Besides costing more (due to required extra/wider ram chips), ECC ram is slower.
This is primary caused by extra read/modify/write cycles done by the memory controller to keep the ECC in sync for short writes. This RMW sequence can cause a fair amount of performance loss (i've seen 8% on a custom application, doing a lot of pointer chasing/updating).
Furthermore, as anyone who monitors a lot of servers with ECC will attest. Its really rare to see a soft ECC correction (I've personally never seen one). If there are bit errors being corrected/detected its always been a full blow hardware failure.
if it were me, I would detect the amount of physical memory available, allocate 80% of it, lock it into physical memory and manage the paging myself.
Are you serious? You can honestly say that your monitoring each pages hit counts and swapping out only the pages with the lowest usage, and your not affecting anything else on the machine? What a load of baloney, just map the whole thing (like the author of the article is talking about) and let the OS paging algorithms make global decisions about the amount of ram and the consumption of the rest of the machine. That way when your on a machine with 64G of ram, your not leaving 12G of it unused, or when your on a 128M machine, crashing because you cannot get 80%.
It could be the war on science too... Wake me up when I can buy some chemicals for my chemistry set , or even for that matter some glassware (http://www.crscientific.com/texas-glassware.html).
copyright BS & stupid car analogy, RMS's argum
on
Time To Dump XP?
·
· Score: 1
Ok, so my truck is a 1999. It runs fine, gets 22MPG. The most expensive thing i've replaced on it are the tires (3rd pair). Yet, I can still go to the Toyota dealer, get parts, have it fixed etc. I expect that this will be the case for another 10 years, and then parts will get harder to find. Same with my house (25 years) and nearly everything I own outside of mass produced consumer electronics trash (much of which is also fairly old).
In the past upgrading a PC provided a real advantage. But today, for a very large percentage of the population there isn't a real need to upgrade from XP, especially if they are running in a limited user account (killing the security argument in vista/7). In just about every other industry, it would be possible to just keep using what is working, and if something fails call in the repair man at some outrageous rate, that's cheaper than buying new.
But instead, microsoft holds the reigns of the OS and forces people to upgrade by cutting off the ability to fix problems. How do they hold these reigns? By hiding behind copyright, and enforcing strict licenses. Instead of advancing the public good, they effectively kill the usefulness of a product 80+ years before it goes into the public domain.
Hence this discussion, XP could be used for another ten years, in fact with just slight changes it probably could last another 20. The fundamental state of PC technology hasn't changed, and a stable driver model (aka microkernel style plugins) allows the product to be extended to use technologies that weren't even on the drawing boards in 2001.
This is why Linux saddens me so much. I'm firmly in the open source camp, and long term (at this rate 50+ years) I don't believe that the closed source vendors have a chance against it. But, I'm not sure its going to happen before i'm dead. The politics and stupid decisions can be recovered from but its just going to take longer. The day that the kernel developers get together and conclude that they core kernel is "done", because the all the refactoring and renaming isn't improving the product in any meaningful way is the day that linux starts to take over. That doesn't mean it can't get better, it just means that upgrading the block scheduler can be done from an install disk and it doesn't require replacing the VMM, or reinstalling all my applications.
Basically I want an OS, that is user customizable/extenable without fscking everything up. I want to be able to put bigger tires on my OS without having to buy a new car. That requires an interface that is standard enough that someone can write a block scheduler without having to worry that some change next week is going to render it unable to function in the latest version of the OS.
And you will never get a ticket for failing to blink, but I just did... Turns out the police in Austin don't have the manpower to patrol neighborhoods where 2/3'rds of the houses have been broken into. They do have the manpower to post cops on a highway, cause a traffic jam, and then pull people over doing.01 miles (per the ticket) in a 60 mile/hour zone for failing to signal sufficiently when changing lanes.
Buy more memory, and put your swap cache on it. There are a number of ramdisks that use the memory above 4G on 32-bit M$ OS's. With a little google searching you can find a few that work well and are free. Sure it won't act as disk cache but it works fantastically for running applications that are willing to consume multiple GB of disk (aka a whole bunch of VMWARE sessions).
Or call M$ and bitch about them restricting the desktop users to 3G of memory, while the 32-bit server versions can access 64.
I swore off buying another blizzard product, when the WC3 expansion pack refused to install on my machine because it couldn't validate that I had a legitimate WC3. Their customer service was downright useless. In the end, after wasting probably 6 hours, I discovered that the initial WC3 install, had been done under another user and the expansion pack didn't have access to some DRM bullshit in that users account. This, even though I had been playing the base version without a problem in my personal account. The whole time I was thinking, "Why did I give them money for this, when I could have just gotten the hacked version, that xxx is running just fine." Plus, debugging DRM failures is a TOTAL PITA, as it just refuses to install, rather than being helpful and saying "registry key xxxx" is not correct, or something useful.
So, while I own multiple legal copies of WC->WC3+expansion packs, SC+expansion. I won't be buying SC2. I've got such a backlog of games to finish there isn't any reason to give them money for some restrictive crap I won't be able to play in 10 years with my friends.
If you do computationally-intensive workstation tasks, like video editing...
I don't know about you, but my video editor, doesn't consume much CPU time. HD speed (or a sh*tload of RAM) makes a much larger diffrence. On the other hand, when I'm done with the editing and I tell it to create the video it can spin for a long time. Generally though, I batch those up and let them run when I'm not at the computer. In that case having them finish at midnight instead of 3 AM, really doesn't matter much.
But, compiling C++ template code on the other hand, that just EATS CPU like no ones business.
They should indicate the number of cleaning cycles you can get out of a given cartridge. I gave up on inkjets a few years ago (color laser for everything except the best photos, which get printed at a retail outlet (its cheaper)). That's because invariably when I went to print a photo one or more of the nozzles would be clogged forcing me to run the cleaning cycle three or four times. By then I would need a new cartridge. I'm betting that most full cleaning cycles use about 1/10th of the total ink in the cartridge. I've spent maybe $80 on toner in 10 years of running laser printers at home. The photo's printed on the inkjet probably cost me close to $10 each.
Most people just don't seem to notice the quality issues when one or two nozzles are clogged (or they just assume thats the way it works). For those people (my wife) the laser prints on glossy paper are just as good. I on the other-hand want the image to be perfectly clean, and for that I use the little online print function my local photoprocessor has. Those prints cost me ~$2 for a 8x10, and the quality is usually the best available as they upgrade their printers every year or so, and they keep them in tiptop shape. For the dozen or so photo quality prints its a no brainer.
We've advanced more technologically in 100 years than Europe advanced in 1000, and it's largely because of patents.
Which is total bs, pretty much everyone agrees that the rate of progress is mostly due to our ability to dedicate more resources to any given problem. Just as our population has grown, so has our ability to dedicate people into narrower and narrower areas of research and development. Sure being able to leverage previous work is a huge advantage, but that knowledge gets spread around via methods other than patents too. Ever read a research journal?
Prediction: "The wireless networks of the future will be faster, but unless there is a major breakthrough, wired networks will have a far greater bandwidth. Mobile devices will be able to send and receive messages, but it will be expensive and unusual to use them to receive an individual video stream."
Sounds about spot on, especially if you consider HD video. Sure wireless is getting better but so are wired networks. I get 30+Mbit on my cable modem, and 10Gbit on my LAN. I can stream full 1080p HD quality compressed content over the internet without a second thought. I haven't seen to many wireless (ignoring 802.11g/n) networks capable of that, in fact its hard to stream any kind of video on any of the phone networks with any reliability.
The problem usually was the GUS's SB emulation. The best bet was simply to disable it, use the GUS in GUS mode for things that support it and use the SB for everything else. I had both cards too, I don't remember anything particularly problematic about getting them both to work.
I have been a little surprised to see the price on a computer go up. I blame it mostly on the big computer manufactures getting into the game and not being able to make a profit on a machine that asus/msi/etc are willing to sell for just a few dollars profit. The key is just to ignore the name brand netbook manufactures, I picked up a MSI L1350, running win7 starter this weekend with a cheezy inkjet for $200 (frys). The "no-name" ones regularly go on sale for less than $230.
My point is that it exists everywhere (cause its a real problem), but that a fair number of people (including apparently yourself) never notice it because the delay is generally insignificant for a casual user with only one tab or two tabs doing basic surfing. People with a lot of tabs or heavy web apps (lots of java-script) tend to notice it more.
I have the same problem in FF, everyone does. Its a question of what is running in those other tabs. I found the problem because one app in particular made it really noticeable. What appears to be happening is every time a javascript timer fires in one of those other tabs FF is off servicing it leaving your UI dead. Normally you wouldn't notice a little glitch like that but if you have a web app that spends a significant amount of time doing ajax calls in timers, or you have a shed-load of tabs open the browser will seem to pause. Youtube/flash and friends also get paused by it, causing video glitching.
If you have a Phenom II black edition, you can have functionality similar to Intel's turbo feature using this (http://sourceforge.net/projects/clockboost/) little utility which monitors core temp, and overclocks the processor if the temp is below a set threshold.
You need to divide those by three, which places the upper bound for a display at about 4-16 megapixel. Anything beyond that is not going to be useful.
Assuming you never move your eyes.. Which isn't true. Your field of vision is fairly narrow, so you want ~6Mp in an area say 8x8" on a monitor 2.5 feet away. Then you want a lot of area, because humans have spatial memory. I'm sitting in front of 4 1600x1200 monitors right now. I find having different documents (source files) on each monitor is far more useful than alt-tabbing between two things i'm looking at. My eyes can switch and locate multiple lines on diffrent monitors with ease. It doesn't require any thought.
I saw it happening (the transition to HD) and snarfed up a couple of the last affordable PVA 1920x1200 monitors. You can still get them from online computer vendors, you just have to pay real money. Its also possible to get higher resolution monitors but you had better have deep pockets.
The majority of children need that repetition to even recall how to do basic addition, subtraction.
That is such BS its not even funny. I have a 3 year old, in a class full of 3s and 4s that can do addition and subtraction without any problem. Her cousin is in 1st grade and can add, subtract, multiply and divide multiple digit numbers no problem. Her father, as an experiment decided to see if he could teach some of the young neighbor kids (4-7) how to divide small numbers using the same method he taught his daughter. He said it took about ten minutes each, and a quart of ice cream (to initially catch their attention), and they were running home showing their parents how they could divide using "lines". Of course it takes them a little while, and they do a lot of counting but they get the right answer and they can compute how to divide things between kids using paper and pencil. When a young kid "gets" something its stuck, they won't forget it. When you try to teach via rout memorization, of course they forget it, so do adults. Make it a game and its stuck forever. The problem with math and small children is that we try to start showing them the "shortcuts" long before they are ready. Forcing them to add two four digit numbers before they totally grasp the concept of a thousand is a sure fire way to bore them.
It's an Intel running windows/linux to host a c64 emulator at 500Mhz. That's because its cheaper to engineer a 500mhz C64 using an x86 and an emulator than design one from scratch. Chuckle...
None of the modern IDE's let you "zero" the disk, or even fully format it. Without the embedded servo tracks the head can't even know if its staying on track.
Well, most of the raid controllers out there can't even keep up with the throughput of a half dozen modern magnetic drives either. Our application eats bandwidth, and its been a real struggle to find controllers that can sustain higher transfer rates. Getting much more than 1GB/sec out of a RAID box is pretty much impossible no matter the cost. In the end we use fairly low end RAID disks ganged together via high end FC switches and let our software do the stripping across them. AKA we gang a bunch of 600MB/sec RAID arrays together to get multiple GB/sec.
Slashdot Mirror
Besides costing more (due to required extra/wider ram chips), ECC ram is slower.
This is primary caused by extra read/modify/write cycles done by the memory controller to keep the ECC in sync for short writes. This RMW sequence can cause a fair amount of performance loss (i've seen 8% on a custom application, doing a lot of pointer chasing/updating).
Furthermore, as anyone who monitors a lot of servers with ECC will attest. Its really rare to see a soft ECC correction (I've personally never seen one). If there are bit errors being corrected/detected its always been a full blow hardware failure.
if it were me, I would detect the amount of physical memory available, allocate 80% of it, lock it into physical memory and manage the paging myself.
Are you serious? You can honestly say that your monitoring each pages hit counts and swapping out only the pages with the lowest usage, and your not affecting anything else on the machine? What a load of baloney, just map the whole thing (like the author of the article is talking about) and let the OS paging algorithms make global decisions about the amount of ram and the consumption of the rest of the machine. That way when your on a machine with 64G of ram, your not leaving 12G of it unused, or when your on a 128M machine, crashing because you cannot get 80%.
It could be the war on science too... Wake me up when I can buy some chemicals for my chemistry set , or even for that matter some glassware (http://www.crscientific.com/texas-glassware.html).
Ok, so my truck is a 1999. It runs fine, gets 22MPG. The most expensive thing i've replaced on it are the tires (3rd pair). Yet, I can still go to the Toyota dealer, get parts, have it fixed etc. I expect that this will be the case for another 10 years, and then parts will get harder to find. Same with my house (25 years) and nearly everything I own outside of mass produced consumer electronics trash (much of which is also fairly old).
In the past upgrading a PC provided a real advantage. But today, for a very large percentage of the population there isn't a real need to upgrade from XP, especially if they are running in a limited user account (killing the security argument in vista/7). In just about every other industry, it would be possible to just keep using what is working, and if something fails call in the repair man at some outrageous rate, that's cheaper than buying new.
But instead, microsoft holds the reigns of the OS and forces people to upgrade by cutting off the ability to fix problems. How do they hold these reigns? By hiding behind copyright, and enforcing strict licenses. Instead of advancing the public good, they effectively kill the usefulness of a product 80+ years before it goes into the public domain.
Hence this discussion, XP could be used for another ten years, in fact with just slight changes it probably could last another 20. The fundamental state of PC technology hasn't changed, and a stable driver model (aka microkernel style plugins) allows the product to be extended to use technologies that weren't even on the drawing boards in 2001.
This is why Linux saddens me so much. I'm firmly in the open source camp, and long term (at this rate 50+ years) I don't believe that the closed source vendors have a chance against it. But, I'm not sure its going to happen before i'm dead. The politics and stupid decisions can be recovered from but its just going to take longer. The day that the kernel developers get together and conclude that they core kernel is "done", because the all the refactoring and renaming isn't improving the product in any meaningful way is the day that linux starts to take over. That doesn't mean it can't get better, it just means that upgrading the block scheduler can be done from an install disk and it doesn't require replacing the VMM, or reinstalling all my applications.
Basically I want an OS, that is user customizable/extenable without fscking everything up. I want to be able to put bigger tires on my OS without having to buy a new car. That requires an interface that is standard enough that someone can write a block scheduler without having to worry that some change next week is going to render it unable to function in the latest version of the OS.
XP will use multiple rows if you resize the task bar.
And you will never get a ticket for failing to blink, but I just did... Turns out the police in Austin don't have the manpower to patrol neighborhoods where 2/3'rds of the houses have been broken into. They do have the manpower to post cops on a highway, cause a traffic jam, and then pull people over doing .01 miles (per the ticket) in a 60 mile/hour zone for failing to signal sufficiently when changing lanes.
Or it could mean that you have designed a system that requires a human in the loop to be reusable. Which is pretty lame too..
Buy more memory, and put your swap cache on it. There are a number of ramdisks that use the memory above 4G on 32-bit M$ OS's. With a little google searching you can find a few that work well and are free. Sure it won't act as disk cache but it works fantastically for running applications that are willing to consume multiple GB of disk (aka a whole bunch of VMWARE sessions).
Or call M$ and bitch about them restricting the desktop users to 3G of memory, while the 32-bit server versions can access 64.
I swore off buying another blizzard product, when the WC3 expansion pack refused to install on my machine because it couldn't validate that I had a legitimate WC3. Their customer service was downright useless. In the end, after wasting probably 6 hours, I discovered that the initial WC3 install, had been done under another user and the expansion pack didn't have access to some DRM bullshit in that users account. This, even though I had been playing the base version without a problem in my personal account. The whole time I was thinking, "Why did I give them money for this, when I could have just gotten the hacked version, that xxx is running just fine." Plus, debugging DRM failures is a TOTAL PITA, as it just refuses to install, rather than being helpful and saying "registry key xxxx" is not correct, or something useful.
So, while I own multiple legal copies of WC->WC3+expansion packs, SC+expansion. I won't be buying SC2. I've got such a backlog of games to finish there isn't any reason to give them money for some restrictive crap I won't be able to play in 10 years with my friends.
If you do computationally-intensive workstation tasks, like video editing...
I don't know about you, but my video editor, doesn't consume much CPU time. HD speed (or a sh*tload of RAM) makes a much larger diffrence. On the other hand, when I'm done with the editing and I tell it to create the video it can spin for a long time. Generally though, I batch those up and let them run when I'm not at the computer. In that case having them finish at midnight instead of 3 AM, really doesn't matter much.
But, compiling C++ template code on the other hand, that just EATS CPU like no ones business.
They should indicate the number of cleaning cycles you can get out of a given cartridge. I gave up on inkjets a few years ago (color laser for everything except the best photos, which get printed at a retail outlet (its cheaper)). That's because invariably when I went to print a photo one or more of the nozzles would be clogged forcing me to run the cleaning cycle three or four times. By then I would need a new cartridge. I'm betting that most full cleaning cycles use about 1/10th of the total ink in the cartridge. I've spent maybe $80 on toner in 10 years of running laser printers at home. The photo's printed on the inkjet probably cost me close to $10 each.
Most people just don't seem to notice the quality issues when one or two nozzles are clogged (or they just assume thats the way it works). For those people (my wife) the laser prints on glossy paper are just as good. I on the other-hand want the image to be perfectly clean, and for that I use the little online print function my local photoprocessor has. Those prints cost me ~$2 for a 8x10, and the quality is usually the best available as they upgrade their printers every year or so, and they keep them in tiptop shape. For the dozen or so photo quality prints its a no brainer.
We've advanced more technologically in 100 years than Europe advanced in 1000, and it's largely because of patents.
Which is total bs, pretty much everyone agrees that the rate of progress is mostly due to our ability to dedicate more resources to any given problem. Just as our population has grown, so has our ability to dedicate people into narrower and narrower areas of research and development. Sure being able to leverage previous work is a huge advantage, but that knowledge gets spread around via methods other than patents too. Ever read a research journal?
Wireless Networks
Prediction: "The wireless networks of the future will be faster, but unless there is a major breakthrough, wired networks will have a far greater bandwidth. Mobile devices will be able to send and receive messages, but it will be expensive and unusual to use them to receive an individual video stream."
Sounds about spot on, especially if you consider HD video. Sure wireless is getting better but so are wired networks. I get 30+Mbit on my cable modem, and 10Gbit on my LAN. I can stream full 1080p HD quality compressed content over the internet without a second thought. I haven't seen to many wireless (ignoring 802.11g/n) networks capable of that, in fact its hard to stream any kind of video on any of the phone networks with any reliability.
The problem usually was the GUS's SB emulation. The best bet was simply to disable it, use the GUS in GUS mode for things that support it and use the SB for everything else. I had both cards too, I don't remember anything particularly problematic about getting them both to work.
I have been a little surprised to see the price on a computer go up. I blame it mostly on the big computer manufactures getting into the game and not being able to make a profit on a machine that asus/msi/etc are willing to sell for just a few dollars profit. The key is just to ignore the name brand netbook manufactures, I picked up a MSI L1350, running win7 starter this weekend with a cheezy inkjet for $200 (frys). The "no-name" ones regularly go on sale for less than $230.
My point is that it exists everywhere (cause its a real problem), but that a fair number of people (including apparently yourself) never notice it because the delay is generally insignificant for a casual user with only one tab or two tabs doing basic surfing. People with a lot of tabs or heavy web apps (lots of java-script) tend to notice it more.
I have the same problem in FF, everyone does. Its a question of what is running in those other tabs. I found the problem because one app in particular made it really noticeable. What appears to be happening is every time a javascript timer fires in one of those other tabs FF is off servicing it leaving your UI dead. Normally you wouldn't notice a little glitch like that but if you have a web app that spends a significant amount of time doing ajax calls in timers, or you have a shed-load of tabs open the browser will seem to pause. Youtube/flash and friends also get paused by it, causing video glitching.
If you have a Phenom II black edition, you can have functionality similar to Intel's turbo feature using this (http://sourceforge.net/projects/clockboost/) little utility which monitors core temp, and overclocks the processor if the temp is below a set threshold.
Hmmm, I guess you didn't notice that the "hardened" Linux firewalls are the ones getting exploited.
You need to divide those by three, which places the upper bound for a display at about 4-16 megapixel. Anything beyond that is not going to be useful.
Assuming you never move your eyes.. Which isn't true. Your field of vision is fairly narrow, so you want ~6Mp in an area say 8x8" on a monitor 2.5 feet away. Then you want a lot of area, because humans have spatial memory. I'm sitting in front of 4 1600x1200 monitors right now. I find having different documents (source files) on each monitor is far more useful than alt-tabbing between two things i'm looking at. My eyes can switch and locate multiple lines on diffrent monitors with ease. It doesn't require any thought.
I saw it happening (the transition to HD) and snarfed up a couple of the last affordable PVA 1920x1200 monitors. You can still get them from online computer vendors, you just have to pay real money. Its also possible to get higher resolution monitors but you had better have deep pockets.
The majority of children need that repetition to even recall how to do basic addition, subtraction.
That is such BS its not even funny. I have a 3 year old, in a class full of 3s and 4s that can do addition and subtraction without any problem. Her cousin is in 1st grade and can add, subtract, multiply and divide multiple digit numbers no problem. Her father, as an experiment decided to see if he could teach some of the young neighbor kids (4-7) how to divide small numbers using the same method he taught his daughter. He said it took about ten minutes each, and a quart of ice cream (to initially catch their attention), and they were running home showing their parents how they could divide using "lines". Of course it takes them a little while, and they do a lot of counting but they get the right answer and they can compute how to divide things between kids using paper and pencil. When a young kid "gets" something its stuck, they won't forget it. When you try to teach via rout memorization, of course they forget it, so do adults. Make it a game and its stuck forever. The problem with math and small children is that we try to start showing them the "shortcuts" long before they are ready. Forcing them to add two four digit numbers before they totally grasp the concept of a thousand is a sure fire way to bore them.
It's an Intel running windows/linux to host a c64 emulator at 500Mhz. That's because its cheaper to engineer a 500mhz C64 using an x86 and an emulator than design one from scratch. Chuckle...
None of the modern IDE's let you "zero" the disk, or even fully format it. Without the embedded servo tracks the head can't even know if its staying on track.
Well, most of the raid controllers out there can't even keep up with the throughput of a half dozen modern magnetic drives either. Our application eats bandwidth, and its been a real struggle to find controllers that can sustain higher transfer rates. Getting much more than 1GB/sec out of a RAID box is pretty much impossible no matter the cost. In the end we use fairly low end RAID disks ganged together via high end FC switches and let our software do the stripping across them. AKA we gang a bunch of 600MB/sec RAID arrays together to get multiple GB/sec.