Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Scrambled eggs & sausage on your P4/Itanium, anyone? :)

Author: Robert Hyatt

Date: 20:01:53 02/21/03

Go up one level in this thread


On February 21, 2003 at 22:00:44, Matt Taylor wrote:

>On February 21, 2003 at 21:54:45, Robert Hyatt wrote:
>
>>On February 21, 2003 at 21:25:23, Aaron Gordon wrote:
>>
>>>On February 21, 2003 at 20:59:40, Matt Taylor wrote:
>>>
>>>>On February 21, 2003 at 14:55:16, Aaron Gordon wrote:
>>>>
>>>>>On February 21, 2003 at 09:47:27, Robert Hyatt wrote:
>>>>>
>>>>>>On February 21, 2003 at 08:27:55, Jeremiah Penery wrote:
>>>>>>
>>>>>>>On February 21, 2003 at 04:42:21, Charles Worthington wrote:
>>>>>>>
>>>>>>>>I am certainly no expert on cpu design and waht you say makes perfect sense from
>>>>>>>>an economic standpoint.
>>>>>>>
>>>>>>>Please explain this to Bob then, because he seems to think it's madness.  When,
>>>>>>>in reality, it is simple economic principle, and widely known as such.
>>>>>>>
>>>>>>>But, if todays chips were honestly capable of a stable
>>>>>>>>4GHz frequency then you could clock them there with no additional cooling
>>>>>>>>required. I do not doubt that todays chips can be taken to 3.2 GHz or perhaps
>>>>>>>>even 3.3 GHz and maintain stability but intel has a safety margin built into the
>>>>>>>>upper end chips to insure reliable performance. But even with little knowledge
>>>>>>>>of processor design I would have to say that Bob's argument makes more sense
>>>>>>>>from a logical standpoint. Intel would_love_to produce 4GHz Xeons today that
>>>>>>>>operate at low temperatures...problem is they simply can't do it. At least in my
>>>>>>>>humble opinion.
>>>>>>>
>>>>>>>I'm not saying that the current chips they sell are capable of 4GHz operation,
>>>>>>>in any way, shape, or form.  I'm saying that Intel, if it wanted, _could_
>>>>>>>release chips that were capable of such thing.  But right now, there's just
>>>>>>>absolutely no reason for them to do it.  For one thing, Intel doesn't want the
>>>>>>>P4 Xeons to be _too_ fast if it can help it, because they don't want to eat into
>>>>>>>Itanium sales.
>>>>>>
>>>>>>That logic is circular.  They can make faster xeons but they can't make faster
>>>>>>Itaniums???
>>>>>
>>>>>This may be true. Intel actually is going to wait a while before they release a
>>>>>faster P4 and most likely the reason I'm going to suggest is why they may not be
>>>>>producing faster Itaniums. Right now the P4-3.06GHz is 110 watts, this is a
>>>>>*LOT* of heat for a heatsink and fan to cope with. Intel has to figure some
>>>>>people that haven't a clue about cooling will take their new dell/gateway/etc
>>>>>and stuff it under their desk, let papers pile up infront of the vents, etc.
>>>>>Never clean the dust out and whatnot. This will most likely result in a cpu temp
>>>>>of at or over 70C with the regular Intel heatsink/fan. Imagine if they dropped a
>>>>>P4-3.2 to 3.4ghz into the market? You'd be hitting cpu temps that'd fry the chip
>>>>>in those situations.
>>>>>
>>>>>About the Itanium, it's even hotter. I saw the Itanium 800, Itanium-2 800, 900,
>>>>>1GHz all listed as 130 watts. This is pretty insane as is. I don't know how the
>>>>>Itanium servers are put together but some of them probably have liquid cooling.
>>>>>If not then you're going to have MAJOR problems with ANY heatsink today. They
>>>>>need to get the wattages down a LOT before they can ramp the clock speeds up.
>>>>
>>>>Intel plans at least a 3.2 GHz by June. I want to say they're hitting 3.6 GHz by
>>>>June. I don't remember.
>>>>
>>>>60 W is "pretty insane" compared to the 486 I have on my desk. I used to leave
>>>>the case off, and it always felt like the CPU was naked sitting there with no
>>>>heatsink and fan. I looked up the wattage at one point; it's under 1 W.
>>>>
>>>>I remember a side project my Dad worked on when I was younger. Our garage door
>>>>controller fried during an electrical surge, so he decided to build his own.
>>>>After he built it, he discovered thermal issues with some of the components, so
>>>>we flattened a penny and attached it for a heatsink. Obviously the heatsinks we
>>>>use on modern processors are much more sophisticated, but I think the cooling
>>>>solutions will improve to meet demand.
>>>>
>>>>-Matt
>>>
>>>That'd me we'll all be going liquid soon then. Heatsinks can only get so
>>>big/bulky. If you get TOO big it'd just be in the way, cause the PC to be too
>>>heavy (imagine a 30lbs copper heatsink..), etc. Liquid cooling is quiet,
>>>reliable if done properly and not heavy at all, especially if you use an inline
>>>system and small radiator.
>>>
>>>What I think would be neat would be a mini freon compressor.. :) If you can
>>>cascade two very small compressors and vent the heat out of the back that'd
>>>probably be the perfect solution for years & years to come. Have some sort of
>>>thermostat to monitor the cpu temp, perhaps keep it at a constant 75F. I'll be
>>>doing something similar but with much larger compressors (2-3hp each) and I
>>>won't be limiting the temperature at all.. =)
>>
>>
>>I dont think liquid will become mainstream.  Supercomputers used to do this,
>>such as Crays.  And cray went to air-cooling because of the many problems that
>>liquid cooling causes.
>>
>>Of course there are machines like the cray-2 and cray-3 where the silicon is
>>directly immersed into an inert fluorocarbon, but that's exotic, expensive,
>>and _horribly_ problematic (how do you debug the hardware when it is in a
>>liquid you can't be exposed to?)
>>
>>I think the trend to lower and lower voltages is likely to be the future...
>
>On the desktop, the opposite has been the trend. Lower voltages, but higher
>wattage and power consumption. Of course this leads to more heat...
>
>Liquid cooling does have problems, but it is becoming more and more popular
>simply because it's quiet.
>
>-Matt


The well-documented problem is reliability.  Spring a leak, buy a computer.

Air-cooling doesn't have that problem...  And it is a serious one, as many
cray folks can tell you...

Peltier works fine, as it is yet another solid state solution that can't
leak, of course. :)



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.