I can't for the life of me figure out all those steps to calibrate the idle.
Can anyone simplify it.
Printable View
I can't for the life of me figure out all those steps to calibrate the idle.
Can anyone simplify it.
Undervolt vcore as low as it will go, underclock FSB to 266, change multi to 6x. Boot windows, turn fan to max speed. Get ambient temp around the cpu (or the fins if you can). Adjust calibration for cores so that they are around 6-7C above the ambient temp.
loonym: That's how I see it. If it wasn't for RealTemp pushing this subject we might not have ever found out.
BeastNotro: You might want to attend the IDF. I heard they might be bringing that subject up there. ;)
Where is it at Unclewebb?
Also ...... i ain't lower my stable OC clock and all that ... to get idle calibration .... Idle between cores are this 32-31-27-29 ... so it ain't that far off.. :)
There must be a away to do so with OC idle.
Sure i can offset it to match my lowest idle core, but would that be correct?
Does the P5Q allow OC profiles? My P5K when I still had it allowed OC profiles. You can save your profile, and then undervolt. Calibrate and load OC profile.
Yeah it has 2 spots for such a thing, but i don't like loading back from profiles as before i tried that and my RIG seemed to run slower, chuggier.
I may try it, but we will see as like i said it is not that far off .. my room temp is around 24C... so if you look at what my OC idle temps were in my above post .. they are like 5-6C above room temp.
So can calibration be done on OC idle temps ..... question is ... why not?
Quote:
Originally Posted by jas420221
Great THANK, that very useful. I will look forward to next weekQuote:
Originally Posted by unclewebb
The reason I recommend low MHz and low voltage is so users have a common baseline to the testing I've done. There are so many variables including things like case air flow that it's very difficult to say what your core temperature is when you are overclocked by just looking at your idle numbers. For me going into the bios and turning on C1E so the multi drops to 6.0 at idle and setting the MHz and core voltage to the lowest amount is not a big deal. I don't have to play with any of the other settings and it's easy to re-boot and go back to my overclocked settings later on.
If you want the easy way then take your previous idle temps and equalize them to about 32C or 34C with the case open and you'll probably be pretty close.
Intel IDF is in San Francisco. I haven't received my invitation yet to ride down on the company jet! :D
Hey Unc!Quote:
Originally Posted by unclewebb
Thought I'd finally report back. Sorry it took so long but I'm in the middle of a 6 out of 7 day work week and 13 hr days just don't leave me enough time to keep up with the folding farm and figure out what I fraked up trying to beta software.
I'm gonna have to chalk this one up to user error LOL... The way I did the "drop in" upgrade was to stop/kill 2.60, rename the exe and unzip the new files in the working directory. I fired up 2.71 and did my folding @ home updates and then checked my logs before I posted. My guess is what happened (I think you were right on the mark) was that with all four cores folding, the processor didn't give up the 2.60 version when I stopped it before I got the 2.71 version restarted. Probably 2 minutes elapsed between me stopping/renaming 2.60 and restarting 2.71.
Either way, the log today looks all good, so I hope you didn't spend much time looking for my "ghost" user induced problem... :shrug:
Yeah.. code tags are :cool:. Takes up a lot less room on the page.
Lovin the new version! :clap:
gbeans99
I started using RealTemp a few days ago, and calibrated it as described in the manual page on the website.
There seems to be a problem with it with wolfdale's imo.
When I calibrated (fsb266, 6.0* multipiler) The ambient was 17C, so I set it for an adjustment of -1.0 which gave core temps of about 24C.
Now today the ambient was 32C and core temps were showing less than ambient being around 30C, so I reset it to no correction (0.00) and then it read core's at about 34... this can't be right so I set it to +1.0 and then it showed temps about 37... all these readings @ idle.
HW+Clocking in my sig.
Fungus: Most of the sensors I've worked with are pretty consistent. They might be wrong but they are consistently wrong and either read too low or too high at lower temperatures. This behavior doesn't tend to change from one day to the next.
Your results don't make any sense to me either. The ambient goes from 17C to 32C which is a change of 15C but your core temps only go from 24C to 30C or a change of 6C. Something definitely isn't right with that. It seems like you might have some sticking sensors. If your sensors were stuck the other day when you first calibrated then the calibration factor you used might have been completely wrong. A stuck sensor simply reaches a fixed value as it is cooling down. As the CPU gets cooler, it will just keep reporting the same number. There is no way to calibrate a sensor if it has these sticking issues.
Set your calibration back to zero and see if there is a core temperature number that your sensors don't ever go below. You might have to wait for some cooler weather. Drop the core voltage and MHz, open up your case, turn the fans on high, stand on your head and do whatever it takes to get your temps as low as possible. Not all sensors can be used to report accurate idle temperatures. If you can find out if your sensors have a sticking problem and what temperature that problem happens at then maybe you can use the RealTemp calibration feature.
In a few days when Intel tells us everything about these sensors there should be lots of software available that can provide you with accurate temperatures, as long as your sensors aren't sticking.
gbeans99: Don't worry. I love hearing about user error! Far better than having to admit that I screwed something up. You had me scratching my head there for a few minutes but I think I figured it out. I like having the date in the log file now. Nice idea.
One of the sensors sticks and the other floats to +3 degrees. The one that sticks sits at 22C and the other will read between 22 and 25C.
Ambient is 21.5 right now, and core readings are 28/28... so I think the +1.0 is correct.
I'll see tomorrow as it is supposed to get up to 32-35 or so.
Something is a miss here.
http://i31.photobucket.com/albums/c3.../Untitled9.png
Well, nothing stands out for me. Other than 3 of your cores appear to be reading the same thing at the time of the screenshot. I have no idea what that other program is, so why don't you spell it out.
The other program is everest's OSD on screen function.
OSD = On Screen Display
Yes, I know that Beast. Thank you though. I was referring to the other program, the intel burn test. I've never used it, therefore I can't really read much from it other than it passed.
Oh well in that case i have used it many times in the past few days.
And yeah all it is a run he did and it passed, but i think he is refering to his temps being 25-25-25-22 ... which would be after the test ran .... not sure if it is a calibration issue or not as his MAX was 39C which is weird if he is on Air Cooling, but i can't see anything refering to what cooling he is using as that program makes mine and many others CPU go to 74-80-88C max and i idle at 32-30-28-29.. air cooled read my specs.
We need him to post what he was refering to and as to why he is?
Maybe he noticed that RealTemp reports 25,25,25,22 and Everest reports 25,25,22,25. RealTemp is one of the few programs that re-orders the cores on Quad core processors into there correct physical order based on APIC ID.
Yes, I noticed that on my machine. I changed the order in Everest to match when I saw that.
If there's anyone here that wont be at IDF to give us the data from their pda/cell, anand will have it up aug 21st.
old news here quoted from anand:
We met with Intel today and have some interesting IDF (Intel Developer Forum) 2008 news. IDF is scheduled this year for August 19 - 21 in San Francisco. It is shaping up to be an exciting event for those tech geeks like us as X58/i7, Larrabee, and System on Chip (SoC) technologies will be discussed in detail among other things. We learned this afternoon that full disclosure of Intel's existing 45nm processor Digital Thermal Sensor (DTS) specification will be presented on Day 3. This 50-minute technical presentation on DTS (course #TMTS001) will start at 1:40 pm. A complete schedule of events can be found by visiting Intel’s official IDF website.
Benson Inkley, a senior power/thermal engineer with Intel, is prepared to address nearly every aspect of DTS functionality for the attendees. However, perhaps the biggest surprise to come out of his presentation will be the first-ever public disclosure of the maximum Tjunction value for all Core 2 Duo/Quad/Extreme desktop processors built on current 45nm-process technology.
Armed with this information, seasoned application developers and amateur coders alike will finally have everything they need to implement the most accurate, real-time core temperature display tool possible. We discussed this topic in our Intel Core 2 Duo E8500 review back in March 2008, except our discussion left quite a few unanswered questions flapping in the breeze. Come next Thursday, anyone walking out of Mr. Inkley's technical session will have all the knowledge needed to lay any longstanding DTS questions to rest.
We applaud Intel for recognizing the enthusiast community’s interest in this subject. Overclockers, performance enthusiasts, and everyday users will finally be able to monitor their CPU’s individual core temperatures without wondering if the reported results are accurate or not. Check back here on Thursday, August 21, for a detailed update on DTS.
jaredpace: This has been the top story at Tom's Hardware this weekend as well. Let's hope that Intel doesn't hold anything back.
It was such a nice weekend that I didn't get as much accomplished as I thought I would. :cool:
Initial testing of the next version is looking good. I've added fully adjustable background and text color for Normal mode as well as Mini Mode. You can leave normal mode as is, if you like, and have a different color scheme for Mini Mode. Whatever you like! I think it should be pretty easy to go from this to a user selectable bitmap for the background in the near future.
http://img246.imageshack.us/img246/3990/rt273hl0.png
Nice look, where's mine? :D
Very nice!