There are X number of data/address pins coming out of the chip. If they stay with the S940 or S939, that number of pins is exactly equal to the number of pins required for one CPU. Adding another core without adding another X pins means that the two cores will share the X pins to the memory. So, unless the memory controller currently has 2X the bandwidth required for a single Opteron to run, then you will have a bottleneck accessing that one bus off the chip in memory intensive applications as the bandwidth available off-chip will be shared by two CPUs instead of one.
Funny thing... Mandrake 10 for AMD64 and SuSE 9.1 for AMD64 are both comparable to the cost of Windows XP. So not only are the system requirements becoming the same or similar, so is the price.
I guess folks realize to do all the bells and whistles, you might actually have to write code. KDE/Gnome are falling to the same issues that folks complain so much about Windows... features bloat.
That'll even make it faster! The other thing they could do is put purposefully loud and angry bee sounding fans in the case, to get the full ricer effect. And a big ass wing on the top of the case.
The G5s are a lot more efficient than G4s and the Pentium4s are much LESS efficient than PentiumIII.... that run at 3X the clock speed as the P3 and thus, give an overall performance increase, especially when coupled with much higher bandwidth memory and the like. I have yet to see a benchmark where any P3 will beat a P4 2.4GHz or faster, and the 3GHz P4s can be more than 3X as fast as a 1GHz P3 on some tasks.
Well... unless you are the sysadmin and can install whatever you want (or you can somehow otherwise dictate policy), there is usually a dance you have to do in order to get some new software installed or for you to be allowed to install some random new piece of software regardless of monitary cost. There are usually good reasons for this.
Just letting anyone download and install anything is how you get caught up in spyware and the like, regardless of the platform you use. Simply saying that because there is no monitary license cost for this software is sufficient reason to not have any "red tape" for installation of software is a mistake.
KOTOR was a fun game for the first couple of hours. I found it to be rather repetitive after that and eventually never got around to finishing it. I played what I did of it on the XBox.
Hi, I recently bought an AMD64 processor and am looking around for a distribution to use. So far, I've actually installed FC2 x86-64 and am looking at Mandrake 10.0 (we use 9.2 i586 at work and I use 9.2 at home already). FC2 is reasonably stable but I've had a few lockup issues, I think I have them solved but I will switch to Mandrake 10.0 if possible (assuming that it is reasonably like 9.2). I haven't heard anything from the SUSE folks about AMD64 support so what's it like?
Synchronization of our Internet bookmarks across all our computers Now wouldn't THAT be nice?
Give http://www.mybookmarks.com a try.
I read the first 100 lines or so then skimmed the rest. It sounds like this guy is running Microsoft ME or something because the things he talks about were mostly during that era. I haven't seen most of the problems he mentions with XP (on many computers).
Judging by the rest of what I skimmed, I wouldn't be surprised if someone found out later that Apple (or some other company) didn't pay him to write that. The other thing to remember is that there is no better zealot that a convert, so I'll file his stuff with the other junk that comes out daily.
That is the thing... The Athlon XP and Athlon 64 are very similar clock-for-clock on x87 codes (and scalar SSE I believe). The Athlon XP 2700+ actually has a higher core frequency than the Athlon 64 3200 (the A-64 3200+ is actually 2.0 GHz, the A-XP 2700+ is 2.167 GHz). The FPUs didn't change that much from the Athlon XP to the Athlon 64. This gives the XP line an actual advantage in many FPU intensive apps that make good use of the cache.
An early one was that open-source software is anti-competitive!
Actually, it sort of is, in a way. At least... when any particular product line starts to "own" a market, as MSOffice has, you can easily bottom out the market by funding or developing an OSS "alternative" to poison that market, thereby collapsing the market to non-profitability. When the market is no longer profitable, interest will leave it and you'll be stuck with software that has no interest in keeping it "current". The only jobs associated with it then would be the equivalent of blue-collar jobs where people just install or uninstall it (not including the people who use it - because they aren't deriving their income directly from the support of the software).
Like this guy, I've worked (actually gotten paid) to develop OSS software in the past and I have also competed head-to-head against it in the commercial space. What I've found is that the majority of people who use OSS use it because it is free (as in beer) rather than because it is necessarily "good". They will settle for inferior products that do as much to hinder their work as improve it simply to save a buck. In many cases, it breeds mediocrity.
I've done that quite a bit. My Windows box doesn't slow down to a crawl before, during, or after playing those games. Of course, I'm picky about my hardware and software and don't buy or use crap hardware so stability isn't a problem. I have 3 Windows XP boxes that run 24/7 at home and only reboot when patches require it. The same can be said for my Linux boxes.
Another big reason is familiarity with Unix and Unix-alike systems already. Most (all?) HPC centers in the past have been Unix or Unix-alike based so they stick with what they know (both the IT group and the programmers/scientists) and have the personnel to manage.
Nah, I'm saying if they want to change it, they have to change away from our terminology. KiloByte remains 1024 bytes. If they don't like our "kilo" being 1024 and want something to mean 1000, then they should use the Kibi to mean 1000, then they should also not use the term "byte" either, as it is a technical reference, so they should replace that with something else, thus, KiloByte = 1024 bytes, KibiFloople = 1000 Flooples.
As to the other comment, the tech language we use was derived in the context of our field. It'd be like a bunch of novices coming in and completely changing the jargon of the plumbing field or medicine based on their uninformed preconceptions. "That's not a crescent wrench! It looks nothing like a crescent. Let's call it a Variable Gap Bolt Loosener and require everyone else in the world to do the same."
Also... Kibi and Mebi are just very unprofessional sounding, like they belong in some Pokemon cartoon. I wonder how many person-hours in committee were required to come up with those terms. I know they preserve the K and the M but this is rediculous. As for me, I will refuse to use "Kibi" and "Mebi".
I will not change the precise language I have been using for 25 years because people outside of my field has problems with field specific math.
If they want to change the terms, then they should not be using byte either. Then they can use KibiFloople to mean 1000 Flooples and MebiFloople to be 1,000,000 Flooples and leave our field specific terminology alone.
Yeah... and because you have no understanding of combinatorial mathmatics, you don't realize the horrible performance and scalability issues that you will be getting because you don't understand your problem or how the tools you are using work. Bad indices, poor caching, crappy queries, and a host of other things cause the project to be a dog and the customer and your boss aren't happy with what you did.
Most code monkeys think along the lines of: if you can't get it to work right, you need a bigger hammer. I've seen a ton of them like you. While having a CS degree doesn't automatically mean that you know what you are doing (I've seen plenty that don't), the odds are much higher that someone with a degree will be more disciplined than someone who doesn't have a degree. Plenty of folks take the stance: "I don't have a degree and look how great I am" and somehow think that they are the norm when, in fact, they are the exception (or just deluded because they've never been around *really* good programmers, almost all of which have higher eduction) when they are really just the big fish in a small pond.
I had a GeForce4Ti4600 in my machine and it died so I replaced it with an ATi 9600 XT. Since then, there are many things that just don't "work right" as they did before. Specifically these:
When playing EverQuest, I can no longer go into windowed mode and then back to full-screen mode. I get to windowed mode fine, but if I try to go back to full screen all I get is a black screen.
Also, when I leave some games now, the signal gets "lost" somehow and my monitor goes black and it puts up a box saying "unsupported video mode" on it until I reset the monitor. This did not happen ever with the nVidia card.
Installing new drivers: For nVidia: Step 1: download new drivers. Step 2: double click on it and go through setup. Step 3: reboot and all is nice.
For ATi: Step 1: download new drivers. Step 2: uninstall old drivers. Step 3: reboot into VGA mode Step 4: install new drivers. Step 5: reboot Step 6: screen is black, something is wrong Step 7: reboot into safe mode Step 8: reinstall drivers again Repeat Steps 5 through 8 about 6 more times for about 30 minutes and eventually get a "good" install and finally get to use my computer again.
Is there any point in getting one of these cards for any reason other than playing the latest games?
nVidia cards tend to have good OpenGL support and OpenGL is used by a number of "high end" CAD and rendering packages. These cards will work well for folks who don't want to spend the $1500 for the high end CAD cards which are almost the same thing (there are some differences but these will do well on a smaller budget, though $500 for a card is pretty pricey to me:)
why high performance computing centers which need reliability and good cost/benefit rations never use Windows.
Having worked in the HPC business for a while (as a developer and vendor of MPI libraries on about 20 different platforms) and having both Unix/Linux and Windows solutions for HPC (as well as a number of embedded systems), this was seldom the "real" reason we saw one chosen over the other. Costs derived from things like having to hire a bunch of new staff to support Windows if they got a Windows cluster when currently they have no Windows boxes in-house were some of the more "real" reasons.
Reasons to choose Unix/Linux over Windows in HPC:
- HPC centers are, and have been, historically Unix/Linux houses so the conclusions from this are pretty obvious - stick with what you know and already have support for.
- HPC people are familiar with Unix/Linux, have used them for years, already have the machines in place, and don't want to have to learn something new or try to figure it out.
- Unix/Linux houses have a built-in distrust/dislike for Windows for whatever reasons (bunches, some of which are simply FUD and some are real).
As far as development of HPC software, these were things I saw:
- Unix/Linux is built from the ground up as a time-sharing, multi-user system typically accessed from across a network. Tools like RSH, SSH, telnet, etc have been in the toolkit for ages and are a big part of clusters. To get Windows to do similar things isn't as easy and frequently you have to turn to third party tools and/or home-grown solutions to just the job startup issues. In many cases, job startup can be an issue just because of the nature of the Windows OS (but it certainly can be done and done well).
- As far as performance of Windows clusters vs. Linux clusters, it's six to one, half-dozen to the other. Compiler technology on the Windows boxes typically make good code. One of the main traps that folks fall into when porting Unix/Linux libraries/code onto Windows is that they use the POSIX compatibility layers to do things like sockets, instead of using higher performance native calls, then complain about the performance of their stuff. When we implemented our libraries using the native libraries on Windows, we saw good performance (same neighborhood as Linux, sometimes better, sometimes worse, but basically the same overall) on the same hardware (dual booting).
When it came down to it, even Unix/Linux clustering process startup using rsh/etc was junk and scaled horribly. In the end, we had to develop custom startups for Unix/Linux as well as for Windows for larger clusters.
In the end, though, Unix/Linux tends to be easier to use in those environments because of its history of being a time-sharing system (which derived the requirements to be easy to access over the network, to be able to switch between users' credentials easy, and the whole sharing of resources such as transparent graphics displays across a net using X) which still is the model in which the HPC centers run because it suits them well. This doesn't mean that Windows clusters aren't viable or desirable in some places or are even inefficient or slow, because they don't have to be any of those if you have people doing the right things to achieve the end goal.
No, it does make a difference; for what you pay Microsoft for the product licensing alone, you can purchase the 'product' (which is more-or-less free) and awesome support from IBM.
Have you ever purchased anything "enterprise-like" from IBM and also payed for the support? or are you just spouting what you "think" it is like?
What city is your local CompUSA in? It's kind of silly because other than a box I bought years ago simply to try to show some public, local support for Linux. If I were to buy a distro now I'd buy it directly from the distro provider and download everything. I call BS to your post.
4 - Invented outside of MS? C'mon Tom, you should know better!:)
Funny... almost all you ever hear on/. is that Microsoft steals all their technology from poor helpless small companies before either buying them or crushing them into oblivion and sacrificing their babies to pagan gods in bloody rituals. According to that, everything they have would be invited outside of MS;)
Slashdot Mirror
There are X number of data/address pins coming out of the chip. If they stay with the S940 or S939, that number of pins is exactly equal to the number of pins required for one CPU. Adding another core without adding another X pins means that the two cores will share the X pins to the memory. So, unless the memory controller currently has 2X the bandwidth required for a single Opteron to run, then you will have a bottleneck accessing that one bus off the chip in memory intensive applications as the bandwidth available off-chip will be shared by two CPUs instead of one.
I haven't had my GUI or IE on Windows take my whole machine down since around 1999.
Funny thing... Mandrake 10 for AMD64 and SuSE 9.1 for AMD64 are both comparable to the cost of Windows XP. So not only are the system requirements becoming the same or similar, so is the price.
I guess folks realize to do all the bells and whistles, you might actually have to write code. KDE/Gnome are falling to the same issues that folks complain so much about Windows... features bloat.
That'll even make it faster! The other thing they could do is put purposefully loud and angry bee sounding fans in the case, to get the full ricer effect. And a big ass wing on the top of the case.
The G5s are a lot more efficient than G4s and the Pentium4s are much LESS efficient than PentiumIII. ... that run at 3X the clock speed as the P3 and thus, give an overall performance increase, especially when coupled with much higher bandwidth memory and the like. I have yet to see a benchmark where any P3 will beat a P4 2.4GHz or faster, and the 3GHz P4s can be more than 3X as fast as a 1GHz P3 on some tasks.
Good one :) wish I was moderating :)
Well... unless you are the sysadmin and can install whatever you want (or you can somehow otherwise dictate policy), there is usually a dance you have to do in order to get some new software installed or for you to be allowed to install some random new piece of software regardless of monitary cost. There are usually good reasons for this.
Just letting anyone download and install anything is how you get caught up in spyware and the like, regardless of the platform you use. Simply saying that because there is no monitary license cost for this software is sufficient reason to not have any "red tape" for installation of software is a mistake.
KOTOR was a fun game for the first couple of hours. I found it to be rather repetitive after that and eventually never got around to finishing it. I played what I did of it on the XBox.
Hi, I recently bought an AMD64 processor and am looking around for a distribution to use. So far, I've actually installed FC2 x86-64 and am looking at Mandrake 10.0 (we use 9.2 i586 at work and I use 9.2 at home already). FC2 is reasonably stable but I've had a few lockup issues, I think I have them solved but I will switch to Mandrake 10.0 if possible (assuming that it is reasonably like 9.2). I haven't heard anything from the SUSE folks about AMD64 support so what's it like?
Synchronization of our Internet bookmarks across all our computers
Now wouldn't THAT be nice?
Give http://www.mybookmarks.com a try.
I read the first 100 lines or so then skimmed the rest. It sounds like this guy is running Microsoft ME or something because the things he talks about were mostly during that era. I haven't seen most of the problems he mentions with XP (on many computers).
Judging by the rest of what I skimmed, I wouldn't be surprised if someone found out later that Apple (or some other company) didn't pay him to write that. The other thing to remember is that there is no better zealot that a convert, so I'll file his stuff with the other junk that comes out daily.
That is the thing... The Athlon XP and Athlon 64 are very similar clock-for-clock on x87 codes (and scalar SSE I believe). The Athlon XP 2700+ actually has a higher core frequency than the Athlon 64 3200 (the A-64 3200+ is actually 2.0 GHz, the A-XP 2700+ is 2.167 GHz). The FPUs didn't change that much from the Athlon XP to the Athlon 64. This gives the XP line an actual advantage in many FPU intensive apps that make good use of the cache.
An early one was that open-source software is anti-competitive!
Actually, it sort of is, in a way. At least... when any particular product line starts to "own" a market, as MSOffice has, you can easily bottom out the market by funding or developing an OSS "alternative" to poison that market, thereby collapsing the market to non-profitability. When the market is no longer profitable, interest will leave it and you'll be stuck with software that has no interest in keeping it "current". The only jobs associated with it then would be the equivalent of blue-collar jobs where people just install or uninstall it (not including the people who use it - because they aren't deriving their income directly from the support of the software).
Like this guy, I've worked (actually gotten paid) to develop OSS software in the past and I have also competed head-to-head against it in the commercial space. What I've found is that the majority of people who use OSS use it because it is free (as in beer) rather than because it is necessarily "good". They will settle for inferior products that do as much to hinder their work as improve it simply to save a buck. In many cases, it breeds mediocrity.
I've done that quite a bit. My Windows box doesn't slow down to a crawl before, during, or after playing those games. Of course, I'm picky about my hardware and software and don't buy or use crap hardware so stability isn't a problem. I have 3 Windows XP boxes that run 24/7 at home and only reboot when patches require it. The same can be said for my Linux boxes.
Another big reason is familiarity with Unix and Unix-alike systems already. Most (all?) HPC centers in the past have been Unix or Unix-alike based so they stick with what they know (both the IT group and the programmers/scientists) and have the personnel to manage.
Nah, I'm saying if they want to change it, they have to change away from our terminology. KiloByte remains 1024 bytes. If they don't like our "kilo" being 1024 and want something to mean 1000, then they should use the Kibi to mean 1000, then they should also not use the term "byte" either, as it is a technical reference, so they should replace that with something else, thus, KiloByte = 1024 bytes, KibiFloople = 1000 Flooples.
As to the other comment, the tech language we use was derived in the context of our field. It'd be like a bunch of novices coming in and completely changing the jargon of the plumbing field or medicine based on their uninformed preconceptions. "That's not a crescent wrench! It looks nothing like a crescent. Let's call it a Variable Gap Bolt Loosener and require everyone else in the world to do the same."
Also... Kibi and Mebi are just very unprofessional sounding, like they belong in some Pokemon cartoon. I wonder how many person-hours in committee were required to come up with those terms. I know they preserve the K and the M but this is rediculous. As for me, I will refuse to use "Kibi" and "Mebi".
KB = 1024 bytes
MB = (1024 * 1024) bytes
I will not change the precise language I have been using for 25 years because people outside of my field has problems with field specific math.
If they want to change the terms, then they should not be using byte either. Then they can use KibiFloople to mean 1000 Flooples and MebiFloople to be 1,000,000 Flooples and leave our field specific terminology alone.
Thanks, I'm fixing to move up to that area anyways
:)
Spoken like a true Southerner (I am one too, btw)
Yeah... and because you have no understanding of combinatorial mathmatics, you don't realize the horrible performance and scalability issues that you will be getting because you don't understand your problem or how the tools you are using work. Bad indices, poor caching, crappy queries, and a host of other things cause the project to be a dog and the customer and your boss aren't happy with what you did.
Most code monkeys think along the lines of: if you can't get it to work right, you need a bigger hammer. I've seen a ton of them like you. While having a CS degree doesn't automatically mean that you know what you are doing (I've seen plenty that don't), the odds are much higher that someone with a degree will be more disciplined than someone who doesn't have a degree. Plenty of folks take the stance: "I don't have a degree and look how great I am" and somehow think that they are the norm when, in fact, they are the exception (or just deluded because they've never been around *really* good programmers, almost all of which have higher eduction) when they are really just the big fish in a small pond.
Actually, they have problems even today.
I had a GeForce4Ti4600 in my machine and it died so I replaced it with an ATi 9600 XT. Since then, there are many things that just don't "work right" as they did before. Specifically these:
When playing EverQuest, I can no longer go into windowed mode and then back to full-screen mode. I get to windowed mode fine, but if I try to go back to full screen all I get is a black screen.
Also, when I leave some games now, the signal gets "lost" somehow and my monitor goes black and it puts up a box saying "unsupported video mode" on it until I reset the monitor. This did not happen ever with the nVidia card.
Installing new drivers:
For nVidia:
Step 1: download new drivers.
Step 2: double click on it and go through setup.
Step 3: reboot and all is nice.
For ATi:
Step 1: download new drivers.
Step 2: uninstall old drivers.
Step 3: reboot into VGA mode
Step 4: install new drivers.
Step 5: reboot
Step 6: screen is black, something is wrong
Step 7: reboot into safe mode
Step 8: reinstall drivers again
Repeat Steps 5 through 8 about 6 more times for about 30 minutes and eventually get a "good" install and finally get to use my computer again.
Is there any point in getting one of these cards for any reason other than playing the latest games?
:)
nVidia cards tend to have good OpenGL support and OpenGL is used by a number of "high end" CAD and rendering packages. These cards will work well for folks who don't want to spend the $1500 for the high end CAD cards which are almost the same thing (there are some differences but these will do well on a smaller budget, though $500 for a card is pretty pricey to me
The drug induced haze of visiting Xerox PARC and seeing their GUI works?
why high performance computing centers which need reliability and good cost/benefit rations never use Windows.
Having worked in the HPC business for a while (as a developer and vendor of MPI libraries on about 20 different platforms) and having both Unix/Linux and Windows solutions for HPC (as well as a number of embedded systems), this was seldom the "real" reason we saw one chosen over the other. Costs derived from things like having to hire a bunch of new staff to support Windows if they got a Windows cluster when currently they have no Windows boxes in-house were some of the more "real" reasons.
Reasons to choose Unix/Linux over Windows in HPC:
- HPC centers are, and have been, historically Unix/Linux houses so the conclusions from this are pretty obvious - stick with what you know and already have support for.
- HPC people are familiar with Unix/Linux, have used them for years, already have the machines in place, and don't want to have to learn something new or try to figure it out.
- Unix/Linux houses have a built-in distrust/dislike for Windows for whatever reasons (bunches, some of which are simply FUD and some are real).
As far as development of HPC software, these were things I saw:
- Unix/Linux is built from the ground up as a time-sharing, multi-user system typically accessed from across a network. Tools like RSH, SSH, telnet, etc have been in the toolkit for ages and are a big part of clusters. To get Windows to do similar things isn't as easy and frequently you have to turn to third party tools and/or home-grown solutions to just the job startup issues. In many cases, job startup can be an issue just because of the nature of the Windows OS (but it certainly can be done and done well).
- As far as performance of Windows clusters vs. Linux clusters, it's six to one, half-dozen to the other. Compiler technology on the Windows boxes typically make good code. One of the main traps that folks fall into when porting Unix/Linux libraries/code onto Windows is that they use the POSIX compatibility layers to do things like sockets, instead of using higher performance native calls, then complain about the performance of their stuff. When we implemented our libraries using the native libraries on Windows, we saw good performance (same neighborhood as Linux, sometimes better, sometimes worse, but basically the same overall) on the same hardware (dual booting).
When it came down to it, even Unix/Linux clustering process startup using rsh/etc was junk and scaled horribly. In the end, we had to develop custom startups for Unix/Linux as well as for Windows for larger clusters.
In the end, though, Unix/Linux tends to be easier to use in those environments because of its history of being a time-sharing system (which derived the requirements to be easy to access over the network, to be able to switch between users' credentials easy, and the whole sharing of resources such as transparent graphics displays across a net using X) which still is the model in which the HPC centers run because it suits them well. This doesn't mean that Windows clusters aren't viable or desirable in some places or are even inefficient or slow, because they don't have to be any of those if you have people doing the right things to achieve the end goal.
No, it does make a difference; for what you pay Microsoft for the product licensing alone, you can purchase the 'product' (which is more-or-less free) and awesome support from IBM.
Have you ever purchased anything "enterprise-like" from IBM and also payed for the support? or are you just spouting what you "think" it is like?
What city is your local CompUSA in? It's kind of silly because other than a box I bought years ago simply to try to show some public, local support for Linux. If I were to buy a distro now I'd buy it directly from the distro provider and download everything. I call BS to your post.
4 - Invented outside of MS? C'mon Tom, you should know better! :)
/. is that Microsoft steals all their technology from poor helpless small companies before either buying them or crushing them into oblivion and sacrificing their babies to pagan gods in bloody rituals. According to that, everything they have would be invited outside of MS ;)
Funny... almost all you ever hear on