Reading through these comments, I have a better understanding of the limitations of SSD technology. But I really don't get all the hand-wringing about lifespan/durability, or of the NAND chips being soldered to the motherboard.
1) This is a rumor, so we don't yet know what is actually happening.
2) This is a rumor about the MacBook Air, not to be confused with the MacBook Pro line. The MacBook Air is targeted towards consumers, not pros, and also, is at the low end of the "spec" spectrum. I don't think that the engineers were expecting MBA users to run FCPX (oh dear.), or Logic or Adobe CSx, and anyone who expects to run their MBA that hard is asking for trouble, IMHO. I believe the MBA was designed for people who don't have heavy computing requirements, who just need something streamlined and lightweight to write papers on, surf the web, keep in touch w/ friends & family, etc.
3) If we remove price from the analogy, I see the MBA as a Ferrari, and the MBP as a Landrover. The Ferrari is sleek, smooth, and super cool, but has severe limitations on where it can go, how many passengers it can carry, how much trunk space it has. If you take the Ferrari off-road, it probably won't last very long. The Landrover, on the other hand is very durable, can travel over most kinds of terrain, has a lot more storage and passenger capacity. All these complaints about the SSD on the MBA seem to me to be complaining that the Ferrari won't last as long in off-road terrain as the Landrover. To which I say, "DUH!!!"
Point being, if your needs require the kind of read/write usage that will wear out a SSD in a year or less, then the MBA is probably not for you. Get a MBP. There. Problem solved.
Now, if Apple announced plans to solder flash memory to the motherboards on the MacBook Pro line as well, then I would join in the uproar, because that would be a dumb move on Apple's part. But, as long as flash memory remains on a removable card on the MBP line, then users will have options.
I love the vehicle analogy. I have to agree whole heartedly. However, I would argue that there is a method of integrating this on the MBP that makes sense and would be awesome. Put the OS on a soldered SSD allowing only the OS any write permissions. This would be a pretty small SSD, like only 40GB or less and move any heavy write portions, such as temp and swap, out to the standard storage whether that's SSD or HDD. This would give you the best of both worlds are a pretty affordable price. The majority of the writes to the soldered SSD would simply be OS updates and these are most definitely not frequent enough to present a problem.
In effect 8bit "characters" get mapped to 10 bit values.
Quote:
Originally Posted by ash471
Thanks for the heads up on max read rates for drives that are SATA II. I don't know what 8/10b encoding is, but I'll take your word for it. I'm a biochemist with some III-V semiconductor manufacturing experience, so data encoding is completely out of my expertise.
The point here is that 400 MB/s read is really fast. The reference to 3.0 Gbps or 6.0 Gbps is obviously someone misunderstanding the difference between read speeds and what it means to be SATA II or SATA III compliant.
Yes it is fast relative to old drive technology. However it is extremely slow when compared to what happens in a CPU chip. In reality any improvement helps. Even then part of the SSD experience comes from the massively reduced latency.
As to mSATA their is a card format called mSATA, however it has nothing in common with Apples blade SSD frOm what I can see.
Welcome to Apple. The only guarantee you can ever have is to pony up for the 3-year AppleCare and know your system is covered for 3 years. Beyond that, it's anyone's guess.
In all honesty no one should be buying Mac stuff and not have their 2nd and 3rd year covered. Even if one thing goes wrong like the screen, motherboard or storage, you've got back the cost of your AppleCare. This isn't sales spin, I'm sure many will tell you it's experience.
The failure rate in computers isn't that significant to justify for mist consumers. Those using their computers for school or business have a separate set of issues to deal with. The reality is Apple pushes AppleCare to make money.
When it comes to iPhone you again have a different set of parameters where AppleCare might be justified. As to the rest of the "I" lineup again it is likely an expense that is better if avoided.
For the most part people with multiple Apple hardware would be better off putting the cost of AppleCare in the bank. Especially if you have more than a couple of devices. Many Apple customers would end up with a grand or more in the bank if they did that.
Students and business people can more easily justify the expense of these contracts. Mainly because of the need for rapid turnarounds and the freedom from the DIY effort. Even so a lot of businesses are realizing that sometimes a spare machine makes more sense.
So what I'm saying is that a consumer needs to rationally question the value of these contracts for their personal needs. Many times they are a very bad deal. In my case I've avoided all Apple extended warranties except for on my iPhone as that has become critical for me.
As to mSATA their is a card format called mSATA, however it has nothing in common with Apples blade SSD frOm what I can see.
mSATA is not a card format, it's an interface format which looks like a miniPCIe connector but offers SATA wiring. It's used in Toshiba X-blades (Apple's MBA), Intel SSD 310 series, and plenty other small format SSD units from various manufacturers.
2) This is a rumor about the MacBook Air, not to be confused with the MacBook Pro line. The MacBook Air is targeted towards consumers, not pros, and also, is at the low end of the "spec" spectrum. I don't think that the engineers were expecting MBA users to run FCPX (oh dear.), or Logic or Adobe CSx, and anyone who expects to run their MBA that hard is asking for trouble, IMHO. I believe the MBA was designed for people who don't have heavy computing requirements, who just need something streamlined and lightweight to write papers on, surf the web, keep in touch w/ friends & family, etc.
...
Point being, if your needs require the kind of read/write usage that will wear out a SSD in a year or less, then the MBA is probably not for you. Get a MBP. There. Problem solved.
I thought the myth of the MBA as a netbook equivalent had been busted pretty convincingly. The current crop of 13" MBAs will run a small Lightroom library just fine, as an example.
Reading through these comments, I have a better understanding of the limitations of SSD technology. But I really don't get all the hand-wringing about lifespan/durability, or of the NAND chips being soldered to the motherboard.
This is simple for some people it could be a very big deal. Current flash tech that Apple is using is pretty durable and will work fine for most users. There are more unknowns for the new tech when it come to durability, it may result in fewer people being happy with storage lifespan.
In any case above though soldered in flash would such royally for those that know they will be putting a lot of stress on the SSD. What would be a simple upgrade/repair becomes a motherboard swap out.
Quote:
1) This is a rumor, so we don't yet know what is actually happening.
Yep
Quote:
2) This is a rumor about the MacBook Air, not to be confused with the MacBook Pro line. The MacBook Air is targeted towards consumers, not pros,
I believe this is a mistake many pro users have adopted the AIR. For many business users it is a very good machine.
Quote:
and also, is at the low end of the "spec" spectrum. I don't think that the engineers were expecting MBA users to run FCPX (oh dear.), or Logic or Adobe CSx, and anyone who expects to run their MBA that hard is asking for trouble, IMHO.
That is overly critical if you ask me. The AIR can be used effectively by many professionals. The big problem is that the CPU can fall on it's face if thrown CPU bound apps.
Quote:
I believe the MBA was designed for people who don't have heavy computing requirements, who just need something streamlined and lightweight to write papers on, surf the web, keep in touch w/ friends & family, etc.
It is a general purpose computer nothing more and more importantly nothing less.
Quote:
3) If we remove price from the analogy, I see the MBA as a Ferrari, and the MBP as a Landrover. The Ferrari is sleek, smooth, and super cool, but has severe limitations on where it can go, how many passengers it can carry, how much trunk space it has. If you take the Ferrari off-road, it probably won't last very long. The Landrover, on the other hand is very durable, can travel over most kinds of terrain, has a lot more storage and passenger capacity. All these complaints about the SSD on the MBA seem to me to be complaining that the Ferrari won't last as long in off-road terrain as the Landrover. To which I say, "DUH!!!"
They are nothing of the sort. The legitimate complaint here is that soldered in flash changes the repair upgrade equation significantly. Both the Landrover and the Farrari break down from time to time, you wouldn't be to happy if those parts where all welded in place.
Quote:
Point being, if your needs require the kind of read/write usage that will wear out a SSD in a year or less, then the MBA is probably not for you. Get a MBP. There. Problem solved.
No the problem is that if you wear out the SSD on the current AIR you can plug a new one in. On the speculated AIR that might be impossible.
Quote:
Now, if Apple announced plans to solder flash memory to the motherboards on the MacBook Pro line as well, then I would join in the uproar, because that would be a dumb move on Apple's part. But, as long as flash memory remains on a removable card on the MBP line, then users will have options.
Why is it dumb on one machine and not another?
You seem to imply that the AIR is some sort of toy not suitable for serious use. This is not the case at all. AIR can actually be impressive given that it does work so well for many users.
The second point is true, the first isn't necessarily. If the processor finishes a task faster and gets to idle longer, battery life could be positively impacted vs it processing at a lower TDP for a longer time.
I come from the "there is no free lunch" school of engineering. It is certainly possible that a clock boost will result in less overall power usage but I don't think that is a common outcome.
Quote:
For longer term tasks like encoding I can see the second point becoming an issue, but TB should help out with brief bursts of processing, for example web browsing.
Yes there will certainly be many cases where Turbo Boost will help out. The problem is all of the unknowns. For example how long the boost can be maintained before thermal loading causes throttling. At this point we don't even know if Apple will allow for a clock boost nor by how much. (it would be stupid not to but you never know)
Turbo boost will be a positive for the AIR I'm just not sure how positive it will be. Currently turbo boost has a significant impact on power usage on Intel hardware.
I look at this way; Sandy Bridge should improve the AIRs CPU performance. I don't think anybody would argue against that. The question is how much of an improvement.
I see comparisons being made to MBPs which I think is just a bit overly optimistic. I just don't see sustained performance coming anywhere near the MBPs. Of course I'm certain that somebody will come up with a single threaded bench mark that never throttles to prove their point. In the more general sense of modern day users where multithreaded apps and numerous apps are running I'm not so convinced.
Do you expect to do home repairs on your $600 iPhone or iPad? If you need ultraportability then you sacrifice ease of servicing, that's just how it is. If you want something that you can service yourself then'll always be best with a full sized laptop.
Unfortunately, the serviceability does not apply when going from a full sized laptop to a desktop like the iMac. In fact, the MacBook Pro is more accessible than the iMac.
Why would Apple use Toggle Mode DDR 2.0 if it is only 400Mbps? Wouldn't that be slower than the ssd in the current MBA? Maybe it gives a faster seek time????
P.S. There was no need to post the link to wikipedia. I purposely said I would take your word for it. Your contribution to the thread was that read speeds are even slower than one might think after correcting for bits and bytes. That was enough for me.
The Link was there in case anyone or you would want to learn more.
400Mbps was per NAND, SSD are made of Multiple NAND Chip, you could make an 400MBps if you want by using 10+ of these with a controller capable of doing it.
So even if Apple did use these new fast NAND, without an decent controller it will still be as fast / slow as the current one.
The point is, the flash memory, NAND, is 400Mbps, not 400MBps.
What I'm concerned about with flash drives is that the methods used today to even out the writes will begin to lose effectiveness in later drives.
Right now, the firmware for this purpose is pretty sophisticated, and does a good job of evening out the writes to all cells so as to increase the effective lifetime of the drives.
But we've seen a chips' life drop from 10,000 writes, to 5,000 writes, and now to 3,000 writes. At a point in time, if this drop continues unabated with smaller and faster chips, there will be a time where the firmware won't be able to keep up. Firmware is just software encoded on the chip itself. While amazing efficiencies have been managed here, once you get to a certain point, there's little more to be gained. Once the chip life drops below what firmware gains can be eked out, drive life will drop.
We are no more than two process cycles away from that point.
Several years ago, Samsung announced that they had pretty much solved that problem with flash memory with their own development, and that theirs was faster, and had an order of magnitude longer life. I haven't heard much about it for a couple of years now. I wonder if they've finished it. It was supposed to be on the market in 2010. The page I bookmarked for that is giving me a "404" error, so I can't find that article.
What I'm concerned about with flash drives is that the methods used today to even out the writes will begin to lose effectiveness in later drives.
Right now, the firmware for this purpose is pretty sophisticated, and does a good job of evening out the writes to all cells so as to increase the effective lifetime of the drives.
But we've seen a chips' life drop from 10,000 writes, to 5,000 writes, and now to 3,000 writes. At a point in time, if this drop continues unabated with smaller and faster chips, there will be a time where the firmware won't be able to keep up. Firmware is just software encoded on the chip itself. While amazing efficiencies have been managed here, once you get to a certain point, there's little more to be gained. Once the chip life drops below what firmware gains can be eked out, drive life will drop.
We are no more than two process cycles away from that point.
Several years ago, Samsung announced that they had pretty much solved that problem with flash memory with their own development, and that theirs was faster, and had an order of magnitude longer life. I haven't heard much about it for a couple of years now. I wonder if they've finished it. It was supposed to be on the market in 2010. The page I bookmarked for that is giving me a "404" error, so I can't find that article.
The problem has been greatly exaggerated. Under normal usage, you willNEVER be able to hit the current limit within the lifespan of your mac or your pc. To give you an example, on a 16 NAND devices 100 GB SSD, if you write in 2GB per day, in the WORST case scenario it will take 20 years for you to fill the 3000 write cycle.
20nm is projected to have 3000 write cycle as well, so two node down we could have 1000 cycle and we should still manage, another node down we could go back to SLC. That is still 4 generation node span over the next decade within the current technology possibilities, not taken into account any other improvement made during this time.
The problem has been greatly exaggerated. Under normal usage, you willNEVER be able to hit the current limit within the lifespan of your mac or your pc. To give you an example, on a 16 NAND devices 100 GB SSD, if you write in 2GB per day, in the WORST case scenario it will take 20 years for you to fill the 3000 write cycle.
20nm is projected to have 3000 write cycle as well, so two node down we could have 1000 cycle and we should still manage, another node down we could go back to SLC. That is still 4 generation node span over the next decade within the current technology possibilities, not taken into account any other improvement made during this time.
That's not how it works. You don't have to fill each block completely, you can write as little as 1 byte and a whole 128kb block will have to be erased and rewritten. As I posted a while ago over at MacRumors, for a 128gb drive, thats just one million 1 byte writes across the disk. 1 million bytes is about 1 megabyte. So that's just 1mb written, and you've used up one write cycle on every block on the drive. Of course that's an extreme case, there are hefty caches to prevent small writes, and you don't generally go around writing millions of single bytes across the whole disk.
Even so, as I said above, I've had my SSD for around a year, and i've already used up 20% of its life. Again at the risk of repeating myself, that's with virtual memory off, and my media on a separate drive. I'm careful not to write to the SSD too much, but the average user wouldn't know to be concerned about that.
The problem has been greatly exaggerated. Under normal usage, you willNEVER be able to hit the current limit within the lifespan of your mac or your pc. To give you an example, on a 16 NAND devices 100 GB SSD, if you write in 2GB per day, in the WORST case scenario it will take 20 years for you to fill the 3000 write cycle.
20nm is projected to have 3000 write cycle as well, so two node down we could have 1000 cycle and we should still manage, another node down we could go back to SLC. That is still 4 generation node span over the next decade within the current technology possibilities, not taken into account any other improvement made during this time.
That's not true. It also depends on how full your drive is. This is a complex subject.
What if your 100GB drive has 40GB data on it? how about 60GB? What about 75GB?
SSD's are estimated to have about a TEN year life, as I mentioned before. But we really don't know what we'll see several years from now.
That's not true. It also depends on how full your drive is. This is a complex subject.
It gets even deeper than that as the file system and the OS affect wear. Plus the user has considerable impact on wear.
Quote:
What if your 100GB drive has 40GB data on it? how about 60GB? What about 75GB?
SSD's are estimated to have about a TEN year life, as I mentioned before. But we really don't know what we'll see several years from now.
That 10 year life is for drives built on 40nm tech. We really don't know how well chips built on sub 20nm tech will hold up. Beyond that these high density chips do lose their memory
That's not how it works. You don't have to fill each block completely, you can write as little as 1 byte and a whole 128kb block will have to be erased and rewritten. As I posted a while ago over at MacRumors, for a 128gb drive, thats just one million 1 byte writes across the disk. 1 million bytes is about 1 megabyte. So that's just 1mb written, and you've used up one write cycle on every block on the drive. Of course that's an extreme case, there are hefty caches to prevent small writes, and you don't generally go around writing millions of single bytes across the whole disk.
Even so, as I said above, I've had my SSD for around a year, and i've already used up 20% of its life. Again at the risk of repeating myself, that's with virtual memory off, and my media on a separate drive. I'm careful not to write to the SSD too much, but the average user wouldn't know to be concerned about that.
Yes but as you have said it your self, your extreme cases will never happen in real life due to memcache for small write.
Of coz it depends on usage. With 8GB RAM. No Pagefile. No Index, I am just over 10% after a year. And that is with some low usage of BT on my SSD.
That's not true. It also depends on how full your drive is. This is a complex subject.
What if your 100GB drive has 40GB data on it? how about 60GB? What about 75GB?
SSD's are estimated to have about a TEN year life, as I mentioned before. But we really don't know what we'll see several years from now.
Your SSD Controller Software is suppose to shift through files for maximum write cycle. So your "constant" data wont be just sitting on a single NAND block for its entire life.
Yes but as you have said it your self, your extreme cases will never happen in real life due to memcache for small write.
The extreme cases no, but it was just to illustrate that it's not as clear cut as "you have to fill the drive up 3000 times to wear it out". The OS does write small bits and pieces to the drive all the time, which soon add up. Especially when you can round up a significant number of writes to 128kb, the erase block size.
Quote:
Originally Posted by ksec
Of coz it depends on usage. With 8GB RAM. No Pagefile. No Index, I am just over 10% after a year. And that is with some low usage of BT on my SSD.
You contradict yourself, earlier you said:
Quote:
Originally Posted by ksec
In the WORST case scenario it will take 20 years for you to fill the 3000 write cycle.
If that's the case, then why have you already used 10% of your erase cycles after a year? That's assuming you have a SSD with 3000 erase cycles. If it's 5000 and you've used 10%, the problem's even worse, as the number of erase cycles seems to be reducing as technology progresses, accelerating the wear. After 10 years your SSD would be defunct, even though you've turned off some of the features that cause wear. You can't have 8GB RAM in a MBA, and as standard you only get 2GB. There's going to be a lot of pageouts, and that'll cause tremendous wear on the SSD. Most users aren't tech-savvy, so they won't turn off indexing and VM. They'll just use the computer like a HDD based one, editing movies, downloading photos, music etc.
It gets even deeper than that as the file system and the OS affect wear. Plus the user has considerable impact on wear.
That 10 year life is for drives built on 40nm tech. We really don't know how well chips built on sub 20nm tech will hold up. Beyond that these high density chips do lose their memory
Yup. It's possible that things could get worse. I'm trying to be optimistic.
Your SSD Controller Software is suppose to shift through files for maximum write cycle. So your "constant" data wont be just sitting on a single NAND block for its entire life.
But all that means is that there is more info that's got to be moved, which means written. Wear leveling isn't perfect, but it's gotten pretty good. You have to understand that if it becomes close to perfect, and write life continues to shrink, there's nothing they can do other than to keep adding extra NAND to the drive, and that brings the price up.
Comments
Reading through these comments, I have a better understanding of the limitations of SSD technology. But I really don't get all the hand-wringing about lifespan/durability, or of the NAND chips being soldered to the motherboard.
1) This is a rumor, so we don't yet know what is actually happening.
2) This is a rumor about the MacBook Air, not to be confused with the MacBook Pro line. The MacBook Air is targeted towards consumers, not pros, and also, is at the low end of the "spec" spectrum. I don't think that the engineers were expecting MBA users to run FCPX (oh dear.), or Logic or Adobe CSx, and anyone who expects to run their MBA that hard is asking for trouble, IMHO. I believe the MBA was designed for people who don't have heavy computing requirements, who just need something streamlined and lightweight to write papers on, surf the web, keep in touch w/ friends & family, etc.
3) If we remove price from the analogy, I see the MBA as a Ferrari, and the MBP as a Landrover. The Ferrari is sleek, smooth, and super cool, but has severe limitations on where it can go, how many passengers it can carry, how much trunk space it has. If you take the Ferrari off-road, it probably won't last very long. The Landrover, on the other hand is very durable, can travel over most kinds of terrain, has a lot more storage and passenger capacity. All these complaints about the SSD on the MBA seem to me to be complaining that the Ferrari won't last as long in off-road terrain as the Landrover. To which I say, "DUH!!!"
Point being, if your needs require the kind of read/write usage that will wear out a SSD in a year or less, then the MBA is probably not for you. Get a MBP. There. Problem solved.
Now, if Apple announced plans to solder flash memory to the motherboards on the MacBook Pro line as well, then I would join in the uproar, because that would be a dumb move on Apple's part. But, as long as flash memory remains on a removable card on the MBP line, then users will have options.
I love the vehicle analogy. I have to agree whole heartedly. However, I would argue that there is a method of integrating this on the MBP that makes sense and would be awesome. Put the OS on a soldered SSD allowing only the OS any write permissions. This would be a pretty small SSD, like only 40GB or less and move any heavy write portions, such as temp and swap, out to the standard storage whether that's SSD or HDD. This would give you the best of both worlds are a pretty affordable price. The majority of the writes to the soldered SSD would simply be OS updates and these are most definitely not frequent enough to present a problem.
Thanks for the heads up on max read rates for drives that are SATA II. I don't know what 8/10b encoding is, but I'll take your word for it. I'm a biochemist with some III-V semiconductor manufacturing experience, so data encoding is completely out of my expertise.
The point here is that 400 MB/s read is really fast. The reference to 3.0 Gbps or 6.0 Gbps is obviously someone misunderstanding the difference between read speeds and what it means to be SATA II or SATA III compliant.
Yes it is fast relative to old drive technology. However it is extremely slow when compared to what happens in a CPU chip. In reality any improvement helps. Even then part of the SSD experience comes from the massively reduced latency.
As to mSATA their is a card format called mSATA, however it has nothing in common with Apples blade SSD frOm what I can see.
Welcome to Apple. The only guarantee you can ever have is to pony up for the 3-year AppleCare and know your system is covered for 3 years. Beyond that, it's anyone's guess.
In all honesty no one should be buying Mac stuff and not have their 2nd and 3rd year covered. Even if one thing goes wrong like the screen, motherboard or storage, you've got back the cost of your AppleCare. This isn't sales spin, I'm sure many will tell you it's experience.
The failure rate in computers isn't that significant to justify for mist consumers. Those using their computers for school or business have a separate set of issues to deal with. The reality is Apple pushes AppleCare to make money.
When it comes to iPhone you again have a different set of parameters where AppleCare might be justified. As to the rest of the "I" lineup again it is likely an expense that is better if avoided.
For the most part people with multiple Apple hardware would be better off putting the cost of AppleCare in the bank. Especially if you have more than a couple of devices. Many Apple customers would end up with a grand or more in the bank if they did that.
Students and business people can more easily justify the expense of these contracts. Mainly because of the need for rapid turnarounds and the freedom from the DIY effort. Even so a lot of businesses are realizing that sometimes a spare machine makes more sense.
So what I'm saying is that a consumer needs to rationally question the value of these contracts for their personal needs. Many times they are a very bad deal. In my case I've avoided all Apple extended warranties except for on my iPhone as that has become critical for me.
As to mSATA their is a card format called mSATA, however it has nothing in common with Apples blade SSD frOm what I can see.
mSATA is not a card format, it's an interface format which looks like a miniPCIe connector but offers SATA wiring. It's used in Toshiba X-blades (Apple's MBA), Intel SSD 310 series, and plenty other small format SSD units from various manufacturers.
2) This is a rumor about the MacBook Air, not to be confused with the MacBook Pro line. The MacBook Air is targeted towards consumers, not pros, and also, is at the low end of the "spec" spectrum. I don't think that the engineers were expecting MBA users to run FCPX (oh dear.), or Logic or Adobe CSx, and anyone who expects to run their MBA that hard is asking for trouble, IMHO. I believe the MBA was designed for people who don't have heavy computing requirements, who just need something streamlined and lightweight to write papers on, surf the web, keep in touch w/ friends & family, etc.
...
Point being, if your needs require the kind of read/write usage that will wear out a SSD in a year or less, then the MBA is probably not for you. Get a MBP. There. Problem solved.
I thought the myth of the MBA as a netbook equivalent had been busted pretty convincingly. The current crop of 13" MBAs will run a small Lightroom library just fine, as an example.
Reading through these comments, I have a better understanding of the limitations of SSD technology. But I really don't get all the hand-wringing about lifespan/durability, or of the NAND chips being soldered to the motherboard.
This is simple for some people it could be a very big deal. Current flash tech that Apple is using is pretty durable and will work fine for most users. There are more unknowns for the new tech when it come to durability, it may result in fewer people being happy with storage lifespan.
In any case above though soldered in flash would such royally for those that know they will be putting a lot of stress on the SSD. What would be a simple upgrade/repair becomes a motherboard swap out.
1) This is a rumor, so we don't yet know what is actually happening.
Yep
2) This is a rumor about the MacBook Air, not to be confused with the MacBook Pro line. The MacBook Air is targeted towards consumers, not pros,
I believe this is a mistake many pro users have adopted the AIR. For many business users it is a very good machine.
and also, is at the low end of the "spec" spectrum. I don't think that the engineers were expecting MBA users to run FCPX (oh dear.), or Logic or Adobe CSx, and anyone who expects to run their MBA that hard is asking for trouble, IMHO.
That is overly critical if you ask me. The AIR can be used effectively by many professionals. The big problem is that the CPU can fall on it's face if thrown CPU bound apps.
I believe the MBA was designed for people who don't have heavy computing requirements, who just need something streamlined and lightweight to write papers on, surf the web, keep in touch w/ friends & family, etc.
It is a general purpose computer nothing more and more importantly nothing less.
3) If we remove price from the analogy, I see the MBA as a Ferrari, and the MBP as a Landrover. The Ferrari is sleek, smooth, and super cool, but has severe limitations on where it can go, how many passengers it can carry, how much trunk space it has. If you take the Ferrari off-road, it probably won't last very long. The Landrover, on the other hand is very durable, can travel over most kinds of terrain, has a lot more storage and passenger capacity. All these complaints about the SSD on the MBA seem to me to be complaining that the Ferrari won't last as long in off-road terrain as the Landrover. To which I say, "DUH!!!"
They are nothing of the sort. The legitimate complaint here is that soldered in flash changes the repair upgrade equation significantly. Both the Landrover and the Farrari break down from time to time, you wouldn't be to happy if those parts where all welded in place.
Point being, if your needs require the kind of read/write usage that will wear out a SSD in a year or less, then the MBA is probably not for you. Get a MBP. There. Problem solved.
No the problem is that if you wear out the SSD on the current AIR you can plug a new one in. On the speculated AIR that might be impossible.
Now, if Apple announced plans to solder flash memory to the motherboards on the MacBook Pro line as well, then I would join in the uproar, because that would be a dumb move on Apple's part. But, as long as flash memory remains on a removable card on the MBP line, then users will have options.
Why is it dumb on one machine and not another?
You seem to imply that the AIR is some sort of toy not suitable for serious use. This is not the case at all. AIR can actually be impressive given that it does work so well for many users.
The second point is true, the first isn't necessarily. If the processor finishes a task faster and gets to idle longer, battery life could be positively impacted vs it processing at a lower TDP for a longer time.
I come from the "there is no free lunch" school of engineering. It is certainly possible that a clock boost will result in less overall power usage but I don't think that is a common outcome.
For longer term tasks like encoding I can see the second point becoming an issue, but TB should help out with brief bursts of processing, for example web browsing.
Yes there will certainly be many cases where Turbo Boost will help out. The problem is all of the unknowns. For example how long the boost can be maintained before thermal loading causes throttling. At this point we don't even know if Apple will allow for a clock boost nor by how much. (it would be stupid not to but you never know)
Turbo boost will be a positive for the AIR I'm just not sure how positive it will be. Currently turbo boost has a significant impact on power usage on Intel hardware.
I look at this way; Sandy Bridge should improve the AIRs CPU performance. I don't think anybody would argue against that. The question is how much of an improvement.
I see comparisons being made to MBPs which I think is just a bit overly optimistic. I just don't see sustained performance coming anywhere near the MBPs. Of course I'm certain that somebody will come up with a single threaded bench mark that never throttles to prove their point. In the more general sense of modern day users where multithreaded apps and numerous apps are running I'm not so convinced.
Do you expect to do home repairs on your $600 iPhone or iPad? If you need ultraportability then you sacrifice ease of servicing, that's just how it is. If you want something that you can service yourself then'll always be best with a full sized laptop.
Unfortunately, the serviceability does not apply when going from a full sized laptop to a desktop like the iMac. In fact, the MacBook Pro is more accessible than the iMac.
Why would Apple use Toggle Mode DDR 2.0 if it is only 400Mbps? Wouldn't that be slower than the ssd in the current MBA? Maybe it gives a faster seek time????
P.S. There was no need to post the link to wikipedia. I purposely said I would take your word for it. Your contribution to the thread was that read speeds are even slower than one might think after correcting for bits and bytes. That was enough for me.
The Link was there in case anyone or you would want to learn more.
400Mbps was per NAND, SSD are made of Multiple NAND Chip, you could make an 400MBps if you want by using 10+ of these with a controller capable of doing it.
So even if Apple did use these new fast NAND, without an decent controller it will still be as fast / slow as the current one.
The point is, the flash memory, NAND, is 400Mbps, not 400MBps.
Right now, the firmware for this purpose is pretty sophisticated, and does a good job of evening out the writes to all cells so as to increase the effective lifetime of the drives.
But we've seen a chips' life drop from 10,000 writes, to 5,000 writes, and now to 3,000 writes. At a point in time, if this drop continues unabated with smaller and faster chips, there will be a time where the firmware won't be able to keep up. Firmware is just software encoded on the chip itself. While amazing efficiencies have been managed here, once you get to a certain point, there's little more to be gained. Once the chip life drops below what firmware gains can be eked out, drive life will drop.
We are no more than two process cycles away from that point.
Several years ago, Samsung announced that they had pretty much solved that problem with flash memory with their own development, and that theirs was faster, and had an order of magnitude longer life. I haven't heard much about it for a couple of years now. I wonder if they've finished it. It was supposed to be on the market in 2010. The page I bookmarked for that is giving me a "404" error, so I can't find that article.
What I'm concerned about with flash drives is that the methods used today to even out the writes will begin to lose effectiveness in later drives.
Right now, the firmware for this purpose is pretty sophisticated, and does a good job of evening out the writes to all cells so as to increase the effective lifetime of the drives.
But we've seen a chips' life drop from 10,000 writes, to 5,000 writes, and now to 3,000 writes. At a point in time, if this drop continues unabated with smaller and faster chips, there will be a time where the firmware won't be able to keep up. Firmware is just software encoded on the chip itself. While amazing efficiencies have been managed here, once you get to a certain point, there's little more to be gained. Once the chip life drops below what firmware gains can be eked out, drive life will drop.
We are no more than two process cycles away from that point.
Several years ago, Samsung announced that they had pretty much solved that problem with flash memory with their own development, and that theirs was faster, and had an order of magnitude longer life. I haven't heard much about it for a couple of years now. I wonder if they've finished it. It was supposed to be on the market in 2010. The page I bookmarked for that is giving me a "404" error, so I can't find that article.
The problem has been greatly exaggerated. Under normal usage, you will NEVER be able to hit the current limit within the lifespan of your mac or your pc. To give you an example, on a 16 NAND devices 100 GB SSD, if you write in 2GB per day, in the WORST case scenario it will take 20 years for you to fill the 3000 write cycle.
20nm is projected to have 3000 write cycle as well, so two node down we could have 1000 cycle and we should still manage, another node down we could go back to SLC. That is still 4 generation node span over the next decade within the current technology possibilities, not taken into account any other improvement made during this time.
The problem has been greatly exaggerated. Under normal usage, you will NEVER be able to hit the current limit within the lifespan of your mac or your pc. To give you an example, on a 16 NAND devices 100 GB SSD, if you write in 2GB per day, in the WORST case scenario it will take 20 years for you to fill the 3000 write cycle.
20nm is projected to have 3000 write cycle as well, so two node down we could have 1000 cycle and we should still manage, another node down we could go back to SLC. That is still 4 generation node span over the next decade within the current technology possibilities, not taken into account any other improvement made during this time.
That's not how it works. You don't have to fill each block completely, you can write as little as 1 byte and a whole 128kb block will have to be erased and rewritten. As I posted a while ago over at MacRumors, for a 128gb drive, thats just one million 1 byte writes across the disk. 1 million bytes is about 1 megabyte. So that's just 1mb written, and you've used up one write cycle on every block on the drive. Of course that's an extreme case, there are hefty caches to prevent small writes, and you don't generally go around writing millions of single bytes across the whole disk.
Even so, as I said above, I've had my SSD for around a year, and i've already used up 20% of its life. Again at the risk of repeating myself, that's with virtual memory off, and my media on a separate drive. I'm careful not to write to the SSD too much, but the average user wouldn't know to be concerned about that.
The problem has been greatly exaggerated. Under normal usage, you will NEVER be able to hit the current limit within the lifespan of your mac or your pc. To give you an example, on a 16 NAND devices 100 GB SSD, if you write in 2GB per day, in the WORST case scenario it will take 20 years for you to fill the 3000 write cycle.
20nm is projected to have 3000 write cycle as well, so two node down we could have 1000 cycle and we should still manage, another node down we could go back to SLC. That is still 4 generation node span over the next decade within the current technology possibilities, not taken into account any other improvement made during this time.
That's not true. It also depends on how full your drive is. This is a complex subject.
What if your 100GB drive has 40GB data on it? how about 60GB? What about 75GB?
SSD's are estimated to have about a TEN year life, as I mentioned before. But we really don't know what we'll see several years from now.
That's not true. It also depends on how full your drive is. This is a complex subject.
It gets even deeper than that as the file system and the OS affect wear. Plus the user has considerable impact on wear.
What if your 100GB drive has 40GB data on it? how about 60GB? What about 75GB?
SSD's are estimated to have about a TEN year life, as I mentioned before. But we really don't know what we'll see several years from now.
That 10 year life is for drives built on 40nm tech. We really don't know how well chips built on sub 20nm tech will hold up. Beyond that these high density chips do lose their memory
That's not how it works. You don't have to fill each block completely, you can write as little as 1 byte and a whole 128kb block will have to be erased and rewritten. As I posted a while ago over at MacRumors, for a 128gb drive, thats just one million 1 byte writes across the disk. 1 million bytes is about 1 megabyte. So that's just 1mb written, and you've used up one write cycle on every block on the drive. Of course that's an extreme case, there are hefty caches to prevent small writes, and you don't generally go around writing millions of single bytes across the whole disk.
Even so, as I said above, I've had my SSD for around a year, and i've already used up 20% of its life. Again at the risk of repeating myself, that's with virtual memory off, and my media on a separate drive. I'm careful not to write to the SSD too much, but the average user wouldn't know to be concerned about that.
Yes but as you have said it your self, your extreme cases will never happen in real life due to memcache for small write.
Of coz it depends on usage. With 8GB RAM. No Pagefile. No Index, I am just over 10% after a year. And that is with some low usage of BT on my SSD.
That's not true. It also depends on how full your drive is. This is a complex subject.
What if your 100GB drive has 40GB data on it? how about 60GB? What about 75GB?
SSD's are estimated to have about a TEN year life, as I mentioned before. But we really don't know what we'll see several years from now.
Your SSD Controller Software is suppose to shift through files for maximum write cycle. So your "constant" data wont be just sitting on a single NAND block for its entire life.
That is overly critical if you ask me. The AIR can be used effectively by many professionals.
Pro = what I do for a living
Non-Pro = what everyone else does for a living
You seem to imply that the AIR is some sort of toy not suitable for serious use.
Serious machine = what I just shelled out $$$ to buy
Toy = what other folks bought
Once you have the decoder ring it's kinda easy to figure out where 90% of the arguments are about.
Yes but as you have said it your self, your extreme cases will never happen in real life due to memcache for small write.
The extreme cases no, but it was just to illustrate that it's not as clear cut as "you have to fill the drive up 3000 times to wear it out". The OS does write small bits and pieces to the drive all the time, which soon add up. Especially when you can round up a significant number of writes to 128kb, the erase block size.
Of coz it depends on usage. With 8GB RAM. No Pagefile. No Index, I am just over 10% after a year. And that is with some low usage of BT on my SSD.
You contradict yourself, earlier you said:
In the WORST case scenario it will take 20 years for you to fill the 3000 write cycle.
If that's the case, then why have you already used 10% of your erase cycles after a year? That's assuming you have a SSD with 3000 erase cycles. If it's 5000 and you've used 10%, the problem's even worse, as the number of erase cycles seems to be reducing as technology progresses, accelerating the wear. After 10 years your SSD would be defunct, even though you've turned off some of the features that cause wear. You can't have 8GB RAM in a MBA, and as standard you only get 2GB. There's going to be a lot of pageouts, and that'll cause tremendous wear on the SSD. Most users aren't tech-savvy, so they won't turn off indexing and VM. They'll just use the computer like a HDD based one, editing movies, downloading photos, music etc.
It gets even deeper than that as the file system and the OS affect wear. Plus the user has considerable impact on wear.
That 10 year life is for drives built on 40nm tech. We really don't know how well chips built on sub 20nm tech will hold up. Beyond that these high density chips do lose their memory
Yup. It's possible that things could get worse. I'm trying to be optimistic.
Your SSD Controller Software is suppose to shift through files for maximum write cycle. So your "constant" data wont be just sitting on a single NAND block for its entire life.
But all that means is that there is more info that's got to be moved, which means written. Wear leveling isn't perfect, but it's gotten pretty good. You have to understand that if it becomes close to perfect, and write life continues to shrink, there's nothing they can do other than to keep adding extra NAND to the drive, and that brings the price up.