Skeptical about claim that stamp creation burns out modern CPUs

[Tim May]

On Jan 1, 2004, at 2:35 PM, Eric S. Johansson wrote:

> Tim May wrote:
>
>> I'm skeptical of this claim. A lot of Intel and AMD and similar 
>> machines are running full-tilt, "24/7." To wit, Beowulf-type 
>> clusters, the Macintosh G5 cluster that is now rated third fasted in 
>> the world, and so on. None of these machines is reported to be 
>> burning up literally. Likewise, a lot of home and corporate users are 
>> running background tasks which are at 100% CPU utilization.
>
> I will admit to a degree of skepticism myself even though I am 
> describing overheating as a likely outcome.

But what is your actual evidence, as opposed to your belief that 
overheating is a likely outcome? I have said that I know of many 
machines (tens of thousands of CPUs, and probably many more I don't 
know about directly) which are running CPU-bound applications 24/7. I 
have heard of no "burning up literally" cases with the many Beowulf 
clusters, supercomputers, and 24/7 home or business screensavers and 
crunching apps, so I suspect they are not common.

If you have actual evidence, as opposed to "likely outcome" 
speculations, please present the evidence.
>

> First, if you lose a fan on an Intel CPU of at least Pentium III 
> generation or an AMD equivalent, you will lose your CPU to thermal 
> overload.  This is a well-known and well-documented problem.  One 
> question is can stamp work thermally overload and damage a CPU.  
> Second question is how much stamp work  can you do without thermally 
> overloading the CPU.

This is true whether one is running Office or a stamp program. You are 
just repeating a general point about losing a fan, not about stamp 
generation per se. Boxer fan lifetimes are usually about comparable to 
hard drive lifetimes, which also kill a particular machine. You are not 
presenting anything new here, and the association with stamp generation 
is nonexistent.
>
> Large clusters have more careful thermal engineering applied to them 
> than probably most of the zombies out there.  I have seen one Beowulf 
> cluster constructed out of standard 1U chassis, motherboards, fans 
> etc. and frequently 10 percent of the systems are down at any one 
> time.  The vast majority of the failures have been due to thermal 
> problems.

Most clusters use exactly the same air-cooled machines as are available 
from Dell, Sun, Apple, etc. In fact, the blades and rackmount systems 
are precisely those available from Dell, Sun, Apple, etc.

You are presenting no evidence, just hypothesizing that your stamp 
protocol somehow burns out more CPUs than render farms do, than 
Mersenne prime apps to, than financial simulations do, etc. Yet you 
present no actual numbers.
>
> so, will we see a Pentium IV spontaneously ignite like a third tier 
> heavy-metal group in a Rhode Island nightclub?  No, you're right, we 
> won't.  I think it's safe to say we will see increasing unreliability, 
> power supply failures, and failures of microelectronics due to 
> increased thermal load.  Which is good enough for my purposes.

Evidence is desirable, belief is just belief.



--Tim May
"That government is best which governs not at all." --Henry David 
Thoreau