Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Technology Stocks : Advanced Micro Devices - Moderated (AMD) -- Ignore unavailable to you. Want to Upgrade?

To: Not a Short who wrote (238994)	8/17/2007 10:36:57 AM
From: wbmw	Read Replies (3) \| Respond to of 275872

Re: I'm just trying to understand the use of the acronym TDP. I thought TDP was a spec not a measurement. You seem to be using it the way someone would use MPG (Miles Per Gallon) as a way to state observations. I thought TDP was a fixed value, once you had a CPU in hand it would always have the same TDP no matter what heatsink you put on it. TDP (the Thermal Design Point) is somewhat empirical, rather than an exact specification. In order to derive it, you have to test how real applications dissipate power. The end result is that if a system designer constructs a cooling solution that can dissipate that many watts, while keeping case temperature (also called junction temperature, or Tj) below a stated specification (usually in the range of 50-100C degrees). Of course, most of the time, the cooling solution is below Tj, and cooler temperatures lower one very important element of power - leakage. You would generally expect that across a number of various chips a frequency F and temperature T, the vast majority will dissipate power well below TDP, which is consistent with most CPU measurements. In the Pentium 4 and Pentium D era, many noticed that Intel's TDP value was far too aggressive, and that the CPU generally dissipated more power under normal conditions than what was rated as TDP. Of course, the processor is designed to throttle in these cases, but the consumer certainly doesn't want throttling to occur under normal workloads. It appears that besides a few early chips that exhibited this behavior, many newer chips became far more conservative. And by the time Intel launched Core 2, their chips were measuring far below TDP, just like AMD's chips. For what it's worth, here is AMD's definition of TDP, word for word: TDP. Thermal Design Power. The thermal design power is the maximum power a processor can draw for a thermally significant period while running commercially useful software. The constraining conditions for TDP are specified in the notes in the thermal and power tables. amd.com Notice that the notes usually mention which temperature the measurement assumes. Intel summarizes their definition of TDP in 5.1.1 of this specification: download.intel.com Additional details on the measurement of TDP can be found here, too (specifically, note the text in section 4.2.6, starting on page 37): download.intel.com

To: Not a Short who wrote (238994)	8/17/2007 12:48:14 PM
From: pgerassi	Read Replies (1) \| Respond to of 275872

Dear Not_a_short: TDP means thermal dissipation power and yes, I am using it as a shorthand for thermal output like many on this thread. It has the same acronym as TDP, thermal design power, commonly used by Intel. It is used as a marketing dodge by Intel. Thermal dissipation power is how its historically perceived and how Intel used to specify their CPUs in the old days. We have had a lot of debates revolve around the difference in these two definitions. Intel's TDP was defined like typical power was used in the past and AMD defined it more like maximum possible thermal dissipation power for a family of CPUs, a bounded power constraint never to be exceeded. Thus TDP(intel) < TDP(AMD) which I shortened to TDPtyp and TDPmax. You can find both powers at a place like www.sandpile.org. Intel supporters hated the fact that when using the same historical standards, their CPUs had a higher TDPmax than their design power rating. So their 80W design power 3GHz@1333MHz FSB C2D actually carries a TDPmax of 124.83W. That is higher than the TDPmax of 119.2W of the 3GHz Opteron 8222SE. TDPs as specified by AMD are all absolute upper bounds on actual dissipation of any given CPU. Many CPUs may be much lower than the amounts specified. TDP does change depending on the conditions used. There is a TDP for each power state. On my 90nm SC A64 3500+ (Venus), there are 4 P (power) states, 2.2Ghz, 2.0GHz, 1.8GHz and 1GHz each with its own required voltage defined by the VID. In addition there are sleep states. Mobile chips have more sleep states than desktop or server chips. These have various circuits or voltage planes turned off or idle. One used is S1 IIRC where the CPU is halted in a stop grant state, the HTT ports, NB and DDR controllers are still active. It can be done at every P state having a TDP for each. Of course the lowest on mine would be P4S1, 1Ghz. There is sleep state S3 where the CPU is turned off having saved its state to DRAM, the HTT ports are off and the DDR memory is turned to self refresh. On mine, it is constrained to not use more than 350mW. My series of A64s had variable TDP specified as each chip had its own max TDP. Where the family allowed up to 68W, my CPU had a TDPmax of 50W. Some had one as low as 35W. And thermal output does vary with time. Most of the time it gets worse as time goes on, but occasionally it gets better. Mine is in the latter group. Pete