| Author |
Message |
Rupert Pigott
Guest
|
 Posted:
Tue Aug 02, 2005 5:22 am Post subject:
Re: Itanium versus Others |
|
|
John Savard wrote:
| Quote: | On 31 Jul 2005 17:39:08 -0700, "Rupert Pigott" <darkboong@hotmail.com
wrote, in part:
Hardly exotic. Multi-core chips have been with us for a while now, and
it sells more significantly more units than Itanic does (if you believe
the market analysts). These days you can buy multi-core chips over the
counter for your PC (AMD X2s and the new Opterons). :)
That wasn't quite what I meant.
The POWER5 is simply 'exotic' because it is less easy to find over the
counter than the Itanium.
|
True enough. It appeared that you can buy POWER5s from IBM's website
last
time I checked. I have not seen many Itanics on offer at my usual OTC
suppliers.
<SARCASM>
Maybe that's down to the "lack of demand" that Dell seems so worried
about
with chips like the AMD64 & Opteron.
</SARCASM>
| Quote: | As for multi-core, what I have against it is simply that it isn't really
better - except for Microsoft licensing policies - than just having
several chips on the same motherboard. Which avoids yield issues (such
as, of course, bedevil the Itanium).
|
It does make for far cheaper boards, lower latency on CC ops (very
important for scaling), plus it allows you to pack more into a smaller
space. I am not convinced that it helps the heat situation at all
though.
| Quote: | If one looks at *latency* *instead* of throughput, because throughput
can be increased without limit by using more of the same kind of chip -
whatever kind of chip - then it only stands to naive reason that if chip
A gets the same throughput as chip B with half the cores... it likely
has half the latency.
|
It does not stand to reason at all. Latency and Throughput are entirely
orthogonal to each other. In fact if I was going to pick a "naive
reason"
for the same throughput I'd put it down to better optimised code or
perhaps the core simply fits the task at hand better. A prime example
of
that kind of architecture fitting the task would be a C6000 running a
DSP
app.
| Quote: | Is the Itanium the answer? If so, Intel is right to believe in it.
|
Itanic is not the only answer, and it appears that it is not a
stunningly
good answer for many people at that (YMMV).
FWIW I do have some affection for unusual architectures. My current
interest lies in threaded architectures that present a uniform latency
to the programmer. I really like the idea of being able to prove that
my code will always respond in exactly x nanoseconds. Deterministic
Execution could simplify a lot of sychronisation, comms and I/O guff.
True on-chip DE with a MTA may actually allow you to ditch sync HW and
SW altogether, thereby eliminating inter-thread comms latency for zero
cost. Definitely not something that Netburst, Itanic or the K8 core
are going to do anytime soon. :)
Cheers,
Rupert |
|
| Back to top |
|
 |
John Savard
Guest
|
| Posted:
Wed Aug 03, 2005 6:14 am Post subject:
Re: Itanium versus Others |
|
|
On 31 Jul 2005 21:41:01 -0700, "David Kanter" <dkanter@gmail.com> wrote,
in part:
| Quote: | Could you define your terms a little specifically? What are throughput
and latency measured in? TPC-C transactions/sec and the latency to
finish a transaction?
|
In how many months can nine women, working together, have one baby?
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 |
|
|
Rupert Pigott
Guest
|
| Posted:
Fri Aug 05, 2005 12:15 am Post subject:
Re: Itanium versus Others |
|
|
John Savard wrote:
| Quote: | On 31 Jul 2005 21:41:01 -0700, "David Kanter" <dkanter@gmail.com> wrote,
in part:
Could you define your terms a little specifically? What are throughput
and latency measured in? TPC-C transactions/sec and the latency to
finish a transaction?
In how many months can nine women, working together, have one baby?
|
The man was talking about Transactions, not Babies. :)
Cheers,
Rupert |
|
| Back to top |
|
|
CJT
Guest
|
| Posted:
Fri Aug 05, 2005 6:37 am Post subject:
Re: Itanium versus Others |
|
|
Rupert Pigott wrote:
| Quote: | John Savard wrote:
On 31 Jul 2005 21:41:01 -0700, "David Kanter" <dkanter@gmail.com> wrote,
in part:
Could you define your terms a little specifically? What are throughput
and latency measured in? TPC-C transactions/sec and the latency to
finish a transaction?
In how many months can nine women, working together, have one baby?
The man was talking about Transactions, not Babies. :)
Cheers,
Rupert
Delivery seems like a transaction to me. |
--
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 |
|
|
Hank Oredson
Guest
|
| Posted:
Fri Aug 05, 2005 8:15 am Post subject:
Re: Itanium versus Others |
|
|
"CJT" <abujlehc@prodigy.net> wrote in message
news:42F2C2DD.2080005@prodigy.net...
| Quote: | Rupert Pigott wrote:
John Savard wrote:
On 31 Jul 2005 21:41:01 -0700, "David Kanter" <dkanter@gmail.com> wrote,
in part:
Could you define your terms a little specifically? What are throughput
and latency measured in? TPC-C transactions/sec and the latency to
finish a transaction?
In how many months can nine women, working together, have one baby?
The man was talking about Transactions, not Babies. :)
Cheers,
Rupert
Delivery seems like a transaction to me.
|
As is conception.
There is no commit ("Name of new baby is ...") until both
transactions have completed.succesfully.
--
... Hank
http://home.earthlink.net/~horedson
http://home.earthlink.net/~w0rli |
|
| Back to top |
|
|
Dan Koren
Guest
|
| Posted:
Sat Aug 06, 2005 8:15 am Post subject:
Re: Itanium versus Others |
|
|
"Rupert Pigott" <darkboong@hotmail.com> wrote in message
news:1123189702.890063.182360@g43g2000cwa.googlegroups.com...
| Quote: | John Savard wrote:
On 31 Jul 2005 21:41:01 -0700, "David Kanter" <dkanter@gmail.com> wrote,
in part:
Could you define your terms a little specifically? What are throughput
and latency measured in? TPC-C transactions/sec and the latency to
finish a transaction?
In how many months can nine women, working together, have one baby?
The man was talking about Transactions, not Babies. :)
|
Aren't babies just another kind of transaction?
dk |
|
| Back to top |
|
|
Mark Hahn
Guest
|
| Posted:
Sat Aug 06, 2005 4:15 pm Post subject:
Re: Itanium versus Others |
|
|
Robert Myers <rbmyersusa@gmail.com> wrote:
| Quote: | From that single measure of performance, the statement that the
enormous cache size of Itanium makes a big difference is incorrect, as
no. it2's enormous cache does clearly make a huge difference. if you
don't believe, just recalculate it2's specFP results after having omitted
the obsolete-because-they-ridiculously-tiny components. think of it as
an outlier-rejection procedure.
I did this late last year, and it2 winds up barely faster than opteron.
From figure of merit I cited, the statement I made is correct. The
only benchmark that matters is your own application.
|
that's true but useless for most people, who are not buying It2's as embedded
processors. benchmarks like Spec exist because people want a more general
answer. the problem is that Spec2000 is, in some cases, dominated by
performance of certain components which no longer run as intended.
| Quote: | But which are you saying? That, if you remove the "outliers"
(definition: data that don't meet your preconceptions), the importance
|
bullshit. the outliers are codes which run drastically faster because
the It2 cache is comparable in size to a whole machine's core when spec2000
was designed. this makes the It2 seem fast in general, when in fact it's
basically only fast for either hand-tuned codes or codes that fit entirely
in cache.
| Quote: | of the cache becomes apparent or that Itanium is no faster than Opteron
or both?
|
it's not a matter of performance, but of predictive power or generality
of the benchmark. the art component is a real problem these days
because It2 break is by cache size and Sun breaks it by a unique loop
transformation. |
|
| Back to top |
|
|
Robert Myers
Guest
|
| Posted:
Sat Aug 06, 2005 4:15 pm Post subject:
Re: Itanium versus Others |
|
|
Mark Hahn wrote:
| Quote: | Robert Myers <rbmyersusa@gmail.com> wrote:
From that single measure of performance, the statement that the
enormous cache size of Itanium makes a big difference is incorrect, as
no. it2's enormous cache does clearly make a huge difference. if you
don't believe, just recalculate it2's specFP results after having omitted
the obsolete-because-they-ridiculously-tiny components. think of it as
an outlier-rejection procedure.
I did this late last year, and it2 winds up barely faster than opteron.
From figure of merit I cited, the statement I made is correct. The
only benchmark that matters is your own application.
that's true but useless for most people, who are not buying It2's as embedded
processors. benchmarks like Spec exist because people want a more general
answer. the problem is that Spec2000 is, in some cases, dominated by
performance of certain components which no longer run as intended.
But which are you saying? That, if you remove the "outliers"
(definition: data that don't meet your preconceptions), the importance
bullshit. the outliers are codes which run drastically faster because
the It2 cache is comparable in size to a whole machine's core when spec2000
was designed. this makes the It2 seem fast in general, when in fact it's
basically only fast for either hand-tuned codes or codes that fit entirely
in cache.
|
Spare me the locker room talk, okay? I can sound like an ignorant lout
without using that kind of language, and I'll bet you can, too.
I think it's pretty well understood that getting good scheduling for
Itanium is a challenge to compiler developers and to developers in
general. One way around the scheduling problem, if the problems are
due to memory stalls, is to fit the entire problem into cache. If the
problem can't be fit into cache, then you need to solve the scheduling
problem some other way.
As it is, changing the size of the L2 cache between 3M and 9M changes
the overall score negligibly. If your explanation is correct, it's
apparently because the problems will fit into 3M, not because Itanium
needs 9M to get decent performance, which was the original claim.
| Quote: | of the cache becomes apparent or that Itanium is no faster than Opteron
or both?
it's not a matter of performance, but of predictive power or generality
of the benchmark. the art component is a real problem these days
because It2 break is by cache size and Sun breaks it by a unique loop
transformation.
|
Uh-huh. John McCalpin complained about that here, too. Optimization
is about being clever. Your favorite processor (let me guess,...) puts
minimal emphasis on cleverness. That's a plus.
RM |
|
| Back to top |
|
|
Rupert Pigott
Guest
|
| Posted:
Sun Aug 07, 2005 5:37 am Post subject:
Re: Itanium versus Others |
|
|
Dan Koren wrote:
| Quote: | "Rupert Pigott" <darkboong@hotmail.com> wrote in message
news:1123189702.890063.182360@g43g2000cwa.googlegroups.com...
John Savard wrote:
On 31 Jul 2005 21:41:01 -0700, "David Kanter" <dkanter@gmail.com> wrote,
in part:
Could you define your terms a little specifically? What are throughput
and latency measured in? TPC-C transactions/sec and the latency to
finish a transaction?
In how many months can nine women, working together, have one baby?
The man was talking about Transactions, not Babies. :)
Aren't babies just another kind of transaction?
|
That only really applies if you are a vendor who fucks customers.
Cheers,
Rupert |
|
| Back to top |
|
|
David Kanter
Guest
|
| Posted:
Sun Aug 07, 2005 6:01 am Post subject:
Re: Itanium versus Others |
|
|
| Quote: | it's not a matter of performance, but of predictive power or generality
of the benchmark. the art component is a real problem these days
because It2 break is by cache size and Sun breaks it by a unique loop
transformation.
Uh-huh. John McCalpin complained about that here, too. Optimization
is about being clever. Your favorite processor (let me guess,...) puts
minimal emphasis on cleverness. That's a plus.
|
Well, I think that John was absolutely right that SPEC was in need of
an overhaul...
I guess we'll just have to wait for SPEC_2004...
David |
|
| Back to top |
|
|
Colonel Forbin
Guest
|
| Posted:
Sun Aug 07, 2005 7:26 am Post subject:
Re: Itanium versus Others |
|
|
In article <1123375052.984554.176140@g43g2000cwa.googlegroups.com>,
Rupert Pigott <darkboong@hotmail.com> wrote:
| Quote: |
Dan Koren wrote:
"Rupert Pigott" <darkboong@hotmail.com> wrote in message
news:1123189702.890063.182360@g43g2000cwa.googlegroups.com...
John Savard wrote:
On 31 Jul 2005 21:41:01 -0700, "David Kanter" <dkanter@gmail.com> wrote,
in part:
Could you define your terms a little specifically? What are throughput
and latency measured in? TPC-C transactions/sec and the latency to
finish a transaction?
In how many months can nine women, working together, have one baby?
The man was talking about Transactions, not Babies. :)
Aren't babies just another kind of transaction?
That only really applies if you are a vendor who fucks customers.
|
I'm not sure what transpired between you two in the powder room, but
you have to admit that dk's "Rhubarb Spigot" jab was at least grossly
humorous in a Mel Brooks sense.
The "nine women and a baby" problem is the stereotypical definition
of pipelining. |
|
| Back to top |
|
|
Rupert Pigott
Guest
|
| Posted:
Sun Aug 07, 2005 4:15 pm Post subject:
Re: Itanium versus Others |
|
|
Colonel Forbin wrote:
[SNIP]
| Quote: | I'm not sure what transpired between you two in the powder room, but
|
FWIW that sardonic comment was not specifically aimed at DK.
Nothing happened in the Powder room. IIRC DK ran out of technical
arguments so he resorted to name-dropping and name-calling. AFAIK
it is all in the public domain.
| Quote: | you have to admit that dk's "Rhubarb Spigot" jab was at least grossly
humorous in a Mel Brooks sense.
|
Sure, it was much higher quality than his on-topic ravings.
| Quote: | The "nine women and a baby" problem is the stereotypical definition
of pipelining.
|
Yup, aware of it, accredited to Fred Brooks.
Cheers,
Rupert |
|
| Back to top |
|
|
Dan Koren
Guest
|
| Posted:
Tue Aug 09, 2005 12:16 am Post subject:
Re: Itanium versus Others |
|
|
"Rupert Pigott" <darkboong@hotmail.com> wrote in message
news:1123375052.984554.176140@g43g2000cwa.googlegroups.com...
| Quote: |
Dan Koren wrote:
"Rupert Pigott" <darkboong@hotmail.com> wrote in message
news:1123189702.890063.182360@g43g2000cwa.googlegroups.com...
John Savard wrote:
On 31 Jul 2005 21:41:01 -0700, "David Kanter" <dkanter@gmail.com
wrote,
in part:
Could you define your terms a little specifically? What are
throughput
and latency measured in? TPC-C transactions/sec and the latency to
finish a transaction?
In how many months can nine women, working together, have one baby?
The man was talking about Transactions, not Babies. :)
Aren't babies just another kind of transaction?
That only really applies if you are a vendor who fucks customers.
|
Are you familiar with the concept of "immaculate conception"?
dk |
|
| Back to top |
|
|
Dan Koren
Guest
|
| Posted:
Tue Aug 09, 2005 12:16 am Post subject:
Re: Itanium versus Others |
|
|
"Colonel Forbin" <forbin@dev.nul> wrote in message
news:kjeJe.65328$yC5.9544@tornado.ohiordc.rr.com...
| Quote: | In article <1123375052.984554.176140@g43g2000cwa.googlegroups.com>,
Rupert Pigott <darkboong@hotmail.com> wrote:
Dan Koren wrote:
"Rupert Pigott" <darkboong@hotmail.com> wrote in message
news:1123189702.890063.182360@g43g2000cwa.googlegroups.com...
John Savard wrote:
On 31 Jul 2005 21:41:01 -0700, "David Kanter" <dkanter@gmail.com
wrote,
in part:
Could you define your terms a little specifically? What are
throughput
and latency measured in? TPC-C transactions/sec and the latency to
finish a transaction?
In how many months can nine women, working together, have one baby?
The man was talking about Transactions, not Babies. :)
Aren't babies just another kind of transaction?
That only really applies if you are a vendor who fucks customers.
I'm not sure what transpired between you two in the powder room, but
you have to admit that dk's "Rhubarb Spigot" jab was at least grossly
humorous in a Mel Brooks sense.
|
I am deeply touched and flattered.
dk |
|
| Back to top |
|
|
Rupert Pigott
Guest
|
| Posted:
Thu Aug 11, 2005 1:47 pm Post subject:
Re: Itanium versus Others |
|
|
Dan Koren wrote:
| Quote: | "Rupert Pigott" <darkboong@hotmail.com> wrote in message
news:1123375052.984554.176140@g43g2000cwa.googlegroups.com...
Dan Koren wrote:
|
[SNIP]
| Quote: | Aren't babies just another kind of transaction?
That only really applies if you are a vendor who fucks customers.
Are you familiar with the concept of "immaculate conception"?
|
Sure, but I don't buy it because I'm an Aethiest.
AFAICT it works roughly the same way but has only been corroborated
by prejudiced witnesses. Furthermore the support operation requires
millions of people, and it costs far more in terms of blood and
money.
Cheers,
Rupert |
|
| Back to top |
|
|
|
|
|
|
FAQ
Memberlist
Register
Profile
Log in to check your private messages
Log in