Via vs. Intel on power

[Felger Carbon]

I've been following the vigorous debate with considerable amusement.
That's because I think the debaters can't see the tree for the forest.
;-)

There was a time when I used a computer primarily as a text processor.
Had power consumption been critical to me, I would _not_ have asked,
"what processor has the best power-performance?". I'd have asked "of
the computers that will do my job, which has the lowest power" and
then I would have bought that computer.

Contrariwise, if a particular CPU/computer has the finest
power-performance in the entire galaxy, but it is not fast enough to
do the job I currently need done, I'll pass it over and buy one (with
higher power consumption) that _will_ do the job I need done.

The concept of selecting a CPU/computer based on power-performance is
ludicrous! ;-)

 

[Robert Myers]

I've been following the vigorous debate with considerable amusement.
That's because I think the debaters can't see the tree for the forest.
;-)

There was a time when I used a computer primarily as a text processor.
Had power consumption been critical to me, I would _not_ have asked,
"what processor has the best power-performance?". I'd have asked "of
the computers that will do my job, which has the lowest power" and
then I would have bought that computer.

Contrariwise, if a particular CPU/computer has the finest
power-performance in the entire galaxy, but it is not fast enough to
do the job I currently need done, I'll pass it over and buy one (with
higher power consumption) that _will_ do the job I need done.

The discussion wasn't about selecting a CPU. The discussion was about
the state of the technology and a claim that Via had the edge over
Intel.

Performance/watt is indisputably important now and going to be
increasingly important wherever there is lots of muscle in a small
space (servers, HPC) and for mobile applications. For applications
like gaming machines, the cooling envelope is going to determine how
much power you can stuff into one box. What does that leave? Business
desktop and home entertainment centers, where fanless operation is
highly desirable. Performance/watt is where it's at.

RM

 

[Tony Hill]

I've been following the vigorous debate with considerable amusement.
That's because I think the debaters can't see the tree for the forest.
;-)

There was a time when I used a computer primarily as a text processor.
Had power consumption been critical to me, I would _not_ have asked,
"what processor has the best power-performance?". I'd have asked "of
the computers that will do my job, which has the lowest power" and
then I would have bought that computer.

The fallacy here is that a modern processor doesn't consume maximum
power if it's not in it's maximum performance state. Dynamic power
adjustment tends to dictate that the processor with the best
performance/watt at it's maximum level of performance will also have
the lower power consumption for any other level of performance you
need. This isn't always 100% accurate, but it's a close enough
estimate to go by.

Contrariwise, if a particular CPU/computer has the finest
power-performance in the entire galaxy, but it is not fast enough to
do the job I currently need done, I'll pass it over and buy one (with
higher power consumption) that _will_ do the job I need done.

Certainly, otherwise we would all be using super-low-power embedded
ARM chips for everything. However where in the past there were LOTS
of applications that require maximum performance regardless of power
consumption, going forward the quantity of applications seems to be
dropping.

The concept of selecting a CPU/computer based on power-performance is
ludicrous! ;-)

As a sole metric, certainly. However it is an increasingly important
measure when comparing processors. Whether you need a low power
consumption chip to fit into the small confines of a notebook or a
low-noise system to sit in your living room, hitting a certain
performance level for a given amount of power consumption (or hitting
a certain power consumption point for a given level of performance)
can definitely play a role in deciding things.

 

[Kevin C.]

The concept of selecting a CPU/computer based on power-performance is
ludicrous! ;-)

Not any more ludicrous than the "silent pc" builders, who are really just
reacting to the increasingly large and noisy cooling devices necessary to
counter the sloppy heat orgies that have been growing ever more common in
consumer desktop PCs. Do we really want to continue with this? Compare the
size of a modern CPU die to its accompanying heatsink. What's the point of a
"micro"processor if you have to stick a borg cube on top of it? Taken to its
logical conclusion, one day we will end up back where we started --
computers will take up entire rooms. Though this time because of the
requisite liquid nitrogen refrigeration system instead of vacuum tubes.

No, I think it's high time that Intel, AMD, and the others took a good long
look at factors other than the raw performance of their chips. They need to
get back to engineering a consumer *product* and not just manufacturing
usable prototypes. GPU makers should also take note before we're forced to
go 3-wide on the slots.

 

[George Macdonald]

Not any more ludicrous than the "silent pc" builders, who are really just
reacting to the increasingly large and noisy cooling devices necessary to
counter the sloppy heat orgies that have been growing ever more common in
consumer desktop PCs. Do we really want to continue with this? Compare the
size of a modern CPU die to its accompanying heatsink. What's the point of a
"micro"processor if you have to stick a borg cube on top of it? Taken to its
logical conclusion, one day we will end up back where we started --
computers will take up entire rooms. Though this time because of the
requisite liquid nitrogen refrigeration system instead of vacuum tubes.

Computers took up whole rooms long after vacuum tubes for the same reasons
as today... and which you mention: cooling requirements was one important
and enduring reason. There was, of course, also the size of peripheral
devices, like disk drive housings the size of a smallish truck, rows of
tape drives each the size of a large refrigerator, card readers, printers
and power units and the cabling of it all together. The cooling method
varied from water circulation to direct contact freon. Keith knows more
about details here.

No, I think it's high time that Intel, AMD, and the others took a good long
look at factors other than the raw performance of their chips. They need to
get back to engineering a consumer *product* and not just manufacturing
usable prototypes. GPU makers should also take note before we're forced to
go 3-wide on the slots.

People buy what they need and I "need" the most power I can get,
fortunately today in a box the size of a modern PC - raw performance
matters, whether it be for the avid gamer or the business decision maker.
Quite simply, there is no such thing as too much compute power and I for
one am disappointed that the performance "wall" has been hit. In fact I'm
hoping that there just might be some genius sitting in a lab somewhere who
can bend/break the laws of physics a bit more.

If you want low power, fanless operation and the resulting performance
deficit, you can still get it: buy a system with a VIA based CPU or a low
power Celeron or even the very low power Pentium-M. Me?... I don't care
for that and am very happy with the "borg cube" which keeps my CPU cool.