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: ted burton who wrote (35991)	4/17/2001 8:15:59 AM
From: Bill Jackson	Respond to of 275872

Ted, The question is will the P4 ever go into this mode in a well designed system with good forced air cooling and good thermal contact between the CPU and heat sink? If it never goes there = moot question. If it goes there all the time as the die hots up then it will in fact be derated by 50%. Bad thermal design will make it go there more often. As it is just an alternative to BSD or chip decay by migration I do not see it as a feature. Any CPU maker can downclock to save the CPU, the trick is to make it work without needing to downclock. Intel was unable to make it work without downclocking the CPU at peak times. This is a feature? More like a creature from the blue lagoon. Bill

To: ted burton who wrote (35991)	4/17/2001 11:09:15 AM
From: pgerassi	Respond to of 275872

Dear Ted Burton: What you say goes against Intel published specs in the side note. You ask for proof, well you must now supply proof this does not occur. I am asking you for a power vs idle curve when running a multiuser multitasking server where you can easily estimate the load (Linux allows one to show CPU usage in % for the user, system, and idle. Just add the first two for % not in idle. Now run in single user mode with the minimum running. What is the temperature of the ambient, inside the case at the intake of CPU cooling air, the die temperature measured by the die. The critical measurements are the difference between the die and the case, and the die and the ambient. These are known as the case delta (mostly the HSF) and the ambient delta (the whole case and HSF performance). Now start multiuser with a simple "init 5" (starts the Xwindow server). Measure again. You now have two points. Run something like Primordia or Moldyn continuously (you can write a shell script that runs each bench right after it starts and because the system is multitasking both at the same time to as many copies as you want). Measure again. You now have three points and more if you measure different loads. You can measure intensive I/O loads vs Memory loads vs Number Crunching loads. I have achieved 100% usage with these varying types. Now you can see that the maximum delta gets quite large, but the real minimum delta quinescent system is not that far from full bore in a ratio. You will see that Intel says is pure bull. Even for RISC processors, the idle current of a running box is not 10% or 5% but more like 30% or 40%. It is only in stop grant mode and other such "sleep" modes that wattage at idle is significantly lower. Otherwise, P4 would not need such modes. As to the case of 50C being unrealistic, one computer used by a business as a server operated in a closet where temperatures reached 62C (at least in measurements I took from a strip recorder) (they did it for security and out of the way reasons (or so they said, it was a noisy box)). They were experiencing failures every few days during the summer but none during the winter. A air conditioner installed in the back wall of the closet solved that problem. So 50C ambient is not that unusual. Industrial grade boxes must run in environments to 55C (military ones even higher). So experience shows that you are misinformed at least. I have been to some stores in northern Ontario where the store temps must have been in the 90's. It does get to 140 in some unconditioned house in the desert. So 60C is not out of the question either. Now show proof that it doesn't in worst case conditions. Pete

To: ted burton who wrote (35991)	4/17/2001 2:45:18 PM
From: Joe NYC	Respond to of 275872

Ted, The 72C is a misunderstanding on your part. The thermal management kicks in at a die temp around 90C. Can you please explain? I looked at the spec a while ago and was a little confused. Where do you get 90C? 72C is mentioned as a temperature of case of the chip (heat spreader on the die?) that the deisgner should try not to exceed, or else, the thermal management may kick in. Is this interpretation incorrect? If it is, can you explain (if it can be explained in 1 paragraph or 2). Thanks Joe