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: wbmw who wrote (239019)	8/18/2007 8:36:10 AM
From: Reseller Mike	Read Replies (2) \| Respond to of 275872

WOW! I mean just, WOW. I remember something from one of my engineering classes maybe 30 years ago (I am not a thermal design engineer) and point it out that what you posted seems wrong and you choose to pontificate to me about it? And as for educating... My opinion of you is that you have great resources to find information on your own (evidence the reply to the message). Just look at the great lengths you go to on your posts. I like the fact that you track down all of the little pieces of fact and fiction posted here and set the record straight, and I respect your termacity, but don't expect me to research something I have no deep interest in other than the 'does that look right?' test. Engineers, like me, and I take it you are one from your responses, are trained to use the 'gut' test on things told to them. If I see something and it doesn't correlate to what I remember or think it should be I'll respond, usually with a short and succint statement because I have other things to do all day long than post on boards. 'Nuff said. Regards.

To: wbmw who wrote (239019)	8/20/2007 6:29:03 PM
From: Petz	Read Replies (3) \| Respond to of 275872

Since you keep repeating this nonsense, I have to respond -- Tj would presumably be larger than Tc, since some of the heat is presumably lost through the edges of the material, or from the resistance of the interface itself. In order to measure Tj, one would need some kind of digital sensor in the silicon of the die itself, while measuring Tc would require any kind of temperature probe attached to the IHS of any Intel or AMD CPU. It just so happens that today, both Intel and AMD standardize around Tc, so it's irrelevant to have any kind of Tc vs. Tj debate. download.intel.com It is not true that Tj cannot be measured. The problem with measuring Tj is that it varies all over the die. Tc varies somewhat, but not near as much. The FPU might be a hot spot, or the L1 or an internal crossbar. Both AMD and Intel embed sensors on the die to detect unsafe junction temperatures in specific "hot spots." In Intel's case, for the quad core X5300 series 120w (145w max) Xeons, Intel states categorically that a heatsink designed to allow Tcase to reach 70C when dissipating 120w WILL result in "noticeable performance loss due to increased TCC activation [thermal throttling]." They state this is true even if the actual Tcase temp doesn't come close to 70C! Sounds contradictory, but it is not. The Tc will barely exceed 63C under any circumstances, because the internal thermal sensors measuring Tj will trigger throttling, probably at an internal temp in the 85C range. Even if the Tcase is kept below 63C there will still be some throttling for brief periods, according to the above document. In fact, the throttling above Tcase=63 is so severe and effective that even a crappy heatsink won't result in any Tcase measurement much above 63C. The datasheet states that the throttling is accomplished first by gating the clock and second by calling for a lower power state, i.e., lower frequency and voltage. AMD uses internal thermal sensors much less aggressively. Paragraph 5.2 of AMD NPT Family 0Fh Processor Electrical Data Sheet states that an "internal temperature sensor independent of the thermal diode" will result in THERMTRIP_L being asserted if the internal termperature reaches 125C. The motherboard must shut down power within .5s to prevent damage. There is no mention of any throttling under normal operation. The THERMTRIP is really intended to prevent damage when the heatsink clip (or the assembler's brain!) fails. It just so happens that today, both Intel and AMD standardize around Tc, so it's irrelevant to have any kind of Tc vs. Tj debate. AMD specifies that Tcmax is 67-72C. I wasn't able to find the doc that gives a specific value for each DC Opteron, but I strongly suspect that AMD will not lower the required Tcmax for the Barcelona. The higher the better, as it makes the thermal solution smaller and lighter. Given that ambient inside a server case is at least 45C, it is very much harder to dissipate 120w allowing only an 18C temperature rise (45 to 63) than it is to dissipate 95-120w with a 22-27C temp rise. The former (Intel) requires a HSF with 0.15degC/w, almost unattainable. The latter requires a HSF with 0.183 to 0.225 degC/w performance. And I'm assuming 120w for an SE version Barcelona, which I suspect will not exist until the 3 GHz stepping is released. If you read nothing else, read this: You have stated several times that Pete "invented" TDPmax. You stated that he invented it as the product of voltage times current. Both statements are incorrect. Table 6.3 of the above document clearly defines "Maximum Power" of 145 watts for the quad core Xeons. It is not power consumed, it is power dissipated as heat, which of course means that max power consumed is even larger. Also, when stating that Intel should be commended for measuring TDP at Vccmax, recognize that Vccmax is not a fixed number, but varies depending on current draw. At max power, Vccmax is always less than Vid. Petz