All, when measuring core temperatures, do you average the core temps? Or do you simply use the highest core temperature? My core 0 on an i5-2500K is some 10 degrees lower than core 1, which runs hottest. I was just wondering how to compute my temperatures for o/c purpose.
![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
Core Temp Readings
- Thread starter aoya
- Start date
In general I think most people (OCing wise) go based off the single highest temperature reading you get even if it's for a brief second the entire time during a test run, as far as measuring for the sake of safe temps.
Running IBT on maximum w/ an OC of 3.8 (stock 2.8) I was getting on a single core 75 degrees while the others never got higher than 70, so I say I get a max of 75 assuming all 4 cores even though that's not necessarily true. Just gives me that leeway in case.
Running IBT on maximum w/ an OC of 3.8 (stock 2.8) I was getting on a single core 75 degrees while the others never got higher than 70, so I say I get a max of 75 assuming all 4 cores even though that's not necessarily true. Just gives me that leeway in case.
drescherjm
[H]F Junkie
- Joined
- Nov 19, 2008
- Messages
- 14,933
I would go for the highest reading and keep that under 90C when you run a program like IBT. Are you running all 4 cores at 100% CPU usage?
IBT is usually considered more stressful so errors tend to pop up a lot sooner which saves time. However its ideal to eventually run both to ensure stability.
drescherjm
[H]F Junkie
- Joined
- Nov 19, 2008
- Messages
- 14,933
No. 79C is fine.
dragonstongue
2[H]4U
- Joined
- Nov 18, 2008
- Messages
- 3,162
Highest if you want to quote "hottest my cpu got was" or if you are looking for yourself the high average is fine as well. I use openhardware monitor or hardware monitor for my temp testing.
Each core is slightly different, so this obviously can lead to some temp/voltage differences, also even though these chips are wicked tolerances, when you are talking hundreds of millions of anything, even a very small difference can add up quick, part of the reason the cores are slightly skewed from each other is to keep temp loading/voltage loading under control.. From my experience 2-7c difference idleload can be expected between cores in a very efficent multi-core type program/app, others tend to be very one sided that they hit the cache or procesing of the cpu harder on a single core, but that doesnt matter as much
Each core is slightly different, so this obviously can lead to some temp/voltage differences, also even though these chips are wicked tolerances, when you are talking hundreds of millions of anything, even a very small difference can add up quick, part of the reason the cores are slightly skewed from each other is to keep temp loading/voltage loading under control.. From my experience 2-7c difference idleload can be expected between cores in a very efficent multi-core type program/app, others tend to be very one sided that they hit the cache or procesing of the cpu harder on a single core, but that doesnt matter as much
Are you manually setting the number of threads? It won't work properly with chips that have hyper-threading unless you manually set the number.
I remember when I got my first AMD 486/DX4 100mhz processor it was so fast. I bet the temps back then was probably like room temp at full load. I dont even think it had a fan on it, I think just a heatsink.
Now chips are hitting 90C sigh haha.
Usually just go by the highest core. They will never be the same. Each core has a different load values and TIM application is about as uneven as can be unless applied by some expensive machine in a factory.
Now chips are hitting 90C sigh haha.
Usually just go by the highest core. They will never be the same. Each core has a different load values and TIM application is about as uneven as can be unless applied by some expensive machine in a factory.
No, it tells me something about memory use being limited to less than 2gb because of Win32.
Should those progs be creating more heat than Prime95's "maximum heat / power" test?
Yeah, that's the one thing I don't get. I mean chips shrink, require less power, produce less heat, etc. but you'd figure they'd give off radically different temps when you move from say 45nm to 32nm at the same clock speed and near identical voltage, but they often just kind of sit around the same, boggles me as to why that is.
One thing I do kinda miss in the winter though is that even though decade old slower chips didn't require a heatsink/fan, they would heat my room up like crazy, modern ones don't do that on the other hand which seems counter to the argument that chips now run hotter. I really wish I was an engineer or understood why that is lol.
Part of that is that the chips are smaller, so there is less surface area to dissipate the heat that is produced. Power/mm^2 is actually going up, not down. Transistor density has gone up significantly.
Last edited: