Re: 4 Thermal Design Power (TDP) is measured under the conditions of Tcase Max, IDD Max, and VDD=VID_VDD,
and include all power dissipated on-die from VDD, VDDIO, VLDT, VTT, and VDDA.
Your link proves that AMD has always used the acronym for Thermal Design Power, and never Thermal Dissipation Power.
Obviously, AMD's *definition* of the term has changed from max theoretical power, to a standard definition of highest power while running real commercial applications for thermally significant time periods. This is a relaxation of the specs, in response to Intel's more relevant system design standard.
Re: They still keep that no processor of AMD's will exceed their specified thermal design power, a claim Intel does not make.
No, they don't. In fact, they specifically state that the TDP is only applicable for a "thermally significant period", which implies that instantaneous power draw may exceed the TDP. This is exactly how Intel defines it. They also specifically state that the TDP is only applicable "while running commercially useful software", which implies that a thermal virus (non-commercially useful software) could be engineered to exceed TDP. Intel's wording more explicitly states the same thing.
Re: The way that they (Intel) slip it in is that given a HSF that will push the CPU to the TMA2 trip point at the specified TDP and the BIOS enabled TMA2, then and only them, will the TDP be a strict upper bound.
Another gross misunderstanding. Read this article for more information:
digit-life.com
An application can exceed TDP for short amounts of time, as long as the processor does not exceed Tj. Only if the thermal trip is activated will the Intel TM2 engine throttle down the processor. The independent tests confirm this.
It is very difficult to get the CPU to throttle on its own. With the stress test thermal virus application, they were able to see a little bit of throttling, but only after running it continuously until Tj was breached. And furthermore, TM2 didn't really kick in until they slowed down their fan rotation to 1000 RPM. Even further, extended throttling didn't kick in until they manually stopped their fan by wedging something in the mechanism. And even after that, the system continued to run at a slower speed.
It's actually a very efficient throttling mechanism that maintains very good performance, even under thermally constrictive situations. AMD would be lucky to have such a mechanism. You can complain about the Intel retail heatsink all you want. The fact is that consumers aren't running power viruses for sustained periods of time, so thermal throttling should never actually happen under real world conditions. |
