IT is often looked at with a frown because of what it tends to be in most organizations: a very expensive division whose responsibility is to provide critical tools to perform business, with the ultimate goal being to produce a result, typically profit.
Additionally, failures in IT tend to be very very visible. IT is where people get overly excited often, especially after recieving new marketing materials from the "vendor", and expensive projects get created to only fizzle out after they realize there was no need in the first place.
For example, why spend garganuan amounts of money for integrated voice-mail and e-mail, etc. etc. on a single-point-of-failure "cluster", when $30 answering machines are probably more reliable and fit the need like a glove, and are disposable.
In many cases, IT projects end up being cost shifting rather than cost savings.
Or, they turn into classic and memorable "White Elephants", where the cost increases are so embarassing that management simply stops talking about it.
Often, some processes simply cannot be efficiently automated. For example, sometimes an entrenched and very manual government office gets in the way (automating on either side of them seems to be accepted practice, but I may be wrong).
There is appears to be a fundamental flaw in accounting methods when they totally miss productivity loss due to having crap like Windows 9x or NT and PCs with double-digit failure rates.
It is not uncommon to see someone with 32MB (or even 128MB) of RAM in their ancient PC, swapping all to hell, wasting hours upon hours of time, when "there is no budget" for a new PC. It really is sick.
Not only that, after five years, stuff wears out. Replacements? "There is no budget for replacements."
I wonder how close you could get to a working linux machine with nothing offically from GNU?
I should have written this in my other reply:
If you look at OpenBSD, for example, that will be about as close as you can get, where the remaining GNU-branded things are not optional without tremendous effort.
For example, OpenBSD relys on GCC, where doing a BSD-only compiler simply to get rid of "GNU" would not be worth it.
This needs to be qualified with how you use Linux.
If you are a average desktop user (Red Hat, GNOME, Mozilla, OpenOffice.org, etc.), then, in fact, Linux, itself, is really not visible in your day-to-day work. If Linux were spontaneously removed, and the NetBSD kernel or HURD took its place (with necessary libraries, etc.), odds are you would never notice.
If you use Linux directly for creating custom kernels for various applications (e.g., you are a kernel developer for the Sharp Zaurus), then, of course, Linux disappearing would have a very large impact on your life. Starting from scratch and reworking the FreeBSD kernel for the Zaurus would definitely be a setback.
From an end-user's perspective, GNOME on Solaris 9, GNOME on Red Hat 9, GNOME on OpenBSD 3.3, and GNOME on Mac OS X (for those who do it because they can) don't present much of a change among them. The fact that each OS has adopted different conventions for package management and/etc directory structure, for example, is irrelevant, because these things are ultimately independent of the kernel (except, perhaps, the kernel parameter tuning files under/etc and device names under/dev).
Re:Creation of a blue collar computing segment
on
More Cheap Linux PCs
·
· Score: 1
While I dislike the possibility of computer expertise segmenting along economic lines (for social reasons), I do see some benefits: clearer cut job descriptions and areas of expertise...
I think this is inevitable as the mist from the bursted bubble settles. In the last ten years, we experienced a rediculous phenonmenon, where computers progressed from 30MHz CPUs to 3000MHz CPUs, the WWW grew like weeds, and we are now fully bombarded from all sides with techno-buzz gadgets. As society adjusts and people find their niches, I wouldn't be at all suprised if the IT industry begins to look like other older and more established industries.
Perhaps I am naive, but it seems that technology companies have a pretty flat social hierarchy when compared to other industries like construction and health care. For example, there are no physicians (M.D.) or Professional Engineers (P.E.) in the IT industry (i.e., we're really all a bunch of people who essentially made our own job titles and hope the people with money believe what we say--sort of like physicians and engineers a few hundred years ago).
Re:Linare is quite full of themselves .....
on
More Cheap Linux PCs
·
· Score: 1
"(company name) is the worldâ(TM)s premier technology system integrator for (technology name) in the enterprise"
I think this phrase is hard-coded into business and marketing majors' brains before graduation.
The art of saying nothing in so many words seems to be the next growth industry. Maybe I should get a lobotomy and jump on board...
I recall that Lindows PCs from Walmart had a huge shipping markup (like $100).
I think this one was debunked as poor--and irresponsible--reporting. A follow-up post to that Lindows review a while ago said the wal-mart shipping is more like $25 or so.
I know from experience that shipping a Sun workstation (no monitor) via the least-expensive UPS option is under $20. Add handling, and 25 to 30 dollars isn't impossible.
Trust me, a lot of times when you donate to a charity, all you are doing is paying for some aging hippie to afford his apartment on the Upper West Side.
Or, paying the professional telemarketing or junk-mailing firm doing the charity's dirty work.
Most charities suck. Many are crooked. That's why I give only to charities I seek out and am already familiar with. For example, it is pretty clear that OpenBSD CD-ROM purchases go to Theo, that Calgary book shop, some of the other developers, etc (i.e., I already know what I'm getting into).
1) Perhaps people see in Java some of the reasons they would rather choose Lisp but can't, because Lisp is still fragmented and not widely understood.
2) From a high-level design point of view, it's better than C but less complex than C++ (although interfaces and the lack of multiple inheritence still nag at me a bit).
3) It is possible to isolate platform dependencies to configuration files and distribute the same compiled binaries to all platforms (think: compile once, run anyhere--if programmed properly, that is). This one is actually pretty darn good reason.
4) The immensity of the Java APIs is overwhelming, but the capabilities in there are impressive. A Java programmer that finds themselves writing utility classes for common operations probably missed something.
5) Although the "null pointer exception" took over where "segmentation fault" left off, there really are classes of bugs that don't occur in Java, such as buffer overflows and random memory corruption. In C, only a good debugger, such as Sun's dbx, can catch some bugs (dbx's run-time checking is very very useful at times).
6) Java applications are not necessarily more reliable than C programs, but it seems that Java programs are more easily debugged (however, it seems J2EE apps might be harder due to the edit-compile-deploy-test cycle).
7) Java is stable, from the point of view that the language and APIs don't often change in incompatible ways. Java inherited Sun's conservatism. Compare this to Microsoft's history of API replacements and extensions (and to the volatility present in some OSS projects, such as GNOME).
8) The Java runtime is, itself, pretty reliable. I don't recall seeing the JVM crash or running into bugs in the APIs. I'm sure they are there, but it appears they are not common.
8) While it is true that C is much more widely ported than J2SE, kludges like autoconf scripts aren't needed for Java applications, which reflects well on Java's architecture. I swear I waste more time fighting with autoconf, libtool, etc. than any other aspect of managing applications on my systems. I don't even know what purpose libtool serves other than to break builds on Solaris. Libtool sucks. (okay that was a rant)
9) Java isn't Microsoft. I know this is flame-bait, but the sad thing is that this reason is as big is any other I listed above. Who knows where C#/.NET is going, but, if Microsoft goes down, C#/.NET is going with them. Sun Microsystems is smart enough to realize companies don't like to put all their eggs in one basket. Thus, the JCP, J2EE licensing and certification, etc. Other than J2SE, Sun is actually a small player in the Java market (they are dwarfed by BEA, for example, in the J2EE market). Sun also sets other examples that reinforce that they aren't complete ass holes (OpenOffice.org, for one; Liberty Alliance, for another)
In general, I think there are a lot of thing that Java does really well, but most people completely take them for granted. It is easy to forget how frustrating memory bugs in C are or how damn complex C++ is, for example.
... developers who are migrating now from Visual Basic are moving to Java and C# in roughly equal numbers."
That isn't so bad. It means that there are a non-trivial number of developers leaving Microsoft and not coming back. Also, the ones moving to C# probably didn't think much about the alternatives--they were captured by default.
A few years ago I was contemplating a new ISP, so I called several in the DC area (Erol's, ATT, etc.) and asked the "order takers" if their systems supported UNIX and/or Linux.
It seems the UNIX-friendliness of an ISP is made clear when their installation instructions take you the Windows-Dial-Up-Adapter route. Modern PPP programs (at least under OpenBSD) understand Microsoft's proprietary "extensions", so basically anything that can be setup under the Windows dial-up configuration can also be set up under UNIX.
Once the PPP connection is made, finding a DNS server isn't too hard, and there are certainly UNIX-friendly e-mail providers out there (POP/IMAP, etc).
As far as broad-band goes, doesn't the ISP stop at that little router-gadget? I'm not too familiar with broad-band, unfortunately.
...I think the best CAD packages are currently for PC.
Not necessarily. The cheapest ones are certainly for the PC but definitely not the best.
One thing that is important is that PCs configured to produce similar display quality relative to, for example, higher-end Sun graphics begin to approach the cost of Sun workstations ( $1000+ video cards, etc.). One thing in particular is the quality of the anti-aliased display in a CAD system, which many low to mid-range PC cards simply don't do well.
If I have to spend hours a day staring at a CAD model, I would want to, first, do it using the best graphics and monitor combinations available, and, second, use the fastest CPUs/RAM available.
I still dont understand. I can claim my rusty old 8086 machine to be the "fastest", and post incomplete performance graphs that dont even conclusively prove that it is "fastest".
Each time a manufacture claims "fastest", they usually back it up with some vendor-produced SPEC scores that are complete and quantitative at first glance. Sun did it with the Blade 1000/2000 workstations, and Apple is doing it with the G5. Of course the SPEC scores are always debatable; even Sun was accused of tweaking their compiler when they were making their claims.
Apple is living high, right now, due to the G5 and the buzz surrounding it. By saying we should give them a pat on the back, I am saying we should recognize that they did do some pretty darn impressive work, regardless of the ultimate conclusion of the benchmarking debate.
Besides, I think most people are willing to give Apple a chance, even if they are a little mischievous. It's not like they're Microsoft:)
He is simply presenting it in a readable fashion...
No, not really. As many others have pointed out, the article suffers from its own fallacies.
The only conclusion we can draw is that Apple's claims, while great marketing, need to be indpendently verified. Big deal.
Every computer maker needs their turn in the spot light. Hell, a while ago, the "fastest" workstation was the 1GHz Sun Blade 2000 (see how long that lasted). For a while after that the "fastest" was the Pentium 4/Xeon. Now, the "fastest" is the G5/POWER4. Does HP sell any Alpha-based desktops, anymore?
Basically, "fastest" is at best temporary; perhaps we should give Apple a pat on the back and move on.
I can hardly detect the difference between a 1.5GHz machine and a 3.0GHz machine without using a benchmark.
A lot of this has to do with most computers shipping with 5400 RPM and 7200 RPM IDE disks, because anything faster is too expensive. Disks are slow, and, every time the computer boots or a new application runs, you are constantly waiting on the disk.
This is why an OS like Solaris 8 or 9 uses all free RAM as one big-ass file cache. It makes a very noticable difference, too, when launching applications several times in a row or doing repeated file system operations.
I would bet that most computer users spend much more time waiting on disk operations than any other CPU-oriented operation, except, perhaps, when using computers more than four years old (sub-400MHz territory). On sufficiently old computers, launching apps like Mozilla becomes more CPU-bound.
Just wait and see if Apple releases benchmark numbers to spec.org. There, they would have to pull out all stops (not use GCC, etc.) to get as high a number as they possibly can.
Considering the IBM pSeries benchmarks already trounce the P4 and Xeon using 1.7GHz POWER4 CPUs, it would be interesting to see how the G5 does with its smaller cache but at 2GHz (don't forget the 1GHz bus, either).
I think we would find the benchmarks at Apple.com were off, but probably not by much. Another thing that is not denyable is that the G5 scaled to two CPUs much much better than the Xeon (look at the rate numbers--this is unsuprising given the POWER4 heritage).
...the regular 'ol Xeon is commonly available at 3+ Ghz. That's still hella faster than the Mac any way you look at it.
Forget the clue stick, this guy needs a salami!
The balance of SPEC scores is pretty typical for the powerful RISC cpus like POWER 4. Did you notice how the PPC 970 scaled from one to two CPUs? Suddenly the Xeon fell behind?
The one thing that machines like the new G5s have in common with high-end UNIX workstations is that they will scale like mad. They have bandwidth, SMP-friendly architectures, much less historical baggage holding them back, etc.
Any more, Xeons are best suited for people who are stuck with Windows Server. I predict that the PowerPC 970 and the Opteron will be wildly successful, much to the chagrin of more than one currently established market leader.
Out of interest, how much is it for the 106-CPU version of the Mac, again?
For the workstation market, it doesn't matter very much.
I have been a Sun fan for some time, now, but I see the dual PowerPC 970 Mac and dual Opteron workstations coming down the line and wonder. These, feature-for-feature, make Sun Blade, IBM RS-6000, SGI Fuel, etc., much harder sells.
The G5 is gorgeous and powerful (like a Bond girl). The Opteron will be white-box and powerful (like the neighborhood geek-girl:). There's probably something for everyone, here. For completeness, I suppose Windows on Itanium would be like some sort of beast woman who still gets guys, because she is easy (blecch).
Sun and Apple are targeting completely different markets.
This was very true three years ago. However, what would happen if PTC released Pro/E for OS X? It's really a matter of the applications. Not only that, but I would bet getting Motif on OS X isn't too hard (suddenly lots of UNIX applications on OS X becomes plausible).
The workstation is going through serious evolution, right now. 64-bits is no longer the domain of the "big guys." The next two years will be very interesting.
Reading the review makes me wonder if this Unleashed book is as bad as the Linux Unleashed book I wasted $50 on several years ago.
Linux Unleashed was little more than a re-printing of the Linux man pages and other trivially-accessible information. It was so little help beyond what I already had that I regretted buying it, and it proved to me that the publisher wanted to put forth as little effort as possible for my money.
Ever since, I have regarded the "Unleashed" books as an insult to the trees used in making them.
Slashdot Mirror
IT is often looked at with a frown because of what it tends to be in most organizations: a very expensive division whose responsibility is to provide critical tools to perform business, with the ultimate goal being to produce a result, typically profit.
Additionally, failures in IT tend to be very very visible. IT is where people get overly excited often, especially after recieving new marketing materials from the "vendor", and expensive projects get created to only fizzle out after they realize there was no need in the first place.
For example, why spend garganuan amounts of money for integrated voice-mail and e-mail, etc. etc. on a single-point-of-failure "cluster", when $30 answering machines are probably more reliable and fit the need like a glove, and are disposable.
I know all businesses are not like this...
You know it?!? You've actually seen one? I thought they were a myth!
In many cases, IT projects end up being cost shifting rather than cost savings.
Or, they turn into classic and memorable "White Elephants", where the cost increases are so embarassing that management simply stops talking about it.
Often, some processes simply cannot be efficiently automated. For example, sometimes an entrenched and very manual government office gets in the way (automating on either side of them seems to be accepted practice, but I may be wrong).
There is appears to be a fundamental flaw in accounting methods when they totally miss productivity loss due to having crap like Windows 9x or NT and PCs with double-digit failure rates.
It is not uncommon to see someone with 32MB (or even 128MB) of RAM in their ancient PC, swapping all to hell, wasting hours upon hours of time, when "there is no budget" for a new PC. It really is sick.
Not only that, after five years, stuff wears out. Replacements? "There is no budget for replacements."
I wonder how close you could get to a working linux machine with nothing offically from GNU?
I should have written this in my other reply:
If you look at OpenBSD, for example, that will be about as close as you can get, where the remaining GNU-branded things are not optional without tremendous effort.
For example, OpenBSD relys on GCC, where doing a BSD-only compiler simply to get rid of "GNU" would not be worth it.
"my" community is linux.
/etc directory structure, for example, is irrelevant, because these things are ultimately independent of the kernel (except, perhaps, the kernel parameter tuning files under /etc and device names under /dev).
This needs to be qualified with how you use Linux.
If you are a average desktop user (Red Hat, GNOME, Mozilla, OpenOffice.org, etc.), then, in fact, Linux, itself, is really not visible in your day-to-day work. If Linux were spontaneously removed, and the NetBSD kernel or HURD took its place (with necessary libraries, etc.), odds are you would never notice.
If you use Linux directly for creating custom kernels for various applications (e.g., you are a kernel developer for the Sharp Zaurus), then, of course, Linux disappearing would have a very large impact on your life. Starting from scratch and reworking the FreeBSD kernel for the Zaurus would definitely be a setback.
From an end-user's perspective, GNOME on Solaris 9, GNOME on Red Hat 9, GNOME on OpenBSD 3.3, and GNOME on Mac OS X (for those who do it because they can) don't present much of a change among them. The fact that each OS has adopted different conventions for package management and
While I dislike the possibility of computer expertise segmenting along economic lines (for social reasons), I do see some benefits: clearer cut job descriptions and areas of expertise...
I think this is inevitable as the mist from the bursted bubble settles. In the last ten years, we experienced a rediculous phenonmenon, where computers progressed from 30MHz CPUs to 3000MHz CPUs, the WWW grew like weeds, and we are now fully bombarded from all sides with techno-buzz gadgets. As society adjusts and people find their niches, I wouldn't be at all suprised if the IT industry begins to look like other older and more established industries.
Perhaps I am naive, but it seems that technology companies have a pretty flat social hierarchy when compared to other industries like construction and health care. For example, there are no physicians (M.D.) or Professional Engineers (P.E.) in the IT industry (i.e., we're really all a bunch of people who essentially made our own job titles and hope the people with money believe what we say--sort of like physicians and engineers a few hundred years ago).
"(company name) is the worldâ(TM)s premier technology system integrator for (technology name) in the enterprise"
I think this phrase is hard-coded into business and marketing majors' brains before graduation.
The art of saying nothing in so many words seems to be the next growth industry. Maybe I should get a lobotomy and jump on board...
I recall that Lindows PCs from Walmart had a huge shipping markup (like $100).
I think this one was debunked as poor--and irresponsible--reporting. A follow-up post to that Lindows review a while ago said the wal-mart shipping is more like $25 or so.
I know from experience that shipping a Sun workstation (no monitor) via the least-expensive UPS option is under $20. Add handling, and 25 to 30 dollars isn't impossible.
Trust me, a lot of times when you donate to a charity, all you are doing is paying for some aging hippie to afford his apartment on the Upper West Side.
Or, paying the professional telemarketing or junk-mailing firm doing the charity's dirty work.
Most charities suck. Many are crooked. That's why I give only to charities I seek out and am already familiar with. For example, it is pretty clear that OpenBSD CD-ROM purchases go to Theo, that Calgary book shop, some of the other developers, etc (i.e., I already know what I'm getting into).
I happen to like the refined flavor of potted meat
What is, exactly, "Potted Meat Food Product", anyway?
Here's another one I saw recently on a vending-machine sandwich: "Partially-defatted beef tissue". What the hell is that?!?
You still have to test everything on all platforms.
However, significantly less redundancy is a big win for long-term maintainence. Also, not having to learn three GUI toolkits is very important.
99% of our apps are the back end anyway.
This is significant. I would guess if you were writing IDEs, for example, the multiple front ends would become a burden.
Why does anyone use Java, ever?
1) Perhaps people see in Java some of the reasons they would rather choose Lisp but can't, because Lisp is still fragmented and not widely understood.
2) From a high-level design point of view, it's better than C but less complex than C++ (although interfaces and the lack of multiple inheritence still nag at me a bit).
3) It is possible to isolate platform dependencies to configuration files and distribute the same compiled binaries to all platforms (think: compile once, run anyhere--if programmed properly, that is). This one is actually pretty darn good reason.
4) The immensity of the Java APIs is overwhelming, but the capabilities in there are impressive. A Java programmer that finds themselves writing utility classes for common operations probably missed something.
5) Although the "null pointer exception" took over where "segmentation fault" left off, there really are classes of bugs that don't occur in Java, such as buffer overflows and random memory corruption. In C, only a good debugger, such as Sun's dbx, can catch some bugs (dbx's run-time checking is very very useful at times).
6) Java applications are not necessarily more reliable than C programs, but it seems that Java programs are more easily debugged (however, it seems J2EE apps might be harder due to the edit-compile-deploy-test cycle).
7) Java is stable, from the point of view that the language and APIs don't often change in incompatible ways. Java inherited Sun's conservatism. Compare this to Microsoft's history of API replacements and extensions (and to the volatility present in some OSS projects, such as GNOME).
8) The Java runtime is, itself, pretty reliable. I don't recall seeing the JVM crash or running into bugs in the APIs. I'm sure they are there, but it appears they are not common.
8) While it is true that C is much more widely ported than J2SE, kludges like autoconf scripts aren't needed for Java applications, which reflects well on Java's architecture. I swear I waste more time fighting with autoconf, libtool, etc. than any other aspect of managing applications on my systems. I don't even know what purpose libtool serves other than to break builds on Solaris. Libtool sucks. (okay that was a rant)
9) Java isn't Microsoft. I know this is flame-bait, but the sad thing is that this reason is as big is any other I listed above. Who knows where C#/.NET is going, but, if Microsoft goes down, C#/.NET is going with them. Sun Microsystems is smart enough to realize companies don't like to put all their eggs in one basket. Thus, the JCP, J2EE licensing and certification, etc. Other than J2SE, Sun is actually a small player in the Java market (they are dwarfed by BEA, for example, in the J2EE market). Sun also sets other examples that reinforce that they aren't complete ass holes (OpenOffice.org, for one; Liberty Alliance, for another)
In general, I think there are a lot of thing that Java does really well, but most people completely take them for granted. It is easy to forget how frustrating memory bugs in C are or how damn complex C++ is, for example.
... developers who are migrating now from Visual Basic are moving to Java and C# in roughly equal numbers."
That isn't so bad. It means that there are a non-trivial number of developers leaving Microsoft and not coming back. Also, the ones moving to C# probably didn't think much about the alternatives--they were captured by default.
Windows: MFC front end, C++ back end
Linux: gtk front end, same C++ back end
OS X: obj c front end, same C++ back end
Perhaps
Windows: Qt front end, C++ back end
Linux: Qt front end, C++ back end
OS X: Qt front end, C++ back end
might be a more acceptable work load?
What's the Japanese way of laying off someone?
Revoke their smart-card rights to the karaoke machine...they'll get the message.
A few years ago I was contemplating a new ISP, so I called several in the DC area (Erol's, ATT, etc.) and asked the "order takers" if their systems supported UNIX and/or Linux.
It seems the UNIX-friendliness of an ISP is made clear when their installation instructions take you the Windows-Dial-Up-Adapter route. Modern PPP programs (at least under OpenBSD) understand Microsoft's proprietary "extensions", so basically anything that can be setup under the Windows dial-up configuration can also be set up under UNIX.
Once the PPP connection is made, finding a DNS server isn't too hard, and there are certainly UNIX-friendly e-mail providers out there (POP/IMAP, etc).
As far as broad-band goes, doesn't the ISP stop at that little router-gadget? I'm not too familiar with broad-band, unfortunately.
...I think the best CAD packages are currently for PC.
Not necessarily. The cheapest ones are certainly for the PC but definitely not the best.
One thing that is important is that PCs configured to produce similar display quality relative to, for example, higher-end Sun graphics begin to approach the cost of Sun workstations ( $1000+ video cards, etc.). One thing in particular is the quality of the anti-aliased display in a CAD system, which many low to mid-range PC cards simply don't do well.
If I have to spend hours a day staring at a CAD model, I would want to, first, do it using the best graphics and monitor combinations available, and, second, use the fastest CPUs/RAM available.
I still dont understand. I can claim my rusty old 8086 machine to be the "fastest", and post incomplete performance graphs that dont even conclusively prove that it is "fastest".
:)
Each time a manufacture claims "fastest", they usually back it up with some vendor-produced SPEC scores that are complete and quantitative at first glance. Sun did it with the Blade 1000/2000 workstations, and Apple is doing it with the G5. Of course the SPEC scores are always debatable; even Sun was accused of tweaking their compiler when they were making their claims.
Apple is living high, right now, due to the G5 and the buzz surrounding it. By saying we should give them a pat on the back, I am saying we should recognize that they did do some pretty darn impressive work, regardless of the ultimate conclusion of the benchmarking debate.
Besides, I think most people are willing to give Apple a chance, even if they are a little mischievous. It's not like they're Microsoft
He is simply presenting it in a readable fashion...
No, not really. As many others have pointed out, the article suffers from its own fallacies.
The only conclusion we can draw is that Apple's claims, while great marketing, need to be indpendently verified. Big deal.
Every computer maker needs their turn in the spot light. Hell, a while ago, the "fastest" workstation was the 1GHz Sun Blade 2000 (see how long that lasted). For a while after that the "fastest" was the Pentium 4/Xeon. Now, the "fastest" is the G5/POWER4. Does HP sell any Alpha-based desktops, anymore?
Basically, "fastest" is at best temporary; perhaps we should give Apple a pat on the back and move on.
I can hardly detect the difference between a 1.5GHz machine and a 3.0GHz machine without using a benchmark.
A lot of this has to do with most computers shipping with 5400 RPM and 7200 RPM IDE disks, because anything faster is too expensive. Disks are slow, and, every time the computer boots or a new application runs, you are constantly waiting on the disk.
This is why an OS like Solaris 8 or 9 uses all free RAM as one big-ass file cache. It makes a very noticable difference, too, when launching applications several times in a row or doing repeated file system operations.
I would bet that most computer users spend much more time waiting on disk operations than any other CPU-oriented operation, except, perhaps, when using computers more than four years old (sub-400MHz territory). On sufficiently old computers, launching apps like Mozilla becomes more CPU-bound.
Just wait and see if Apple releases benchmark numbers to spec.org. There, they would have to pull out all stops (not use GCC, etc.) to get as high a number as they possibly can.
Considering the IBM pSeries benchmarks already trounce the P4 and Xeon using 1.7GHz POWER4 CPUs, it would be interesting to see how the G5 does with its smaller cache but at 2GHz (don't forget the 1GHz bus, either).
I think we would find the benchmarks at Apple.com were off, but probably not by much. Another thing that is not denyable is that the G5 scaled to two CPUs much much better than the Xeon (look at the rate numbers--this is unsuprising given the POWER4 heritage).
...the regular 'ol Xeon is commonly available at 3+ Ghz. That's still hella faster than the Mac any way you look at it.
Forget the clue stick, this guy needs a salami!
The balance of SPEC scores is pretty typical for the powerful RISC cpus like POWER 4. Did you notice how the PPC 970 scaled from one to two CPUs? Suddenly the Xeon fell behind?
The one thing that machines like the new G5s have in common with high-end UNIX workstations is that they will scale like mad. They have bandwidth, SMP-friendly architectures, much less historical baggage holding them back, etc.
Any more, Xeons are best suited for people who are stuck with Windows Server. I predict that the PowerPC 970 and the Opteron will be wildly successful, much to the chagrin of more than one currently established market leader.
Out of interest, how much is it for the 106-CPU version of the Mac, again?
For the workstation market, it doesn't matter very much.
I have been a Sun fan for some time, now, but I see the dual PowerPC 970 Mac and dual Opteron workstations coming down the line and wonder. These, feature-for-feature, make Sun Blade, IBM RS-6000, SGI Fuel, etc., much harder sells.
The G5 is gorgeous and powerful (like a Bond girl). The Opteron will be white-box and powerful (like the neighborhood geek-girl:). There's probably something for everyone, here. For completeness, I suppose Windows on Itanium would be like some sort of beast woman who still gets guys, because she is easy (blecch).
Sun and Apple are targeting completely different markets.
This was very true three years ago. However, what would happen if PTC released Pro/E for OS X? It's really a matter of the applications. Not only that, but I would bet getting Motif on OS X isn't too hard (suddenly lots of UNIX applications on OS X becomes plausible).
The workstation is going through serious evolution, right now. 64-bits is no longer the domain of the "big guys." The next two years will be very interesting.
Reading the review makes me wonder if this Unleashed book is as bad as the Linux Unleashed book I wasted $50 on several years ago.
Linux Unleashed was little more than a re-printing of the Linux man pages and other trivially-accessible information. It was so little help beyond what I already had that I regretted buying it, and it proved to me that the publisher wanted to put forth as little effort as possible for my money.
Ever since, I have regarded the "Unleashed" books as an insult to the trees used in making them.