Hi all,
For a long time i was running my i7 920 on the stock cooler at stock clocks. Recently i decided to plump for a decent cooler so i could get back into oc'ing.
My i7, idle, has always exhibited varying temps from core to core. After the install of the new HSF (Noctua vertical one with two 120mm fans, think it's called an NH-U12P SE2) 2 cores idled at 45, 2 at 40. I figured this was a little warm, even ocd to 3.4Ghz. Under load - 1 hr of prime95 - every core was up at 82/83 ish. Again, a little hotter than i'd like under ideal circumstances. I figured I'd screwed up the thermal paste, so resolved to have another crack. I went with the "single blob in centre of heat spreader" method initially, as this seems to be the recommended method these days, according to various articles i read. Sure enough though, when i took the HSF off, there was a load of gunk over one side, and none over the other. I decided to go back to the way we USED to apply thermal paste back in the day, and spread a very thin, even covering over the whole CPU. Back into windows and my idle temps are now a much more respectable 42 and 37. A 3 degree saving - nice!
Under load, the story is similar. My max temps are hitting 80 now. Still seems hot to me, but not unmanageable.
However, here's the rub. This time, under load, one core is significantly cooler than the others - WTF? Looking at coretemp as i write this, running prime95, 3 cores are 80, 80, 79 max temp, the other is 76. Plus, looking at the way the temps are bouncing around now, two are 79/80, one is 78/79, the other is 75/76. Doesn't seem right to me - any ideas? Prime hasn't made any mistakes yet btw.
Another bit of info: temps quickly hit mid 70s when i start prime blend, but creep up to 80 ish over a period of 20-30 mins. I had read that thermal saturation is reached after 8-10 mins, and that if temps continue to rise after this, it's caused by heat buildup rather than poor mounting of HSF. I'm using the lower fan speed adapters for the fans on the Noctua - not ultra low noise, just low noise - worth switching this out and going for max speed to keep temps down?
Finally, I'm currently letting my mobo (gigbyte UD5) do the vcore by itself. I'd read that leaving auto vcore on is good for preventing vdroop. Interestingly though, it seems to choose frequencies unde rload far in excess of what various tutorials think i should be using for the oc i'm on. For example, it's currently on 1.312v at 3.4ghz, whereas tuts online reckoned i should be using a vcore of 1.1v for that sort of OC. When i tried using a vcore of 1.1, the pc hung on me trying to start the prime blend. CPUz however, reported the vcore as 1.075 - not quite 1.1 - maybe that's why? What do people here recommend voltage wise?
Sorry for the epic post, and thanks in advance for any help you can offer
For a long time i was running my i7 920 on the stock cooler at stock clocks. Recently i decided to plump for a decent cooler so i could get back into oc'ing.
My i7, idle, has always exhibited varying temps from core to core. After the install of the new HSF (Noctua vertical one with two 120mm fans, think it's called an NH-U12P SE2) 2 cores idled at 45, 2 at 40. I figured this was a little warm, even ocd to 3.4Ghz. Under load - 1 hr of prime95 - every core was up at 82/83 ish. Again, a little hotter than i'd like under ideal circumstances. I figured I'd screwed up the thermal paste, so resolved to have another crack. I went with the "single blob in centre of heat spreader" method initially, as this seems to be the recommended method these days, according to various articles i read. Sure enough though, when i took the HSF off, there was a load of gunk over one side, and none over the other. I decided to go back to the way we USED to apply thermal paste back in the day, and spread a very thin, even covering over the whole CPU. Back into windows and my idle temps are now a much more respectable 42 and 37. A 3 degree saving - nice!
Under load, the story is similar. My max temps are hitting 80 now. Still seems hot to me, but not unmanageable.
However, here's the rub. This time, under load, one core is significantly cooler than the others - WTF? Looking at coretemp as i write this, running prime95, 3 cores are 80, 80, 79 max temp, the other is 76. Plus, looking at the way the temps are bouncing around now, two are 79/80, one is 78/79, the other is 75/76. Doesn't seem right to me - any ideas? Prime hasn't made any mistakes yet btw.
Another bit of info: temps quickly hit mid 70s when i start prime blend, but creep up to 80 ish over a period of 20-30 mins. I had read that thermal saturation is reached after 8-10 mins, and that if temps continue to rise after this, it's caused by heat buildup rather than poor mounting of HSF. I'm using the lower fan speed adapters for the fans on the Noctua - not ultra low noise, just low noise - worth switching this out and going for max speed to keep temps down?
Finally, I'm currently letting my mobo (gigbyte UD5) do the vcore by itself. I'd read that leaving auto vcore on is good for preventing vdroop. Interestingly though, it seems to choose frequencies unde rload far in excess of what various tutorials think i should be using for the oc i'm on. For example, it's currently on 1.312v at 3.4ghz, whereas tuts online reckoned i should be using a vcore of 1.1v for that sort of OC. When i tried using a vcore of 1.1, the pc hung on me trying to start the prime blend. CPUz however, reported the vcore as 1.075 - not quite 1.1 - maybe that's why? What do people here recommend voltage wise?
Sorry for the epic post, and thanks in advance for any help you can offer
Last edited:
No doubt water would do a great job, but right now i'm just planning on sticking with what i've got. Can't afford anything else really 