Interesting read about the P4 Thermal Solution.

[jdurg]

Well, I got really bored tonight so I decided to do some reading over at Intel. While there, I came upon the technical datasheet for the Northwood P4 processors. The first thing that struck me as odd was that they didn't list the 1.6A and 1.8A processors. The lowest speed Northwood that was mentioned in the datasheet was the 2.0A GHz CPU. So I think this fully proves that all of the 1.6A and 1.8A processors are just marked down 2.0A processors. (This would just be a Captain Obvious statement as to why just about everyone is able to get a minimum of 2.0 GHz out of their CPUs.)

The most odd thing that I read was about the temperature specifications of the CPUs. As the clock speed of the CPU goes up, the maximum allowable temperature goes up as well. For instance, in section 5.1 table 22, they list the maximum operating temperature of the 2.53 GHz P4 as 71 degrees Celsius. The 2.0's maximum temperature is 68 degrees Celsius. However, this isn't the temperature when the clock throttling takes effect. This is just the maximum temperature at which the processor is guaranteed to perform properly. The entire thermal solution is explained in detail in section 6.3 of the datasheet. It is there where they say that the CPU will completely shut down if the internal diode measures a temperature of 135 degrees Celsius or higher! :Q Otherwise, the chip will just throttle off and on to cool things down if need be. Finally, the datasheet lists the maximum allowable voltage on the Northwoods to be 1.75 actual volts reaching the CPU.

It seems as if some of this data has changed from the original P4's datasheet. It seems that the Northwoods are able to withstand higher temperatures since the 2.53 GHz chip is able to run at up to 71 degrees with guaranteed reliability. After reading this, I am no longer worried about my CPU running too hot because my idle temperatures of 50 degrees and my load temperatures of 67 degrees are within Intel's specs. Also, my Vcore of 1.71 applied volts is within spec as well. Reading that document put a lot of my fears to rest.

 

[THUGSROOK]

nice - finally puts some facts into our "myths"

 

[jdurg]

Yeah. Although the datasheet was very technical, and a few things were WAY over my head, enough of it was easy to understand. I thought it was nice seeing the maximum allowable voltage to be 1.75 volts. This relaxes me quite a bit because I now know that my 1.71 applied volts is within Intel specifications. I was also very happy to see the temperature limits. I no longer am worrying about how hot my CPU is running at. I know very well that my highest load temperatures of 67-68 degrees fall within Intel's specs. I'm beginning to think that throttling may not kick in until somewhere in the 80 degree mark since complete shutdown doesn't occur until 135 degrees Celsius. So I think that everyone who is worrying about whether or not their CPU is too hot shouldn't worry too much. The voltage limit of 1.75 volts is where most of the attention should be put.

(BTW, since my CPU needs the 1.71 volts to be stable, it pretty much assures me that the 2.52 GHz I have right now is the chip's limit as far as overclocking. Can't complain, however, since I only paid for a 1.8 GHz CPU and I'm actually getting a 2.52 GHz CPU that runs faster than the 600 dollar ones Intel is selling simply because of the faster front side bus. )

 

[Wingznut]

Kudos to someone who isn't afraid to dig into a datasheet.

 

[jdurg]

Thanks. I just wanted some reassurance that my cpu was running within the specs that Intel set forth, and what better way then to look at the source.

 

[Duvie]

Damn you....I just bought a new HSF!!!!:|


I didn't like seeing 58c with my chip 1.6a@2.72ghz and 1.68v actual applied.....


Makes since though...As soon as I went over the 1.75v barrier to 1.83v atual applied it didn't last more then a week (my old 1.8a that is)..now it will only do 2.25ghz but at a very nice 1.58v...still respectable...think I will build my parents a machine....heck they would just love the 1.8 default...

good thread jdurg!!!

 

[mechBgon]

Next thing to do is for someone with an Alpha 8942 and a variable-speed fan to do benchmarks with the CPU at various temperatures, to identify the temperature at which the throttling begins to occur. Anyone?

 

[jdurg]

I too would like to know exactly when the throttling kicks in. I couldn't find that information anywhere in the data sheet, although I'll look through it some more to see if I can find out where. I know the CPU will shut itself off at 135 degrees Celsius, but I'm assuming that it will throttle before that point. Also, the maximum temperatures allowable for the CPU listed in the datasheet were for how hot the chips could run and still be guaranteed by Intel to run at that speed. So for the 2.53 GHz chip, Intel will guarantee its reliability up to 71 degrees Celsius. (Which also makes me assume that an overclocked chip can run up to that speed without throttling.) Just from personal hunches, and from what I could gather from the specifications, I don't think clock throttling kicks in until the chip reaches the upper seventies.

Also, the temperatures listed by Intel weren't taken from the internal diode. These were temperatures measured on top of the CPU right on the integrated heat spreader. So I'm then assuming that the 71 degree temperature would actually relate to a 75 degree internal temperature. So these Northwoods are built pretty tough. It's amazing how much information is in those datasheets.

 

[Duvie]

Hmmm...well I guess this puts the old temp fears out of most ppl's minds as even the retail on my 2.67ghz would likley keep it to under 70c load....

The voltage seems to be the main concern....Was there any other voltage tidbits other then max allowable 1.75v....That would seem to me that it should be safe to run higher but from that point how fast will it degrade on the cpus life??? According to that 1.75v then should be able to run fine and not cause the warrantied or life expenctancy of the chip any harm....


Are we sure you were looking at the p4a specs not the p4 willamette??? Just asking....1.75v is default for willamette chip, and I am surprised Intel would say anything above that default especially .25v more would cause no harm to the more fragile .13 micron parts...

 

[mechBgon]

Here's a link to the .pdf's. I'm scoping out the voltage question.

 

[mechBgon]

Holy convoluted wording, Batman!

I think the 1.75V mentioned in Table 6 on page 20 is meant to be a momentary maximum value, not sustained (note that they give a minimum voltage of -0.3V there as well, and we all know we don't run CPUs on negative voltages). You'll note in Table 7, Table 8 and Figure 4 that the voltage "window" varies with the current being drawn by the CPU. As Fig. 4 shows, 1.50V is the max, and as the Icc (processor current) approaches 60-70 amps, the voltage should be down to about 1.4V.

 

[jdurg]

See what I mean? The wording in that technical datasheet is about as clear as mud. I now notice a little blurb above table 6 that says "Extended exposure to the maximum ratings may affect device reliability." Also, table 6 shows a maximum storage temperature of 85 degrees Celsius. So perhaps it's at 85 degrees Celsius where possible chip failure could begin, and perhaps that is where the clock throttling would start kicking in? (Why couldn't they have released a layman's version of this datasheet? lol) So from what I can gather, the chip will be able to run at a voltage of 1.75, but in the long term it should be kept below this value. (Mine runs at a constant of 1.71 volts with a maximum jump to 1.73 volts. It generally tends to fluctuate between 1.66 and 1.71 volts when under load. Only an occasional jump up to 1.73 volts.)

 

[Duvie]

That is what I figured....Now we are back to what we already believed and that is run it at spec 1.5v and it will run the life expentancy as laid out in INtel specs....The question is how much for how long??? I guess once once we start seeing them dropping like flies we will know. for the most part there are many who have been running them in the 1.7v or less range for over 4 months now...I would be happy for 1 year cause it is highly unlikely I will keep it that long and hell for 130 bucks I can throw it away after a year....


Mine runs at 1.66v currently and drop a bit under load but not too much....

 

[mechBgon]

The question is how much for how long??? I guess once once we start seeing them dropping like flies we will know. for the most part there are many who have been running them in the 1.7v or less range for over 4 months now...I would be happy for 1 year cause it is highly unlikely I will keep it that long and hell for 130 bucks I can throw it away after a year....

That's my outlook on my home systems too... darned if I would want a state-of-the-art computer from 1999 today! For corporate use, well, a five-year life cycle may be realistic, but those systems are not going to be overclocked/overvoltaged.

BTW Duvie, did that P4 power adapter end up being useful?

 

[SteelCityFan]

kick that P4 up another Ghz or 2 and let us know what temp it gets to when it starts to throttle.. hehe

 

[jdurg]

The one for sure thing that we can gather from that datasheet is that you don't want to crank the voltage past 1.75 volts. As to how long it will last at 1.75 volts, I'm not sure. But at least we know that 1.75 is the absolute limit that we should be pushing these things to.

 

[Duvie]

Well if my experience can lend some insight...

1.8@2.52ghz with 1.8v set in bios plus board overvolted .03v for 1.83v....ran it under load heavy for a week doing multiple divx encoding and prime95 for 12 hours at the onset...chip lasted one week!!! Max temp ever seen was 52c with asus pcprobe...hardly a high temp by any means...

I had to back chip down to 2.4ghz and 1.65v. It was prime stable there for 1 week and then that became unstable....most errors were program errors with page faults....I then backed it down to 2.25ghz and 1.58v...It ran there for 10weeks solid (heavy use) until I put it into retirement with the arrival of my 1.6.....

Damn chip made me buy a new power supply (old sparkle was perfectly fine), nearly buy new ram but I thoroughly checked it out and it ran fine at 400mhz ddr but even at 266mhz it wouldn't run trying to get the chip to run 133fsb, then I bought a new mobo....Slapped chip into new mobo same problem...get 1.6 slap that in with all the same components no issues and look where I am now...

So it runs fine at default and up to 2.25ghz...Think I will build the parents a machine to replace their P-Pro 200. I will run it at default speed with retail HSF. Should last for years....

 

[jdurg]

Bump, in case someone else wanted to read it but didn't feel like going to page 2.

 

[Jwyatt]

Great info there Jburg.

 

[Jwyatt]

jburg. I just noticed in your system rig your running the msi 845 ultra aru. Im running the same MB! Stability is not a problem what so ever! The temps you mentioned are exactly like mine. Runs low 50s idle and mid to high 60s full load! I think realistically were not running those temps. MSI over compensated for the low temps being reported with the initial bios release.

Finding your reads makes me feel a bit more foolish for trading to get a vapochil oh well hopefully I'll have some fun with it

 

[rifken1]

I think you may want to take another look at your datasheet. I think your 71 C was the max for the case temp. I think you are looking at a max of 59.3 C for the CPU itself with a max case temp of 71 C. I believe you are correct about the Vcc Max Voltage though. I would like another set of eyes to look at this though, I am not an expert. It has been several years since I had these classes in college.

 

[jdurg]

Yeah. I've heard from a lot of people that the MSI 845 board has their temperatures running pretty high. But again, the thing is so rock stable that I'm not worrying about the temperatures. Plus, after reading the info from Intel's datasheet, I'm no longer worried about the temperature. Now I'm just enjoying having a 2.52 GHz CPU even though I only paid for a 1.8 GHz CPU. I no longer go to bed at night thinking of how I can lower that temperature. lol. I just wish that MSI would release a BIOS that would give us some more memory options. I have PC2400 RAM so I should be able to take it up higher than the fsb of 140 MHz. I'd like some more RAM options so I can max out the RAM I have. Right now, I'm only getting memory scores of about two gigabytes/second in SiSoft. That's about the only slow running thing in my system.

 

[jdurg]

I think your 71 C was the max for the case temp.

I saw that too, but if you look a little further down, they show you where that temperature is measured. It's measured right on top of the Integral Heat Spreader of the CPU. So if the temperature right on the CPU is able to be 71 degrees Celsius, then the internal temperature could actually be a little bit higher than that. (It took me a while before I figured that out, but when I looked a little further down at the diagram they had, it made sense. Intel has no specifications for the system temperature since with proper cooling, the CPU can be run in really hot systems.)

 

[Duvie]

Oh don't think so....

71c for a case temp!!! nothing could run in that environment...HDD would fail, chipsets would fail and the hsf would be blowing warmer air on the cpu then likely the cpu is producing....

I don't have time to read it....but jdurg follow this up!!!

 

[paralazarguer]

According to the new sharky extreme northwood overclocker's guide we're not supposed to go over 1.65 volts for the long term
http://www.sharkyextreme.com/guides/hwGuides/article.php/10709_1380951__1