Author: Robert Hyatt
Date: 07:32:07 03/18/03
Go up one level in this thread
On March 18, 2003 at 00:38:31, Aaron Gordon wrote: >On March 18, 2003 at 00:06:36, Robert Hyatt wrote: > >>On March 17, 2003 at 22:37:33, Aaron Gordon wrote: >> >>>On March 17, 2003 at 15:26:15, Sakkas Takis wrote: >>> >>>>The AMD are geting very hot when you use chessprograms >>>>Go for Intel >>> >>>How do you figure? Look at both AMD and Intel technical documents (or test one >>>yourself). AMD's run quite a bit cooler (by some 20-30 watts). >> >> >>Just for the record, "watts dissipated" is not the same thing as "running hot or >>cool." The Cray-3 for example radiates 125 kilowatts. Yes, that is kilowatts. >>It runs about -70C or so, immersed in liquid. A chip can convert more watts to >>heat, but _still_ run cooler if it has a solid path for the heat to escape >>quickly (as does the Cray with copper rods and plates everywhere). >> >>That is the limiting factor on speed, in fact. Ramp up the voltage to ramp up >>the clock and eventually you reach the point where the heat can't get out as >>quickly as it builds up, and the thing melts. > >If you have the exact same cooler (lets say 0.15 c/w) with the exact same >thermal compound (Arctic silver 3 for example) then more watts means more heat. > Let's define heat. 1. Energy produced by current leaking to ground thru resistive elements. 2. Energy that can not escape the CPU core due to lack of a good heat- transmission pathway. 1. Is not a problem, so long as 2 is not a problem as well. IE I can, in a single chip, radiate 125 kilowatts, so long as I can also conduct 125 kilowatts of heat _out_ to something that can absorb it. The cooler is only part of the equation. You have to conduct the heat away from where it is produced and get it to a place where the cooler can conduct it to whatever medium it uses to dissipate the heat into the atmosphere. >I'm going to round down the P4 to 100 watts (I think it's closer to 110 watts). >Anyway. A cooler with a 0.15 c/w rating and a 27c ambient temperature would keep >the P4 at 42C. Now, lets try that with an AthlonXP 2800+ (rounding to 75 watts). >39C cpu temp. Again, that is only _part_ of the equation. Your "cooler" can only dissipate that heat that gets efficiently conducted to the surface of the chip. If one chip conducts heat better than the other, that chip is going to run cooler, until it runs up against the limit of the cooling device. If one chip conducts worse than another, the cooler will _not_ dissipate its rated heat, because the chip can't get the heat to where the cooler can pick it up. > >As long as the cooling remains the same P4's are going to run hotter. No doubt >about that. More watts means more heat. Doesn't matter if you're putting a >cooling block on the core thats room temp or absolute zero. Heat is heat. >1 watt > 2 watts. So? If a processor only produces 1 watt of heat, it can run hotter than a processor that produces 2 watts, if the 1-watt processor can't transfer the heat to a point where it can be removed effectively. > >Lets try a 'normal' heatsink/fan c/w rating. Say, 0.35c/w. 27C ambient, 100 watt >cpu, thats 62C full load. Pretty toasty. Again, that is ignoring heat conduction. You have to deliver the heat to the cooler before the cooler can remove it. >Just for the fun of it lets throw in my particular setup. AthlonXP 2512MHz, >1.98v, 137.849 watts. If you think THAT is hot.. you should see my friends 200 >watt P4 @ 4GHz. Anyway, 15C ambient air temp on average, 16C water temp, liquid >cooler c/w is 0.0766c/w and the cpu temp is 26.56C full load. :) Same comment. Cooler capacity is only _part_ of the equation. If the heat can't travel to the surface of the chip, no cooler in the world is going to help. I mentioned the cray-3 earlier. Chips have no "carriers" at all to increase density. Chips do have small gaps between them for the connections to be robot-welded. And for liquid to circulate right on the chip itself to extract the heat.
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.