And don't forget the Kin (Microsoft's first forary into building phones themselves), which got dropped pretty quickly.
First cellular phone, you mean. I had the first phone made by Microsoft and it was lack of support that killed it, too.
The answering features on it were phenomenal for the day (almost Asterix level), and unlike many answering machines at that time, you had essentially unlimited recording time (even a 20GB hard disk with 32Kbps audio files is a lot) for incoming and outgoing messages.
Please demonstrate how I'm wrong. You just became a personal project of mine.
The problem was that the data released in the FOIA response had personally-identifiable information (the mediallion number) replaced with something that could be used to re-generate the PII without any information that isn't public.
The GP's scheme was to replace the PII with a number that cannot be used to re-generate the PII with just the FOIA response. The PII could be re-generated if you had some kind of extra knowledge (e.g., the mapping used, or knowledge of when a particular cab was at a particular location), but this is still the best you can do with this sort of information release.
The quality of analog vinyl vs digital format audio is hotly debated, and vinyl has a strong following among audiophiles.
There actually isn't any debate over properly mastered vinyl vs. properly mastered digital...digital wins every time in a double blind test for accurate reproduction. For people who like vinyl (and analog amplifiers) because the sound is "warmer", etc., that's fine for their personal taste, but it's not an accurate reproduction.
In the same vein, with many current digital recordings having extremely limited dynamic range, vinyl of the same music with correct range is a much more accurate representation of the original music. This is one of the big reasons that vinyl is having a resurgence among people who want accurate reproduction of their music.
Frame rate is for gamers. Programmers need pixels.
The mouse lag on a 24 or 30Hz display will drive you nuts when you are trying to select a block of text.
If you are a keyboard-only editor, it's not as bad, but even highlighting text or trying to page down quickly will likely send you back to a high-speed multi-monitor setup.
You should use ECC RAM if you are doing any form of disk IO no matter which file system you're using, or you are under the risk of data loss.
I agree. Unfortunately, no Intel desktop CPU/chipset supports ECC, and many AMD desktop chipsets are castrated by the board manufacturer to not allow ECC.
When RAM was slow and 4GB was huge and expensive, this wasn't as big a deal, but now that 8GB is the reasonable starting point and 32GB is quite affordable, Intel especially needs to step up and add ECC support to their desktop CPUs/chipsets.
If you consider an IP address, a port number, a timestamp and an account number to be insanely detailed then I can't wait to see what you're going to say when you discover all the information Facebook, Google and others keep about you!
Google, et.al., keep their data because they can monetize it. Comcast will have to keep this data despite the fact that it not only won't make them money, but will cost them money since they will have to have people to search it for the legal requests.
Plus, I can generate thousands of connections per second and Comcast will have to log them all. If my both my neighbor and I have Comcast, I might just set up a box to do nothing but flood his WiFi with connections, just to cost Comcast money.
Why didn't you just refer to the LHC web page and imply that you are writing at that same data rate to a single SSD...it would have exactly the same value as an argument.
btw, If you're complaint about the Prius appearance - what's the drag coefficient of your car? Is it as good as my 10 year old Prius? 'Cause that's why it looks like it does - it's part of it's design elegance.
It's also why the Prius would get about 40mpg even if it had no hybrid features. If Toyota sold that car, they'd really corner the market, as it would have a better ROI than the hybrid Prius, and wouldn't have any risk about battery replacement (which isn't always covered by warranty).
You have much bigger problems if you have memory corruption.
If you don't use ECC memory, you will have memory corruption. Even if you do use ECC memory, you might have corruption, and it might even go unnoticed, but the odds are far less likely.
"Corruption" in this sense doesn't mean that whole DIMMs are broken...it just means that one bit has changed in a way that the user/OS/CPU didn't want it to. In many cases, this can be completely harmless (e.g., graphical data used in-memory only has one bit wrong...you might not even notice a color shift if it's the LSB), a little annoying (e.g., unexpected program termination), or very annoying (e.g., BSOD). But, if this happens in memory used for disk write buffers, then you get the issue that the GP had you Google for.
Besides it further assumes that they are not using Carrier Grade NAT which is exactly how Free, a French ISP that has been doing the same thing for years, is handling this.
Even better, as now all the WiFi users appear to come from a single IP as far as the MPAA/RIAA is concerned, which means the only way they can get more info is if Comcast keeps insanely detailed records about every one of these connections. Keeping normal accounting information won't be enough to identify a copyright infringer...Comcast would also have to keep the IP/port connection logs from the NAT device.
Cash takes considerably longer to tender than credit. The customer takes time selecting the bills and coins, the cashier takes time counting it, then enters the amount in the cash register, and after the till opens, they have to count out the customer's change.
The one assumption you make here is that the credit card user is on the ball, and swipes either before the final total is rung or immediately after. I have seen many customers stand there until the cashier tells them the total, then reach into their wallet/purse and hunt the credit card, swipe it the wrong way, finally get it right, then hit "debit" on a card that is credit only.
Granted, these same people would likely take even longer to pay with cash, but I can see why some people think that cash is faster, based on anecdotes.
They are using the wifi and completely segregating traffic. It appears with a distiinct SSID and on a different IP. The capacity is on a different channel, so gain the host user isn't affected.
It's a completely separate segment of the private IP space, but once it heads through the router, every other device on the Internet (including the MPAA/RIAA scanners) will see the exact same public IP as the customer is given.
I'm assuming that Comcast doesn't have 50,000 spare routable IP addresses, but that's not a bad assumption.
It means the iPhone won't provide its MAC address *until* it finds a recognised network to connect to - it won't be broadcasting it constantly while you are out traveling or shopping.
This problem is easily solved by not turning on WiFi unless you know you are going to connect.
For me, I use an app on Android that keeps WiFi off unless I am in a location where I have already said I want to auto-connect to a specific network. It uses cell tower IDs and GPS to determine where I am.
Super computers are EXPENSIVE. A super computer is not just a tower with 30 gigs of ram and 10 processors, this is a building full of wires and computer components.
At this point, you can get a "supercomputer" in a rack for around US$500K, and it uses about 20kW of power.
Sure, it's not going to set any records, but with 500-1000 cores and 5-10TB of RAM, it's a lot more than most users will ever see. Heck, we have 40-core systems with over 2TB of RAM that fit in 2U. Again, not a supercomputer, but certainly a lot more power than in most single systems.
Based on TFA, I suspect the "cost" of $150K was what the time on the computer might sell for if somebody outside the project wanted to use it, not actual cost of electricity used. Most of these projects (like where I work) have computer resources that are free to use for all researches directly associated with the project, but could have a charge back for others. And, as others have stated, general purpose computers are pretty bad at mining, so it might have taken a lot of CPU-hours, and that's what the accountants are looking at. Here, we do have systems that would do fairly well at bitcoin mining (8 nVidia GPUs per 2U box), but they (like the other systems) are so busy doing actual work that nobody could possibly get enough idle cycles to do anything. That's why I put "cost" in quotes...if a system that is supposed to be used for research isn't doing enough that somebody could find that much idle time, then the system was either overbuilt or under-advertised.
Those are 240-256GB drives that have all had 600TB written to them, and many haven't even started to dip into the protected sectors. That's already up to 2400 P/E cycles on drives with 3000 or less on their specs. To put it into real-world terms, that's the equivalent of completely erasing the drive every day for 6.5 years. I think they'll last the same 7 to 8 years as your hard drives when used more realistically.
Since you brought it up: if you were building a heavy-use machine for code development today, would you go with 6-or-more-core Xeon, or a newer Haswell 4-core chip and board?
To me, Xeon vs. desktop chip comes down to how much memory you are going to want. Trusting 32GB or more to not have a random memory error isn't something I want to do, so I went with a Xeon and ECC. In addition, if you want more than 32GB, that means an LGA2011 socket for Intel...none of their other sockets have chipsets that can handle more. If you go LGA2011, then your best bet is a workstation board, since you could eventually put a Xeon and 256GB of RAM on it, but could start with a desktop CPU and whatever RAM you can afford. Workstation typically means no on-board graphics and plenty of PCIe slots available.
Your use case sounds like it doesn't really need the extra CPU threads, but does need disk speed. For that, I say SSD for your boot plus a real RAID controller (preferably with battery backup) with spinning disks for bulk storage and lots of RAM for cache. The same system with the E5-2630 v2 has 64GB of RAM, a 500GB Samsung 840 EVO and five 2TB WD Red drives in RAID-5 connected to an Adaptec 6805 with battery backup. Reads and writes to the SSD and the array are at similar speeds (between 400MB/sec and 600MB/sec), and more than one app hitting the disk at the same time doesn't make things slow to a crawl.
Essentially, if your disk can't support the I/O for all those services, going from 8 to 12 CPU threads won't really buy you much. You can then add any GPU you want (I have a Radeon R9 270).
That is so if there's a fire, the firemen can get to it.
Maybe it's just different places, but if that's why you can't park in front of a hydrant, then I couldn't park in my own driveway, as there is a hydrant less than 3 feet from the left edge (in the easement section between the sidewalk and the street).
Parking my car in my driveway puts my car closer to that hydrant than parking in the street would (it's about 5 feet from the street).
3x4 Is a very good format for reading, at any rate clearly better than any more elongated format like 9x16.
Based on the scans of my book covers, most books are about 1.5:1 (height/width), and looking at the text on the page, the actual used area is very similar. Most dedicated eBook readers have similar ratios.
So, it appears that both 1.78:1 and 1.33:1 aren't quite the same as historical reading aspect ratios, and both are off by about the same amount. I would suspect that 1.78:1 would look better, though, as it's much closer to the golden ratio than 1.33:1 is, and about as different as 1.5:1 is.
Personally, that's the jump I'm really looking forward to: Intel's first 8-core desktop chip.
With Google's stats on how often ECC in their servers correct errors, I'm not sure why anyone is anxious for an 8-core chip that can't use ECC memory.
My guess is that if you need 8 cores, you need more memory, and that means you are more likely to have errors. That plus the fact that the only current Intel desktop chipset supports more than 32GB of RAM has an LGA2011 socket is why I went with a Xeon...if I'm going to pay for a more expensive motherboard and processor, I might as well pay just a little more and get ECC protection on the memory. Also, the Xeon boards support 256GB instead of the 64GB that the desktop board support.
Although, to be fair, Intel's CPU "turbo" is probably a bit more of a car analogy because it's a temporary speed boost, not a permanent one.
For Ivy Bridge chips with decent cooling, it's basically permanent. Every Ivy Bridge chip has the base clock speed (the number on the box), the "Turbo" speed (which is available for a single core only), and the "multi-core turbo" that isn't advertised very well, but sites like CPUBenchmark will give you the details.
For example, I have an E5-2630 (yes, it's a Xeon, but there are fairly equivalent desktop chips) with a base clock of 2.6GHz, a single-core turbo of 3.1GHz, and a multi-core turbo of 2.9GHz. Since the chip stays cool enough, it runs at least at 2.9GHz whenever it needs speed.
Except only those making less than $15/hr will see that increase. As someone who earns more than minimum wage it won't benefit me, and in fact will take away much of my buying power.
It will especially hurt those currently making less than about 1.5x of the new minimum wage. People making more than that should be able to absorb the price increases without too much pain, but those in the 1x to 1.5x range will not get the 20-30% raise they need to make up the difference.
Slashdot Mirror
And don't forget the Kin (Microsoft's first forary into building phones themselves), which got dropped pretty quickly.
First cellular phone, you mean. I had the first phone made by Microsoft and it was lack of support that killed it, too.
The answering features on it were phenomenal for the day (almost Asterix level), and unlike many answering machines at that time, you had essentially unlimited recording time (even a 20GB hard disk with 32Kbps audio files is a lot) for incoming and outgoing messages.
Useless when the "failure" is: dd if=/dev/urandom of=/dev/sda
Even /dev/zero as the source is good enough.
Please demonstrate how I'm wrong. You just became a personal project of mine.
The problem was that the data released in the FOIA response had personally-identifiable information (the mediallion number) replaced with something that could be used to re-generate the PII without any information that isn't public.
The GP's scheme was to replace the PII with a number that cannot be used to re-generate the PII with just the FOIA response. The PII could be re-generated if you had some kind of extra knowledge (e.g., the mapping used, or knowledge of when a particular cab was at a particular location), but this is still the best you can do with this sort of information release.
The quality of analog vinyl vs digital format audio is hotly debated, and vinyl has a strong following among audiophiles.
There actually isn't any debate over properly mastered vinyl vs. properly mastered digital...digital wins every time in a double blind test for accurate reproduction. For people who like vinyl (and analog amplifiers) because the sound is "warmer", etc., that's fine for their personal taste, but it's not an accurate reproduction.
In the same vein, with many current digital recordings having extremely limited dynamic range, vinyl of the same music with correct range is a much more accurate representation of the original music. This is one of the big reasons that vinyl is having a resurgence among people who want accurate reproduction of their music.
Frame rate is for gamers. Programmers need pixels.
The mouse lag on a 24 or 30Hz display will drive you nuts when you are trying to select a block of text.
If you are a keyboard-only editor, it's not as bad, but even highlighting text or trying to page down quickly will likely send you back to a high-speed multi-monitor setup.
You should use ECC RAM if you are doing any form of disk IO no matter which file system you're using, or you are under the risk of data loss.
I agree. Unfortunately, no Intel desktop CPU/chipset supports ECC, and many AMD desktop chipsets are castrated by the board manufacturer to not allow ECC.
When RAM was slow and 4GB was huge and expensive, this wasn't as big a deal, but now that 8GB is the reasonable starting point and 32GB is quite affordable, Intel especially needs to step up and add ECC support to their desktop CPUs/chipsets.
If you consider an IP address, a port number, a timestamp and an account number to be insanely detailed then I can't wait to see what you're going to say when you discover all the information Facebook, Google and others keep about you!
Google, et.al., keep their data because they can monetize it. Comcast will have to keep this data despite the fact that it not only won't make them money, but will cost them money since they will have to have people to search it for the legal requests.
Plus, I can generate thousands of connections per second and Comcast will have to log them all. If my both my neighbor and I have Comcast, I might just set up a box to do nothing but flood his WiFi with connections, just to cost Comcast money.
See for yourself.
Why didn't you just refer to the LHC web page and imply that you are writing at that same data rate to a single SSD...it would have exactly the same value as an argument.
btw, If you're complaint about the Prius appearance - what's the drag coefficient of your car? Is it as good as my 10 year old Prius? 'Cause that's why it looks like it does - it's part of it's design elegance.
It's also why the Prius would get about 40mpg even if it had no hybrid features. If Toyota sold that car, they'd really corner the market, as it would have a better ROI than the hybrid Prius, and wouldn't have any risk about battery replacement (which isn't always covered by warranty).
You have much bigger problems if you have memory corruption.
If you don't use ECC memory, you will have memory corruption. Even if you do use ECC memory, you might have corruption, and it might even go unnoticed, but the odds are far less likely.
"Corruption" in this sense doesn't mean that whole DIMMs are broken...it just means that one bit has changed in a way that the user/OS/CPU didn't want it to. In many cases, this can be completely harmless (e.g., graphical data used in-memory only has one bit wrong...you might not even notice a color shift if it's the LSB), a little annoying (e.g., unexpected program termination), or very annoying (e.g., BSOD). But, if this happens in memory used for disk write buffers, then you get the issue that the GP had you Google for.
Besides it further assumes that they are not using Carrier Grade NAT which is exactly how Free, a French ISP that has been doing the same thing for years, is handling this.
Even better, as now all the WiFi users appear to come from a single IP as far as the MPAA/RIAA is concerned, which means the only way they can get more info is if Comcast keeps insanely detailed records about every one of these connections. Keeping normal accounting information won't be enough to identify a copyright infringer...Comcast would also have to keep the IP/port connection logs from the NAT device.
Cray CS series.
Cash takes considerably longer to tender than credit. The customer takes time selecting the bills and coins, the cashier takes time counting it, then enters the amount in the cash register, and after the till opens, they have to count out the customer's change.
The one assumption you make here is that the credit card user is on the ball, and swipes either before the final total is rung or immediately after. I have seen many customers stand there until the cashier tells them the total, then reach into their wallet/purse and hunt the credit card, swipe it the wrong way, finally get it right, then hit "debit" on a card that is credit only.
Granted, these same people would likely take even longer to pay with cash, but I can see why some people think that cash is faster, based on anecdotes.
They are using the wifi and completely segregating traffic. It appears with a distiinct SSID and on a different IP. The capacity is on a different channel, so gain the host user isn't affected.
It's a completely separate segment of the private IP space, but once it heads through the router, every other device on the Internet (including the MPAA/RIAA scanners) will see the exact same public IP as the customer is given.
I'm assuming that Comcast doesn't have 50,000 spare routable IP addresses, but that's not a bad assumption.
It means the iPhone won't provide its MAC address *until* it finds a recognised network to connect to - it won't be broadcasting it constantly while you are out traveling or shopping.
This problem is easily solved by not turning on WiFi unless you know you are going to connect.
For me, I use an app on Android that keeps WiFi off unless I am in a location where I have already said I want to auto-connect to a specific network. It uses cell tower IDs and GPS to determine where I am.
Super computers are EXPENSIVE. A super computer is not just a tower with 30 gigs of ram and 10 processors, this is a building full of wires and computer components.
At this point, you can get a "supercomputer" in a rack for around US$500K, and it uses about 20kW of power.
Sure, it's not going to set any records, but with 500-1000 cores and 5-10TB of RAM, it's a lot more than most users will ever see. Heck, we have 40-core systems with over 2TB of RAM that fit in 2U. Again, not a supercomputer, but certainly a lot more power than in most single systems.
Based on TFA, I suspect the "cost" of $150K was what the time on the computer might sell for if somebody outside the project wanted to use it, not actual cost of electricity used. Most of these projects (like where I work) have computer resources that are free to use for all researches directly associated with the project, but could have a charge back for others. And, as others have stated, general purpose computers are pretty bad at mining, so it might have taken a lot of CPU-hours, and that's what the accountants are looking at. Here, we do have systems that would do fairly well at bitcoin mining (8 nVidia GPUs per 2U box), but they (like the other systems) are so busy doing actual work that nobody could possibly get enough idle cycles to do anything. That's why I put "cost" in quotes...if a system that is supposed to be used for research isn't doing enough that somebody could find that much idle time, then the system was either overbuilt or under-advertised.
And I don't use Samsung drives if I can help it. I know Apple uses them a lot, but they aren't really celebrated for their high performance.
For spinning disks, maybe not, but for SSDs, Samsung is king, especially their new M.2 XP941 model with 4 PCIe lanes.
They have always had issues with write endurance, since Day One. It has been their Achilles Heel.
Not according to real world testing.
Those are 240-256GB drives that have all had 600TB written to them, and many haven't even started to dip into the protected sectors. That's already up to 2400 P/E cycles on drives with 3000 or less on their specs. To put it into real-world terms, that's the equivalent of completely erasing the drive every day for 6.5 years. I think they'll last the same 7 to 8 years as your hard drives when used more realistically.
Since you brought it up: if you were building a heavy-use machine for code development today, would you go with 6-or-more-core Xeon, or a newer Haswell 4-core chip and board?
To me, Xeon vs. desktop chip comes down to how much memory you are going to want. Trusting 32GB or more to not have a random memory error isn't something I want to do, so I went with a Xeon and ECC. In addition, if you want more than 32GB, that means an LGA2011 socket for Intel...none of their other sockets have chipsets that can handle more. If you go LGA2011, then your best bet is a workstation board, since you could eventually put a Xeon and 256GB of RAM on it, but could start with a desktop CPU and whatever RAM you can afford. Workstation typically means no on-board graphics and plenty of PCIe slots available.
Your use case sounds like it doesn't really need the extra CPU threads, but does need disk speed. For that, I say SSD for your boot plus a real RAID controller (preferably with battery backup) with spinning disks for bulk storage and lots of RAM for cache. The same system with the E5-2630 v2 has 64GB of RAM, a 500GB Samsung 840 EVO and five 2TB WD Red drives in RAID-5 connected to an Adaptec 6805 with battery backup. Reads and writes to the SSD and the array are at similar speeds (between 400MB/sec and 600MB/sec), and more than one app hitting the disk at the same time doesn't make things slow to a crawl.
Essentially, if your disk can't support the I/O for all those services, going from 8 to 12 CPU threads won't really buy you much. You can then add any GPU you want (I have a Radeon R9 270).
That is so if there's a fire, the firemen can get to it.
Maybe it's just different places, but if that's why you can't park in front of a hydrant, then I couldn't park in my own driveway, as there is a hydrant less than 3 feet from the left edge (in the easement section between the sidewalk and the street).
Parking my car in my driveway puts my car closer to that hydrant than parking in the street would (it's about 5 feet from the street).
Troll is the mod for the comment that you dislike which is neither flamebait nor offtopic, duh.
I thought that's what "overrated" was for.
3x4 Is a very good format for reading, at any rate clearly better than any more elongated format like 9x16.
Based on the scans of my book covers, most books are about 1.5:1 (height/width), and looking at the text on the page, the actual used area is very similar. Most dedicated eBook readers have similar ratios.
So, it appears that both 1.78:1 and 1.33:1 aren't quite the same as historical reading aspect ratios, and both are off by about the same amount. I would suspect that 1.78:1 would look better, though, as it's much closer to the golden ratio than 1.33:1 is, and about as different as 1.5:1 is.
Personally, that's the jump I'm really looking forward to: Intel's first 8-core desktop chip.
With Google's stats on how often ECC in their servers correct errors, I'm not sure why anyone is anxious for an 8-core chip that can't use ECC memory.
My guess is that if you need 8 cores, you need more memory, and that means you are more likely to have errors. That plus the fact that the only current Intel desktop chipset supports more than 32GB of RAM has an LGA2011 socket is why I went with a Xeon...if I'm going to pay for a more expensive motherboard and processor, I might as well pay just a little more and get ECC protection on the memory. Also, the Xeon boards support 256GB instead of the 64GB that the desktop board support.
Although, to be fair, Intel's CPU "turbo" is probably a bit more of a car analogy because it's a temporary speed boost, not a permanent one.
For Ivy Bridge chips with decent cooling, it's basically permanent. Every Ivy Bridge chip has the base clock speed (the number on the box), the "Turbo" speed (which is available for a single core only), and the "multi-core turbo" that isn't advertised very well, but sites like CPUBenchmark will give you the details.
For example, I have an E5-2630 (yes, it's a Xeon, but there are fairly equivalent desktop chips) with a base clock of 2.6GHz, a single-core turbo of 3.1GHz, and a multi-core turbo of 2.9GHz. Since the chip stays cool enough, it runs at least at 2.9GHz whenever it needs speed.
Except only those making less than $15/hr will see that increase. As someone who earns more than minimum wage it won't benefit me, and in fact will take away much of my buying power.
It will especially hurt those currently making less than about 1.5x of the new minimum wage. People making more than that should be able to absorb the price increases without too much pain, but those in the 1x to 1.5x range will not get the 20-30% raise they need to make up the difference.