Ali-69C max case temperature, I think Intel has got some problems...
As stated before, the Tj_max and Tc_max can be established for two different reasons, long term reliability or stable operation. For most semiconductor chips, it's all about long term reliability. So if you consistently exceed the max junction temperature, the chip will still operate but will only last 10 years instead of 20 years.
But Intel microprocessors, it seems that stable operation is the key factor. This seems like the only possible reason for a strange Tc_max like 69C. Above that temperature, the chip goes unstable (or throttling kicks in). If Intel was worried about long term reliability, they would just set the Tc_max at 70C. I personally doubt that the Tc_max can even be measured to within +/-2C, just too hard to place the thermocouple accurately and there is also thermocouple wire error, etc. But Intel doesn't worry about that too much because they have thermal throttling to fall back on. Here's what they say in the pdf: "In all cases the Thermal Monitor feature must be enabled for the processor to be operating within the published specification.".
Now it seems that some testers are able to overclock easily. But this is not a typical sample that we're seeing. Intel sends out the real good chips to testers, not the marginal ones. I think Intel has a yield/binsplit problem. The good ones can be overclocked but there are many that can't go past 2.0 or 2.2. That's what I think. So it won't be easy for Intel to push frequency.
>> Maximum Tcase=68 C for 2GHz, and 69C for 2.2GHz, strangely enough.
Does seem strange. I think these are two different steppings. Note that for a 10% increase in frequency, there is only a 5% increase in power dissipation. So it looks like these are different chips.
Thanks for providing the link for the pdf. Your thoughts on the above? |
