The opposite is true, and the specific ratio is about 40:1 reads to writes. The reason is obvious: data is written to the page file in large mostly sequential writes, but read in small chunks as needed. From the horse's mouth:
IIRC the G2 had a die shrink from 50nm to 24nm, so it dropped the rating down from 10K writes to 5K writes. If you've got a 160GB, it should be good for 800000 GB written, and you've therefore used 0.7% of the write lifespan on your drive. That should mean you have another ~280 years, not 9 years.
Sorry, you're right, I misplaced a decimal there. Doesn't change my point, and on top of the write buffer (as has been pointed out), newer SSDs do internal work to reduce write amplification (the phenomenon you're talking about). Many SSDs have an onboard cache, for example, and any 20GB torrent would have a block size larger than an SSD's page size anyhow.
There's the question too about why you'd be downloading a different 20GB torrent to an SSD drive every day anyhow, since most SSDs would fill up in just a few days at that rate.
What SSD did you manage to burn out in 18 months? What you're claiming is pretty extreme, since a 160GB first-gen Intel x25-m would require you to write roughly 3TB per day to a 160GB drive, an average sustained write speed of 35MB/s, which is very close to the maximum write speed of that drive anyhow. You'd be more or less maxing out the write speeds of the drive 24/7 for 18 months. I rather suspect that you had a really crappy MLC drive that predates that, before companies cared about write amplification.
About the x25-e, you're right, I meant to say the x25-m. I just pulled up the smart data (It's a 160GB x25-M first-gen), and after heavy use on the thing for over three years, including lots of OS reinstalls, secure erase and re-imaging, and game installation/uninstallations, I'm seeing 12.48 TB written (and this would include wasted writes from amplification) during 25289 power on hours. The theoretical lifespan is 1600 TB (I'm pretty sure about that math this time, 160 GB * 10000 = 1600000GB, / 1000 = 1600TB), so even with heavy use this drive should be lasting me about another... 380 years or so.
Basically, you are either using bargain bin SSDs from nobody companies, or you're doing something wrong...
My problem isn't really physical space. If I just replace the 3.5" to 2.5" adapters that came with my Intel 330 SSDs with something like a Scythe BayRafter, I could get four SSDs in the system. The problem is that I only have four internal SATA ports, since the Shuttle SZ77R5 relies only on the Intel chipset's internal SATA channels. It's got six total: two SATA3 and four SATA2. Four of those are exposed on the mobo, one is exposed via the mSATA slot, one via an eSATA jack on the back.
For the internal SATA ports, I've populated the two SATA3 ones with the SSDs, and one of the two SATA2 ones with a bluray burner. That leaves only one SATA2 port, which would bottleneck the drive. However, it is still an option, since the performance hit isn't that severe, and I also can get an mSATA SSD, although they cost a lot more than 2.5" drives in the same product line.
But really, I don't need to. I've got 2x180GB Intel 330 SSDs in this thing now, which is plenty for now, considering I've got a file server with 20TB of currently installed capacity for all my bulk storage needs:)
I used a first-gen Intel x25-e as my only drive for several years, and when I finally retired it (because I bought larger and faster drives), it had barely scratched the surface of its lifespan. A modern SSD will last for years, or even decades, before it wears out.
Look at it this way: even with the reduced lifespan of high density NAND, you get something like 3000 writes out of them (used to be 10k for the 45nm stuff, but write amplification is below 1 these days due to compression). On a 180GB drive, that will get you a lifetime write count of 540 PB. To hit that writing 20GB of fresh data every single day (which is probably way more than what actually happens in practice, even with swapping, which is predominantly read-heavy, not write-heavy), the drive would last roughly 74 years...
Laptops that only have room for an mSATA port have all gone SSD-only, and laptops that have both an mSATA port and 2.5" bay can use a caching SSD drive in the mSATA port and a large traditional drive in the 2.5" bay.
Heck, even my desktop is set up for that: I bought a Shuttle XPC, which has room for two 3.5" HDDs and an mSATA port intended for use with Intel's caching system. Their intention is for you to stick two multi-terabyte 3.5" disks in there and then put a 40GB SSD in the mSATA port or something, although you could also just put a 300GB SSD in there and use that for storage.
Me? I just filled the 3.5" bays with 2.5" Intel SSDs and did without any traditional storage:)
It's unfortunate in that Seagate (and most of the traditional HDD companies) have basically failed to break into the SSD market, which is rapidly replacing a good chunk of the HDD market. I think Samsung and Toshiba are the only traditional HDD manufacturers who have managed to significantly break into the SSD market, although they both largely do that via OEM stuff (Samsung does have some retail models which are pretty good, but review poorly). Instead, companies like Seagate and Western Digital just sat around and pish-poshed the SSD market while RAM and flash manufacturers manufacturers like Intel, OCZ, Crucial, Kingston, Samsung, and Toshiba broke into the storage market and captured all the SSD marketshare for themselves. And the only traditional HDD manufacturers who did manage to make the transition also happen to fab their own NAND flash (Samsung and Toshiba)...
You could argue that fabbing their own flash memory gave Intel (IMFT), Crucial (IMFT), Samsung, and Toshiba an advantage, but OCZ has managed to capture a large portion of the market despite not having any fabs themselves.
I miss the simpler days when software claiming to double your computers speed wasn't a scam, and was actually from a reputable company (Connectix, for example, before they were bought out by Microsoft). SpeedDoubler (for Mac) back then did logical things. It and its sister products did stuff like replacing the OS's interpretive 68k emulator with a much faster dynamically recompiling 68k emulator (much software and large parts of the OS were emulated during the 68k -> PPC transition), or replacing the OS virtual memory subsystem with one that used compression (disks were way slower back then), or replacing the OS disk cache with a more efficient one, or a better file copy routine... These days, any software claiming to speed up your computer is generally a scam. But back then, they weren't, and their effectiveness was almost magical...
You've replied to the wrong post, I didn't say that.
It may well be verboten for one device to consume both channels, it seems like that might break daisy chaining or something. In any event, I'm curious about where the bottlenecks on a tbolt-based video solution are. A lot of work is done on-card now, so once you've got all your assets uploaded to the card (which these days seem to have gigabytes of RAM), there shouldn't be much traffic flying around. On the other hand, Windows more recently switched to a virtual memory architecture for VRAM, which would imply a great deal more traffic.
Canadians can still fly on plenty of domestic airlines, and I doubt an Air Canada flight that passes over, say, Alaska on the way to Tokyo, is going to divert to the US to arrest somebody.
In terms of if the FDA is overreacting, I'd suggest that if Canadian authorities aren't arresting the guy in Canada, then whatever he is doing is probably not bad enough that the FDA should get involved when he visits the US.
I can't say if they let you buy them from outside of Australia (I'd probably have to order one and pay for it to find out), but voip.ms does have a few international DIDs:
Australia is a heck of a lot more expensive with VoIP.ms ($7/mth compared to $5/yr), but the voip.ms one is unlimited while the mynetphone one is $0.10 per call, if I'm understanding right.
I'm not saying the VoIP.ms one is necessarily a better deal, only that it proves that the option is available, that there are companies that sell Australian DIDs that include enough built-in features (VoIP.ms basically gives you the feature set of a full hosted PBX for free) to do the forwarding and routing rules entirely server-side.
If MyNetPhone supports at least call forwarding to an international number at reasonable rates, it could very likely end up a bunch cheaper for light to moderate use.
You don't need an external device or a foreign landline to do it with pure SIP either...
For example, say you're in the US, but you want calls to a Canadian number to be routed to your PC/cellphone/landline/whatever. You'd pay $1/mth to a company like voip.ms for a DID (Direct Inward Dialing, basically a phone number), and set it up to forward calls to either an existing telephone number (cell, landline, etc) or some SIP software client. You'd pay something like a cent a minute.
The same principal applies overseas; get a DID with a company, set it up to forward to a US phone number or SIP address.
It's an app on the phone. You really think that Google's internal "delete app" functionality is somehow limited to apk installed only through the app store? This assumes that you're on an Android device that even allows sideloading without jailbreaking.
Good point. Of course, you should also not use Android devices for anything important, nor should you use BlackBerry devices for anything important, nor should you use Windows Phone devices for anything important. Because, of course, they all have the same ability to pull stuff.
I'm not convinced about any of that. I know a number of old Android apps won't work on new versions of Android
That's not what the OP meant. They meant that a new OS release that the user DOES NOT install won't cause apps to stop working (because it won't be installed). It's true that future versions of an app might break support for older OS, but that's not a problem in this case since it was pulled from the app store.
Basically, the complaint is that Apple pulled the app from the store at the start of the dispute instead of the end of it. Well, there's nothing stopping anybody from continuing to use this app until the dispute is resolved. That's not forever: either the app is restored, or it wasn't legal to begin with.
7 devices per port? The 7 device limit is a daisy-chaining limit, as far as I can tell, not a maximum number of devices. How many devices currently hooked up to a typical computer are daisy-chained? The number of ports would likely be far more important than the maximum number of daisy-chain hops. Let's assume audio will always remain on 3.5mm jacks for headphone compatibility and assume that any thunderbolt monitor is going to have a daisy chain thunderbolt port at the very least.
On a desktop computer, the minimum number of tbolt connectors would be two. Desktop -> monitor -> keyboard, desktop -> mouse, that would get you all your core functionality (audio, again, being assumed to be inappropriate for tbolt). For a laptop computer, the keyboard is built-in, but people often like a real mouse, so you can get away with a single port.
The problem is, of course, that these scenarios would leave no extra connectivity for expansion. It would seem that a desktop would need at least three or four tbolt ports before it could conceivably replace all other digital connections on a computer. Unless you want to consider stuff like thunderbolt docks (or displays with thunderbolt docks built in), in which case the computer itself needs only one or two ports, although you're not really going thunderbolt-only that way.
At this point, most assumptions of graphics card performance are likely in terms of PCIe 2.0. So if you need a 4x 2.0 slot for optimal performance, you're probably still going to get pretty decent performance out of an effectively 2x slot.
ViDock has been doing this stuff over ExpressCard, which is 1x PCIe 1.0... That works OK with low-end cards, and with thunderbolt pushing four times the bandwidth, there should be a rather big difference there.
I also wonder if somebody will come up with a solution that takes advantage of both thunderbolt channels for more bandwidth. It'd probably require some tricky driver work, but it's theoretically possible. Failing that, there's always SLI (or equivalent). Two midrange graphics cards on independent thunderbolt channels could produce some pretty decent performance. Some variations of SLI have a direct interconnect between the cards, and other variations are entirely software based (like the Lucid Logix stuff).
There's also something to be said for a universal port, one that can be used to connect anything. You get audio and video via the DisplayPort aspect, and generic connectivity via the PCIe aspect. The question is, will the cost of implementation (expensive chipsets and cables) make it useless for cheap peripherals (like a mouse, keyboard, microphone, etc). Perhaps the cost of those will eventually come down enough that it won't matter, or perhaps we'll always see a mix of USB and Thunderbolt on computers, or perhaps all the low-bandwidth devices will migrate to wireless.
Slashdot Mirror
Ikea? Planned obsolescence? It's furniture, you don't need to apply software updates to a bookshelf.
The opposite is true, and the specific ratio is about 40:1 reads to writes. The reason is obvious: data is written to the page file in large mostly sequential writes, but read in small chunks as needed. From the horse's mouth:
http://blogs.msdn.com/b/e7/archive/2009/05/05/support-and-q-a-for-solid-state-drives-and.aspx
MLC started at 10k, we're down to 3k, but increases in capacity and algorithms to reduce write amplification have mitigated this.
IIRC the G2 had a die shrink from 50nm to 24nm, so it dropped the rating down from 10K writes to 5K writes. If you've got a 160GB, it should be good for 800000 GB written, and you've therefore used 0.7% of the write lifespan on your drive. That should mean you have another ~280 years, not 9 years.
Sorry, you're right, I misplaced a decimal there. Doesn't change my point, and on top of the write buffer (as has been pointed out), newer SSDs do internal work to reduce write amplification (the phenomenon you're talking about). Many SSDs have an onboard cache, for example, and any 20GB torrent would have a block size larger than an SSD's page size anyhow.
There's the question too about why you'd be downloading a different 20GB torrent to an SSD drive every day anyhow, since most SSDs would fill up in just a few days at that rate.
What SSD did you manage to burn out in 18 months? What you're claiming is pretty extreme, since a 160GB first-gen Intel x25-m would require you to write roughly 3TB per day to a 160GB drive, an average sustained write speed of 35MB/s, which is very close to the maximum write speed of that drive anyhow. You'd be more or less maxing out the write speeds of the drive 24/7 for 18 months. I rather suspect that you had a really crappy MLC drive that predates that, before companies cared about write amplification.
About the x25-e, you're right, I meant to say the x25-m. I just pulled up the smart data (It's a 160GB x25-M first-gen), and after heavy use on the thing for over three years, including lots of OS reinstalls, secure erase and re-imaging, and game installation/uninstallations, I'm seeing 12.48 TB written (and this would include wasted writes from amplification) during 25289 power on hours. The theoretical lifespan is 1600 TB (I'm pretty sure about that math this time, 160 GB * 10000 = 1600000GB, / 1000 = 1600TB), so even with heavy use this drive should be lasting me about another... 380 years or so.
Basically, you are either using bargain bin SSDs from nobody companies, or you're doing something wrong...
My problem isn't really physical space. If I just replace the 3.5" to 2.5" adapters that came with my Intel 330 SSDs with something like a Scythe BayRafter, I could get four SSDs in the system. The problem is that I only have four internal SATA ports, since the Shuttle SZ77R5 relies only on the Intel chipset's internal SATA channels. It's got six total: two SATA3 and four SATA2. Four of those are exposed on the mobo, one is exposed via the mSATA slot, one via an eSATA jack on the back.
For the internal SATA ports, I've populated the two SATA3 ones with the SSDs, and one of the two SATA2 ones with a bluray burner. That leaves only one SATA2 port, which would bottleneck the drive. However, it is still an option, since the performance hit isn't that severe, and I also can get an mSATA SSD, although they cost a lot more than 2.5" drives in the same product line.
But really, I don't need to. I've got 2x180GB Intel 330 SSDs in this thing now, which is plenty for now, considering I've got a file server with 20TB of currently installed capacity for all my bulk storage needs :)
I used a first-gen Intel x25-e as my only drive for several years, and when I finally retired it (because I bought larger and faster drives), it had barely scratched the surface of its lifespan. A modern SSD will last for years, or even decades, before it wears out.
Look at it this way: even with the reduced lifespan of high density NAND, you get something like 3000 writes out of them (used to be 10k for the 45nm stuff, but write amplification is below 1 these days due to compression). On a 180GB drive, that will get you a lifetime write count of 540 PB. To hit that writing 20GB of fresh data every single day (which is probably way more than what actually happens in practice, even with swapping, which is predominantly read-heavy, not write-heavy), the drive would last roughly 74 years...
Laptops that only have room for an mSATA port have all gone SSD-only, and laptops that have both an mSATA port and 2.5" bay can use a caching SSD drive in the mSATA port and a large traditional drive in the 2.5" bay.
Heck, even my desktop is set up for that: I bought a Shuttle XPC, which has room for two 3.5" HDDs and an mSATA port intended for use with Intel's caching system. Their intention is for you to stick two multi-terabyte 3.5" disks in there and then put a 40GB SSD in the mSATA port or something, although you could also just put a 300GB SSD in there and use that for storage.
Me? I just filled the 3.5" bays with 2.5" Intel SSDs and did without any traditional storage :)
It's unfortunate in that Seagate (and most of the traditional HDD companies) have basically failed to break into the SSD market, which is rapidly replacing a good chunk of the HDD market. I think Samsung and Toshiba are the only traditional HDD manufacturers who have managed to significantly break into the SSD market, although they both largely do that via OEM stuff (Samsung does have some retail models which are pretty good, but review poorly). Instead, companies like Seagate and Western Digital just sat around and pish-poshed the SSD market while RAM and flash manufacturers manufacturers like Intel, OCZ, Crucial, Kingston, Samsung, and Toshiba broke into the storage market and captured all the SSD marketshare for themselves. And the only traditional HDD manufacturers who did manage to make the transition also happen to fab their own NAND flash (Samsung and Toshiba)...
You could argue that fabbing their own flash memory gave Intel (IMFT), Crucial (IMFT), Samsung, and Toshiba an advantage, but OCZ has managed to capture a large portion of the market despite not having any fabs themselves.
I miss the simpler days when software claiming to double your computers speed wasn't a scam, and was actually from a reputable company (Connectix, for example, before they were bought out by Microsoft). SpeedDoubler (for Mac) back then did logical things. It and its sister products did stuff like replacing the OS's interpretive 68k emulator with a much faster dynamically recompiling 68k emulator (much software and large parts of the OS were emulated during the 68k -> PPC transition), or replacing the OS virtual memory subsystem with one that used compression (disks were way slower back then), or replacing the OS disk cache with a more efficient one, or a better file copy routine... These days, any software claiming to speed up your computer is generally a scam. But back then, they weren't, and their effectiveness was almost magical...
You've replied to the wrong post, I didn't say that.
It may well be verboten for one device to consume both channels, it seems like that might break daisy chaining or something. In any event, I'm curious about where the bottlenecks on a tbolt-based video solution are. A lot of work is done on-card now, so once you've got all your assets uploaded to the card (which these days seem to have gigabytes of RAM), there shouldn't be much traffic flying around. On the other hand, Windows more recently switched to a virtual memory architecture for VRAM, which would imply a great deal more traffic.
Canadians can still fly on plenty of domestic airlines, and I doubt an Air Canada flight that passes over, say, Alaska on the way to Tokyo, is going to divert to the US to arrest somebody.
In terms of if the FDA is overreacting, I'd suggest that if Canadian authorities aren't arresting the guy in Canada, then whatever he is doing is probably not bad enough that the FDA should get involved when he visits the US.
I can't say if they let you buy them from outside of Australia (I'd probably have to order one and pay for it to find out), but voip.ms does have a few international DIDs:
http://voip.ms/intldids.php
Australia is a heck of a lot more expensive with VoIP.ms ($7/mth compared to $5/yr), but the voip.ms one is unlimited while the mynetphone one is $0.10 per call, if I'm understanding right.
I'm not saying the VoIP.ms one is necessarily a better deal, only that it proves that the option is available, that there are companies that sell Australian DIDs that include enough built-in features (VoIP.ms basically gives you the feature set of a full hosted PBX for free) to do the forwarding and routing rules entirely server-side.
If MyNetPhone supports at least call forwarding to an international number at reasonable rates, it could very likely end up a bunch cheaper for light to moderate use.
Microsoft's contributions are GPL'd.
Best avoid the Linux kernel, then; there's Microsoft code in it.
You don't need an external device or a foreign landline to do it with pure SIP either...
For example, say you're in the US, but you want calls to a Canadian number to be routed to your PC/cellphone/landline/whatever. You'd pay $1/mth to a company like voip.ms for a DID (Direct Inward Dialing, basically a phone number), and set it up to forward calls to either an existing telephone number (cell, landline, etc) or some SIP software client. You'd pay something like a cent a minute.
The same principal applies overseas; get a DID with a company, set it up to forward to a US phone number or SIP address.
It's an app on the phone. You really think that Google's internal "delete app" functionality is somehow limited to apk installed only through the app store? This assumes that you're on an Android device that even allows sideloading without jailbreaking.
Good point. Of course, you should also not use Android devices for anything important, nor should you use BlackBerry devices for anything important, nor should you use Windows Phone devices for anything important. Because, of course, they all have the same ability to pull stuff.
As has been pointed out, the solution to that is to simply not install any iOS updates.
I'm not convinced about any of that. I know a number of old Android apps won't work on new versions of Android
That's not what the OP meant. They meant that a new OS release that the user DOES NOT install won't cause apps to stop working (because it won't be installed). It's true that future versions of an app might break support for older OS, but that's not a problem in this case since it was pulled from the app store.
Basically, the complaint is that Apple pulled the app from the store at the start of the dispute instead of the end of it. Well, there's nothing stopping anybody from continuing to use this app until the dispute is resolved. That's not forever: either the app is restored, or it wasn't legal to begin with.
7 devices per port? The 7 device limit is a daisy-chaining limit, as far as I can tell, not a maximum number of devices. How many devices currently hooked up to a typical computer are daisy-chained? The number of ports would likely be far more important than the maximum number of daisy-chain hops. Let's assume audio will always remain on 3.5mm jacks for headphone compatibility and assume that any thunderbolt monitor is going to have a daisy chain thunderbolt port at the very least.
On a desktop computer, the minimum number of tbolt connectors would be two. Desktop -> monitor -> keyboard, desktop -> mouse, that would get you all your core functionality (audio, again, being assumed to be inappropriate for tbolt). For a laptop computer, the keyboard is built-in, but people often like a real mouse, so you can get away with a single port.
The problem is, of course, that these scenarios would leave no extra connectivity for expansion. It would seem that a desktop would need at least three or four tbolt ports before it could conceivably replace all other digital connections on a computer. Unless you want to consider stuff like thunderbolt docks (or displays with thunderbolt docks built in), in which case the computer itself needs only one or two ports, although you're not really going thunderbolt-only that way.
At this point, most assumptions of graphics card performance are likely in terms of PCIe 2.0. So if you need a 4x 2.0 slot for optimal performance, you're probably still going to get pretty decent performance out of an effectively 2x slot.
ViDock has been doing this stuff over ExpressCard, which is 1x PCIe 1.0... That works OK with low-end cards, and with thunderbolt pushing four times the bandwidth, there should be a rather big difference there.
I also wonder if somebody will come up with a solution that takes advantage of both thunderbolt channels for more bandwidth. It'd probably require some tricky driver work, but it's theoretically possible. Failing that, there's always SLI (or equivalent). Two midrange graphics cards on independent thunderbolt channels could produce some pretty decent performance. Some variations of SLI have a direct interconnect between the cards, and other variations are entirely software based (like the Lucid Logix stuff).
There's also something to be said for a universal port, one that can be used to connect anything. You get audio and video via the DisplayPort aspect, and generic connectivity via the PCIe aspect. The question is, will the cost of implementation (expensive chipsets and cables) make it useless for cheap peripherals (like a mouse, keyboard, microphone, etc). Perhaps the cost of those will eventually come down enough that it won't matter, or perhaps we'll always see a mix of USB and Thunderbolt on computers, or perhaps all the low-bandwidth devices will migrate to wireless.
Well, that and the liability that comes with widescale commercial copyright infringement...
Isn't that what they're being asked to do? The Oatmeal asks them to stop posting his comics, the logical solution is to blog the hostname.