128G MLC drive, with.5M blocks --> 256k blocks. assume endurance is 10k. so after 2.5G erases the disk is at it's endurance limit. got to erase the block every time you modify it. 32M buffer can buffer the writes - assume writes are evenly spread over the blocks, and ignore write locations, so 32M buffer / 256k blocks --> on average 128 bytes per erase. 2.5G * 128 --> 320G of data. and the first write is free (since the MLC is blank). so 448G can be written before the drive is at it's endurance limit. 448G at 70M/s --> 6400 seconds of write at full speed, about 2 hours
The first error in your calculations is based on your assumption that "modify" = "erase." There does not need to be a 1:1 correspondence between writes and erases. Those 256K erase blocks are just that, ERASE blocks. They are not WRITE blocks. Writes happen at a much finer level than that (64 pages of 4,096+128 bytes each for a block size of 256KB)
Another error in your calculation is that you are conflating post-amplified-write-limits with pre-amplified-write-rates.
But let me spare you some of the details that are causing you problems. The minimum time to kill the drive does not involve write amplification because erase cycles are no faster than writes of the same size. A 128GB MLC with 10,000 write cycles and 70MB/sec sequential transfer rate cannot be killed any faster than 18,285,714 seconds (211 days.) Its basically performing perpetual erase-cycles at that point.
The failure mode that everyone is worried about is exceeding the write endurance rating of the NAND, which you cannot accelerate by buying more drives. You just have to wait... and wait... and wait some more...
If you "wait... and wait... and wait some more..." with a standard econo-platter, you are sure to lose the data.
We can attack this backwards to see how realistic it is to worry about write endurance. Lets pick a single sized drive of 80GB, a horrible write-amplification value of 200%, and a desire for the drive to live for 1825 days (5 years.)
The question then becomes how much data can the owner write per day and attain those 1825 days. A 80GB drive with a 10,000 erase cycle limit has a theoretical maximum write capacity of 800 terabytes. Factoring in the liberal 2x write amplification, thats 400 terabytes over its lifetime.
The user would have to write 220GB per day, every day, for 5 years in order to kill it at the end of the 5th year. Keep in mind that the entire drive itself is only 80GB, so the user would have to replace the entire contents nearly 3 times per day, every day.
As you see, when you do the math, those irrational fears about write limits simply evaporate when considering the larger capacity drives available today. Those fears about write limits apply only when there is no wear leveling (CF cards, etc..), or when the drives are exceptionally small (thumb drives...) They have no part of the modern SSD, not even the "small" 80GB ones.
You dont see the failure mode of these modern SSD's because they are even hard to kill when you are intentionally trying to kill them.
I'm not sure that this is quite right. I think the downfall of SLC is because MLC densities are approaching if not already surpassed the point where SLC makes sense.
Part of what made SLC so attractive is better performance, but the current crop of MLC drives leading the market are banging their heads on performance bottlenecks external to the drive (SATA 2.0 is fully and easily saturated, and drives are now appearing that are saturating SATA 3.0) and this is due to densities. The more flash chips they can pack on, the faster they can make the device as a whole, and its at the point now where they arent concerned about how they will get to specific performance numbers, but rather they are concerned with weather they should.
The other part of what made SLC so attractive is their write limits, but again densities are increasing undermining the concern. 100,000 writes per cell on a 64GB SLC device is about as useful as 10,000 writes per cell on a 640GB MLC device. In both cases the max is 6400000000000000 bytes written (obviously write amplification plays a role to reduce the effective limit, but does so in both SLC and MLC)
At this point I consider any claims that SSD's are less reliable to simply be a myth derived from dishonest reporting.
Furthermore, there are published studies detailing how unreliable traditional magnetic platter drives are.
Do they have write limits? Yes. Can other parts of the device fail? Yes. Are they more expensive than economy platters? Yes. Is there real world data showing that they are less reliable as claimed? Apparently not.
It makes for a self-documenting keyboard interface, as is shown in the video where the guy hits the Windows Key (it could have been ALT, CTRL, or CTRL-SHIFT, etc..) and all the keys but those tied to commands go dark, and the ones tied to commands label themselves with what the command available actually is.
Hah. Your most reminds me of a story about a school janitor that had problems with the school girls putting lipstick on and kissing the bathroom mirrors (leaving lipstick kiss marks.)
No amount of telling the girls not to do it stopped them, so one day they brought the girls into the bathroom in small groups and the janitor says "let me show you how hard it is to clean lipstick off mirrors" and then promptly sticks a mop into one of the toilets, takes it out dripping wet, and then proceeds to scrub the mirrors with the toilet-water.
After that day, the girls never kissed the mirrors again.
So there we have a SIGNIFICANTLY better benchmark on AMD for $5 more, and SLIGHTLY better benchmark for $18 less.
..and unlike most benchmarks, these
(A) include benchmarks of many systems across many motherboards and memory sticks,
(B) do not include any overclocking (the site keeps overclocked benchmarks separate and those can be viewed too),
(C) the benchmarks are very detailed and if you have the passmark benchmark program, can compare systems in all catagories to see whats better and whats not, and most importantly
(D) these systems are not chips "donated" by manufacturers, so no conflicts of interest.. they are chips from stocks available to and sold to consumers, and benchmarked by consumers.
Blast out 6 cores has a lot different meaning than blasting out 4 cores. Please check out the tech Intel has in place to deliver good performance for a single-threaded app (overclocking the core on which the thread is running on).
Thats funny fan boy, because AMD has the same automatic over-clocking tech for low-thread workloads, IN THE 6 CORE CHIPS THAT YOU ARE DISMISSING.
AMD calls it Turbo Core, and the consumer versions of the 6 core chips are named "1055T" and "1090T" where the T signifies this very tech you so ignorantly didnt know the 6 core chips had. We know you are a fanboy because you insist on commenting on AMD's 6 core chips while knowing literally nothing about them other than that they are 6 core, and from AMD.
For one thing, low end motherboards SUCK REALLY BADLY. You are essentially LUCKY if you get a $50 Motherboard that doesnt have quirky issues or downright failures.
Also there is the feature-issue. My latest MoBo purchase included USB 3.0 and SATA 3.0 even though I strictly have no need for such things yet. This forward thinking means that when I finally pick up a high performance SSD, I am not going to be stuck choosing from or settle for the lower 3Gbit tier.
MSI makes some quality boards in the $100 range (find their "Military Grade" lines), and ASUS makes some quality ones in the $125 range. I can't really comment on the others.
I wish Soyo was still around making motherboard. They made some really good ones in the $50 range.. but perhaps thats why they went bankrupt.
So you haven't really done any research there? Intel's i5 750 and 760 "steamroll" all the Phenom II X4 CPUs in the price range. Don't trust me, trust benchmarks.
Phenom II X6 chips with Turbo Core in the same price range would like to have a word with you about you cherry picking old X4 chips.
What are you talking about? AM2 boards support AM3 chips.
You also present a false dichotomy, because upgrading isnt ONLY about buying suboptimal hardware and then upgrading it later. Anyone who purchased bleeding edge AM2 gear when it was introduced can get a bios update and then socket an AM3 Phenom II chip. They still only have DDR2, but amazingly Phenom II's support both DDR2 on AM2 and DDR3 on AM3.
So that guy who purchased a dual-core AM2 Phenom when they were cutting edge can now socket a hexa-core AM3 Phenom II.
Its amazing what designing for the future gives your customers. Intel users have only rarely had the chance to substantially upgrade CPU's.
The regulation supports will just blame capitalist interventions into the legislative process when the regulation ends up screwing us all over.
Meanwhile those of us against net neutrality regulation at this time are shaking our heads wondering why so many people want to forever trade their freedom of choice to a bunch of politicians that are sure to meddle with the trust they have placed in them a thousand times over.
Then with this connection you have a 100Mbits line...
Without competition, they have no reason to give you a 100Mbit/sec line.
America is far behind in the bandwidth race with an average of only 3.9Mbit/sec. We rank 18th in the world and there are seemingly no major plans to increase bandwidth further. We are now at a standstill because the monopolies have no reason to improve service.
You are never going to get 100Mbit/sec service from your current providers as things stand right now. Only allowing extensive competition will get us there, and that also moots the whole net neutrality thing.
First, Net Neutrality would do no such thing to entrench a monopoly. And it's a fact that these ISPs often do have either a monopoly or duopoly in the areas.
You don't seem to get it. The government has already entrenched these monopolies, and now instead of fixing the problem, net neutrality proposes to eliminate favoritism. So you can have shitty monopoly service that favors no one. You are forgetting that ITS STILL SHITTY SERVICE.
I'd choose 100MBit broadband with 10MBit traffic shaping any day of the week over the current 10MBit service. Net Neutrality prevents the traffic shaping, but doesnt give you 100MBit service, and in fact makes for a good excuse to never allow any competition.
Do you have the same objections to the Automatic Updates feature in Windows?
yes. end of story.
Distributed.net isn't making noticeable progress cracking a 72-bit RC5 symmetric key. What makes you think a black-hat botnet will be any more efficient?
You mean besides the fact that each time mozilla ships an update, there is more information available to crack the private key?
It's safer in the case where a computer's administrator may be physically absent for substantial periods of time.
The absolute last thing an administrator does when (s)he leaves for "substantial periods of time" is leave any sort of automatic updating enabled. If the automatic update breaks something, thats the end game nightmare of zero uptime "for a substantial period of time." With firefox this problem is confounded by the fact that it isnt easy to remotely manage the installations (Active Directory) so they couldn't reliably and easily disable such silent updating prior to leaving for that "substantial period of time."
The 'Administrator Isn't Around' card doesnt apply here because of thats the case, the administrator is a fool if he lets things update at all, let alone silently do so.
On the other hand, game developers would have a semi-solid set of numbers to go by, so they can assess the size of their potential market. As it is, there really aren't good numbers on Linux adoption, because no one has to buy a license, everyone can freely copy it from each other, and people collect distro discs that never get used.
I think that you are missing a key element here. Linux users didn't pay anything for their OS. Meanwhile.. Windows users are known to pay for software.
Even if it was a 50/50 split of Linux/Windows, they would STILL target Windows first.
Hello, my name is [competing product to the company I represent] and I dont like the results that come up when people search for my name. You say that you can change the results for a small fee. Where do we...I mean I... send the check?
Slashdot Mirror
Citation needed.
128G MLC drive, with .5M blocks --> 256k blocks. assume endurance is 10k. so after 2.5G erases the disk is at it's endurance limit. got to erase the block every time you modify it. 32M buffer can buffer the writes - assume writes are evenly spread over the blocks, and ignore write locations, so 32M buffer / 256k blocks --> on average 128 bytes per erase. 2.5G * 128 --> 320G of data. and the first write is free (since the MLC is blank). so 448G can be written before the drive is at it's endurance limit. 448G at 70M/s --> 6400 seconds of write at full speed, about 2 hours
The first error in your calculations is based on your assumption that "modify" = "erase." There does not need to be a 1:1 correspondence between writes and erases. Those 256K erase blocks are just that, ERASE blocks. They are not WRITE blocks. Writes happen at a much finer level than that (64 pages of 4,096+128 bytes each for a block size of 256KB)
Another error in your calculation is that you are conflating post-amplified-write-limits with pre-amplified-write-rates.
But let me spare you some of the details that are causing you problems. The minimum time to kill the drive does not involve write amplification because erase cycles are no faster than writes of the same size. A 128GB MLC with 10,000 write cycles and 70MB/sec sequential transfer rate cannot be killed any faster than 18,285,714 seconds (211 days.) Its basically performing perpetual erase-cycles at that point.
The failure mode that everyone is worried about is exceeding the write endurance rating of the NAND, which you cannot accelerate by buying more drives. You just have to wait... and wait... and wait some more...
If you "wait... and wait... and wait some more..." with a standard econo-platter, you are sure to lose the data.
We can attack this backwards to see how realistic it is to worry about write endurance. Lets pick a single sized drive of 80GB, a horrible write-amplification value of 200%, and a desire for the drive to live for 1825 days (5 years.)
The question then becomes how much data can the owner write per day and attain those 1825 days. A 80GB drive with a 10,000 erase cycle limit has a theoretical maximum write capacity of 800 terabytes. Factoring in the liberal 2x write amplification, thats 400 terabytes over its lifetime.
The user would have to write 220GB per day, every day, for 5 years in order to kill it at the end of the 5th year. Keep in mind that the entire drive itself is only 80GB, so the user would have to replace the entire contents nearly 3 times per day, every day.
As you see, when you do the math, those irrational fears about write limits simply evaporate when considering the larger capacity drives available today. Those fears about write limits apply only when there is no wear leveling (CF cards, etc..), or when the drives are exceptionally small (thumb drives...) They have no part of the modern SSD, not even the "small" 80GB ones.
You dont see the failure mode of these modern SSD's because they are even hard to kill when you are intentionally trying to kill them.
I'm not sure that this is quite right. I think the downfall of SLC is because MLC densities are approaching if not already surpassed the point where SLC makes sense.
Part of what made SLC so attractive is better performance, but the current crop of MLC drives leading the market are banging their heads on performance bottlenecks external to the drive (SATA 2.0 is fully and easily saturated, and drives are now appearing that are saturating SATA 3.0) and this is due to densities. The more flash chips they can pack on, the faster they can make the device as a whole, and its at the point now where they arent concerned about how they will get to specific performance numbers, but rather they are concerned with weather they should.
The other part of what made SLC so attractive is their write limits, but again densities are increasing undermining the concern. 100,000 writes per cell on a 64GB SLC device is about as useful as 10,000 writes per cell on a 640GB MLC device. In both cases the max is 6400000000000000 bytes written (obviously write amplification plays a role to reduce the effective limit, but does so in both SLC and MLC)
I keep hearing people claim reliability issues when SSD articles come along to slashdot.
I have never seen a citation, so I went looking for them via Google but could only find citations attesting to the high reliability of these devices.
Dell's Lionel Menchaca stated in 2008, when it was reported by Avian Securities that Dell was having SSD reliability issues, "Our global reliability data shows that SSD drives [that we shipped] are equal to or better than traditional hard disk drives we've shipped." He further notes that Avian Securities never contacted them and that their numbers were a complete fabrication.
At this point I consider any claims that SSD's are less reliable to simply be a myth derived from dishonest reporting.
Furthermore, there are published studies detailing how unreliable traditional magnetic platter drives are.
Do they have write limits? Yes. Can other parts of the device fail? Yes. Are they more expensive than economy platters? Yes. Is there real world data showing that they are less reliable as claimed? Apparently not.
It makes for a self-documenting keyboard interface, as is shown in the video where the guy hits the Windows Key (it could have been ALT, CTRL, or CTRL-SHIFT, etc..) and all the keys but those tied to commands go dark, and the ones tied to commands label themselves with what the command available actually is.
Looks like the future to me.
What Microsoft claims are you talking about?
Microsofts involvement here is simply that they provided hardware and some funding. The UIST is an initiative of the ACM.
He probably didnt know that it had a lifetime warranty.
Hah. Your most reminds me of a story about a school janitor that had problems with the school girls putting lipstick on and kissing the bathroom mirrors (leaving lipstick kiss marks.)
No amount of telling the girls not to do it stopped them, so one day they brought the girls into the bathroom in small groups and the janitor says "let me show you how hard it is to clean lipstick off mirrors" and then promptly sticks a mop into one of the toilets, takes it out dripping wet, and then proceeds to scrub the mirrors with the toilet-water.
After that day, the girls never kissed the mirrors again.
i5 750 non-gaming benchmarks?
..and unlike most benchmarks, these
The Phenom II x6 1055T scores 5153, sells for $200 (491 systems benchmarked)
The i5-750 scores 4211, sells for $195 (3098 systems benchmarked)
The Phenom II x4 965 scores 4261, sells for $177 (2574 systems benchmarked)
So there we have a SIGNIFICANTLY better benchmark on AMD for $5 more, and SLIGHTLY better benchmark for $18 less.
(A) include benchmarks of many systems across many motherboards and memory sticks,
(B) do not include any overclocking (the site keeps overclocked benchmarks separate and those can be viewed too),
(C) the benchmarks are very detailed and if you have the passmark benchmark program, can compare systems in all catagories to see whats better and whats not,
and most importantly
(D) these systems are not chips "donated" by manufacturers, so no conflicts of interest.. they are chips from stocks available to and sold to consumers, and benchmarked by consumers.
Blast out 6 cores has a lot different meaning than blasting out 4 cores. Please check out the tech Intel has in place to deliver good performance for a single-threaded app (overclocking the core on which the thread is running on).
Thats funny fan boy, because AMD has the same automatic over-clocking tech for low-thread workloads, IN THE 6 CORE CHIPS THAT YOU ARE DISMISSING.
AMD calls it Turbo Core, and the consumer versions of the 6 core chips are named "1055T" and "1090T" where the T signifies this very tech you so ignorantly didnt know the 6 core chips had. We know you are a fanboy because you insist on commenting on AMD's 6 core chips while knowing literally nothing about them other than that they are 6 core, and from AMD.
For one thing, low end motherboards SUCK REALLY BADLY. You are essentially LUCKY if you get a $50 Motherboard that doesnt have quirky issues or downright failures.
Also there is the feature-issue. My latest MoBo purchase included USB 3.0 and SATA 3.0 even though I strictly have no need for such things yet. This forward thinking means that when I finally pick up a high performance SSD, I am not going to be stuck choosing from or settle for the lower 3Gbit tier.
MSI makes some quality boards in the $100 range (find their "Military Grade" lines), and ASUS makes some quality ones in the $125 range. I can't really comment on the others.
I wish Soyo was still around making motherboard. They made some really good ones in the $50 range.. but perhaps thats why they went bankrupt.
So you haven't really done any research there? Intel's i5 750 and 760 "steamroll" all the Phenom II X4 CPUs in the price range. Don't trust me, trust benchmarks.
Phenom II X6 chips with Turbo Core in the same price range would like to have a word with you about you cherry picking old X4 chips.
What are you talking about? AM2 boards support AM3 chips.
You also present a false dichotomy, because upgrading isnt ONLY about buying suboptimal hardware and then upgrading it later. Anyone who purchased bleeding edge AM2 gear when it was introduced can get a bios update and then socket an AM3 Phenom II chip. They still only have DDR2, but amazingly Phenom II's support both DDR2 on AM2 and DDR3 on AM3.
So that guy who purchased a dual-core AM2 Phenom when they were cutting edge can now socket a hexa-core AM3 Phenom II.
Its amazing what designing for the future gives your customers. Intel users have only rarely had the chance to substantially upgrade CPU's.
Just to be accurate here, ATT was the one that got bought out.
SBC acquired ATT in 2005, and to prove that they did it just for the name, they re-branded themselves as ATT.
This corruption is legalized. Its called lobbying.
The regulation supports will just blame capitalist interventions into the legislative process when the regulation ends up screwing us all over.
Meanwhile those of us against net neutrality regulation at this time are shaking our heads wondering why so many people want to forever trade their freedom of choice to a bunch of politicians that are sure to meddle with the trust they have placed in them a thousand times over.
Then with this connection you have a 100Mbits line...
Without competition, they have no reason to give you a 100Mbit/sec line.
America is far behind in the bandwidth race with an average of only 3.9Mbit/sec. We rank 18th in the world and there are seemingly no major plans to increase bandwidth further. We are now at a standstill because the monopolies have no reason to improve service.
You are never going to get 100Mbit/sec service from your current providers as things stand right now. Only allowing extensive competition will get us there, and that also moots the whole net neutrality thing.
According to wikipedia, USA already has a solar plants in the 250 MW to 500 MW range.
First, Net Neutrality would do no such thing to entrench a monopoly. And it's a fact that these ISPs often do have either a monopoly or duopoly in the areas.
You don't seem to get it. The government has already entrenched these monopolies, and now instead of fixing the problem, net neutrality proposes to eliminate favoritism. So you can have shitty monopoly service that favors no one. You are forgetting that ITS STILL SHITTY SERVICE.
I'd choose 100MBit broadband with 10MBit traffic shaping any day of the week over the current 10MBit service. Net Neutrality prevents the traffic shaping, but doesnt give you 100MBit service, and in fact makes for a good excuse to never allow any competition.
Unregulated markets tend to function more like a cartel than a true open market. Limiting choices and competition instead of enhancing it.
It is regulation that is preventing competition.
You can bet that there would be a military coupe if that were the case.
Also, the illusion of democracy works better if they keep rotating the figureheads every few years.
Do you have the same objections to the Automatic Updates feature in Windows?
yes. end of story.
Distributed.net isn't making noticeable progress cracking a 72-bit RC5 symmetric key. What makes you think a black-hat botnet will be any more efficient?
You mean besides the fact that each time mozilla ships an update, there is more information available to crack the private key?
It's safer in the case where a computer's administrator may be physically absent for substantial periods of time.
The absolute last thing an administrator does when (s)he leaves for "substantial periods of time" is leave any sort of automatic updating enabled. If the automatic update breaks something, thats the end game nightmare of zero uptime "for a substantial period of time." With firefox this problem is confounded by the fact that it isnt easy to remotely manage the installations (Active Directory) so they couldn't reliably and easily disable such silent updating prior to leaving for that "substantial period of time."
The 'Administrator Isn't Around' card doesnt apply here because of thats the case, the administrator is a fool if he lets things update at all, let alone silently do so.
On the other hand, game developers would have a semi-solid set of numbers to go by, so they can assess the size of their potential market. As it is, there really aren't good numbers on Linux adoption, because no one has to buy a license, everyone can freely copy it from each other, and people collect distro discs that never get used.
I think that you are missing a key element here. Linux users didn't pay anything for their OS. Meanwhile.. Windows users are known to pay for software.
Even if it was a 50/50 split of Linux/Windows, they would STILL target Windows first.
Hello, my name is [competing product to the company I represent] and I dont like the results that come up when people search for my name. You say that you can change the results for a small fee. Where do we...I mean I... send the check?