| |
Calculating computer performance is a tricky issue, and there are a variety of methods that attempt to do this. Most
of these focus on very narrow tests, like floating-point calculations, etc. A major limitation of many of these tests is
that they are actually more dependent on the software used in the test than the hardware of the computer itself. To
date, there is NO industry standard measurement method that gives a real world indication for an average user as to which
computer is faster or more powerful than an other.
[Here are some reasoned observation about benchmarking, and a
discussion at USA Today.]
However, since we need to get some reasonable indication, one accepted way to measure computer
processing power and performance is by looking at MTOPS (Millions of Theoretical Operations Per Second).
You may not have heard, but when Apple’s G4 machines were first introduced it almost caused a major international incident.
The root of the problem was that the US government has severe restrictions about exporting Super Computers to certain other countries.
The government says (not Apple Computer) that what determines whether a computer is categorized as a “Super Computer”
is its MTOPS performance.
Apple’s G4 was the first desktop computer to break the Super Computer barrier. (By the way,
the government’s solution to the export issue was to rewrite the specs to raise the limit.)
If you check out Apple’s
specs and Intel’s
specs, and then AMD’s
specs you will see that there is a surprisingly BIG difference in MTOPS performance.
Pretty obvious from this perspective which is more powerful.
| Microprocessor |
MTOPS |
| Intel 3.66 GHz Xeon (or Pentium 4) |
14,994 |
| AMD dual 250 Opteron |
16,400 |
| AMD quad 850 Opteron |
32,000 |
Apple 1.25 GHz G4 Power PC
(e.g. eMac) |
17,917 |
Apple 1.67 GHz G4 Power PC
(e.g. PowerBook) |
23,941 |
| Apple dual 1.42 GHz G4 Power PC |
38,340 |
Apple 1.8 GHz G5 Power PC
(e.g. iMac) |
21,753 |
| Apple dual 2.7 GHz G5 Power PC |
60,750 |
---------------------------
Here is a relatively new standard you might not have heard about. NewsFactor reports that “Apple
is getting some help from the Embedded
Microprocessor Benchmark Consortium (EEMBC), which in 2002 published statistics
for the MPC7455 — Motorola’s latest incarnation of the G4 — that
is used in the dual 1 GHz Power Mac. The EEMBC tested 46 different kernels and found
Motorola’s G4 to be faster than all other contenders across all five specific
target markets.”
Also included in the information Apple provided to NewsFactor were details of a test involving BLAST (Basic
Local Alignment Search Tool), an open source biotechnology application used to find similarities in DNA and protein sequences.
Apple compared the performance of its dual 1 GHz Power Mac G4 running A/G BLAST to that of a Linux workstation with a
2 GHz Pentium 4 processor running NCBI BLAST. Both computers given the task of seeking similarities between human and
mouse chromosomes.
Depending on the type of search performed the Power Mac delivered anywhere from 3 to 50 times the performance
of the Linux workstation.
Don’t be fooled by the “speed” (i.e. MHz) numbers posted by Intel and the PC assemblers! Many
tests show that Macs have superior performance to Pentiums with three
times the MHz rating.
Why might this be? Consider this example:
two cars are traveling from Los Angeles to San Francisco.
Car A will average 65 MPH, while car B will average 50 MPH.
Assume they start together and neither car makes any stops.
Which one will get there first?
If you subscribe to the misconception that speed is the most important specification, then you’d pick car A — and you would be wrong. The variable we did not discuss is the route each car would take. In our example, Car B is going the most direct route, while car A is going through San Diego (i.e. quite indirect).
I know, you ASSUMED that both cars would take the most straightforward route. Many people have undoubtedly assumed the same thing when comparing Pentiums/Athlons to PowerPC microprocessors — and that assumption would be incorrect. The fact is that for several reasons (e.g. that they were designed from day one to operate a Graphical User Interface) that PowerPC microprocessors are more efficiently designed.
In 2001 MacCentral published a VERY detailed write-up about the Megahertz
Myth that is still well worth reading. (Make sure to go back to the Part
I and Part II links.) Here are some additional articles (from other sources) on the
same subject: Washington Post (9/02): Processor
Speeds Almost Passe, and TechNews: Does
MHz Really Matter Anymore?, and iGeek: MHz
and GHz, and MHz
and GHZ Have Lost Some Meaning, and Mr.
Software Fights the MHz Myth.
Another person wrote “RPM is a better automotive analogy for MHz: a Viper at 2000 RPM smokes a
Neon at 5000 RPM”... Apple also has posted an eight minute QuickTime movie demonstrating how a 1.7 GHz Pentium 4 is considerably slower than several lower MHz G4 versions. Check it out!
Still another writer proposed: “If Intel wants to be the de-facto standard then let them. We’ll call
that standard ‘IntelHertz’. On each new box leaving Apple, there will be a sticker next to the standard
MHz telling the consumer that this 1000 MHz Mac is equal to a 3000 MHz Pentium or 3.0 ‘IntelHertz’.
In every advertisement and promotional material Apple produces it will be there, plain as day, it’s chips equal
to ‘IntelHertz’. This new standard will bring us on par with Intel and show the world how the Mac is
more than equal to the PC offerings.”
This article explains
AMD’s new plan in its battle against the PC megahertz myth. (AMD makes
the Athlon microprocessor chips that are the prime PC competitors to Intel’s
Pentium microprocessors.) It seems that the new AMD Athlons will be specified by model rather than by GHz.
Model A1600, for instance, is a 1.4GHz Athlon, which AMD views as equivalent in performance to an Intel P4 1.6GHz. AMD now demands that no motherboard/BIOS maker ever reveal the actual clockspeed of the chip, and even goes as far as to forbid the printing of the CPU’s clockspeed in the motherboard’s reference manual.
Hmmm. Seems like some PC people are now also catching on to the Megahertz myth...
|
In mid-2002, because Gateway ads were touting its Profile 4 as being much faster than an
iMac (i.e. the low end of Apple’s computer line), Gannett News Service tested a loaded Gateway Profile 4 XL, and compared it to a high end iMac, computers that were similarly priced. Their conclusion: “The comparison chart says it all. While the Gateway toasted the iMac in 3-D game frame rates at a resolution of 640x480 using Quake III Arena, the scores were much closer when the crisper, more realistic resolution of 1,024x768 was used. The Gateway registered 66.2 frames per second, while the iMac scored 62.5 frames per second, an insignificant difference for casual gamers.”
They also “applied test routines employing nine common Adobe Photoshop functions. Here the huge
megahertz gap was mostly surmounted. The iMac’s 800-megahertz G4 sprinted through the tests in 49.6 seconds.
The Gateway’s 2.8-gigahertz Pentium 4 completed the same tasks in 44.7 seconds.”
Obviously these performances in no way correlate to a 2800 MHz vs 800 MHz difference. It should also be clear to see that
a more powerful Mac (e.g. a dual 1250-megahertz G4) would beat the 2.8-gigahertz Pentium 4.
---------------------------
Popular Mechanics has a well-written June
2002 article
that says for a computer salesperson (or anyone else) who tells a typical user (i.e. one who does “basic word processing,
e-mail, Web browsing, maybe some digital camera stuff”) that a higher MHz Pentium will be any more suitable than an
iMac is a “classic case of speed bigotry in which computers are judged not by the contents of their systems, but by the
speed of their CPUs. This is an incredibly limited view on how to choose the right computer system. The
dirty ‘secret’ of the computer industry is this: Chip speed doesn’t matter much anymore.”
They go on to say that there are several items that are more important to the performance than the microprocessor’s
MHz rating. “RAM is much more important than chip speed for almost everything your computer does...
The speed of your hard drive is the next choke point on system performance, and it’s a critical component for digital video and
audio because writing to or from the disk always takes time.... Like RAM, video cards are vital... Finally, the speed at which you connect
to a network can affect how fast your computer feels.” They conclude by saying that “Despite all this, chip speed
remains central to computer manufacturers’ marketing plans. That’s because chip speed is the easiest way to
catch your attention: misinformed consumers figure the faster, the better.”
All true. In addition Popular Mechanics could also have mentioned that the speed of your connection to peripherals
has a big influence on your perception of your computer’s performance. In an attempt to compete with Apple’s wildly
successful Firewire (IEEE-1394) Intel has been promoting USB 2.0 as a higher speed alternative. As usual, when you look below the
surface, this is just not so.
In the July 2002 article entitled
The Numbers Game
the author explains “The crucial difference between USB 2.0 and FireWire is this: FireWire is faster and
more reliable. The reason is subtle, and not mentioned at all by the proponents of USB 2.0. Moving data around using USB
demands the involvement of a processor. FireWire doesn’t. If you’re trying to download a digital video from your
camcorder on to your hard disk to edit it, then with USB 2.0 each chunk of data has to pass to your processor and then to the hard disk.
With FireWire, the hard disk and the camcorder talk directly to each other.”
The bottom line is that Macs are very competitive in any performance test, and typically perform equal to or better than
PCs advertising twice the MHz rating.
---------------------------
With all the advertising we see, it’s not surprising that the typical consumer might conclude
that the Pentium 4 is a superior microprocessor. Don’t be taken in by Madison
Avenue. Here’s some technical observations you may not have heard before:
A PC
company’s CEO (in an extraordinarily detailed analysis)
concludes that “the new Intel architecture has serious fatal flaws that in some cases
can throttle the speed of a 1.5 GHz Pentium 4 chip down to the equivalent speed
of a mere 200 MHz Pentium MMX chip of 4 years ago, even slower than the
level of any Celeron, Pentium II, or Pentium III chip ever released! It’s
a huge setback for Intel...”
A Techweb analysis says: “For
today’s buyer, the Pentium 4 simply doesn’t make sense. It’s
slower than the competition (including a Pentium III) in just about every area;
it’s more expensive; it’s using an interface that won’t be the flagship interface in six
to nine months; and it requires a considerable investment outside of the price
of the CPU itself...”.
An August 2002 PCWorld
article reports that there is a class action against Intel (and selected
PC assemblers) about them over-stating performance specifications. The plaintiffs
claim the companies deceived the public when marketing Intel’s flagship
processor and allege that it is a "material fact that there is no benefit to
consumers in choosing the Pentium 4 over the Pentium III, as the Pentium
4 is less powerful and slower than the Pentium III...".
There is considerable excitement about Apple’s latest release: the G5
computer, which uses a 64-bit microprocessor from IBM: the 970. [This January
2004 article
gives a bit of history here.]
Here is a sample release by
IBM about the 970. This gives
more details how IBM’s 970 compares to it’s larger Power4 chip
sibling. More good news is that IBM
is well along on the next generation, the Power5.
Here is a late 2003 writeup on
the Power5. IBM is saying that
the Power5+ will be shipping in 2005 and Power6
in 2006. (Here is their official roadmap.)
All of these advances will be reflected in upgrades to the 970 version.
[Where does all this leave Motorola, you ask. Good question. On the one hand
there are articles like this. On
the other hand there is this.
My guess is that due to Motorola’s slowness in updating the G4 plus their continued
business problems companywide, that the latter may be more likely.]
Not familiar with the ins-and-outs
for the 64 bit architecture — then read this PCWorld piece Are
You Ready For A 64-Bit PC?.
This September 2002 eWeek story goes
into some detail about the IBM 970 chip, and the connection with Apple. iGeek gives
even more info on the IBM chip, as does Real
World Tech. This analysis of Apple’s G5 says "Suddenly The Leader", and,
in a similar vein, this January 2004 article calls
the G5 "The Hummer of the Computer World." This
story call it the "Ferrari of Computers." Take your pick.
How fast is the G5? Naturally there are some conflicting reports. Apple says
that it is the world’s fastest PC, and backs this up with some impressive test results.
Here is an independent snapshot (as of February 2005). As you might expect, some PC ubergeeks are distraught with this situation, and have
been looking hard for stones to throw. (For example, here is the perspective from PCWorld.)
In "The
Smell Of Fear" the author makes some good points. Note that usually
only in the small print do you see mentioned the fact that the applications
often used on a comparison test have not been optimized for a G5, while they
typically have been for a Pentium 4. Read
"Benchmark
Silliness" for another discussion. Maybe you’d like to see what NASA
says.
Here is a lengthy, technically detailed, and reasoned discussion of
the situation by InfoWorld, published on December 31, 2003. This independent December
2003 analysis states
that "the G5 is notably faster than high-end
PCs." And this January 2004 independent testing
concluded that "Apple G5
Smokes Intel Competition." This
series of tests shows how the G5 compares favorably to the AMD 64 bit Opteron. Sounds like a consensus to me.
[An amazing, lesser known feature of the latest G5 is that it uses only 25% of
the power that Intel’s state-of-the-art Prescott processor uses!]
This January 2004 Forbes story says
“When noted biologist David Botstein was lured from
Stanford University to head the Lewis-Sigler Institute for Integrative Genomics at Princeton, he had his
choice of computing systems. But Botstein says he outfitted the entire center
with Apple computers and servers, which are used for everything from desktop
applications to comparing lengths of genetic code.” His explanation for
choosing Macs was that he wants maximum power for minimum cost.
If that is not enough to convince you, then you won’t know what to make of these
next stories. "November
20, 2003--Engineered Intelligence (EI) introduced its new Parallel
Programming Software offering at the Supercomputing 2003 conference. EI’s
CxC(R) parallel programming software allows scientists and engineers to easily
create programs for Apple’s Power Mac G5, the world’s fastest personal computer."
(Related to this, there is still another outstanding Apple leadership effort
in XGrid,
which some are calling "The
Future of Computing". There is also exciting work being done in a similar way
with Pooch.)
Or the undisputed FACT that Virginia Tech has stunned the supercomputer world
building the third fastest computer in the world, within a
few months, and at about 10% the cost of a typical
supercomputer — using only stock Apple G5s. Here is a super inside summary which
includes photos, a slide show, etc.. With literally hundreds of reports about
this amazing accomplishment, it is difficult to select something representative,
but here is one story,
and another.
Here is what their Dean of Engineering wrote in January 2004. Apologies
to the rest.
Less well known is that Dell is trying to compete in this area, and supported a similar effort at the University of
Texas. This report is
about their $38M Dell/Linux cluster that will achieve 3.7 Tflops. Compare that to
Virginia Tech’s $5.2M Apple/Mac OS X cluster that achieves 17.6 Tflops. Cost per Tflop: Dell = $10.3 million,
Apple = $295,000. Conclusion: Apple is the fastest and
lowest cost.
|
This is an excellent collection of
articles concerning microprocessors, entitled “Let The Chips Fall Where
They May! — Mac Processors & Wintel
Processors: A Resource”
---------------------------
In another progressive effort to improve performance, Apple is one of the founding
members of the HyperTransport Consortium
Consortium of companies backing this new standard.
Apple will use HyperTransport
as a high-speed link between the two processors that make up the chipset in new desktop Macintosh systems, sources
said. A chipset
is a group of chips that manages the internal functions of a computer. (The
companies include AMD, Cisco, Sun Microsystems — but not Intel.)
---------------------------
In a related area, there is periodic speculation about whether Apple will
port Mac OS X to run on Intel or AMD type microprocessors.
For years this has been a favorite rumor, as there is just enough plausibility
to make it interesting. Here is a good discussion of
the topic, and another..
There are a variety of reasons why this change would be 1) difficult,
and 2) undesirable. Here is a sample from
that side of the fence, plus a follow-up.
To my knowledge, these are Apple’s latest comments
(November 2003). Stay tuned.
One contributor accurately pointed out that “It makes absolutely no difference if your machine is 'faster' if
you are blankly looking at the screen, trying to figure out what to do next. For most users in most tasks, a consistent interface, ease of use, and easy
to learn software are much more important than raw horsepower.
“The minute differences in time taken to do average tasks between 'fast' computers and not-so-fast computers are not important for most people in
most applications. Who cares if a task is performed in the blink of an eye or three quarters
of the blink of an eye? I am writing this on a 250 MHz machine. Can I tell that it is not a 450 MHz machine? No.”
Another author observed “You can have the fastest computer on the planet, but if you type 50 words a minute,
it’ll be the same 50 words a minute regardless of the potential of your PC.” Here is a very informative explanation of
how OSX has been optimized for speed.
A third person wrote in: “Were there enough time, I could spend an entire month with all the anecdotes
of suffering I experienced as a PC user. I wasted a good part of 13 years, using 22 PC operating systems — all
in the name of trying to find a computer that I hoped would do the same things my Mac now does effortlessly.
“Regardless of issues of personal preferences and priorities, my bottom line for getting a Macintosh
was that I was losing time by not being productive. This is not a matter of a few hours here and there — it
is much more a matter of many months, if not years trying to accomplish with
Windows what the Mac has made a matter of due course.
“There is no comparison study or benchmark that adequately
equates the amount of lost time, frustration and utter despair I experienced
as a PC user compared to the joy, satisfaction and relaxed peace of mind I’ve
been able to receive as an owner of a brand new Mac. At
every opportunity I point out how Macs save time,
money, and effort in every task a person would
seek to undertake using a computer. Those
numbers tell a story that leaves a lasting impression on everyone I’ve
spoken to. In short, there are a lot of people who have switched to Macs and a lot more
who want to when the time is right for their particular financial situation.
“I never would have considered myself someone to evangelize, but I now
do so at each and every chance I get, with all the vigor and passion I can muster. Macs
are that good — and better. I have now gotten my life back
and I can sleep without any of the anxieties I used to face with Windows
staring back at me.”
Still another said “The measure of 'power' is
not absolute speed, but rather the net combination of the positive element
of productivity plus the negative element of lost/down time.
“Productivity is enhanced by a consistent and powerful interface design, multiple
applications being readily available, doing more than one thing at a time
in a true multitasking environment, and the ability to integrate a diverse
tool set through scripting, programming, etc.
“Negative elements include reboots due to system crashes, lost data due to
crashes, time spent wasted trying to figure out what settings will allow
me do accomplish a task (printer selection, network interfaces), and how
I go about controlling those settings, time lost due to viruses, downtime
for maintenance and reconfigurations, etc.
“Apple’s Macintosh computers do a much better job than Wintel
PC’s in terms of personal productivity. With OS X, I can have dozens of applications
simultaneously open, the system never crashes or hangs, and things 'just
work'. An these are just some of the free productivity
tools I have access to: Apple script, the Cocoa Development Environment (C,C++,
Java, Objective-C, or Applescript Studio), Perl, and a Unix Shell.”
I couldn’t have phrased it better! My observation, putting it again into
the car analogy, would be to compare the Pentium/Athlon to a Corvette, while
the Mac is a Lexus. The Corvette may have faster 0 to 60 acceleration or higher
top end speed, but what does this really mean to the everyday driver? Furthermore,
the trade-offs for owning the Corvette are many: higher gas mileage, much
harsher ride, less trunk space, more frequent repairs, etc., etc. So it is in the computer world.
|
Although we have talked a lot about hardware as it relates to speed, the operating
system itself is, of course, a major contributor to performance. You should know
that, in addition to offering more features, Apple’s OS revisions are also aimed
at improved performance (e.g. by making more things 'native'). Here are some 2004 tests
results
that show their progress.
Please read this 2002 Wired
article. The whole thing.
Particularly interesting are the remarks by physicists at UCLA, who explain
that they have been grouping Macs together (clustering) to get an even more
powerful machine to perform complex scientific calculations. They use Macs
because “Not only was the
performance faster than the Pentiums but it was comparable to the performance
achieved on some Crays.”
This is a simply astounding observation! A Cray Super Computer (costing in
the millions) is a true no-holds-barred super computer...
The article continues by saying “Most clusters are
based on Pentium machines that run Linux. But according to these UCLA Physicists,
Linux clusters require a PhD to set up and to run. By contrast, Mac clusters are
so easy to make, even teenagers can do it.
“There’s a book called How to Build a Better Beowulf that’s
230 pages long and tells you how to set up clusters with Linux. We have a one-page manual that
shows you how to do it on PowerMacs. We’ve had
high school students do it. We’ve had junior high school students do
it. We even had a sixth grader in Hawaii do it.
“It took NASA’s Jet Propulsion Laboratory two weeks to put together
a 16-node Linux cluster. We could do the same thing with Macs in less than an
hour.”
They go on to say that “Linux clusters are also extremely fragile: If all the machines in the cluster aren’t
running the same version of the kernel, everything grinds to a halt. By contrast,
a Macintosh cluster can be made from a mix of
G3 and G4 Macs running Mac OS 9 or X.”
This August 2002 article in
the Boston Globe makes the same point when it interviews several different
engineers. For example: “...Craig Hunter, an aerospace engineer at NASA’s
Langley Research Center in Virginia, is replacing Silicon Graphics workstations with Macintoshes. Hunter uses computational fluid
dynamics software to simulate airflow over the wings of an airplane. It’s extremely complex work that he sometimes
runs on a Cray supercomputer with 64 processor chips. But Hunter says the PowerPC G4 processors used by the Mac are as
powerful as those in the Cray. ‘If I put together 64 G4s, we’ll get the same performance as the Cray for a lot
less money’.”
If you want to see how easy it is to set up your own Mac cluster, go
here.
---------------------------
Here is somebody else that picked up on Mac performance. Terra Soft co-founder
and CEO Kai Staats told NewsFactor (March
2002) “in real-world applications, PowerPC processors are better
equipped for Linux than their Intel counterparts. When considering
the price/performance/power consumption/footprint ratio, PowerPC wins nearly
every time.”
---------------------------
In another technology area, Jeff posted this on Java.Net in
November of 2003: "I run JBuilder and CVS on both my Mac laptop and my
Wintel box, and use CVS to move code between them quickly and easily... The
Mac VM in unique among Java VMs currently
in that it shares static data between versions of the VM. Thus IME once you
have a first VM up, launching subsequent ones is like greased lightning..
In short I think Apple’s got a lot of support for their claim that they are
the 'best Java desktop environment around.'"
---------------------------
Awhile back I received an email from the Department of Physics and Astronomy at Oak Ridge,
Tennessee. It said, “We are a group of
computational astrophysicists here at Oak Ridge National Laboratory and UTK. We
all use three or four of the top 20 most powerful computers on the planet almost
daily. We are all, to lesser and much greater degrees, computer experts. We
are computational, not computer scientists. By that I mean we USE computers
to get complicated tasks done and hard problems solved.
And... we are completely a Mac shop — including laptops and home
machines, throughout.
”
If Macs perform well enough to satisfy the demanding users of some of the
most powerful heavyweight computers in the universe, they should do just fine for YOU...
--------------------------------------
Conclusion: If you want the most powerful, productive, and easiest to use computer available:
buy a Mac.
Download a printable pdf version of this
document (rev: 3/06/04).
If you have any constructive comments or suggestions about this page,
please email
John.
This section’s Haiku
(see the bottom of our Intro page for more explanation):
Aborted effort:
Close all that you have worked on.
You ask way too much.
rev: August 6, 2007
— Section #9 —
|
