Will you asshats STFU with the ECC bullsh*t? If Apple required ECC memory in the G5, everyone would be blasting them for forcing users to use expensive, "non-standard" components.
I guess you aren't capable of comprehending the difference between "option" and "required". Many PC motherboards support both registered and unregistered and non/ECC memory. Understand now?
There is NO reason a consumer desktop/workstation needs ECC memory -- Servers, yes; but the G5 is a friggin' desktop!
Uh, "consumer workstation" is much closer to an oxymoron than "gaming workstation". Workstations are for professional use, doing mission-critical things like designing buildings. Would you rather have your engineer use a computer with definitely reliable memory, or memory that has some finite chance (most likely fairly low) of having an undetected flipped bit (read: incorrect number) in the data somewhere? It's very inconvenient and expensive when those buildings just fall down after you build them - even if it only happens once in a while. This explains why all the major workstation (not just server) manufacturers use ECC memory in their machines.
The very minor cost increase and speed hit associated with ECC is nothing by comparison.;-)
All the people screaming that "gaming workstations are an oxymoron" are really missing some important points.
Game developers need to test on their development boxes.
Today's development box is tomorrow's mainstream gaming box (this may not be true of dual Opteron workstations for awhile;).
Games are the some of the most intensive non-pro apps out there and it's silly for the fastest hardware not to do both.
One other point that the author missed: the new dual G5 PowerMac is also a very nice candidate (especially with the 9800 Pro). The authors have declined to provide pricing for anything AFAIK, but I'm pretty sure the Mac will come in less expensive for similar features - and it runs MacOS X among many other advantages.:-)
A whole lot of the free software the author is enjoying on Linux also runs on MacOS X. There is way more commercial software and games for MacOS X than for Linux (less than for Windows, but then you'd have to run...Windows). The G5s should be ideally suited to scientific computing with the Altivec vector instruction set. The only nit with the G5s is not supporting ECC memory. Apple should do that, as an option.
Yeah, show me where in the U.S. teachers are being overpaid. Only neo-con, big business puppets suggest that teachers are paid too much. Sure, the state of education in the US has resulted in underqualifed or just pain bad teachers in some areas, but generally only because those districs are so dangerous and hopeless that better teachers get discouraged and quit.
Well, we are touching heavily on the issue of whether our current economic system does a sufficient job of rewarding critical members of society. Doctors, teachers, and engineers are all critical members of society, in terms of providing crucial services. Actors, professional athletes and (sorry) attorneys are not (judges are necessary). The jury system should "metamoderate" the judges.
You decide whether or not the right people are making the big money.
An interesting point about our current economy is that monetary compensation is now largely arbitrary, and in many cases no tangible good is exchanged (see "CEO"). We clearly produce far more than we need, so production is no longer the sole basis for wealth. This trend will only continue and accelerate until we end with no human labor being expended for any necessity of life (see "Riders of the Purple Wage by Phillip Jose Farmer). So, what we see is a decreasing need for human labor, and a rapidly increasing population of humans...sounds like trouble to me.;-)
Hello? People at war kill civilians all the time. It is only a very recent phenomenon that military commanders have taken care to not kill civilians, and this has only occured during conflicts where they have an overwhelming military advantage and so can afford such luxuries. Remember World War II? Cities (containing mostly civillians!) were regularly attacked, even with nuclear weapons! (although more civilians were killed by conventional bombs) Try going around telling people that the allies in WWII were terrorists and see how popular you are.
The key difference between "terrorists" and "troops" (other than the fact that the terrorists are irregular forces) is that the "troops" have an identifiable state associated with them. That means we can attack them in the same way they attack us. Terrorists are mighty inconvenient in terms of conventional warfare. Some might argue that our current foray into Iraq is the result of not having an identifiable state associated with Al Qaida (or however it's spelled today). What's that saying? "If all you've got is a hammer, everything starts looking like a nail."
With current technology, there is no need to kill civilians to win wars. Killing the opposing leadership, forces and infrastructure is sufficient.
And of course it is illegal as well as wrong to make prank phone calls.
Absolutely.
However, in my case I simply have a burning desire to let the American Telemarketers Association know exactly what I think of it's business practices.;-)
The price/performance difference between an Itanium and an "enterprise" Opteron 800-series is nil.
Sure. We'll see at what price point the first 8p Opteron systems ship.;-)
Re:Sun is just following SGI down the tubes
on
Merrill Lynch Rips Sun
·
· Score: 2, Interesting
Man, the chip doesn't suck, it just came too late to take away much from the big iron. Itanium actually did beat most of the big iron 64 bit chips, and without shortcutting to 32 bit mode, like most of the others do in benchmarks, except Alpha, which is another 64 bit only chip.
Itanium *does* (finally, at this point, six years or so late) beat most of the 64 bit big-iron competition.
However, what it does *not* beat is Opteron, which is about as fast (faster on int, slower on fp), much less expensive, runs your legacy software well, and uses less memory and power. BTW, before you amplify on your chestnut above, Opteron is not "shortcutting to 32 bit mode" to get it's SPEC results.
In short, Itanium is in the same position as Sun - a dinosaur that's too big and slow to adapt. The Opteron mammals are about to take over. (Yes, it's ironic that the "mammals" are running the venerable x86 ISA.)
It also has lock-stepping, which is important for computation checking for true high availability systems (think Himalaya systems), few architectures have that.
And only 1/10 of 1% of customers want it...how much do you think this will help Itanic? There are also other routes to high availability/reliability.
Did you know AMD has already sold more Opterons than Intel has Itaniums? Itanium has been around much longer.
It _is_ expensive and hot though, and that is what really sucks.
It is expensive and hot due to it's enormous die size. A big part of that is an enormous on-chip cache (6 MB) which is necessary due to EPIC's poor code density.
(BTW all my positive comments regarding the Opteron pretty much apply to the G5 as well, another true 64-bit chip. It looks like the Opteron is a little faster (though I'd like to see more benchmarking with better compilers), but the G5 has Altivec, which provides FP functionality to rival the Itanium. IBM has a strong background in scientific computing.)
the Japanese Earth Simulator supercomputer is producing results showing that it is possible to model climate down to the level of severe weather events
First of all, that is already happening with current weather models...those are the ones that predict hurricane paths and such. There were already predictions that this would be an unusually heavy hurricane season before it started - those were due to climate models that showed the ocean area responsible would be warmer than normal.
Predicting where hurricanes will appear and where they will go ahead of time (that is without looking at the current weather patterns while it is happening) involves that pesky chaos thing and good luck with that.
Perhaps what the person was trying to say is that this is the first time researchers have been able to run 10 km. (or 5, or 1) resolution models on a global scale all at once - and that is quite an achievement if so.
BTW, the point of all this is not to predict individual hurricanes or tracks. It is primarily to identify long-term climate trends. From the article:
"This means that we potentially have the capability to predict whether storms like Hurricane Isabel will be on the increase in future." - Professor Julia Slingo. (Hmmm, I guess she's from Soviet Russia;)
The Web cannot be beat for current events. It's also a great source for directory information: phone numbers, locations, maps, and the like. But it falls flat on its face for in-depth information, unless you're looking for computer and related geekery in all 31 flavors.
Hi Carolyn. How are things?:-)
There is real content on the web, especially if you're willing to pay. Encyclpedia Britannica, for instance. I also ran across a URL the other day that was free, but seemed to have a lot of interesting information (if perhaps sometimes of dubious origin. This is the first article I read, regarding aerospike engines..
There is plenty of good scientific content (I think it's a bit much to lump it in as geekery related to computers;) on the web at sites like fas.org, newscientist.com and many more technical examples (spie.org for instance).
There is a ton of sports, travel and hobby related info on the web. It pays to do a little research before believing any given opinion, but you can usually shortly arrive at a pretty good understanding of a given topic.
What the f***? How is that supposed to help reverse falling unemployment?
Exactly. "I'm unemployed. I know, I'll feed my family by working free for a few years!"
Sounds like a poor business model.;-)
Entrepreneurial innovation (you know, so you make money) sounds like a much better idea. Put some grunts out of work with your software, and get rich. The grunts would be better off doing something else anyway.
Windows 2003 has 185K total installs so far. That is next to nothing (100 million PCs will be sold this year, probably 10 million PC architecture servers).
To extrapolate anything from 185K installs is silly.
Further, the opposite statistic should be considered...the number of Win 98, Win NT, Win 2000, and Win XP boxes being converted to Linux. I'm pretty sure the rate will end up much higher than 5%.;-) And that will be applied to the hundreds of millions of existing machines out there.
Certainly not time to cut and run, Taco.:-P
(Maybe I should set my house on fire today...nah.)
Hey there I know this discussion is over but you might read this...
The obivious difficulty with RPN syntax in everyday calculations is that you have to consioucly keep the stack shallow since you can't remember many numbers at the same time. While with normal algebraic notation you have to consioucly add parentheses when you must keep more than 1 extra number in mind. What I mean is, in RPN on paper you might do something like
4 2 7 4 9 6 + + + + +
No, you would do 4 [ent] 2 + 7 + 4 + 9 + 6 +. RPN feels very natural after just a few days of using it, plus ALL operators are immediate. 2 or more operand operators use additional stack slots.
Keeping the stack shallow isn't normally a problem, it's the way it naturally works...plus you can see intermediate results (subexpressions). The HP calculators display the first 4 stack entries, which is plenty in almost all circumstances (internally the stack is quite large).
RPN is worthwhile, I'm glad to see it hanging in there.
RPN may not save that much time, but it is very clean and teaches important CS concepts if nothing else.
I've been using the free HP48 emulators for the Palm devices, and they are very cool. And free (did I mention that?). It is reported that on the Arm based devices the emulators run faster than the native 48xx calculators. Here they are. Mine is a 66 MHz. Dragonball and it runs about the same speed as my HP48. You will need a Palm device with at least 320x320 display, like the Sonys.
It has no place whatsoever if you have to store data long term and the size of data will exceed the size of RAM on the machine.
One acronym: VM (memory not machine).
I'm actually rather curious how this would perform vs. a traditional DBMS, simply using the page file to replace all the traditional filehandling a DBMS does. You could of course specify a 100 GB page file for your 16 GB RAM Opteron box (presumably with a 64-bit IBM Java VM). A 116 GB database is not inconsiderable.
What you are missing is that Prevalayer is very desirable in it's simplicity and low software development overhead. That same simplicity also gives it speed and inherent ACID properties.
You are also being very shortsighted in another regard. It wouldn't take too much to park a layer on top of Prevalayer (no, not called Prevalayerlayer) that would distribute objects across a farm of Prevalayer boxes (high-reliability boxes with ECC, redundant power supplies, and so on)...then you have a very fast, large, scalable database.
With sane, reliable software (like Java) such things are easier. I find Prevalayer to be very interesting.:-)
Well...no. C is used for system development (i.e. hardware banging). Java cannot be used in that environment without direct support of the libraries and the JVM (or by JNI calls to libraries written in C/C++).
This depends on which flavor of Java you use. Your statement is correct for J2SE. There has been quite a bit of work done on embedded/realtime Java, the whole point of which is touching the hardware.
There is also gcj, which provides CNI, a much thinner calling layer for C and C++ code.
the provider we chose (mind you...we were working for a small client, building a small website) allocated us XXX amount of heap for our processes. We ran over that a number of times with our system because of the nature of JSP (issues with includes and resource allocation).
We didn't thrash the GC ever, but other people might have. When you share a JVM, everyone's process on that box run under the same JVM. That means if one person thrashes it, everyone's processes go down. It sucks, try to avoid it if possible in your projects (we were held to a given budget, and therefore couldn't spend the money on private JVM support...yada yada).
So basically due to not spending an extra $100 or so a month on ISP resources you spent untold thousands debugging and testing...and ended up deploying an unsatisfactory product regardless.
And the thing at fault in all this is...Java? Not hardly.
11) This is pseudo relevant and somewhat justified. Java is controlled by SUN, a company with a history of abandoning products depending upon which way the wind blows. There is a certain risk in using Java, as SUN may at any time decide that Java is no longer a strategic product. However, gcc provides support for the core language and libraries. Only GUI applications run the risk of immediate obsolescence. Most others will merely turn into natively compiled applications, so the risk is mitigated.
I don't really see this risk. If Sun were to go away tomorrow, that would leave Java in the hands of it's biggest proponent - IBM. IBM has over 10 times the market capitalization that Sun does, and is vastly more profitable. IBM could very easily take over the JCP. Rumor also has it that Sun is finally ready to open-source JSE.
There are also open alternatives that don't look too bad. gcj has always been interesting, although it won't run Java 2 apps completely yet. However, it will build SWT based apps including the Eclipse IDE. I still think natively compiled Java holds a lot of promise, especially coupled with the real time extensions.
Slashdot Mirror
I guess you aren't capable of comprehending the difference between "option" and "required". Many PC motherboards support both registered and unregistered and non/ECC memory. Understand now?
There is NO reason a consumer desktop/workstation needs ECC memory -- Servers, yes; but the G5 is a friggin' desktop!
Uh, "consumer workstation" is much closer to an oxymoron than "gaming workstation". Workstations are for professional use, doing mission-critical things like designing buildings. Would you rather have your engineer use a computer with definitely reliable memory, or memory that has some finite chance (most likely fairly low) of having an undetected flipped bit (read: incorrect number) in the data somewhere? It's very inconvenient and expensive when those buildings just fall down after you build them - even if it only happens once in a while. This explains why all the major workstation (not just server) manufacturers use ECC memory in their machines.
The very minor cost increase and speed hit associated with ECC is nothing by comparison. ;-)
I hope this cleared things up.
- Game developers need to test on their development boxes.
- Today's development box is tomorrow's mainstream gaming box (this may not be true of dual Opteron workstations for awhile;).
-
Games are the some of the most intensive non-pro apps out there and it's silly for the fastest hardware not to do both.
One other point that the author missed: the new dual G5 PowerMac is also a very nice candidate (especially with the 9800 Pro). The authors have declined to provide pricing for anything AFAIK, but I'm pretty sure the Mac will come in less expensive for similar features - and it runs MacOS X among many other advantages.A whole lot of the free software the author is enjoying on Linux also runs on MacOS X. There is way more commercial software and games for MacOS X than for Linux (less than for Windows, but then you'd have to run...Windows). The G5s should be ideally suited to scientific computing with the Altivec vector instruction set. The only nit with the G5s is not supporting ECC memory. Apple should do that, as an option.
Well, we are touching heavily on the issue of whether our current economic system does a sufficient job of rewarding critical members of society. Doctors, teachers, and engineers are all critical members of society, in terms of providing crucial services. Actors, professional athletes and (sorry) attorneys are not (judges are necessary). The jury system should "metamoderate" the judges.
You decide whether or not the right people are making the big money.
An interesting point about our current economy is that monetary compensation is now largely arbitrary, and in many cases no tangible good is exchanged (see "CEO"). We clearly produce far more than we need, so production is no longer the sole basis for wealth. This trend will only continue and accelerate until we end with no human labor being expended for any necessity of life (see "Riders of the Purple Wage by Phillip Jose Farmer). So, what we see is a decreasing need for human labor, and a rapidly increasing population of humans...sounds like trouble to me. ;-)
The key difference between "terrorists" and "troops" (other than the fact that the terrorists are irregular forces) is that the "troops" have an identifiable state associated with them. That means we can attack them in the same way they attack us. Terrorists are mighty inconvenient in terms of conventional warfare. Some might argue that our current foray into Iraq is the result of not having an identifiable state associated with Al Qaida (or however it's spelled today). What's that saying? "If all you've got is a hammer, everything starts looking like a nail."
With current technology, there is no need to kill civilians to win wars. Killing the opposing leadership, forces and infrastructure is sufficient.
Absolutely.
However, in my case I simply have a burning desire to let the American Telemarketers Association know exactly what I think of it's business practices. ;-)
No prank there.
Busy signal at 7:31 PST on Sunday. ;-)
Sure. We'll see at what price point the first 8p Opteron systems ship. ;-)
Itanium *does* (finally, at this point, six years or so late) beat most of the 64 bit big-iron competition.
However, what it does *not* beat is Opteron, which is about as fast (faster on int, slower on fp), much less expensive, runs your legacy software well, and uses less memory and power. BTW, before you amplify on your chestnut above, Opteron is not "shortcutting to 32 bit mode" to get it's SPEC results.
In short, Itanium is in the same position as Sun - a dinosaur that's too big and slow to adapt. The Opteron mammals are about to take over. (Yes, it's ironic that the "mammals" are running the venerable x86 ISA.)
It also has lock-stepping, which is important for computation checking for true high availability systems (think Himalaya systems), few architectures have that.
And only 1/10 of 1% of customers want it...how much do you think this will help Itanic? There are also other routes to high availability/reliability.
Did you know AMD has already sold more Opterons than Intel has Itaniums? Itanium has been around much longer.
It _is_ expensive and hot though, and that is what really sucks.
It is expensive and hot due to it's enormous die size. A big part of that is an enormous on-chip cache (6 MB) which is necessary due to EPIC's poor code density.
(BTW all my positive comments regarding the Opteron pretty much apply to the G5 as well, another true 64-bit chip. It looks like the Opteron is a little faster (though I'd like to see more benchmarking with better compilers), but the G5 has Altivec, which provides FP functionality to rival the Itanium. IBM has a strong background in scientific computing.)
First of all, that is already happening with current weather models...those are the ones that predict hurricane paths and such. There were already predictions that this would be an unusually heavy hurricane season before it started - those were due to climate models that showed the ocean area responsible would be warmer than normal.
Predicting where hurricanes will appear and where they will go ahead of time (that is without looking at the current weather patterns while it is happening) involves that pesky chaos thing and good luck with that.
Perhaps what the person was trying to say is that this is the first time researchers have been able to run 10 km. (or 5, or 1) resolution models on a global scale all at once - and that is quite an achievement if so.
BTW, the point of all this is not to predict individual hurricanes or tracks. It is primarily to identify long-term climate trends. From the article:
"This means that we potentially have the capability to predict whether storms like Hurricane Isabel will be on the increase in future." - Professor Julia Slingo. (Hmmm, I guess she's from Soviet Russia;)
Hi Carolyn. How are things? :-)
There is real content on the web, especially if you're willing to pay. Encyclpedia Britannica, for instance. I also ran across a URL the other day that was free, but seemed to have a lot of interesting information (if perhaps sometimes of dubious origin. This is the first article I read, regarding aerospike engines..
There is plenty of good scientific content (I think it's a bit much to lump it in as geekery related to computers;) on the web at sites like fas.org, newscientist.com and many more technical examples (spie.org for instance).
There is a ton of sports, travel and hobby related info on the web. It pays to do a little research before believing any given opinion, but you can usually shortly arrive at a pretty good understanding of a given topic.
I was also pleased to see the recent release of a bunch of MIT courseware to the web. More colleges and universities should follow that lead!
Apparently they just wanted a better kernel than the WinCE kernel. Not hard.
Java is already on a ton of competing devices, including Linux based ones.
- Gray Davis: Dead in the water - obviously a fool and liar.
- Bustamente: Gray Davis reincarnated, only even oilier. Certainly no solution.
- McClintock: Too socially conservative for many people, even if his fiscal policies might be great.
- Schwarzenegger: Political Terminator.
:-)
Arnold is simply the right man for the times.Ah, moderation as punishment for a viewpoint with which you don't agree. Very intellectually honest...
Would any moderators care to state why the parent was modded to -1 Offtopic when the "Georgy" and "Arnold vs. Gary Coleman" weren't modded down?
And of course, this wasn't off-topic at all - it is an explanation of why the "mob" is actually smart.
I'd argue that the parent wasn't "funny", but I suppose it is - to most non-Californians. To us, unfortunately, the whole thing is deadly serious.
Oh, and don't think that trying to censor me via moderation will work - I have plenty of karma to burn. ;-)
What the f***? How is that supposed to help reverse falling unemployment?
Exactly. "I'm unemployed. I know, I'll feed my family by working free for a few years!"
Sounds like a poor business model. ;-)
Entrepreneurial innovation (you know, so you make money) sounds like a much better idea. Put some grunts out of work with your software, and get rich. The grunts would be better off doing something else anyway.
Because he is the best fit for most people of the available candidates.
- Gray Davis: Dead in the water - obviously a fool and liar.
- Bustamente: Gray Davis reincarnated, only even oilier. Certainly no solution.
- McClintock: Too socially conservative for many people, even if his fiscal policies might be great.
- Schwarzenegger: Political Terminator.
:-)
Arnold is simply the right man for the times.Here we are, in 2003, and the most appropriate language in which to write Y is...C?!?
The mind boggles.
I'm beginning to think the solution is the industry-wide adoption of the current flavor of FORTRAN. Now there's a timeless language. ;-)
Very good idea! This was an alarmist article about nothing much.
To extrapolate anything from 185K installs is silly.
Further, the opposite statistic should be considered...the number of Win 98, Win NT, Win 2000, and Win XP boxes being converted to Linux. I'm pretty sure the rate will end up much higher than 5%. ;-) And that will be applied to the hundreds of millions of existing machines out there.
Certainly not time to cut and run, Taco. :-P
(Maybe I should set my house on fire today...nah.)
The obivious difficulty with RPN syntax in everyday calculations is that you have to consioucly keep the stack shallow since you can't remember many numbers at the same time. While with normal algebraic notation you have to consioucly add parentheses when you must keep more than 1 extra number in mind. What I mean is, in RPN on paper you might do something like 4 2 7 4 9 6 + + + + +
No, you would do 4 [ent] 2 + 7 + 4 + 9 + 6 +. RPN feels very natural after just a few days of using it, plus ALL operators are immediate. 2 or more operand operators use additional stack slots.
Keeping the stack shallow isn't normally a problem, it's the way it naturally works...plus you can see intermediate results (subexpressions). The HP calculators display the first 4 stack entries, which is plenty in almost all circumstances (internally the stack is quite large).
RPN is worthwhile, I'm glad to see it hanging in there.
It should be:
25;46*254+2462-645;2453-/
25 keys vs. 26. RPN wins again. :^)
RPN may not save that much time, but it is very clean and teaches important CS concepts if nothing else.
I've been using the free HP48 emulators for the Palm devices, and they are very cool. And free (did I mention that?). It is reported that on the Arm based devices the emulators run faster than the native 48xx calculators. Here they are. Mine is a 66 MHz. Dragonball and it runs about the same speed as my HP48. You will need a Palm device with at least 320x320 display, like the Sonys.
One acronym: VM (memory not machine).
I'm actually rather curious how this would perform vs. a traditional DBMS, simply using the page file to replace all the traditional filehandling a DBMS does. You could of course specify a 100 GB page file for your 16 GB RAM Opteron box (presumably with a 64-bit IBM Java VM). A 116 GB database is not inconsiderable.
What you are missing is that Prevalayer is very desirable in it's simplicity and low software development overhead. That same simplicity also gives it speed and inherent ACID properties.
You are also being very shortsighted in another regard. It wouldn't take too much to park a layer on top of Prevalayer (no, not called Prevalayerlayer) that would distribute objects across a farm of Prevalayer boxes (high-reliability boxes with ECC, redundant power supplies, and so on)...then you have a very fast, large, scalable database.
With sane, reliable software (like Java) such things are easier. I find Prevalayer to be very interesting. :-)
This depends on which flavor of Java you use. Your statement is correct for J2SE. There has been quite a bit of work done on embedded/realtime Java, the whole point of which is touching the hardware.
There is also gcj, which provides CNI, a much thinner calling layer for C and C++ code.
We didn't thrash the GC ever, but other people might have. When you share a JVM, everyone's process on that box run under the same JVM. That means if one person thrashes it, everyone's processes go down. It sucks, try to avoid it if possible in your projects (we were held to a given budget, and therefore couldn't spend the money on private JVM support...yada yada).
So basically due to not spending an extra $100 or so a month on ISP resources you spent untold thousands debugging and testing...and ended up deploying an unsatisfactory product regardless.
And the thing at fault in all this is...Java? Not hardly.
This is Moore's law.
Actually, he said transistor density will double every 18 months. Not quite the same thing.
BTW, computer power could continue doubling every X amount of time for quite a while - once parallelism is exploited more fully.
My favorite variation is:
"Nature sides with the hidden flaw."
I don't really see this risk. If Sun were to go away tomorrow, that would leave Java in the hands of it's biggest proponent - IBM. IBM has over 10 times the market capitalization that Sun does, and is vastly more profitable. IBM could very easily take over the JCP. Rumor also has it that Sun is finally ready to open-source JSE.
There are also open alternatives that don't look too bad. gcj has always been interesting, although it won't run Java 2 apps completely yet. However, it will build SWT based apps including the Eclipse IDE. I still think natively compiled Java holds a lot of promise, especially coupled with the real time extensions.