Get them while they're hot

Tue, Jul 12, 2005 at 11:35:22PM -0400, Atom wrote:
Jul 12, 2005, at 6:51 PM, Bill Bradford wrote:
Get them while they're hot
Are you running the optimized builds?

1.0.4 was somehow unstable the last few days i am wondering if that
is why they went to 1.0.5.

I had been running the G4/G5 optimized builds, but ended up going back
to the official releases as I had too much trouble with the optimized
builds (and got tired of updating every 2-3 days).

Bill

Wed, Jul 13, 2005 at 10:11:43PM +0200, Jonathan Groll wrote:
Did you actually notice any difference using the optimised builds? It
does have me confounded as to why it requires optimisation. Surely, web
browsing is not too system intensive, and this is firefox not mozilla
after all?

They were a little faster (builds were optimized for the G4 or G5
rather than the G3, which is the target the "official releases" are
built for). However, there's been enough speed increase in the mainstream
code by this point that I dont really notice the difference between
the official releases and the optimized ones anymore.

Another feature of the optimized builds was tweaked prefs (pipelining,
max simultaneous connections, etc). I take care of that with the Tweak
Network Settings extension, or by twiddling about:config by hand.

I just wish Firefox on the mac was as snappy as it is on the PC though 8-(

Bill

Jul 13, 2005, at 6:09 PM, Jonathan Groll wrote:

a similar note, it will be enlightening to read what speed
improvements there will be when os x runs fully natively on intel
(without rosetta), logically it just makes sense that the fastest Ghz
intel chip is faster than the fastest 2.7? Ghz G5

That's a troll, right?

MHz is not a valid indicator of speed, capability, or ability to
gracefully handle load. A 350Z makes 287 horsepower, my truck (when I
owned it) made 240. Does that mean that a 350Z can tow as much, haul
as much, and get more people to their destination?

Wed, 13 Jul 2005, Kurt Huhn wrote:

MHz is not a valid indicator of speed, capability, or ability to
gracefully handle load. A 350Z makes 287 horsepower, my truck (when I
owned it) made 240. Does that mean that a 350Z can tow as much, haul
as much, and get more people to their destination?

, hey, a Honda Civic has a redline of 8600 rpm, so it must kick the
teeth out of a Corvette that has a redline of only 6500 rpm, right?

Thu, Jul 14, 2005 at 12:09:07AM +0200, Jonathan Groll wrote:
Probably because your mac, like mine is an aged G4 in need of an WC
processor upgrade perhaps?

Wha?

1.8Ghz iMac G5 with 1G of RAM.

I see similar "slower than Intel" performance in Firefox on my Mac mini
(1.25Ghz, 512M). The Firefox developers have admitted that performance
on the SX port isn't as fast as the other platforms (something they hope
to fix by the 1.1 release).

Bill

Jul 14, 2005, at 4:28 AM, Jonathan Groll wrote:
Seriously not intended as a troll. Granted, many, many factors come
into
play when doing comparisons, only one of which is clock speed. However,
many of us, and for too long now have blithely accepted the apple
gospel
that PPC is the inherently "superior" architecture. Surely, though,
there must come a time when a processor with enough pure 'grunt' can
outperform one that is slower but more able?

MHz != speed. improvments may bring the Intel processors closer
to being acceptable (faster pipelining, faster RAM, switched bus, etc)
but "speed" as measured in MHz is not one of them. I will grant that
AMD and Intel are bringing their Desktop processors closer to
acceptable, but they will never get there unless they're willing to
dump the legacy crap that surrounds them and start designing them to be
put into elegant and well-designed systems.

Take a look at today's slashdot:
(Intel Developer Macs G5s)

tid=174&tid=3

I see nothing there indicating that the developer macs are
outperforming G5s. I see a vague reference to "as little as 10
seconds" to boot to the desktop. Now there's a reliable benchmark!!
Not.

If we're going to claim that the intel systems from Apple are faster
than anything else, why don't we compare them side by side, under equal
environments and conditions, and take real measurements, with real
applications, under heavy load, and performing functions that people
who buy high-end workstations need them for? Until I see such tests,
I'll not even consider "it feels faster" with the smallest grain of
salt.

Wed, 13 Jul 2005 @ 20:10 -0500, Bill Bradford said:

Thu, Jul 14, 2005 at 12:09:07AM +0200, Jonathan Groll wrote:
Probably because your mac, like mine is an aged G4 in need of an WC
processor upgrade perhaps?

Wha?

1.8Ghz iMac G5 with 1G of RAM.

I see similar "slower than Intel" performance in Firefox on my Mac mini
(1.25Ghz, 512M). The Firefox developers have admitted that performance
on the SX port isn't as fast as the other platforms (something they hope
to fix by the 1.1 release).

Is some of that due to the compiler?

Thu, 14 Jul 2005 @ 09:36 -0400, Kurt Huhn said:

Jul 14, 2005, at 4:28 AM, Jonathan Groll wrote:
Seriously not intended as a troll. Granted, many, many factors come
into
play when doing comparisons, only one of which is clock speed. However,
many of us, and for too long now have blithely accepted the apple
gospel
that PPC is the inherently "superior" architecture. Surely, though,
there must come a time when a processor with enough pure 'grunt' can
outperform one that is slower but more able?

MHz != speed.

I don't believe that's what he is saying.

He's saying that the clock speeds of Intel/AMD CPUs are being boosted so
much that they are making up for architectural issues.

AMD is achieving more performance at lower clock speeds, much like other
CPUs. From what I read in the K7 papers, the AMD CPUs are *NT* Intel
architecture CPUs. They are a different beast under the covers, and the
Intel ISA and other bits are more like a layer of emulation.

It seems AMD could design and ship a competitive non-Intel CPU if they
were not tied to the Intel-compatible market.

Clock-for-clock, I don't see the G5 spanking the AMD like it does the
Intel. It seems to me that the AMD64 especially, despite its Intel ISA
ball-and-chain, is quite competitive.

I have a theory about why Apple is going with Intel instead of AMD, but
I'm going to post that in another thread.

improvments may bring the Intel processors closer
to being acceptable (faster pipelining, faster RAM, switched bus, etc)
but "speed" as measured in MHz is not one of them. I will grant that
AMD and Intel are bringing their Desktop processors closer to
acceptable, but they will never get there unless they're willing to
dump the legacy crap that surrounds them and start designing them to be
put into elegant and well-designed systems.

I think AMD is doing that. They are hampered by the need to be
compatible with Intel.

If they were free of that, I think they'd really shine.

I see nothing there indicating that the developer macs are
outperforming G5s. I see a vague reference to "as little as 10
seconds" to boot to the desktop. Now there's a reliable benchmark!!
Not.

Hey, Windows users put quite a lot of value into quick reboots :)

If we're going to claim that the intel systems from Apple are faster
than anything else, why don't we compare them side by side, under equal
environments and conditions, and take real measurements, with real
applications, under heavy load, and performing functions that people
who buy high-end workstations need them for? Until I see such tests,
I'll not even consider "it feels faster" with the smallest grain of
salt.

course, to be completely fair you need to test headless, because the
G5 pays a heavy price for the GUI.

Things like image manipulators, sound encoders, and stuff like that
would make useful test of code that people actually use.

Maybe put Darwin on PPC up against an AMD64 running something like *BSD
or Linux? As long as you avoid system calls, I think that would be
pretty good. No GUI, just some batch jobs with compute intensive tasks
to avoid comparing Mach with a monolithic UNIX.

Thu, Jul 14, 2005 at 11:24:55PM -0400, Patrick Giagnocavo 717-201-3366 wrote:
Thu, 2005-07-14 at 23:08, Nate wrote:
Thu, 14 Jul 2005, Charles Shannon Hendrix wrote:

Thu, 14 Jul 2005 @ 09:36 -0400, Kurt Huhn said:
>I see nothing there indicating that the developer macs are
>outperforming G5s. I see a vague reference to "as little as 10
>seconds" to boot to the desktop. Now there's a reliable benchmark!!
>Not.
>
Hey, Windows users put quite a lot of value into quick reboots :)

" neat thing about Suns is that they really boot fast. You ought to see
one boot, if you haven't already. It's inspiring to those of us whose
LispMs take all morning to boot."

Hey, mine boot into the F real fast :-) did they mean
something other than Forth?

Well, a Sun might take 10 minutes, but how long does a Symbolics, LMI,
or TI take to boot? 45?

Fri, Jul 15, 2005 at 11:07:28AM -0400, Dave Fischer wrote:
jdboyd (AT) jdboyd (DOT) net writes:

>Well, a Sun might take 10 minutes, but how long does a Symbolics, LMI,
>or TI take to boot? 45?

And you were supposed to reboot the Symbolics machines inbetween
logins.

("Login" being conceptually different from most systems - you had
full access to a lisp prompt before logging in - you "logged in"
to get access to your files. If I understand the docs I've read
correctly)

Yikes. That kinda sucks.

Jul 15, 2005, at 10:31 AM, Joshua Boyd wrote:

Thu, Jul 14, 2005 at 11:24:55PM -0400, Patrick Giagnocavo
717-201-3366 wrote:
>>
>Hey, mine boot into the F real fast :-) did they mean
>something other than Forth?
>
Well, a Sun might take 10 minutes, but how long does a Symbolics, LMI,
or TI take to boot? 45?

<horse type=dead action=beating>
Whuh? Which Sun? My Sparcbook boots to a login prompt in about 30
seconds. The Ultra 60 in far less.
</horse>

Fri, Jul 15, 2005 at 11:55:44AM -0400, Kurt Huhn wrote:
Jul 15, 2005, at 10:31 AM, Joshua Boyd wrote:

Thu, Jul 14, 2005 at 11:24:55PM -0400, Patrick Giagnocavo
717-201-3366 wrote:
>>
>Hey, mine boot into the F real fast :-) did they mean
>something other than Forth?
>
Well, a Sun might take 10 minutes, but how long does a Symbolics, LMI,
or TI take to boot? 45?

<horse type=dead action=beating>
Whuh? Which Sun? My Sparcbook boots to a login prompt in about 30
seconds. The Ultra 60 in far less.
</horse>

We are talking about the Sun3 through early Sparc days, and machines
that are very likely being netbooted.

That said, 10 minutes may not be acurate, but then again since in the
quote they were being rebooted because of crashes, you need to account
for fsck time on a non logged FS-check.

Fri, Jul 15, 2005 at 12:12:27PM -0400, Joshua Boyd wrote:
Fri, Jul 15, 2005 at 11:55:44AM -0400, Kurt Huhn wrote:
>Jul 15, 2005, at 10:31 AM, Joshua Boyd wrote:
>>
Thu, Jul 14, 2005 at 11:24:55PM -0400, Patrick Giagnocavo
717-201-3366 wrote:

Hey, mine boot into the F real fast :-) did they mean
something other than Forth?

Well, a Sun might take 10 minutes, but how long does a Symbolics, LMI,
or TI take to boot? 45?
>>
><horse type=dead action=beating>
>Whuh? Which Sun? My Sparcbook boots to a login prompt in about 30
>seconds. The Ultra 60 in far less.
></horse>
>
We are talking about the Sun3 through early Sparc days, and machines
that are very likely being netbooted.

That said, 10 minutes may not be acurate, but then again since in the
quote they were being rebooted because of crashes, you need to account
for fsck time on a non logged FS-check.

Netbooted machines shouldn't need an fsck. Now if the box the diskless
workstations were booting off of crashed, then yeah, everyone suffers.

As far as a Sun taking a long time to boot, try booting some of the
early Ultra Enterprise class boxes with diag-switch? set to true and
diag-level set to max. =)

Might be 5 minutes before you even see something on the console.

But the UNIX Haters' Handbook was refering to UNIX Workstations, so they
were probably refering to Sun3 or early Sparcs, which compared to the
LISP machines they usually replaced, they did boot pretty quickly.

Fri, Jul 15, 2005 at 11:37:16AM -0500, Mike Parson wrote:

Netbooted machines shouldn't need an fsck. Now if the box the diskless
workstations were booting off of crashed, then yeah, everyone suffers.

True.

As far as a Sun taking a long time to boot, try booting some of the
early Ultra Enterprise class boxes with diag-switch? set to true and
diag-level set to max. =)

Might be 5 minutes before you even see something on the console.

I've heard reports that the fastest PWER machines manage to boot more
slowly.

But the UNIX Haters' Handbook was refering to UNIX Workstations, so they
were probably refering to Sun3 or early Sparcs, which compared to the
LISP machines they usually replaced, they did boot pretty quickly.

Now to just get one of those LISP machines in running condition at a
price I can afford.

Fri, Jul 15, 2005 at 02:23:58PM -0400, Dave Fischer wrote:

>Now to just get one of those LISP machines in running condition at a
>price I can afford.

Those SMD drives start to get flakey after a few decades

Well, some of those machines will take SCSI, so there might be hope for
one of those.