I know its not TJ. Tcase is the sensor between the cores, typically referred to as "package temperature" by HWMonitor.
But if running under load for extended periods, the package temp will rise to be close to the core temps.
The OP made it sound like the machine is running at 80C 24/7. Which is too hot for prolonged usage in my opinion.
As I still understand it, there is no "sensor between the cores." TCASE is a spec for a phantom sensor: what would be measured between IHS and die at dead center -- if there WERE a sensor there.
There was a history of contention about the spec. For instance, ShintaiDK (a respected member) doesn't think it's at all relevant to anything. IN his view, anything below the throttling Tj temperature spec is "OK."
The TCASE spec, however, is still published for Intel chips. ShintaiDK had dismissed that -- saying "that's for OEM builders!" I have to ask "Whether OEM builders or others -- why do they publish the spec, when they stopped publishing a spec for 'maximum safe voltage' after Nehalem?"
There had also been some white-papers or articles published at Tom's Hardware or similar repositories during the C2D/C2Q era, noting that TCASE would be about 10C lower than a core-sensor temperature average. But this would likely change from one chip generation to another.
If there's a "controversy," I try to steer clear of it. Instead, I "observe" the published TCASE spec, and try and use it as a target, an objective -- a guideline.
So for instance, with my 2700K system, I'd managed to keep the absolute worst load temperatures at ~71C -- about 2 or 3C below TCASE. Then, I introduced a second graphics card for SLI, and watched my stress-load temperature increase to ~76C. Not much more I can do about it, except to switch over to a good water loop. [I was patting myself on the back for the 71C achievement . . . ]
So I'll close my rant by suggesting that CPUs have two enemies: voltage (or excessive current), and heat. The materials used to make the chips have thermal properties. Heat will slowly degrade the chips; voltage -- electro-migration will also do damage over time. If the processors were made to operate within a voltage range AND a thermal range, their expected lifespan and the cost-accountant's choice of a "warranty period" likely incorporate those factors.
Finally, I'd suggest that there's a point where higher temperature requires higher voltage for stability, but lower temperature doesn't magically offer up a "voltage premium" for an over-clock, unless we're talking about LN2, maybe phase-change.
If it were MY 5960X and MY investment, I'd look at my usage patterns to determine just what percentage of time the CPU spends at that temperature (80C). And I'd either reduce the clock and voltage, or try and figure out a way to make the peak temperature close to or below 70.
Even with water-cooling, that could be a tough nut to crack -- I'd think. Maybe someone has a better understanding of it.
Facebook
Twitter