haw! Yeah, well, since my accidents, I know better what the things can withstand! 1.8125 volts and booted fine, cores idle at 85c, and it LOADED windows? I'd hate to see what the active temps were while loading junk, if Idle was 80+ c, lol!

BUT.. though I meant to set it at 1.3125, I didnt. And it survived. Point, it can take more voltage than ya think! How safe it is? I dont really know that, but both my quads are still here OC'd!

--Lupi!
March 22, 2008 4:12:27 AM

im in uk 5:08am now!feel so sleepy.but would you run your pc when load 85C everyday?where is the person who need help?he is not replying.

i got a q6600 burnin in my room right now.380x8=3.04G@1.33V FSb/VTT=1.425V NB=1.425V.i do that just to test the FSB\NB than next the memory!!!haha

feeling abit hot and dry now!all my windows are shut!!!lol
March 22, 2008 4:13:49 AM

why you not on msn all the time!remember me?i just check my mnsn its you that i added sometime ago!!!
March 22, 2008 4:19:43 AM

He's off line, hehe. And not every day! But I settle for less when stress testing. I go for 75c, not 65, because I am crazy like that! But the TRUE even makes that hard! And at 3.8 no less.

Better his the sack, its like 10 30 here, hehe, and I just got up an hour ago! night shift is awesome!!

--Lupi!
March 22, 2008 4:21:13 AM

Wait, is that what it took to be stable? 1.475 at 3.4? See, thats all VID there. Mine is a 1.2625, and I get 3.6 @ 1.38 under load.

--Lupi
March 22, 2008 4:22:03 AM

you never look at my thread do you?i were asking whats the best software and method to stress the memory?
March 22, 2008 4:23:59 AM

Well, at 3.8 and an fsb of 475, my memory is not a problem at 950 Mhz in sync. And unless I want 3.9, the memory is staying right there, because my next option is something like 1190mhz. Hehe!

--Lupi

Oh, we are spamming the hell outa this guys question! I hope we answered it for him, lol!
March 22, 2008 4:28:14 AM

yeah i hope so.i think he got enough info.he,can always ask if he need to.

whats the memory timing and voltage on your at 950mhz?really need it!!!

can you come on msn and talk?rather then "spamming" in here lol
March 22, 2008 4:30:55 AM

Its the RAM, lol, but I am looking for PC 6400 that will work! I think I found one that should do the job! I hace Some GSkill 1066 5-5-5-15 @ 2.0-2.1 volts. Its at the 5-5-5-15 @ 2.0 @ 951.

So thats underclocked, technically, but in reality, thats faster than a lot of people can reach in sync!

--Lupi!

Let me re DL msn, hehe, on a 64 bit comp. Just a min.
March 22, 2008 4:35:21 AM

alright cool man.i wonder how many comp you have!lol see you on msn
March 22, 2008 7:33:23 AM

4 total, two q6600s, a amd phenom 9500 and a core solo 1ghz puny samsung 7inch vista comp.
March 22, 2008 3:36:08 PM

i got an old celeron 478 on 3Ghz stock voltage!lol
!