Processor throughput has increased tremendously. Clock speed has not increased. Issue widths are wider. Larger, faster, and more effective caches are being used, in addition to the introduction of trace caches. Branch prediction continues to gets better, along with speculation techniques. More physical registers and larger lookahead windows allow modern CPUs to pull more parallelism out of single threaded programs than ever before.
Features like hyper-threading and dual cores give a much greater system wide speedup than simply raising the clock rate, and avoid all the problems of power consumption. Even on single thread performance, having another core to run the OS, so your not constantly context switching, can make a differance.
Reading this article made me sick, because they equate speed with clock rate. This is patently false, as the last two years of computer architecture have shown us.
Allow me to second how horrible Windows Media Player for Mac is. Is there any reason they don't use hardware acceleration? It is a perfect example of typical microsoft, it works just good enough to say it exists, and no better. On Windows you get playlists, CDDB, and hardware acceleration. On Mac I get a window where I have to click on the exact pixel on the progress bar to jump to a later part, it looks nothing like the rest of my desktop, and takes forever to start up. I can only have one copy of it open at a time, and that one copy has no way to queue the next movie.
It just makes me mad everytime I have to use it, but because of conversations like this, people need a cross-platform video solution, and MS makes this piece of shit just to say it is cross-platform.
Please correct this if incorrect, but I was under the impression that gcj output architecture specific code, and that there was no interpreter, thus giving Java a huge speed boost over a normal run-time engine/interpreter. Is this true? And if so, wouldn't we expect code compiled with gcj to not impact startup time much, because it wouldn't be loaded until it was referenced, and we know that the Java code in OO is optional, and not required for basic functionality?
An interesting point about TA's, is that being a TA is the low spot on the totem pole. All the really smart people get fellowships, and then the next group down get research assistantships, and then whoever is left can compete for a TAship. It is skills like networking that help get RAships, and therefore the ones that are the worst communicators end up being TAs.
Its a tradeoff between furthering knowledge, or teaching a bunch of younger engineers. If you get paid to do research, you graduate faster. If you get paid to teach, it is a huge time committment, and you still have to do research on top of it.
You got me. Your onto my plan. Co-op let me realize how the system worked before I graduated, so I didn't even have to leave to come back for grad school. An interesting side note on the GRE. Getting less than a perfect on the math section is a black mark against admission for grad school.
People whining about TA's language skills is a pet-peeve of mine. Im in the middle of Indiana. Outside of Purdue, the population is pretty homogenous. It doesn't matter. Now that I'm in grad school, and I go to conferences, I have to be able to talk intelligently with these same people. To work in any modern corporation, one must interact with many differant langauge backgrounds.
Engineering is hard. It just is. No amount of sugar coating will make it easier. Studying hard, going to office hours, going to class and actually doing the homework, instead of copying, makes one better. I partied my fair share, managed to play an intercollegiate sport, got exceptional grades, co-opped 6 terms, and am involved in many extra-curricular activities. I'm not an exceptionally smart person, I just work hard, and I budget my time.
What more can the government due to encourage higher education? Money is all over the place for qualified candidates. I got a full ride scholarship for a PhD from the National Science Foundation. I didn't get the scholarship because of a physics test score my freshman year, I got it because I was a problem solver, and I got to know many, many professors. Being on a first name basis with a professor is always a good thing. The fact that I can go to a state school, and from the day I step in the door as an undergrad, to having a PhD, only spend $30k in tuition is pretty amazing. And Indiana isn't the only state where deals like this exist. Residents of Michigan, Illinois, Wisconsin, California, Washington, Iowa, Florida, and a whole bunch I'm forgetting have wonderful schools that are really cheap.
This guy washed out because he was looking to make a buck, not because he really wanted it. I'm glad he isn't designing my bridges.
Not to be picky, but $4500 for a summer of work isn't pretty darn good. Most companies factor a summer as 12 weeks * 40 hrs/week = 480 hours. This is $9.37 / hour pretax. If the students were going to do the projects anyways, out of the goodness of their hearts, than this was a nice gesture, but it is not the same as an internship where the pay is normally twice that of google, and an internship which will have mentoring and allow a student to learn a companies culture.
Using the latest in computer aided design and analysis software, Phoenix provides technical support in the design and development of towed, autonomous, and remotely operated systems used to accomplish underwater tasks. Capabilities include concept developments for sea floor installations, deep water moored data collection systems, and all aspects of the deployment and recovery of the installations.
Yeah, but when it's every manufacturer, it starts to be a Linux problem. I did research what hardware I wanted before I bought, and went to the store looking for that specific model number. I couldn't find any other manufacturer that claimed to support an 802.11g wireless usb adapter, and sites like prism54 keep saying things like, worked with older versions, not with newer.
For Linux to be useful, it doesn't matter how this "specs being released and drivers written" problem gets fixed, it just needs to get fixed. I can't wait 6 months to lobby a manufacturer for a driver to not appear. I could write a driver, but I'm not going to, even if I could get the specs. I have real work to do, and Linux is just one of the tools I use to do it. When the solution is to switch distribution for what is essentially the same code base, that is a problem. This is fragmentation, and hurts everyone. ESR has us so entwinned into believing that Linux is the best thing, and anything not GPL is horrible, that we overlook the pile of crap it can be. It does great running webservers, but from a user standpoint, it is like a three-legged dog trying to buy a turd on a frozen pond. When no manufacturer can provide a piece of standard hardware, it stops being a manufacturer problem, and it starts to be a platform problem.
I am forced to use it for school research, and for work. 95% of it is wonderful. Sometimes having all the editors and desktop choices is nice. One thing I spent the weekend working on, is wireless in Linux. My machine (Dell PowerEdge SC420), has only 3 PCI slots, filled up with Video, Audio, and TV Tuner. I wanted to add a USB wireless network adapter. Windows has had these for years. I looked up to see which ones work with Linux. D-Link claims their DWL-G132 works with Linux. I drive over to Best Buy, and purchase one. It is $70, with $20 rebate. Sitting next to it on the shelf is the DWL-G122 for $19 after rebate, but that doesnt work in Linux. I get home, plug it in, it doesn't work. This is Fedora Core 4. I poke around, see that the kernel detects it, but has no driver. I go to D-Link's site, their is no driver download. After cursing some, I download Ndiswrapper, compile that, except that it first needs kernel source, so I download the kernel sources, get them installed, which isn't straightforward like in the older redhat versions, compile Ndiswrapper, install that. I use the Ndiswrapper install script to install my windows drivers where they go. I have to read the INF files to determine which one matches my usb-id. It is here that I learn the DWL-G132 and DWL-G122 are the same chip, it is only the driver that is differant. The G122 is limited to 54 MBps, the G132 runs at 108 MBps. Same freaking hardware. I call Ndiswrapper to load the driver, and get no error. In the kernel log, I see a stack trace from the kernel side thread dying. Y'all still with me? Im a freaking computer engineer working on my PhD, and I can't get this shit to work. How many "normal people" know what a driver, a kernel, a thread, or a usb-id is? How many of them can dig through INF files and find out which one really goes to their device? This is unf*ckingbelievable. Finally, I return the usb adapter, and buy a wireless gaming adapter that converts the on-board ethernet to wireless, for $100. Linux sure as hell ain't free when I got to pay $100 for what costs $20 on Windows. Everytime I talk about loving my two Macs, people whine about how expensive they are. Guess what, my time is worth money. I have to have Linux to do my work, but I sure wish I didn't.
The reverse DNS lookup doesn't tell you whether or not is was through a firewall. Many names could resolve to the same address. Say I have mail.cide1.com and ntp.cide1.com, and both point to 3.57.0.2, and you do a lookup, both will return 3.57.0.2, which is in reality my firewall. My firewall is smart enough to route packets on the mail port to my mailserver, sitting well protected behind the firewall, with only the required ports forwarded to/ from it, while packets received on the ntp port are forwarded to a differant machine, which has only the ntp port open to the rest of the world. You just don't know whether you hit the real machine, or my firewall. Without knowing the company in question, their size, and their expertise, it is hard to make a judgement call. Either way, I would be almost certain it is a hijacked machine, and their admin would be glad to know about it. (After you have documented everything, and discussed this with whoever makes the decisions.)
Umm, I program in Java. I've been a developer for 8 years, in a mix of C, C++, Python, Assembly and Java. For sheer ease of programming, I would have to choose Java, for two simple reasons. One, Javadoc is great. (Pydoc is also great). Having a concise reference of the language is something that is horribly missing in c and c++. Second is Eclipse. Eclipse is constantly evaluating your code, so you fix your typos while you code. You know about type mis-matches, or it will ask you what to import to resolve a dependency. This means that when I write Java code, it normally works the first time.
I do develop for very specific things, such as manufacturing test equipment for medical devices. Java can be very flexible. JNI allows Java to call third party device drivers. I was able to port our proprietary messaging library in about 15 hours time from Solaris C to Windows Java. I can control National Instrument's boards, bar code scanners, all kinds of stuff. I can talk over serial, parallel, usb, ethernet, gpib, etc... Things like database drivers and XML parsers are part of the language standard. Eclipse + SWT Designer allows one to make high quality user interfaces fast, and the resulting code is human readable (Not that I'd compare it to Visual Studio)
Java used to get a bad rap for speed, this is no longer true with the newer VMs. Java used to get a bad rap for being inflexible, but this also has improved.
All languages suck at something, but Java isn't bad. It would be nice if you could overload operators, but I understand the reasoning of why it isn't done. When I pick up someone else's code, it does make it easier to understand. I suspect Java will gain more momentum with the GNU Java Toolchain finally coming together.
Anyway, flame away... These opinions seem to be rather unpopular here.
Why is this modded up? 1.) First of all, the article is ancient (Sept 2003).
2.) Second of all, the revolutionary thing promised was 64 bits, which we have today.
3.) Intel is not behind AMD in 64 bit chips. AMD chose a differant design, which sacrificed a lot of transistors for x86 compatibility, limiting the scalability and performance of their chip. It makes sense now, but it further embeds x86 cruft in the market place. Intel was working on 64 bit chips when AMD's main product was making pentium 1 clones.
4.) 90nm wont allow for gigabytes of memory on the die. Cache SRAM takes 6 transistors per bit. There just arent enough transistor now to do it. In addition, regular SDRAM cells take a transistor and a capacitor. They are the same speed, no matter where you put them. Delays from SDRAM sense amps aren't going away, either. I know it's a nice concept, but the L1 / L2 cache structure won't be changing drastically anytime soon.
5.) The last point just doesn't make sense to me. Backing store is normally a fancy word for a hard-drive, which virtual memory uses to store pages that are not in main memory. RAM and system memory are the same thing. All modern operating systems are smart enough not to cache file's in the on-disk memory backing store, because the same data is already located elsewhere on the drive. Why cache the data twice? To extend this concept further, the user can use mmap to map a file into user space as a memory block, and work with the file as if it where a block of memory.
Altix is paying the bills? I really think they have something going with these large scale, single image linux systems. From a technical viewpoint, they are very well designed. My understanding is that they are priced competitively for what is effectively the modern mainframe. When you need rediculous amounts of memory, that isn't segmented over many differant nodes, and gobs of IO power, these things are the way to go.
I've always always heard it as ah-mish, never the other way.
Or you can just say "Pennsylvania Dutch" and be done with it... *smirk
Not all Pennsylvania Dutch are Amish though. I come from a Pennsylvania Dutch family, and while my great grandparents spoke low-german, and were farmers, my great-grandfather also worked in a commercial dairy for an hourly wage. My great-grandparents was buried in a lutheran church, in York, PA. Amish have much stricter(simpler?) religious convictions, but on the outside, I can see how people confuse the two. My grandfather joined the Navy out of high school, and then went to college, as did most of the males from his community (WWII), and took a government job. He was a member of a lutheran church, and later an alderman of a methodist church for many years. If you met him on the street, you would have no idea, other than he is very careful about how he spends money (not cheap, but thrifty), tends to shun anything more than the basics, and values education over pretty much anything else.
I second this. Schools say they are out of money, because they spend money. Lots of money. Make sure they spend some of it on you. They might have to be creative, they have all kinds of way to hide money and move it around, but I would never take a job without a guaranteed paycheck, unless I was getting a degree, or an owner of the company.
I will second this. My iBook (G3 700 MHz, 14") has gotten faster with each revision of MacOS. This thing came with the original 10.1, which I have continually upgraded. When I went from Panther to Tiger, I upgraded from 256 MB to 640 MB of RAM, and added (finally) an airport card. Besides these additions, this machine would be ancient in the PC world, but is completely supported and runs the most recent stuff really well. Things like good hardware accelerated video and the correct DMA settings can make a differance, and when Apple controls everything, this is easy for the OS to do by default.
This is hard for Linux people to accept, because Apple stuff isn't really about choice. I like having only one of each tool, and having it work well. I realize that some of MacOS is based on free software (apache, gcc, freebsd, konquerer...), but there is only one of everything. I do install Firefox and Thunderbird, but thats probably more out of habit than necessity. When I buy a MIDI keyboard, plug it in, and GarageBand puts out sound, thats really nice. When I install the Airport card, and all it asks is which network to join, thats really nice. I've had similar experiences with external harddrives, floppies, Firewire video cameras, iPod, printer, scanner, media reader and digital camera.
I'm debating buying a webcam. Cheap ones are $30, and require a third party driver. The Apple iSight is $150, but will just work (and continue to work in future versions), and I'm sure has much better image quality. Apple stuff is expensive, but my time is worth money, and it saves me time, and is usually much higher quality.
OS X ships with XCode, which is sortof like visual studio. You can write in Java, C, C++, Objective-C, Python, and a bunch of others I'm forgetting. For C, C++, and Objective-C, the actual compiler is GCC.
While the clockspeed remains the same, these chips are much faster, due to greater parallelism. A chip at 3.8 GHz with no hyperthreading, and one core, can only execute one stream of instructions at a time. Add hyperthreading, and it can execute only one stream at a time, but can switch to a second very quickly. Hyperthreading is basically one core with two sets of registers. In modern computers, the processor is so much faster than memory, that when a processor has a cache miss, hyperthreading gives it the capability of quickly moving to a differant stream of instructions while the missing data is brought from slower main memory into faster cache memory, instead of wasting those cycles. Two cores adds even more parallelism, becuase now two complete units are on the same die. Two complete streams of instructions are processed at the same time. I imagine the next step will be two cores with hyperthreading, allowing four threads on the die at the same time, two executing, and two that can be switched in quite fast.
Parallelism is an easier way to add speed. Every doubling in clock speed is a quadrupling in power consumption. Intel's biggest flaw is their power density, and they have hit the physical limits of silicon. Adding a second core does not increase their power density, just their overall power consumption.
It should be noted, too, that caches use six transistors per bit to store data, and therefore, leakage current from caches are a large amount of a processors power consumption. Features like hyperthreading do not increase the number of transistors dedicated to caching.
I worked in mail order for years. I know the company I worked for could only ship UPS. The post-office wouldn't pick up, and Fed-Ex was too expensive. Can UPS ship where you are?
I imagine the problem you are fighting is that companies just dont know how to ship there, not that they wont do it. If you explained exactly how to address the box and whatnot, and explained that there is no additional cost, I know I would be much more willing to help you.
Slashdot Mirror
Processor throughput has increased tremendously. Clock speed has not increased. Issue widths are wider. Larger, faster, and more effective caches are being used, in addition to the introduction of trace caches. Branch prediction continues to gets better, along with speculation techniques. More physical registers and larger lookahead windows allow modern CPUs to pull more parallelism out of single threaded programs than ever before.
Features like hyper-threading and dual cores give a much greater system wide speedup than simply raising the clock rate, and avoid all the problems of power consumption. Even on single thread performance, having another core to run the OS, so your not constantly context switching, can make a differance.
Reading this article made me sick, because they equate speed with clock rate. This is patently false, as the last two years of computer architecture have shown us.
Allow me to second how horrible Windows Media Player for Mac is. Is there any reason they don't use hardware acceleration? It is a perfect example of typical microsoft, it works just good enough to say it exists, and no better. On Windows you get playlists, CDDB, and hardware acceleration. On Mac I get a window where I have to click on the exact pixel on the progress bar to jump to a later part, it looks nothing like the rest of my desktop, and takes forever to start up. I can only have one copy of it open at a time, and that one copy has no way to queue the next movie.
It just makes me mad everytime I have to use it, but because of conversations like this, people need a cross-platform video solution, and MS makes this piece of shit just to say it is cross-platform.
Please correct this if incorrect, but I was under the impression that gcj output architecture specific code, and that there was no interpreter, thus giving Java a huge speed boost over a normal run-time engine/interpreter. Is this true? And if so, wouldn't we expect code compiled with gcj to not impact startup time much, because it wouldn't be loaded until it was referenced, and we know that the Java code in OO is optional, and not required for basic functionality?
Yeah, it just means his is readable and writable by himself, his group, and others :).
Its a good thing he's not executable, or this joke would be killing him.
An interesting point about TA's, is that being a TA is the low spot on the totem pole. All the really smart people get fellowships, and then the next group down get research assistantships, and then whoever is left can compete for a TAship. It is skills like networking that help get RAships, and therefore the ones that are the worst communicators end up being TAs.
Its a tradeoff between furthering knowledge, or teaching a bunch of younger engineers. If you get paid to do research, you graduate faster. If you get paid to teach, it is a huge time committment, and you still have to do research on top of it.
You got me. Your onto my plan. Co-op let me realize how the system worked before I graduated, so I didn't even have to leave to come back for grad school. An interesting side note on the GRE. Getting less than a perfect on the math section is a black mark against admission for grad school.
People whining about TA's language skills is a pet-peeve of mine. Im in the middle of Indiana. Outside of Purdue, the population is pretty homogenous. It doesn't matter. Now that I'm in grad school, and I go to conferences, I have to be able to talk intelligently with these same people. To work in any modern corporation, one must interact with many differant langauge backgrounds.
Engineering is hard. It just is. No amount of sugar coating will make it easier. Studying hard, going to office hours, going to class and actually doing the homework, instead of copying, makes one better. I partied my fair share, managed to play an intercollegiate sport, got exceptional grades, co-opped 6 terms, and am involved in many extra-curricular activities. I'm not an exceptionally smart person, I just work hard, and I budget my time.
What more can the government due to encourage higher education? Money is all over the place for qualified candidates. I got a full ride scholarship for a PhD from the National Science Foundation. I didn't get the scholarship because of a physics test score my freshman year, I got it because I was a problem solver, and I got to know many, many professors. Being on a first name basis with a professor is always a good thing. The fact that I can go to a state school, and from the day I step in the door as an undergrad, to having a PhD, only spend $30k in tuition is pretty amazing. And Indiana isn't the only state where deals like this exist. Residents of Michigan, Illinois, Wisconsin, California, Washington, Iowa, Florida, and a whole bunch I'm forgetting have wonderful schools that are really cheap.
This guy washed out because he was looking to make a buck, not because he really wanted it. I'm glad he isn't designing my bridges.
Not to be picky, but $4500 for a summer of work isn't pretty darn good. Most companies factor a summer as 12 weeks * 40 hrs/week = 480 hours. This is $9.37 / hour pretax. If the students were going to do the projects anyways, out of the goodness of their hearts, than this was a nice gesture, but it is not the same as an internship where the pay is normally twice that of google, and an internship which will have mentoring and allow a student to learn a companies culture.
Using the latest in computer aided design and analysis software, Phoenix provides technical support in the design and development of towed, autonomous, and remotely operated systems used to accomplish underwater tasks. Capabilities include concept developments for sea floor installations, deep water moored data collection systems, and all aspects of the deployment and recovery of the installations.
All these TLA's are making me think WTF, but as long as I can get the 10-4, I'll be A-OK. AFAIK, all we need is a FAQ, and then I can forget this 404.
Yeah, but when it's every manufacturer, it starts to be a Linux problem. I did research what hardware I wanted before I bought, and went to the store looking for that specific model number. I couldn't find any other manufacturer that claimed to support an 802.11g wireless usb adapter, and sites like prism54 keep saying things like, worked with older versions, not with newer.
For Linux to be useful, it doesn't matter how this "specs being released and drivers written" problem gets fixed, it just needs to get fixed. I can't wait 6 months to lobby a manufacturer for a driver to not appear. I could write a driver, but I'm not going to, even if I could get the specs. I have real work to do, and Linux is just one of the tools I use to do it. When the solution is to switch distribution for what is essentially the same code base, that is a problem. This is fragmentation, and hurts everyone. ESR has us so entwinned into believing that Linux is the best thing, and anything not GPL is horrible, that we overlook the pile of crap it can be. It does great running webservers, but from a user standpoint, it is like a three-legged dog trying to buy a turd on a frozen pond. When no manufacturer can provide a piece of standard hardware, it stops being a manufacturer problem, and it starts to be a platform problem.
I am forced to use it for school research, and for work. 95% of it is wonderful. Sometimes having all the editors and desktop choices is nice. One thing I spent the weekend working on, is wireless in Linux. My machine (Dell PowerEdge SC420), has only 3 PCI slots, filled up with Video, Audio, and TV Tuner. I wanted to add a USB wireless network adapter. Windows has had these for years. I looked up to see which ones work with Linux. D-Link claims their DWL-G132 works with Linux. I drive over to Best Buy, and purchase one. It is $70, with $20 rebate. Sitting next to it on the shelf is the DWL-G122 for $19 after rebate, but that doesnt work in Linux. I get home, plug it in, it doesn't work. This is Fedora Core 4. I poke around, see that the kernel detects it, but has no driver. I go to D-Link's site, their is no driver download. After cursing some, I download Ndiswrapper, compile that, except that it first needs kernel source, so I download the kernel sources, get them installed, which isn't straightforward like in the older redhat versions, compile Ndiswrapper, install that. I use the Ndiswrapper install script to install my windows drivers where they go. I have to read the INF files to determine which one matches my usb-id. It is here that I learn the DWL-G132 and DWL-G122 are the same chip, it is only the driver that is differant. The G122 is limited to 54 MBps, the G132 runs at 108 MBps. Same freaking hardware. I call Ndiswrapper to load the driver, and get no error. In the kernel log, I see a stack trace from the kernel side thread dying. Y'all still with me? Im a freaking computer engineer working on my PhD, and I can't get this shit to work. How many "normal people" know what a driver, a kernel, a thread, or a usb-id is? How many of them can dig through INF files and find out which one really goes to their device? This is unf*ckingbelievable. Finally, I return the usb adapter, and buy a wireless gaming adapter that converts the on-board ethernet to wireless, for $100. Linux sure as hell ain't free when I got to pay $100 for what costs $20 on Windows. Everytime I talk about loving my two Macs, people whine about how expensive they are. Guess what, my time is worth money. I have to have Linux to do my work, but I sure wish I didn't.
The reverse DNS lookup doesn't tell you whether or not is was through a firewall. Many names could resolve to the same address. Say I have mail.cide1.com and ntp.cide1.com, and both point to 3.57.0.2, and you do a lookup, both will return 3.57.0.2, which is in reality my firewall. My firewall is smart enough to route packets on the mail port to my mailserver, sitting well protected behind the firewall, with only the required ports forwarded to/ from it, while packets received on the ntp port are forwarded to a differant machine, which has only the ntp port open to the rest of the world. You just don't know whether you hit the real machine, or my firewall. Without knowing the company in question, their size, and their expertise, it is hard to make a judgement call. Either way, I would be almost certain it is a hijacked machine, and their admin would be glad to know about it. (After you have documented everything, and discussed this with whoever makes the decisions.)
Umm, I program in Java. I've been a developer for 8 years, in a mix of C, C++, Python, Assembly and Java. For sheer ease of programming, I would have to choose Java, for two simple reasons. One, Javadoc is great. (Pydoc is also great). Having a concise reference of the language is something that is horribly missing in c and c++. Second is Eclipse. Eclipse is constantly evaluating your code, so you fix your typos while you code. You know about type mis-matches, or it will ask you what to import to resolve a dependency. This means that when I write Java code, it normally works the first time.
I do develop for very specific things, such as manufacturing test equipment for medical devices. Java can be very flexible. JNI allows Java to call third party device drivers. I was able to port our proprietary messaging library in about 15 hours time from Solaris C to Windows Java. I can control National Instrument's boards, bar code scanners, all kinds of stuff. I can talk over serial, parallel, usb, ethernet, gpib, etc... Things like database drivers and XML parsers are part of the language standard. Eclipse + SWT Designer allows one to make high quality user interfaces fast, and the resulting code is human readable (Not that I'd compare it to Visual Studio)
Java used to get a bad rap for speed, this is no longer true with the newer VMs. Java used to get a bad rap for being inflexible, but this also has improved.
All languages suck at something, but Java isn't bad. It would be nice if you could overload operators, but I understand the reasoning of why it isn't done. When I pick up someone else's code, it does make it easier to understand. I suspect Java will gain more momentum with the GNU Java Toolchain finally coming together.
Anyway, flame away... These opinions seem to be rather unpopular here.
So do I, So do I
Why is this modded up?
1.) First of all, the article is ancient (Sept 2003).
2.) Second of all, the revolutionary thing promised was 64 bits, which we have today.
3.) Intel is not behind AMD in 64 bit chips. AMD chose a differant design, which sacrificed a lot of transistors for x86 compatibility, limiting the scalability and performance of their chip. It makes sense now, but it further embeds x86 cruft in the market place. Intel was working on 64 bit chips when AMD's main product was making pentium 1 clones.
4.) 90nm wont allow for gigabytes of memory on the die. Cache SRAM takes 6 transistors per bit. There just arent enough transistor now to do it. In addition, regular SDRAM cells take a transistor and a capacitor. They are the same speed, no matter where you put them. Delays from SDRAM sense amps aren't going away, either. I know it's a nice concept, but the L1 / L2 cache structure won't be changing drastically anytime soon.
5.) The last point just doesn't make sense to me. Backing store is normally a fancy word for a hard-drive, which virtual memory uses to store pages that are not in main memory. RAM and system memory are the same thing. All modern operating systems are smart enough not to cache file's in the on-disk memory backing store, because the same data is already located elsewhere on the drive. Why cache the data twice? To extend this concept further, the user can use mmap to map a file into user space as a memory block, and work with the file as if it where a block of memory.
Altix is paying the bills? I really think they have something going with these large scale, single image linux systems. From a technical viewpoint, they are very well designed. My understanding is that they are priced competitively for what is effectively the modern mainframe. When you need rediculous amounts of memory, that isn't segmented over many differant nodes, and gobs of IO power, these things are the way to go.
I've always always heard it as ah-mish, never the other way.
Or you can just say "Pennsylvania Dutch" and be done with it... *smirk
Not all Pennsylvania Dutch are Amish though. I come from a Pennsylvania Dutch family, and while my great grandparents spoke low-german, and were farmers, my great-grandfather also worked in a commercial dairy for an hourly wage. My great-grandparents was buried in a lutheran church, in York, PA. Amish have much stricter(simpler?) religious convictions, but on the outside, I can see how people confuse the two. My grandfather joined the Navy out of high school, and then went to college, as did most of the males from his community (WWII), and took a government job. He was a member of a lutheran church, and later an alderman of a methodist church for many years. If you met him on the street, you would have no idea, other than he is very careful about how he spends money (not cheap, but thrifty), tends to shun anything more than the basics, and values education over pretty much anything else.
I got that under firefox, it works fine in IE
it is so very, very OK
I second this. Schools say they are out of money, because they spend money. Lots of money. Make sure they spend some of it on you. They might have to be creative, they have all kinds of way to hide money and move it around, but I would never take a job without a guaranteed paycheck, unless I was getting a degree, or an owner of the company.
I will second this. My iBook (G3 700 MHz, 14") has gotten faster with each revision of MacOS. This thing came with the original 10.1, which I have continually upgraded. When I went from Panther to Tiger, I upgraded from 256 MB to 640 MB of RAM, and added (finally) an airport card. Besides these additions, this machine would be ancient in the PC world, but is completely supported and runs the most recent stuff really well. Things like good hardware accelerated video and the correct DMA settings can make a differance, and when Apple controls everything, this is easy for the OS to do by default.
This is hard for Linux people to accept, because Apple stuff isn't really about choice. I like having only one of each tool, and having it work well. I realize that some of MacOS is based on free software (apache, gcc, freebsd, konquerer...), but there is only one of everything. I do install Firefox and Thunderbird, but thats probably more out of habit than necessity. When I buy a MIDI keyboard, plug it in, and GarageBand puts out sound, thats really nice. When I install the Airport card, and all it asks is which network to join, thats really nice. I've had similar experiences with external harddrives, floppies, Firewire video cameras, iPod, printer, scanner, media reader and digital camera.
I'm debating buying a webcam. Cheap ones are $30, and require a third party driver. The Apple iSight is $150, but will just work (and continue to work in future versions), and I'm sure has much better image quality. Apple stuff is expensive, but my time is worth money, and it saves me time, and is usually much higher quality.
OS X ships with XCode, which is sortof like visual studio. You can write in Java, C, C++, Objective-C, Python, and a bunch of others I'm forgetting. For C, C++, and Objective-C, the actual compiler is GCC.
While the clockspeed remains the same, these chips are much faster, due to greater parallelism. A chip at 3.8 GHz with no hyperthreading, and one core, can only execute one stream of instructions at a time. Add hyperthreading, and it can execute only one stream at a time, but can switch to a second very quickly. Hyperthreading is basically one core with two sets of registers. In modern computers, the processor is so much faster than memory, that when a processor has a cache miss, hyperthreading gives it the capability of quickly moving to a differant stream of instructions while the missing data is brought from slower main memory into faster cache memory, instead of wasting those cycles. Two cores adds even more parallelism, becuase now two complete units are on the same die. Two complete streams of instructions are processed at the same time. I imagine the next step will be two cores with hyperthreading, allowing four threads on the die at the same time, two executing, and two that can be switched in quite fast.
Parallelism is an easier way to add speed. Every doubling in clock speed is a quadrupling in power consumption. Intel's biggest flaw is their power density, and they have hit the physical limits of silicon. Adding a second core does not increase their power density, just their overall power consumption.
It should be noted, too, that caches use six transistors per bit to store data, and therefore, leakage current from caches are a large amount of a processors power consumption. Features like hyperthreading do not increase the number of transistors dedicated to caching.
I worked in mail order for years. I know the company I worked for could only ship UPS. The post-office wouldn't pick up, and Fed-Ex was too expensive. Can UPS ship where you are?
I imagine the problem you are fighting is that companies just dont know how to ship there, not that they wont do it. If you explained exactly how to address the box and whatnot, and explained that there is no additional cost, I know I would be much more willing to help you.
4 years? True students take at least 5, maybe 6.
My second sophmore year was the best.