Message boards :
Number crunching :
CPU Tcase
Message board moderation
Author | Message |
---|---|
![]() Send message Joined: 10 Apr 10 Posts: 14 Credit: 590,352 RAC: 0 ![]() |
Hi guys! Started my reading a bit late, but i googled my cpu, to see what kind of temps it could stand. Since i allready knew that my GPU could stand 105 C.(And sorry i dont know what that is in F.) So i checked my cpu aswell wich is a intel I7 920, and it says that it can only stand a temp. of Tcase of 67 C. But my CPU has been running in BOINC with temps exeding this many times. Is there anything to worry about or? I have already downloaded TThrottle from SETI`s website, and regulated the CPU temp down to 65 C just to be on the safe side, but what do you guys say? |
![]() ![]() Send message Joined: 16 Mar 00 Posts: 634 Credit: 7,246,513 RAC: 9 ![]() |
...my GPU could stand 105 C.(And sorry i dont know what that is in F.) (F-32)/9 = C/5 » 105 °C = 221 °F ![]() |
![]() Send message Joined: 10 Apr 10 Posts: 14 Credit: 590,352 RAC: 0 ![]() |
thanks for the calculation. :D And yes it may be a stupid questin since the manufacture said it should NOT get over 67 C, but i tought they had a buffer of some kind with some Celsius, for example that the cpu wont get "destroyd"/broken if it gets over 67 C but if it gets over a 100 Celsius it wont "survive".... Make any sence at all? |
![]() ![]() Send message Joined: 30 Aug 01 Posts: 1228 Credit: 47,779,411 RAC: 32 ![]() ![]() |
When you say your i7 has been running over the Tcase max, are you measuring Tcase or Tjunction? Most of the common programs I know of (RealTemp, CoreTemp, Speedfan) read the Tjunction, which is usually higher than Tcase. Tjunction is the temperature at the center of each core, Tcase is the over-all processor temperature. If you're seeing one temp per core, you're looking at Tjunction. -Dave ![]() |
![]() Send message Joined: 6 Apr 07 Posts: 7105 Credit: 147,663,825 RAC: 5 ![]() |
For 24/7 crunching: If you know the max Tcase of your CPU, I would stay ~ 10 °C under this temp. For example you could use HWMonitor for to look the CPU (Tcase) temp. If you use CoreTemp and Delta to Tj.Max is > 20 °C you are fine (CPU-Cores). ..I do this like this. If your CPU is hotter, I would think about an other heatsink or better PC case cooling. ![]() |
![]() Send message Joined: 23 Oct 07 Posts: 142 Credit: 27,815,748 RAC: 0 ![]() |
I don't run an I7 but: What program are you checking temps with? I'd try Everest. How hot are we talking? Tcase refers to the temperature at the outside of the CPU case (The metal surface). Core temperatures are generally higher than that depending on cooling solution. Intel spec values are based on their Thermal Test Vehicle (TTV) using their reference cooler design. If you are using a reference cooler you can probably run your core temp about 5-7 degrees hotter without exceeding this specification. (this is an estimate, there is no real way to convert Tcase to Tjunction or Tcore.) If you crunch SETI you might want to upgrade to a higher end cooler (Zalman 9900 or something and use good thermal compound.) This can easily reduce core temp 10 degrees. The I7s can actually run well into the 90s fairly safely (thermal cutoff should kick in though), The 67C limit is needed to make the XXXXXX hour MTBF. running hotter can shorten it's lifespan. Disclaimer: This is based on previous processors. I am not completely sure I7s work this way or what their temp sensor locations are. |
archae86 Send message Joined: 31 Aug 99 Posts: 909 Credit: 1,582,816 RAC: 0 ![]() |
Safety (not destroying the part) and working properly are two separate questions. 1. The manufacturer does speed/voltage testing and uses the results to assign dice to a specific speed grade. They try to assure that (almost) all the parts will work properly at specified speed when the temperature reaches the maximum they say they'll support, with the voltage at least favorable (usually lowest) value. While marketing to some degree massages demand with pricing and other measures to get a rough balance between speed grade sales and the actual output of the factory, in real life many customers happen to get a part that is much faster than specified (I'm speaking of the worst-case temperature/voltage condition--of course ALL of them are much faster at more favorable operating conditions). 2. Reliability--as in the permanent loss of proper function--is another matter altogether. The manufacturer seeks to reach or better a target failure fraction of the whole user population at the actual conditions of use. As higher temperature raises the failure rate for nearly all the mechanisms of practical interest, this fraction looks better the cooler that users run their parts. To that end they seek to influence use at less than the spec maximum, by such methods as TDP postings, advocacy of better cooling methods such as ducting, and the like. But, most of the time, the actual permanent damage temperature for short-term exposure for a particular sample of the part is likely to be very far above the "meets spec" temperature. Were this not true, the total population failure rate would in fact be catastrophically high. [disclosure: I spent 26 years working for a certain large microprocessor manufacture, four of them as a reliability guy, and some others in design work and test work.] |
![]() Send message Joined: 10 Apr 10 Posts: 14 Credit: 590,352 RAC: 0 ![]() |
Well i use both Everest UE, and TThrottle, where both show core temp of each core, and both show CPU temp, where the CPU temp is lower then most of the cores, bot with TThrottle, i can set the temp, so the program slows down the CPU so it wont exceed that temp. Just wondered, what you guys would recommend to set it on, when the "maximum" temp by the manufacture is 67.5 C. should i set it as much as 10 degrees lower, so it would be 57.5 C? All inputs are very very good inputs to me :D :P Thanks guys! |
![]() Send message Joined: 23 Oct 07 Posts: 142 Credit: 27,815,748 RAC: 0 ![]() |
How hot was it getting in both the cores and "Overall"? |
![]() Send message Joined: 10 Apr 10 Posts: 14 Credit: 590,352 RAC: 0 ![]() |
Goes from 55 to 70 Celsius in the 4 cores and have a stabile temp on 50-55 celsius "overall"... but keep in mind that i have set TThrottle to have the temp around 67 celsius, and never exceed that... the basic question is really, how high can go with the temp, cuz i wanna be able to squiz the most power out of my cpu as well, but this thread is basicly to find the balance on this machine.... |
![]() Send message Joined: 6 Apr 07 Posts: 7105 Credit: 147,663,825 RAC: 5 ![]() |
Which kind of PC you have? It's a complete bought, or you built the PC? How is the ambient temp in the room? If you let run the PC at stock - not overclocked - speed and you reach this temps, I think then there is something wrong. It could be also, that the heatsink isn't fixed well on the CPU. Or what ever. I wouldn't go higher than the temps I posted already if you crunch 24/7. ![]() |
![]() Send message Joined: 6 Apr 07 Posts: 7105 Credit: 147,663,825 RAC: 5 ![]() |
BTW. Because of your PCs are hidden.. Which kind of GPU you have? If you say your GPU can reach 105 °C, yes this stay maybe in the specs of nVIDIA. BUT, I would never let run the GPU if 24/7 at this temp. The graphic card RAM is also cooled with the same heatsink. I guess if you reach 95 °C, your GPU start to make CUDA calculation errors. And live max. 1 year if 24/7. My GPUs reach max. ~ 85 °C. Automatically fan speed control. ![]() |
![]() ![]() Send message Joined: 16 Jun 00 Posts: 648 Credit: 228,292,957 RAC: 0 ![]() |
I would highly recommend a different CPU cooler and some additional cooling fans inside of your case to develop better flow to disperse the heat. I use the Corsair H50 watercoolers and they work very well - but they are coming out with a new version. The H70 will offer a bigger radiator and dual fans. Todd |
![]() ![]() Send message Joined: 27 May 07 Posts: 3720 Credit: 9,385,827 RAC: 0 ![]() |
Well i use both Everest UE, and TThrottle, where both show core temp of each core, and both show CPU temp, where the CPU temp is lower then most of the cores, bot with TThrottle, i can set the temp, so the program slows down the CPU so it wont exceed that temp. Just wondered, what you guys would recommend to set it on, when the "maximum" temp by the manufacture is 67.5 C. should i set it as much as 10 degrees lower, so it would be 57.5 C? I think TThrottle measures/shows/uses only the internal CPU cores temperature (measured by internal in-the-CPU-chip diode(s)) Temperature Monitoring Programs http://setiathome.berkeley.edu/forum_thread.php?id=59292 Core i7 (Bloomfield) Socket 1366 Core i7 920 (quad 2.66GHz - 4x 256KB - 8MB L3) Core Voltage (0.8V~1.375V) Vmax 1.55V Max. Core Amp. 145A Thermal Design Power 130W Max. Cover Temp. 67.9°C http://mysite.verizon.net/pchardwarelinks/elec.htm Usually the Max. allowed internal (TJunc) CPU temperature is about 5-10°C higher than Max. Cover (TCase) Temp. (this means temperature of the "Cover" or "Case" of the CPU - not the computer case) So maybe the Max. TJunc temperature of this CPU is e.g. 75°C and TThrottle setting of 65°C for Core(s) temperature is (maybe) safe. (But if this was my CPU I will use 60°C limit in TThrottle and try better cooling to keep it even under 55°C) I wonder what Core Temp http://www.alcpu.com/CoreTemp/ is showing about TJuncMax (or TCaseMax?) and Cores temps of your i7 920 CPU.  ![]() ![]()  |
![]() Send message Joined: 10 Apr 10 Posts: 14 Credit: 590,352 RAC: 0 ![]() |
Antec Nine Hundred Miditower QuadCore Intel Core i7 920, 2666 MHz (20 x 133) MSI X58 Platinum (MS-7522) (2 PCI, 3 PCI-E x1, 2 PCI-E x16, 6 DDR3 DIMM, Audio, Dual Gigabit LAN, IEEE-1394) Kingston 9905403-011.A02LF 2 Gb DDR3-1333 DDR3 SDRAM (8-8-8-22 @ 609 MHz) (7-7-7-20 @ 533 MHz) (6-6-6-17 @ 457 MHz) (Got three of these, so 6 GB ram total) NVIDIA GeForce GTX 285 (1024 Mb) ___________________________________________________________________________ As we talked about temps and so on, I should say, that all the cooling systems/fans, are standards, that followed the different components. And to answer a earlier question, yes, me and my brother built this computer. |
![]() Send message Joined: 6 Apr 07 Posts: 7105 Credit: 147,663,825 RAC: 5 ![]() |
And you used the boxed heatsink for the CPU? Maybe the heatsink or the thermal compound isn't fixed well? I had the boxed heatsink of/on my QX6700 and if the ambient was ~ 30 °C earliest then the autom. fan was on max. (full load SETI@home, optimized CPU app). ![]() |
![]() Send message Joined: 10 Apr 10 Posts: 14 Credit: 590,352 RAC: 0 ![]() |
Yea used the cooling paste. Just sat down and calcuated and counted how many tasks i had and how many were ready for upload. Dont know if this is much or avrage, but i think its pretty good :) 700 SETI@home tasks, give or take a couple... 210 completed, give or take a couple And offcourse about 490 to go... :) And one more thing, i think i "fixed" the temp problem for now, and i took a reagular table fan, and sat it into one of the fan "filters/openings" that i dont have a fan in, and it just sittin there, blowing a cool breeze into the PC, the temps are stabile, at 55 C (CPU) and 65 C (GPU). |
![]() ![]() Send message Joined: 18 Aug 99 Posts: 1432 Credit: 110,967,840 RAC: 67 ![]() ![]() |
Antec Nine Hundred Miditower May I suggest the following: 1) If possible replace your case with a full tower that has a high air flow (Cooler Master 932 HAF) is a good one. This case give you maximum air flow with three incoming 230x30, 700 rpm fans (front & side). The other 230 fan is on top for exhaust. There is also a 140x25 rear fan that is also for exhaust. I don't know what your PSU is, but in this case it will sit on the bottom and will draw incoming air under the case as oppose to using the air inside for cooling. Plus the case will give you excellent cable management which will help with the ait flow. I have two of these cases, one for a system I just built and the other for a system that was in a mid-tower which got too small. When I get the money together I will replace a second mid-tower case with one. 2) As noted in another post, I also recommend the Corsair H50 cooling system or its big brother H70. I have tow of the H50s and will probablly replace the one in the new system with the H70. 3) Check out the downloads section of the Intel site and see the Desktop Control Center is applicable to your board. I have it on the new system and you you can use it, you can control all kinds of things related to your CPU/chipset/board including temps & o/c. ![]() ![]() I don't buy computers, I build them!! |
![]() Send message Joined: 23 Oct 07 Posts: 142 Credit: 27,815,748 RAC: 0 ![]() |
I've found the The Antec 900 has better air cooling than the HAF 932. Stock they have twin 120 inlet fans and a 120 and a 200 outlet which should be plenty for a single or even double GPU application. Those 230s in the HAF are poorly shrouded and not aimed all that well. My gear generally ran a couple degrees cooler in the Antec. Cable management in the Antec sucks though so if you have a mess you might want to fix that. Obviously the cooler master is better for water cooling or really big builds. |
![]() Send message Joined: 6 Apr 07 Posts: 7105 Credit: 147,663,825 RAC: 5 ![]() |
Yea used the cooling paste. Nice you solved the temp prob! :-) But, you need to look because of dust. You will be surprised after ~ 6 months how much dust is in the PC case. I had also temp probs because of a too small and not well cooled PC case. The left door little bit opened. After ~ 1/2 year I re-built/modded the machine and was suprised how much dust was in the graphic card case of my GTX260. I needed to use my vacuum cleaner on lowest power level to get out the dust again. Other take air pressure can for this work. I saw a PC in the internet, the owner took away the left door and placed a ~ 1/2 meter room fan on the left which blow much air into the PC case. He never had temp probs, but collected the room dust like with a vacuum cleaner in the PC case.. X-D ![]() |
©2025 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.