| Author |
Message |
John Savard
Guest
|
 Posted:
Thu Jul 28, 2005 8:15 am Post subject:
Itanium versus Others |
|
|
Intel's Itanium chip is rather expensive, and its MHz speed is not as
high as that of their Pentium 4 chips. Now, of course, we have the AMD
64 chips and Intel's EM64T. Is there a point to Itanium?
Wondering that, I looked up benchmark results, and found that as far as
floating-point performance, at least, is concerned, there certainly is.
I also came across IBM's POWER5. It also performs comparably to an
Itanium, but with a higher MHz speed. And I found out that the POWER5
isn't the same as the PowerPC G5 from IBM used in high-end Macintosh
computers. It's considerably more powerful, and it is a bulky multi-chip
module with four dual-core chips.
Since the POWER5 appears 'exotic', it does seem to me that the Itanium
is just about the only 'supercomputer-like' chip easily available; and,
of course, a single core providing the same throughput as eight cores
will do it with considerably less latency as well.
John Savard
http://www.quadibloc.com/index.html
_________________________________________
Usenet Zone Free Binaries Usenet Server
More than 120,000 groups
Unlimited download
http://www.usenetzone.com to open account |
|
| Back to top |
|
 |
glen herrmannsfeldt
Guest
|
| Posted:
Thu Jul 28, 2005 8:15 am Post subject:
Re: Itanium versus Others |
|
|
John Savard wrote:
| Quote: | Intel's Itanium chip is rather expensive, and its MHz speed is not as
high as that of their Pentium 4 chips. Now, of course, we have the AMD
64 chips and Intel's EM64T. Is there a point to Itanium?
|
Clock speed is never a good comparison between different processor
families, and often not even within a processor family.
Itanium, as a VLIW processor, depends on many things getting done in
a clock cycle, so likely will have a slower clock for equivalent
performance than most other processors.
AMD processor often do more per clock cycle, and have people trying to
compare them using clock rates. They have a system designed to allow
fair comparison independent of clock rate.
-- glen |
|
| Back to top |
|
|
CJT
Guest
|
| Posted:
Thu Jul 28, 2005 8:15 am Post subject:
Re: Itanium versus Others |
|
|
John Savard wrote:
| Quote: | Intel's Itanium chip is rather expensive, and its MHz speed is not as
high as that of their Pentium 4 chips. Now, of course, we have the AMD
64 chips and Intel's EM64T. Is there a point to Itanium?
Wondering that, I looked up benchmark results, and found that as far as
floating-point performance, at least, is concerned, there certainly is.
I also came across IBM's POWER5. It also performs comparably to an
Itanium, but with a higher MHz speed. And I found out that the POWER5
isn't the same as the PowerPC G5 from IBM used in high-end Macintosh
computers. It's considerably more powerful, and it is a bulky multi-chip
module with four dual-core chips.
Since the POWER5 appears 'exotic', it does seem to me that the Itanium
is just about the only 'supercomputer-like' chip easily available; and,
of course, a single core providing the same throughput as eight cores
will do it with considerably less latency as well.
John Savard
http://www.quadibloc.com/index.html
_________________________________________
Usenet Zone Free Binaries Usenet Server
More than 120,000 groups
Unlimited download
http://www.usenetzone.com to open account
|
I'd be inclined to put the Itanium in the "exotic" category. It's a
departure from mainstream architectures and not all that ubiquitous.
--
The e-mail address in our reply-to line is reversed in an attempt to
minimize spam. Our true address is of the form che...@prodigy.net. |
|
| Back to top |
|
|
Andrew Reilly
Guest
|
| Posted:
Thu Jul 28, 2005 2:47 pm Post subject:
Re: Itanium versus Others |
|
|
On Thu, 28 Jul 2005 00:11:47 -0700, glen herrmannsfeldt wrote:
| Quote: | John Savard wrote:
Intel's Itanium chip is rather expensive, and its MHz speed is not as
high as that of their Pentium 4 chips. Now, of course, we have the AMD
64 chips and Intel's EM64T. Is there a point to Itanium?
Clock speed is never a good comparison between different processor
families, and often not even within a processor family.
Itanium, as a VLIW processor, depends on many things getting done in
a clock cycle, so likely will have a slower clock for equivalent
performance than most other processors.
AMD processor often do more per clock cycle, and have people trying to
compare them using clock rates. They have a system designed to allow
fair comparison independent of clock rate.
|
A quick squiz on the SPEC results site, looking for single processor
systems with broadly similar cache sizes (i.e., the Itanium2 systems with
1.5M L3 vs Opteron 145s or so) seem to have fairly similar peak SPEC2000
FPU numbers. I imagine that 9M of on-chip cache on an Opteron might help
it along too, and make it cost a bomb too.
--
Andrew |
|
| Back to top |
|
|
Colonel Forbin
Guest
|
| Posted:
Thu Jul 28, 2005 4:15 pm Post subject:
Re: Itanium versus Others |
|
|
In article <42e84f97.2504039@news.usenetzone.com>,
John Savard <jsavard@excxn.aNOSPAMb.cdn.invalid> wrote:
| Quote: |
Intel's Itanium chip is rather expensive, and its MHz speed is not as
high as that of their Pentium 4 chips. Now, of course, we have the AMD
64 chips and Intel's EM64T. Is there a point to Itanium?
Wondering that, I looked up benchmark results, and found that as far as
floating-point performance, at least, is concerned, there certainly is.
I also came across IBM's POWER5. It also performs comparably to an
Itanium, but with a higher MHz speed. And I found out that the POWER5
isn't the same as the PowerPC G5 from IBM used in high-end Macintosh
computers. It's considerably more powerful, and it is a bulky multi-chip
module with four dual-core chips.
Since the POWER5 appears 'exotic', it does seem to me that the Itanium
is just about the only 'supercomputer-like' chip easily available; and,
of course, a single core providing the same throughput as eight cores
will do it with considerably less latency as well.
|
I wouldn't really call either Itanium or POWER5 "supercomputer chips."
I'd reserve that for things like the NEC SX-6i. The Itanium and
POWER5 are more aptly termed high performance general purpose
processors which can be used in MPP supercomputers. Neither is
specifically targeted at the HPC market.
POWER5 is IBM's current high end RISC architecture. While it can be
used to construct MPP systems, the main focus is on high end SMP
servers with high RAS.
The Itanium is a VLIW design putting the location and exploitation
of concurrency into the compilers.
Given the substantial differences in architecture and implementation
between Itanium, POWER5 and x86-64 products (indeed even between
AMD and Intel x86-64) it is not meaningful to compare 'MHz' clock
rates. The main points of comparison are benchmarks and, of
increasing importance, power dissipation.
Indeed, these days there is much more emphasis on the comparison
of systems over particular CPU architectures as MPP systems place
substantial emphasis on communications latency.
There was a point to Itanium in the beginning, but the initial
vision that the EPIC VLIW approach would prove inherently superior
to other architectures largely failed to be realized, particularly
with rapid evolution of OoO and the appearance of multicore versions
of other architectures. The Itanium advantage was based on limitations
of RISC designs of a decade ago and largely evaporated as development
time far exceeded early estimates.
Early Itanium versions had quite disappointing performance, and power
management has been a real problem. Many people believe the market
opportunity for Itanium is past with the development of relatively
powerful x86-64 products at much lower cost. This is particularly true
given HP's recent end of its central role in Itanium design and
development.
SGI does have a heavy investment in Itanium at present in its Altix
line of ccNUMA MPP supers, but the HPC market isn't sufficient to
sustain the Itanium, particularly in the face of rapid x86-64
development.
With the consolidation of the processor market, it's not clear that
Itanium sales volume will ever become sufficient to sustain the product
in the long term. |
|
| Back to top |
|
|
Zak
Guest
|
| Posted:
Thu Jul 28, 2005 10:14 pm Post subject:
Re: Itanium versus Others |
|
|
John Savard wrote:
| Quote: | Intel's Itanium chip is rather expensive, and its MHz speed is not as
high as that of their Pentium 4 chips. Now, of course, we have the AMD
64 chips and Intel's EM64T. Is there a point to Itanium?
Wondering that, I looked up benchmark results, and found that as far as
floating-point performance, at least, is concerned, there certainly is.
|
Itanium is sold at a very low price compared to the silicon it uses, I
think. Intel finances this from x86 sales, I guess. Intel wants Itanium
to succeed - probably because it has a monopoly on it, while x86
'suffers' competition from AMD and VIA.
Competition is bad for margins but good for consumers. Consumer avoid
Itanium because it is effectively single source - unless it is a short
term ('niche'?) application.
Thomas |
|
| Back to top |
|
|
Robert Myers
Guest
|
| Posted:
Thu Jul 28, 2005 10:40 pm Post subject:
Re: Itanium versus Others |
|
|
Zak wrote:
| Quote: | John Savard wrote:
Intel's Itanium chip is rather expensive, and its MHz speed is not as
high as that of their Pentium 4 chips. Now, of course, we have the AMD
64 chips and Intel's EM64T. Is there a point to Itanium?
Wondering that, I looked up benchmark results, and found that as far as
floating-point performance, at least, is concerned, there certainly is.
Itanium is sold at a very low price compared to the silicon it uses, I
think. Intel finances this from x86 sales, I guess. Intel wants Itanium
to succeed - probably because it has a monopoly on it, while x86
'suffers' competition from AMD and VIA.
Competition is bad for margins but good for consumers. Consumer avoid
Itanium because it is effectively single source - unless it is a short
term ('niche'?) application.
|
The list I love to hate, the "Top" 500, does not support your
prejudices--unless you think machines on that list are for short term
('niche') applications. It's true: Intel Xeon dominates the list, with
placements in 254 systems. One of the possible "second
sources"--Opteron--accounts for 25 of the systems, as opposed to 77
systems for Power processors, 36 for PA-RISC and 79 for Itanium. There
are no Via processors on that list, and I'd be interested to hear if
they are anywhere in use for HPC applications.
As for pricing per sq cm of silicon, well...
But I have a sense you're right about one thing: Intel would like
Itanium to succeed. Carly Fiorina may no longer have her job, but it
remains to be seen as to whether she was wrong that it is not wise to
bet against Intel.
That's a different question, of course, as to whether *Intel* might not
have been better off making different decisions, but the big money is
already spent.
RM |
|
| Back to top |
|
|
Colonel Forbin
Guest
|
| Posted:
Thu Jul 28, 2005 11:17 pm Post subject:
Re: Itanium versus Others |
|
|
In article <1122572454.862339.6690@g44g2000cwa.googlegroups.com>,
Robert Myers <rbmyersusa@gmail.com> wrote:
| Quote: |
systems for Power processors, 36 for PA-RISC and 79 for Itanium. There
are no Via processors on that list, and I'd be interested to hear if
they are anywhere in use for HPC applications.
|
Not in the Top500 class, but Los Alamos did the "Green Destiny" HPC
cluster using Transmeta CPUs:
http://sss.lanl.gov/pubs/ISC_2004_Final.pdf
"Although the performance of supercomputers on our n-body cosmology code
has improved by a factor of nearly 2000 since 1991, the performance per
watt has only improved 300-fold and the performance per square foot only
65-fold. Clearly, we are building less and less efficient
supercomputers, thus resulting in the construction of new machines rooms
1 and even entirely new buildings. Furthermore, as these supercomputers
continue to follow Moore s Law for Power Consumption, the reliability
of these supercomputers continues to plummet, relative to Arrenhius
equation for microelectronics.
"To address these problems, we built a super-efficient supercomputer
dubbed Green Destiny, a 240-processor supercomputer that fits in a
telephone booth (i.e., a footprint of five square feet) and sips less
than 5.2 kW of power at full load [FWW02, WWF02, Feng03]. This
Supercomputer for the Rest of Us - a 2003 R&D 100 award-winning machine
- provided affordable, general-purpose supercomputing to our application
scientists while sitting in an 85-90° F (29-32° C) dusty warehouse at
7,400 feet (2256 meters) above sea level. Furthermore, it delivered
reliable computing cycles without any special facilities, i.e., no air
conditioning, no humidification control, no air filtration, and no
ventilation, and without any unscheduled downtime." |
|
| Back to top |
|
|
Zak
Guest
|
| Posted:
Thu Jul 28, 2005 11:54 pm Post subject:
Re: Itanium versus Others |
|
|
Robert Myers wrote:
| Quote: | The list I love to hate, the "Top" 500, does not support your
prejudices--unless you think machines on that list are for short term
('niche') applications.
|
I think they are: you buy one supercomputer and adapt your software to
it. The next supercomputer will be a different model. So if you get the
Itanium cache for free, by all means get it.
To build a long term strategy on it is follish from a business point of
view. If Itanium were to dominate, the PC will be back to '50 % of the
cost goes to the CPU' pretty soon - just because customers have no
choice where to buy their compatible part.
AMD and Intel do very well together to keep prices low. VIA is not that
important (low end) but is a possible third supplier.
IMO Itanium was set up to become a single source chip. Which explains
why one can buy a 9 MB cache rather affordably - not because it is cheap
to make, but because it has to be sold in volume.
| Quote: | It's true: Intel Xeon dominates the list, with
placements in 254 systems. One of the possible "second
sources"--Opteron--accounts for 25 of the systems, as opposed to 77
systems for Power processors, 36 for PA-RISC and 79 for Itanium. There
are no Via processors on that list, and I'd be interested to hear if
they are anywhere in use for HPC applications.
|
I wouldn't want to create the next Apply using Itanium. Just too big a
chance to send all your profits to a single supplier.
If you bought Xeon and were gouged, you could switch rather more easily.
Thomas |
|
| Back to top |
|
|
Del Cecchi
Guest
|
| Posted:
Fri Jul 29, 2005 7:09 am Post subject:
Re: Itanium versus Others |
|
|
John Savard wrote:
| Quote: | Intel's Itanium chip is rather expensive, and its MHz speed is not as
high as that of their Pentium 4 chips. Now, of course, we have the AMD
64 chips and Intel's EM64T. Is there a point to Itanium?
Wondering that, I looked up benchmark results, and found that as far as
floating-point performance, at least, is concerned, there certainly is.
I also came across IBM's POWER5. It also performs comparably to an
Itanium, but with a higher MHz speed. And I found out that the POWER5
isn't the same as the PowerPC G5 from IBM used in high-end Macintosh
computers. It's considerably more powerful, and it is a bulky multi-chip
module with four dual-core chips.
Since the POWER5 appears 'exotic', it does seem to me that the Itanium
is just about the only 'supercomputer-like' chip easily available; and,
of course, a single core providing the same throughput as eight cores
will do it with considerably less latency as well.
John Savard
http://www.quadibloc.com/index.html
_________________________________________
Usenet Zone Free Binaries Usenet Server
More than 120,000 groups
Unlimited download
http://www.usenetzone.com to open account
|
Power5 is available in various packages in various models of pSeries and
iSeries boxes from IBM. You might want to recheck your benchmarks if
you think it takes 4 chips of Power5 to keep up with an Itanium.
--
Del Cecchi
"This post is my own and doesn’t necessarily represent IBM’s positions,
strategies or opinions.” |
|
| Back to top |
|
|
Maynard Handley
Guest
|
| Posted:
Fri Jul 29, 2005 8:15 am Post subject:
Re: Itanium versus Others |
|
|
In article <mJ6Ge.37995$yC5.37238@tornado.ohiordc.rr.com>,
forbin@dev.nul (Colonel Forbin) wrote:
| Quote: | There was a point to Itanium in the beginning, but the initial
vision that the EPIC VLIW approach would prove inherently superior
to other architectures largely failed to be realized, particularly
with rapid evolution of OoO and the appearance of multicore versions
of other architectures. The Itanium advantage was based on limitations
of RISC designs of a decade ago and largely evaporated as development
time far exceeded early estimates.
Early Itanium versions had quite disappointing performance, and power
management has been a real problem. Many people believe the market
opportunity for Itanium is past with the development of relatively
powerful x86-64 products at much lower cost. This is particularly true
given HP's recent end of its central role in Itanium design and
development.
|
Let me just state that the situation is not nearly as clear-cut as the
above statement implies. The statement above implies that Itanium is
essentially
(1) a vision failure, based on mistaken assumption about where CPU
architecture would be in 10 yrs and
(2) an engineering failure in implementing specific Itanium chips.
There is an alternative viewpoint which would say that the vision remain
essentially correct, (power will, one day, really limit what OoO can do)
that the chips were engineered as best they could be, and that the
essential problem was managerial:
(1) committee design; from a good starting point everyone and his dog
wanted their favorite weird feature added to support future mythical OSs
that do not and will never exist; and no one appears to have been given
uber-architect powers to put their foot down and say no
(2) that edicts were given demanding the chip be constructed in certain
ways that made it very hard to copy (but thus also hard to construct in
the first place) and that demanded integration with x86 in a way that
has probably sold zero Itaniums, but has weighed them all down with
baggage
(3) a pricing model that (regardless of whether it makes money off
Itaniums now or loses it) pretty clearly sends a message that Itanium is
not a mass market chip now and isn't ever going to be.
Maynard Handley |
|
| Back to top |
|
|
Tarjei T. Jensen
Guest
|
| Posted:
Fri Jul 29, 2005 8:15 am Post subject:
Re: Itanium versus Others |
|
|
Andrew Reilly wrote:
| Quote: | A quick squiz on the SPEC results site, looking for single processor
systems with broadly similar cache sizes (i.e., the Itanium2 systems with
1.5M L3 vs Opteron 145s or so) seem to have fairly similar peak SPEC2000
FPU numbers. I imagine that 9M of on-chip cache on an Opteron might help
it along too, and make it cost a bomb too.
|
I believe that the 9M cache is found on the Itanic.
greetings, |
|
| Back to top |
|
|
Zak
Guest
|
| Posted:
Fri Jul 29, 2005 8:15 am Post subject:
Re: Itanium versus Others |
|
|
Tarjei T. Jensen wrote:
| Quote: | Andrew Reilly wrote:
I imagine that 9M of on-chip cache on an Opteron might help
it along too, and make it cost a bomb too.
I believe that the 9M cache is found on the Itanic.
|
I think the point is: with a huge cash flow, you can take something like
an Opteron and sell it at a loss if you want that for strategic reasons.
I have no idea what an Itanium costs compared to, say, a Xeon MP. You'd
expect Itanium to be more expensive per mm2 of silicon (lower volume,
newer, first to be 64 bit, etc - you'd expect the price to be that of a
USIII of a Power5). But I wouldn't be surprised if Itanium is less
expensive per mm2.
Intel wants Itanium to get a monopoly on CPU manufacturing again. That's
why the architecture is funny (so it could be strongly patented) and why
the Patents are not with Intel (who needs to share them) but with a
joint-venture that shares only with Intel and HPaq.
64-bitness was not enough to sell Itanium in the mainstream, as AMD came
along with a mainstream 64 bit arch. So now it is a huge cash being
tacked on to show 'hey it is fast'. But it isn't fast; it is just sold
at way lower margins than needed. Guess what happens if customers cannot
turn back: the margin becomes normal.
Thomas |
|
| Back to top |
|
|
Nick Maclaren
Guest
|
| Posted:
Fri Jul 29, 2005 8:15 am Post subject:
Re: Itanium versus Others |
|
|
In article <name99-0D79E2.20280928072005@localhost>,
Maynard Handley <name99@name99.org> wrote:
| Quote: |
Let me just state that the situation is not nearly as clear-cut as the
above statement implies. The statement above implies that Itanium is
essentially
(1) a vision failure, based on mistaken assumption about where CPU
architecture would be in 10 yrs and
(2) an engineering failure in implementing specific Itanium chips.
There is an alternative viewpoint which would say that the vision remain
essentially correct, (power will, one day, really limit what OoO can do)
that the chips were engineered as best they could be, and that the
essential problem was managerial:
|
It is not clear that the two can easily be distinguished. I take
your point, except in a few matters where the technical assumptions
were known to be wrong at the time that it was designed - at least
one of which I am on record as pointing out back in the middle 1990s!
Regards,
Nick Maclaren. |
|
| Back to top |
|
|
Tom Linden
Guest
|
| Posted:
Fri Jul 29, 2005 9:06 pm Post subject:
Re: Itanium versus Others |
|
|
On Thu, 28 Jul 2005 19:14:19 +0200, Zak <jute@zak.invalid> wrote:
| Quote: | John Savard wrote:
Intel's Itanium chip is rather expensive, and its MHz speed is not as
high as that of their Pentium 4 chips. Now, of course, we have the AMD
64 chips and Intel's EM64T. Is there a point to Itanium?
Wondering that, I looked up benchmark results, and found that as far as
floating-point performance, at least, is concerned, there certainly is.
Itanium is sold at a very low price compared to the silicon it uses, I
think. Intel finances this from x86 sales, I guess. Intel wants Itanium
to succeed - probably because it has a monopoly on it, while x86
'suffers' competition from AMD and VIA.
Competition is bad for margins but good for consumers. Consumer avoid
Itanium because it is effectively single source - unless it is a short
term ('niche'?) application.
Thomas
|
Very likely HP is giving a significant portion of the $3B they claim to be
putting into Itanium to Intel.Seems hard to believe that Intel would
continue with it without such support, afterall, it was HP's project from
the get go. |
|
| Back to top |
|
|
|
|
|
|
FAQ
Memberlist
Register
Profile
Log in to check your private messages
Log in