Intel made a lot of changes to the architecture in the P4, the primary goal was scalability, not efficiency. So, it is no surprise that they are slower than a P3 at the same clock frequency. But the P4 is expected to scale to the 3-4GHz range, while it is doubtful that the P3 will ever even make it to 1.4GHz. So, Intel has given a little at the low end as the cost for being able to go much, much faster in the long run.
Sounds like you work for a normal company. Almost all of the top ten consumers of H1Bs are contracting agencies, Tata (an Indian-government subsidized company) is #1. Sorry I don't have a link for this, you'll just have to trust me on it.
These agencies are a lot more abusive than a normal, non-contract, American employer. They force their H1B people to sign pre-employment contacts that state that if they leave the company, they are responsible for some very trumped up moving/training costs in the range of $10K and up. That, combined with the implicit threat of being deported if they make waves is enough to keep most of them down.
As for being forced to live in 'squalid' housing, it does happen. About a year ago, there was a guy in the mid-west, I think Cincinnati, that had 6 H1Bs cooped up in a 2 bedroom apartment. The kicker was that he was charging each of them $600/month for rent, that's $3600/month for an apartment that cost him only $600/month. Fortunately, he caught and taken to court, don't know what happened to the H1B's, they were probably deported.
Sure, such a case is on the extreme end of the spectrum, but to those of us who have been watching the H1B saga unfold for 2+ years and know some of the people personally kept down, it is clear that your own experience is on the other end of the spectrum. The truth lies somewhere in between and the truth is not good.
Metcalfe is always wrong about the big things
on
Bob Metcalfe On NPR
·
· Score: 1
Sure Metcalfe invented ethernet when he was at PARC and started the spin-off (3com) that was created to commercialize it. But, he also lost control of 3com long ago in a rather ugly battle. Seems to me that he's not the most astute businessman out there.
Meanwhile, he keeps predicting the "megalapse" where the internet will be crushed under the weight of its own traffic. He thinks traffic will grow faster than bandwidth and one day it will all come to a screeching halt, kind of like when TCP's re-xmit algorithm gets out of hand.
The thing is he has been predicting this event every year for something like that last four years. So far he hasn't been anywhere near close to being right, but that hasn't stopped him from blindly continuing with the predictions.
Meanwhile he's using this "megalapse" theory as one reason to argue for metered access. Geezus! Metcalfe may have been at the centre of a fundamental technology in networking, but he's clearly an old-school thinker when it comes to the Net at large.
Jesus! Jack Valenti is still running the MPAA 20 years later. You would think that after being sooo wrong about VCRs, he would have got the boot. It looks like the MPAA is so corrupt that they would rather institutionalize stupidity than learn from their mistakes. If I were a shareholder in any of the MPAA member companies I would be furious.
With all of the funding that linux development is receiving from the likes of HP/SGI/IBM/Redhat/VA/etc you should look into seeing if any of the companies with deep pockets have a documentation budget and if you can get in on the action. I would think that SGI, with their focus on high performance computing, and scaling to large numbers of cpus would have a particular interest in having the network stack well documented so at the very least their own engineers could better work on scaling it.
This guy must have played with Micronauts
on
Personal Helicopter
·
· Score: 1
It is not just a personal helicopter, it is a full set of Acroyear battle-armor, check it out:
Acroyear
Intel is gaining a rep for being especially litigious wrt to former employees leaving for work at other companies. Particularly in the field of vlsi design. In some ways this kind of approach can back-fire, if an engineer thinks he'll be locked into working for Intel (until they lay him off, that is), he may just avoid Intel in the first place.
All of the game consoles are heavily price subsidized by their manufacturers. They take a loss on the console and expect to make it up on the royalities from game sales.
If a substantial portion of people/companies go out and buy 100+ machines each for their own compute farms, the manufacturers are going to eventually notice that the ratio of games bought to consoles sold is declining. At some point they will attempt to prevent non-game (or really, non-royalty generating uses of their hardware. Kind of like the whole i-opener fiasco.
However, there is probably a point up to which the manufacturers are happy to have unix ported because it will increase their raw sales figures, and act as an incentive for game development companies, whose suits would never conceive of the idea that a console might be used for something other than gaming, to port to the console.
Really, all these awards are not about what films or actors are good, they are about hyping films to put more butts in seats in the theaters. That's why a film that wins an award will often be re-relased to theaters, and then release to tape/dvd/pay-per-view will be delayed.
Stuff shown on the internet doesn't put butts in seats in theaters, so there really is no point in wasting the hype of the awards on films that won't garner much more revenue due to an award.
KSR's stuff was COMA, Cache-Only-Memory-Architecture. Similar to NUMA, but each local cpu's (or node's) memory was a big cache, remote accesses became encached in the local cpu's memory so you only took the big latency hit on the first access from that cpu. Now, multiple cpus accessing the same cache line might cause pathological problems.
Sun owns all of KSR's COMA patents. Look for some derivative of it to show up in their 128-way 'Serengetti' which is the expected successor to the UE1000 (which they got from Cray when SGI bought them).
All major vendors are doing some sort of NUMA-ish architecture. DEC just announced their GS-line of boxes that scale up to 32-cpus in a cabinet, in blocks of 4. Theoretically the architecture will scale to 512 cpus, but they are only selling 32 cpu boxes for now. Their memory-bandwidth and fault-tolerant features are nice, their latency, especially off-qbb, sucks. SGI's 3-year old Origin 2000 has better latency.
I agree, unless they are paid for the work they do in their off hours (in which case it wouldn't really be off hours) then MS, or any other employer is going to have a hard time to legally claim ownership.
Now, it would not be unreasonable for their employment agreement to state that they won't work on such in their off hours (a sort of no moonlighting clause, which a lot of employers have and is itself questionable in some US states), but the only result of that would be termination from their day job and then only if the employer were to find out and then be so foolish as to make a big deal of it.
Never assume that HR's job is to help out the employees. In all large companies and many small companies, HR is your enemy. It is their job to make sure that nobody gets paid "too much." Their job to make sure that the cost of benefits programs is minimized. Their job to keep the employees from getting too large a piece of the pie of the company's revenues.
They are not there to help you resolve disputes with your boss and/or co-workers. In many large companies, HR will report the contents of any conversation you have with them right back to your manager, especially if it gives the company some kind of leverage over you.
This kind of BS doesn't fly in a startup or small company, but once they get big and faceless, it becomes HR's job to keep the employees down.
It is this simplistic kind of attitude that makes switching employers the only way to appreciably increase your salary. HR is short-sighted, no doubt about it.
My last experience with a corporate HR group was when I "volunteered" to take a position within the company doing some crap work for a lot of money, effectively $100/hr. They completely vetoed it even though the manager was desperate for someone and they still would have made a decent profit.
Well, I went independent and billed $135/hr for pretty much the same stuff and the company lost money on the deal, but they were contractually obligated to provide the service and I was the only game in town (or actually in the whole country).
HR is not your friend, but that doesn't mean you have to take it.
There is a lot more to it than just set-associativity vs direct-mapped. HP's PA-RISC line recently changed over from 4MB external direct-mapped caches (2MB instruction-cache and 2MB data-cache) on the PA-8200 to on-chip 4-way set-associative 1.5MB (0.5MB instruction-cache and 1MB data-cache) on the PA-8500.
For some application code (specifically *our* custom floating-point intensive code) performance went in the toilet, in some cases the new chips are more than 50% slower than the old ones, despite a doubling of clock from 220MHz to 440MHz. Part of the problem is a poor (simplistic) cache-line replacement policy and the rest may be due to some chip-specific assumptions we made when developing on the PA-8000/PA-8200 that do not apply on the PA-8500.
So, I agree with your ultimate point that "it's hard to make [accurate] generalizations," but I must extend it to all levels because *most* would say the PA-8500 is much better than the PA-8200, but for us it definitely ain't.
Power consumption of an Athlon and/or a PIII are on the order of 20-45 watts. No where near a kilowatt, but still considerably greater than milliwatts.
Don't confuse mainframes with super-computers. Mainframes tend to be targeted towards the commercial market, and supers are sold to the technical market.
Yes, there is a lot of overlap between the two markets nowadays, but they still have two different goals: commercial cares about reliabilty and uptime, technical cares about getting the right answers even if it takes a while (well, not too long).
In the hey-day of supers, the joke was that super-computing was a synonym for unreliable computing. The reason the joke was true was two-fold: a small customer-base means less people to bang on the OS and find bugs, and because reliability was not a high-priority to the market.
Sure, the hardware in supers tends to be over-engineered and rock solid, but in a cheapo-pc grade cluster situation you can easily make up for hardware reliability problems through redundancy, just have a couple or ten hot spares ready to pick up where the last node crashed.
As for your point about problems with high inter-process communication requirements, I totally agree.
The details on the transaction do not seem to be public yet, the news stories I have read make a vague mention of "undisclosed amount of cash, stock and notes for the company."
Given that TERA is one of those almost-on-life-support companies with some neato tech, perhaps this is really just a way for SGI to get access to TERA's technology. TERA certainly does not have the cash to buy Cray, so it seems likely that the majority of the transaction will be for shares and warrants of TERA. At the very least, such a transaction would give SGI a significant minority interest in TERA. On the other side of the equation, SGI gets to write-off a $600M+ tax-loss which should really help them out should they ever be able to sustain profitability.
I was glad to see one of my favorite up-and-coming actresses in Scream 3. Parker played the actress playing courtney cox's character. Although Scream3 is a little more mainstream than her other stuff, it provided good exposure. Allow me to recommend SubUrbia and Basquiat, as well as *almost* everything else imdb lists in her credits.
The problem here is that the case involves people who were real employees and were 'forced' into temp status so that MS could avoid paying benefits, without a concommitant increase in pay. These people were screwed plain and simple.
However, the industry sees only the hype around the case, namely that people who were actual W2 employees of another company (the agency) have been reclassified as employees of the client. The result of this hype is that more and more large companies are becoming more and more wary of using true contractors in any situation.
I myself am a "true" contractor. I make 4x what I made as an employee, I pay for my own benefits ($600/month for medical, dental and disability insurance, and my own retirement plan where I stuff $30k/yr pre-tax away) and I still am way ahead of my previous employee status.
I want the people in the microsoft case to prevail for justice's sake, but unfortunately the side-effects of this case have already made it more difficult for me to work as a contractor.
My bet is that even though Intel could disable SMP capabilities in the new breed of celerons, I bet they won't and here's why: Currently, only Intel processors work in SMP systems. Athlon has support for SMP, but they can't use the de-facto standard because Intel has patented it. Intel can say that celerons are not supported in an SMP system and thus no big name companies (which comprise the bulk of the market) are going to ship celerons in SMP systems.
Meanwhile, the build-it-yourself hobbyists will have the choice of a cheap 1cpu Athlon system, an almost as cheap SMP celeron system, or an expensive SMP Athlon system. For anyone running Win2K, BeOS, Linux, or whatever that is not Win98, it is a no-brainer to pay a couple of bucks extra and get an extra 60-80% performance benefit.
Think of SMP capable, super-overclockable, 100MHz celerons as Intel's stealth weapon against AMD in the hobbyist market.
This is the way competition is suppossed to work, and I sure hope it does because come next year when those.18 micron celerons start to show up, is when I plan on upgrading to a new system and I want it cheap and fast.
Slashdot Mirror
Intel made a lot of changes to the architecture in the P4, the primary goal was scalability, not efficiency. So, it is no surprise that they are slower than a P3 at the same clock frequency. But the P4 is expected to scale to the 3-4GHz range, while it is doubtful that the P3 will ever even make it to 1.4GHz. So, Intel has given a little at the low end as the cost for being able to go much, much faster in the long run.
Sounds like you work for a normal company. Almost all of the top ten consumers of H1Bs are contracting agencies, Tata (an Indian-government subsidized company) is #1. Sorry I don't have a link for this, you'll just have to trust me on it.
These agencies are a lot more abusive than a normal, non-contract, American employer. They force their H1B people to sign pre-employment contacts that state that if they leave the company, they are responsible for some very trumped up moving/training costs in the range of $10K and up. That, combined with the implicit threat of being deported if they make waves is enough to keep most of them down.
As for being forced to live in 'squalid' housing, it does happen. About a year ago, there was a guy in the mid-west, I think Cincinnati, that had 6 H1Bs cooped up in a 2 bedroom apartment. The kicker was that he was charging each of them $600/month for rent, that's $3600/month for an apartment that cost him only $600/month. Fortunately, he caught and taken to court, don't know what happened to the H1B's, they were probably deported.
Sure, such a case is on the extreme end of the spectrum, but to those of us who have been watching the H1B saga unfold for 2+ years and know some of the people personally kept down, it is clear that your own experience is on the other end of the spectrum. The truth lies somewhere in between and the truth is not good.
Sure Metcalfe invented ethernet when he was at PARC and started the spin-off (3com) that was created to commercialize it. But, he also lost control of 3com long ago in a rather ugly battle. Seems to me that he's not the most astute businessman out there.
Meanwhile, he keeps predicting the "megalapse" where the internet will be crushed under the weight of its own traffic. He thinks traffic will grow faster than bandwidth and one day it will all come to a screeching halt, kind of like when TCP's re-xmit algorithm gets out of hand.
The thing is he has been predicting this event every year for something like that last four years. So far he hasn't been anywhere near close to being right, but that hasn't stopped him from blindly continuing with the predictions.
Meanwhile he's using this "megalapse" theory as one reason to argue for metered access. Geezus! Metcalfe may have been at the centre of a fundamental technology in networking, but he's clearly an old-school thinker when it comes to the Net at large.
Jesus! Jack Valenti is still running the MPAA 20 years later. You would think that after being sooo wrong about VCRs, he would have got the boot. It looks like the MPAA is so corrupt that they would rather institutionalize stupidity than learn from their mistakes. If I were a shareholder in any of the MPAA member companies I would be furious.
With all of the funding that linux development is receiving from the likes of HP/SGI/IBM/Redhat/VA/etc you should look into seeing if any of the companies with deep pockets have a documentation budget and if you can get in on the action. I would think that SGI, with their focus on high performance computing, and scaling to large numbers of cpus would have a particular interest in having the network stack well documented so at the very least their own engineers could better work on scaling it.
It is not just a personal helicopter, it is a full set of Acroyear battle-armor, check it out:
Acroyear
Intel is gaining a rep for being especially litigious wrt to former employees leaving for work at other companies. Particularly in the field of vlsi design. In some ways this kind of approach can back-fire, if an engineer thinks he'll be locked into working for Intel (until they lay him off, that is), he may just avoid Intel in the first place.
If a substantial portion of people/companies go out and buy 100+ machines each for their own compute farms, the manufacturers are going to eventually notice that the ratio of games bought to consoles sold is declining. At some point they will attempt to prevent non-game (or really, non-royalty generating uses of their hardware. Kind of like the whole i-opener fiasco.
However, there is probably a point up to which the manufacturers are happy to have unix ported because it will increase their raw sales figures, and act as an incentive for game development companies, whose suits would never conceive of the idea that a console might be used for something other than gaming, to port to the console.
Stuff shown on the internet doesn't put butts in seats in theaters, so there really is no point in wasting the hype of the awards on films that won't garner much more revenue due to an award.
Sun owns all of KSR's COMA patents. Look for some derivative of it to show up in their 128-way 'Serengetti' which is the expected successor to the UE1000 (which they got from Cray when SGI bought them).
All major vendors are doing some sort of NUMA-ish architecture. DEC just announced their GS-line of boxes that scale up to 32-cpus in a cabinet, in blocks of 4. Theoretically the architecture will scale to 512 cpus, but they are only selling 32 cpu boxes for now. Their memory-bandwidth and fault-tolerant features are nice, their latency, especially off-qbb, sucks. SGI's 3-year old Origin 2000 has better latency.
I agree, unless they are paid for the work they do in their off hours (in which case it wouldn't really be off hours) then MS, or any other employer is going to have a hard time to legally claim ownership.
Now, it would not be unreasonable for their employment agreement to state that they won't work on such in their off hours (a sort of no moonlighting clause, which a lot of employers have and is itself questionable in some US states), but the only result of that would be termination from their day job and then only if the employer were to find out and then be so foolish as to make a big deal of it.
Never assume that HR's job is to help out the employees. In all large companies and many small companies, HR is your enemy. It is their job to make sure that nobody gets paid "too much." Their job to make sure that the cost of benefits programs is minimized. Their job to keep the employees from getting too large a piece of the pie of the company's revenues.
They are not there to help you resolve disputes with your boss and/or co-workers. In many large companies, HR will report the contents of any conversation you have with them right back to your manager, especially if it gives the company some kind of leverage over you.
This kind of BS doesn't fly in a startup or small company, but once they get big and faceless, it becomes HR's job to keep the employees down.
It is this simplistic kind of attitude that makes switching employers the only way to appreciably increase your salary. HR is short-sighted, no doubt about it.
My last experience with a corporate HR group was when I "volunteered" to take a position within the company doing some crap work for a lot of money, effectively $100/hr. They completely vetoed it even though the manager was desperate for someone and they still would have made a decent profit.
Well, I went independent and billed $135/hr for pretty much the same stuff and the company lost money on the deal, but they were contractually obligated to provide the service and I was the only game in town (or actually in the whole country).
HR is not your friend, but that doesn't mean you have to take it.
There is a lot more to it than just set-associativity vs direct-mapped. HP's PA-RISC line recently changed over from 4MB external direct-mapped caches (2MB instruction-cache and 2MB data-cache) on the PA-8200 to on-chip 4-way set-associative 1.5MB (0.5MB instruction-cache and 1MB data-cache) on the PA-8500.
For some application code (specifically *our* custom floating-point intensive code) performance went in the toilet, in some cases the new chips are more than 50% slower than the old ones, despite a doubling of clock from 220MHz to 440MHz. Part of the problem is a poor (simplistic) cache-line replacement policy and the rest may be due to some chip-specific assumptions we made when developing on the PA-8000/PA-8200 that do not apply on the PA-8500.
So, I agree with your ultimate point that "it's hard to make [accurate] generalizations," but I must extend it to all levels because *most* would say the PA-8500 is much better than the PA-8200, but for us it definitely ain't.
Power consumption of an Athlon and/or a PIII are on the order of 20-45 watts. No where near a kilowatt, but still considerably greater than milliwatts.
www.mp3changer.com
Don't confuse mainframes with super-computers. Mainframes tend to be targeted towards the commercial market, and supers are sold to the technical market.
Yes, there is a lot of overlap between the two markets nowadays, but they still have two different goals: commercial cares about reliabilty and uptime, technical cares about getting the right answers even if it takes a while (well, not too long).
In the hey-day of supers, the joke was that super-computing was a synonym for unreliable computing. The reason the joke was true was two-fold: a small customer-base means less people to bang on the OS and find bugs, and because reliability was not a high-priority to the market.
Sure, the hardware in supers tends to be over-engineered and rock solid, but in a cheapo-pc grade cluster situation you can easily make up for hardware reliability problems through redundancy, just have a couple or ten hot spares ready to pick up where the last node crashed.
As for your point about problems with high inter-process communication requirements, I totally agree.
The details on the transaction do not seem to be public yet, the news stories I have read make a vague mention of "undisclosed amount of cash, stock and notes for the company."
Given that TERA is one of those almost-on-life-support companies with some neato tech, perhaps this is really just a way for SGI to get access to TERA's technology. TERA certainly does not have the cash to buy Cray, so it seems likely that the majority of the transaction will be for shares and warrants of TERA. At the very least, such a transaction would give SGI a significant minority interest in TERA. On the other side of the equation, SGI gets to write-off a $600M+ tax-loss which should really help them out should they ever be able to sustain profitability.
I was glad to see one of my favorite up-and-coming actresses in Scream 3. Parker played the actress playing courtney cox's character. Although Scream3 is a little more mainstream than her other stuff, it provided good exposure. Allow me to recommend SubUrbia and Basquiat, as well as *almost* everything else imdb lists in her credits.
Borland does know what open software is.
They are licensing Interbase 6 under the Mozilla 1.1 license.
The problem here is that the case involves people who were real employees and were 'forced' into temp status so that MS could avoid paying benefits, without a concommitant increase in pay. These people were screwed plain and simple.
However, the industry sees only the hype around the case, namely that people who were actual W2 employees of another company (the agency) have been reclassified as employees of the client. The result of this hype is that more and more large companies are becoming more and more wary of using true contractors in any situation.
I myself am a "true" contractor. I make 4x what I made as an employee, I pay for my own benefits ($600/month for medical, dental and disability insurance, and my own retirement plan where I stuff $30k/yr pre-tax away) and I still am way ahead of my previous employee status.
I want the people in the microsoft case to prevail for justice's sake, but unfortunately the side-effects of this case have already made it more difficult for me to work as a contractor.
Meanwhile, the build-it-yourself hobbyists will have the choice of a cheap 1cpu Athlon system, an almost as cheap SMP celeron system, or an expensive SMP Athlon system. For anyone running Win2K, BeOS, Linux, or whatever that is not Win98, it is a no-brainer to pay a couple of bucks extra and get an extra 60-80% performance benefit.
Think of SMP capable, super-overclockable, 100MHz celerons as Intel's stealth weapon against AMD in the hobbyist market.
This is the way competition is suppossed to work, and I sure hope it does because come next year when those .18 micron celerons start to show up, is when I plan on upgrading to a new system and I want it cheap and fast.