I read that SSD should evolve at a rate that's slightly faster than Moore's Law which is good. I also read that they will hit a wall at 20nm. The next SSD should be at 34nm from a few vendors like Intel/Nec so I figure they'll end up at something like 28nm for the 2011 batch and then 20nm in 2013?
While some are arguing about the 18-20 nm wall, many researchers are thinking that it may come to pass. not so much because of technical impossibility, but because of cost. It's possible that it will come out, but years after it would be expected, going by, as you say, "Moore's Law".
Maybe we won't see it until 2015.
At any rate, new technologies may come into production by then. There are a few that are looking better every day.
While some are arguing about the 18-20 nm wall, many researchers are thinking that it may come to pass. not so much because of technical impossibility, but because of cost. It's possible that it will come out, but years after it would be expected, going by, as you say, "Moore's Law".
Maybe we won't see it until 2015.
At any rate, new technologies may come into production by then. There are a few that are looking better every day.
My guess is that we'll just see storage branch out horizontally just as CPU did when they hit the wall. Because SSD are power efficient I expect to see something like PCI Express or even some other type of fast bus connect SSD in RAID like configuration.
My guess is that we'll just see storage branch out horizontally just as CPU did when they hit the wall. Because SSD are power efficient I expect to see something like PCI Express or even some other type of fast bus connect SSD in RAID like configuration.
There are a couple like that out now, but they're expensive, and performance is MEH. But it will get better as time goes on.
If it weren't for so many crappy controllers on the market, things would be better now.
But I expect prices at half the levels we see them at today within a year.
My guess is that we'll just see storage branch out horizontally just as CPU did when they hit the wall.
Unfortunately, this analogy doesn't work. "The wall" in CPUs was clock speed - the number of transistors in a given physical area had been doubling approximately every 18 months as Moore had predicted - what Intel had been doing was relying on the smaller size of an individual transistor allowing it to operate at a faster rate, and using a few extra transistors in a CPU to extend functional unit pipelines. The problem was that making pipelines longer and longer meant they took larger and larger hits for mis-predicting code branches, and they happen rather a lot on multi-tasking general computing platforms. Another problem was higher clock speeds increasing power consumption and creating hot spots in the CPU - areas so small that it was hard to extract the heat fast enough to allow reliable operation. So now, instead of making CPUs with a die that gets ever physically smaller and higher clocked (same number of transistors from one generation to the next, just each transistor is smaller), the physical size of the die stays about the same, but the smaller size of each transistor allows for more cores to fit in the same area. The multi-core approach certainly gives the consumer better value for money!
But the key here is to appreciate that whichever path the CPU industry had taken (higher clock speed or more cores), the size of the transistors would be getting smaller.
With SSDs (and CPUs, for that matter), there will be a minimum size that it's possible to make a traditional transistor. Once that wall is hit, there's no "alternative", "horizontal route" that SSDs can take - more capacity requires more transistors, and that's that. The wall exists because at some point the gate-oxide layer will reach one atom thick, and you can't get thinner, and traditional transistors rely on the bulk statistical behaviour of many, many electrons. The smaller a transistor, the smaller the number of electrons there are involved in the operation of the device. As the transistor shrinks, eventually you'll reach the point where there are too few electrons and the transistor no longer works in a predictable, deterministic manner.
Fortunately, there's spintronics or something else equally mind-boggling to save the day when the wall is hit with traditional transistors.
Unfortunately, this analogy doesn't work. "The wall" in CPUs was clock speed - the number of transistors in a given physical area had been doubling approximately every 18 months as Moore had predicted - what Intel had been doing was relying on the smaller size of an individual transistor allowing it to operate at a faster rate, and using a few extra transistors in a CPU to extend functional unit pipelines. The problem was that making pipelines longer and longer meant they took larger and larger hits for mis-predicting code branches, and they happen rather a lot on multi-tasking general computing platforms. Another problem was higher clock speeds increasing power consumption and creating hot spots in the CPU - areas so small that it was hard to extract the heat fast enough to allow reliable operation. So now, instead of making CPUs with a die that gets ever physically smaller and higher clocked (same number of transistors from one generation to the next, just each transistor is smaller), the physical size of the die stays about the same, but the smaller size of each transistor allows for more cores to fit in the same area. The multi-core approach certainly gives the consumer better value for money!
But the key here is to appreciate that whichever path the CPU industry had taken (higher clock speed or more cores), the size of the transistors would be getting smaller.
With SSDs (and CPUs, for that matter), there will be a minimum size that it's possible to make a traditional transistor. Once that wall is hit, there's no "alternative", "horizontal route" that SSDs can take - more capacity requires more transistors, and that's that. The wall exists because at some point the gate-oxide layer will reach one atom thick, and you can't get thinner, and traditional transistors rely on the bulk statistical behaviour of many, many electrons. The smaller a transistor, the smaller the number of electrons there are involved in the operation of the device. As the transistor shrinks, eventually you'll reach the point where there are too few electrons and the transistor no longer works in a predictable, deterministic manner.
Fortunately, there's spintronics or something else equally mind-boggling to save the day when the wall is hit with traditional transistors.
I was going to edit your post down before responding but decided not to.
But I'll respond to one thing, transistor size.
The thing you're forgetting is that there is a "new" thing in chip manufacture that's becoming possible, and that's talked about for use particularly with memory devices, and that's 3D architecture.
We're already seeing it in some, and it will become used more often, and will become more important as time goes on. This will allow a greater number of cells in the same space. Cooling requirements go up, but they have solutions for that as well.
Once that "wall" in size is reached, and until newer technologies are available, we'll continue to see greater amounts of memory per chip for at least a while.
The thing you're forgetting is that there is a "new" thing in chip manufacture that's becoming possible, and that's talked about for use particularly with memory devices, and that's 3D architecture.
We're already seeing it in some, and it will become used more often, and will become more important as time goes on. This will allow a greater number of cells in the same space. Cooling requirements go up, but they have solutions for that as well.
Once that "wall" in size is reached, and until newer technologies are available, we'll continue to see greater amounts of memory per chip for at least a while.
Good point. I've not read much about it but must assume that the heat problem is not at all trivial. Certainly you're not going to be able to stack 20 layers of transistors on top of each other. Hopefully 3D will bridge the gap from electronics to spintronics or whatever.
Yes I keep hearing about "chip stacking" that may alleviate some of the issues.
I think Melgross is talking about something else.
A chip consists of a die inside a plastic package, with hair-thin wires connecting the comparatively giant pins that stick out of the package, to the pads around the edge of the die. Chip stacking is where you take several dice and package them together in a single package - for example I believe the Samsung SOCs in the iPhone have the RAM on one die and the CPU on another. The two dice are stacked on top of each other, and then packaged into the single plastic package you see on the iPhone motherboard.
The 3D that Melgross is talking about is, I believe, about making several layers of transistors on a single die.
Good point. I've not read much about it but must assume that the heat problem is not at all trivial. Certainly you're not going to be able to stack 20 layers of transistors on top of each other. Hopefully 3D will bridge the gap from electronics to spintronics or whatever.
Hopefully, they WILL be able to stack 20 layers of transistors on top of each other. If they can, they've solved the heat problems.
But I've seen several methods for solving this. Very sophisticated. One from IBM involves micro channels between the transistors and levels that pump a liquid through the chip with tiny pumps that are themselves nano technology somewhat similar to TI's DLP technology. Amazing to see the lab demo!
Intel has two different concepts as well.
This is all being done within the chips themselves. It's easier with memory chips because of the regular layouts, as opposed to the complex layouts of CPU's and GPU's, though that's what this is primarily meant for.
But I've seen several methods for solving this. Very sophisticated. One from IBM involves micro channels between the transistors and levels that pump a liquid through the chip with tiny pumps that are themselves nano technology somewhat similar to TI's DLP technology. Amazing to see the lab demo!
No way! Nano-tech like DLP blows my mind. Sounds very cool (oops, pun!) but sadly also like a lot of potential for unreliability and/or very low yield.
No way! Nano-tech like DLP blows my mind. Sounds very cool (oops, pun!) but sadly also like a lot of potential for unreliability and/or very low yield.
DLPs are incredibly reliable. Think of how many times a second they must move in a Tv chip, then multiply that by hours, months, years. It becomes an astounding number.
All the time, it's being bathed in a very hot light from a projection bulb, even with the filtering between. These are reliable because there is no actual "bending" going on that causes cracks or change in the structure. That's the advantage of working at the molecular level rather than at the "bulk" level.
DLPs are incredibly reliable. Think of how many times a second they must move in a Tv chip, then multiply that by hours, months, years. It becomes an astounding number.
All the time, it's being bathed in a very hot light from a projection bulb, even with the filtering between. These are reliable because there is no actual "bending" going on that causes cracks or change in the structure. That's the advantage of working at the molecular level rather than at the "bulk" level.
DLPs are reliable but there's no liquid involved. It's difficult to envisage how such a structure as complex as the one you described would get the same yields as traditional 2D approaches. How do they get the cooling liquid in the right place? Like I said, it's mind-blowing.
DLPs are reliable but there's no liquid involved. It's difficult to envisage how such a structure as complex as the one you described would get the same yields as traditional 2D approaches. How do they get the cooling liquid in the right place? Like I said, it's mind-blowing.
The channels are made from a metal that is deposited as a strand. Then silicon is deposited around that. Then the metal is acid etched out, which leaves a channel. Actually, hundreds of them.
This is old manufacturing methodology for chips.
The "pump" parts are made the same way, on the same chips as the rest of the stuff. TI gets pretty high yields.
Realize that all chips are actually 3D structures. There are as many as eight layers deposited and etched to make a chip now.
This is basically showing cooling of chip stacks, which is the older iteration of the designs, which are now being designed for the cooling of the interior of the chips themselves. This uses an external, but tiny pump (not shown).
This is basically showing cooling of chip stacks, which is the older iteration of the designs, which are now being designed for the cooling of the interior of the chips themselves. This uses an external, but tiny pump (not shown).
This story and your comment partially answer a question I have. I'm planning to get a new 15" MBP 2.8 shortly after Snow comes pre-installed. Clearly, if HDD's can't tax the 1.5 Sata level and SSD's can, they have some advantages. But I'm wondering how telling they are.
So my question was comparing Apple's SSD's with their 7200 rpm 500 GB drive in terms of various aspects of performance. This will be my next main machine for likely 5 years or so, and I want it to be as fast as possible within my budget, with plenty of onboard storage. I've compared the price of upgrading from 5400 to 7200 and thus getting the space I really need compared to spending far more on a much smaller SSD - and if the perf diff from 5400 to 7200 is noticeable, it seems like kind of a sweet spot for me.
Especially as I think in about 2 or maybe 3 years 512MB SSD's should be readily available and more affordable (probably better too) than 256's today - so a few years down the road I'm eyeballing going from 4 to 8 GB RAM (maybe sooner on the RAM), and one of those future SSD babies to juice me up for the second half of my planned use.
Reviews usually stick to Apple default configs, so this is a question I've never heard answered.
So if anyone can enlighten me, how much faster on which tasks ARE today's Apple-supplied SSD's compared to 7200 HDD? And 7200 to 5400?
I was pretty much like you when I got my first PowerBook 17" G4 years and years ago. That workhorse lasted me more than 6 years, and in that time, I made minor upgrades as prices fell. 7200-rpm hard drives, maxed out memory, upgraded dual-layer, dual-format DVD-burner...
And I'm also an amateur photographer. I have a desktop and a NAS server running at home for storage and backup, but I still like to have all my media with me on the go; better to have something and not need it then to need and not have it, right? Unfortunately, the largest SSD with good random read/write performance (that's the most critical aspect of SSD as has been noted a number of times here now), was the new line of OCZ Vertex 256GB SSDs. But those things are expensive, fetching anywhere from $650-$800 depending on where you're buying from.
If all I wanted was performance, I think I could live with a small 80GB Intel X25-M and get unreal speed from it, but without the ability to hold all the things I need. And traveling with an external hard drive just negates the point of easy, light portability. So like you, I am picking up a comparatively cheap Seagate 7200.4 drive for now, which will fit everything I need, while I wait for SSDs to come down in price and increase in capacity over the next 2 or 3 years until it's economically feasible to plunk down money for it. As much as I would love to have the speed now, I just can't give up the storage space in return.
If all I wanted was performance, I think I could live with a small 80GB Intel X25-M and get unreal speed from it, but without the ability to hold all the things I need. And traveling with an external hard drive just negates the point of easy, light portability. So like you, I am picking up a comparatively cheap Seagate 7200.4 drive for now, which will fit everything I need, while I wait for SSDs to come down in price and increase in capacity over the next 2 or 3 years until it's economically feasible to plunk down money for it. As much as I would love to have the speed now, I just can't give up the storage space in return.
I'm going for the hybrid approach. Gonna jettison the optical drive (rarely use it anyway) and move my 500 GB drive to it's place - there are companies making mounting brackets to do just that. Then I will use the Intel 80 GB SSD as my primary drive with my apps.
I'm going for the hybrid approach. Gonna jettison the optical drive (rarely use it anyway) and move my 500 GB drive to it's place - there are companies making mounting brackets to do just that. Then I will use the Intel 80 GB SSD as my primary drive with my apps.
Just be aware of the fact that SSD's slow down considerably as they fill up, much more so than do HDDs, which slow down from rotational issues. The small SSDs slow down quickly if you fill them to a large percentage, and they also slow down from frequent use, which is something thats not an issue with HDDs.
There are several solutions for this slowdown, but right now, none are implemented for regular users.
SSDs are not yet a panacea. I would wait until the new generation comes out in 2010.
Just be aware of the fact that SSD's slow down considerably as they fill up, much more so than do HDDs, which slow down from rotational issues. The small SSDs slow down quickly if you fill them to a large percentage, and they also slow down from frequent use, which is something thats not an issue with HDDs.
There are several solutions for this slowdown, but right now, none are implemented for regular users.
SSDs are not yet a panacea. I would wait until the new generation comes out in 2010.
I would take issue to your statement "SSDs slow down considerably as they fill up, much more than do HDDs"
I'm aware that any storage device that begins to become full will slow down as the controller has to hunt more for empty storage areas but I've seen nothing that shows SSD slow down comparable to HDD.
The good news is that after an initial dip in performance, SSDs tend to level off, according to Eden Kim, chairman of the Solid State Storage Initative's Consumer SSD Market Development Task Force. Even if they do drop in performance over time -- undercutting a manufacturer's claims -- consumer flash drives are still vastly faster than traditional hard drives, because they can perform two to five times the input/ouput operations (I/Os) per second of a hard drive, he said.
The larger the SSD the less likely we'll run into slow down issues (the same could be said for HDD)
RUMOURS ARE growing louder that Intel will be launching new SSDs based on the firm's 34nm NAND chips within just a couple of weeks. Previous reports had said Chipzilla planned to come out with its new 34nm flash memory drives in Q4, but that timetable has been nudged up.
We've been told that with these new 34nm NAND SSDs, users can expect higher performance, higher capacities, and most importantly, lower prices.
Amongst the 34nm offerings hot off the production line will be a 320GB drive, plus 160GB and 80GB capacities too. But there very well could be more. Our sources tell us there will be drives big enough to replace the HDDs in most, if not all
I welcome this move. We need to push down the current pricing and deliver a 320GB SSD drive which is enough space for most people (at least in the notebook realm)
Comments
I read that SSD should evolve at a rate that's slightly faster than Moore's Law which is good. I also read that they will hit a wall at 20nm. The next SSD should be at 34nm from a few vendors like Intel/Nec so I figure they'll end up at something like 28nm for the 2011 batch and then 20nm in 2013?
While some are arguing about the 18-20 nm wall, many researchers are thinking that it may come to pass. not so much because of technical impossibility, but because of cost. It's possible that it will come out, but years after it would be expected, going by, as you say, "Moore's Law".
Maybe we won't see it until 2015.
At any rate, new technologies may come into production by then. There are a few that are looking better every day.
While some are arguing about the 18-20 nm wall, many researchers are thinking that it may come to pass. not so much because of technical impossibility, but because of cost. It's possible that it will come out, but years after it would be expected, going by, as you say, "Moore's Law".
Maybe we won't see it until 2015.
At any rate, new technologies may come into production by then. There are a few that are looking better every day.
My guess is that we'll just see storage branch out horizontally just as CPU did when they hit the wall. Because SSD are power efficient I expect to see something like PCI Express or even some other type of fast bus connect SSD in RAID like configuration.
My guess is that we'll just see storage branch out horizontally just as CPU did when they hit the wall. Because SSD are power efficient I expect to see something like PCI Express or even some other type of fast bus connect SSD in RAID like configuration.
There are a couple like that out now, but they're expensive, and performance is MEH. But it will get better as time goes on.
If it weren't for so many crappy controllers on the market, things would be better now.
But I expect prices at half the levels we see them at today within a year.
My guess is that we'll just see storage branch out horizontally just as CPU did when they hit the wall.
Unfortunately, this analogy doesn't work. "The wall" in CPUs was clock speed - the number of transistors in a given physical area had been doubling approximately every 18 months as Moore had predicted - what Intel had been doing was relying on the smaller size of an individual transistor allowing it to operate at a faster rate, and using a few extra transistors in a CPU to extend functional unit pipelines. The problem was that making pipelines longer and longer meant they took larger and larger hits for mis-predicting code branches, and they happen rather a lot on multi-tasking general computing platforms. Another problem was higher clock speeds increasing power consumption and creating hot spots in the CPU - areas so small that it was hard to extract the heat fast enough to allow reliable operation. So now, instead of making CPUs with a die that gets ever physically smaller and higher clocked (same number of transistors from one generation to the next, just each transistor is smaller), the physical size of the die stays about the same, but the smaller size of each transistor allows for more cores to fit in the same area. The multi-core approach certainly gives the consumer better value for money!
But the key here is to appreciate that whichever path the CPU industry had taken (higher clock speed or more cores), the size of the transistors would be getting smaller.
With SSDs (and CPUs, for that matter), there will be a minimum size that it's possible to make a traditional transistor. Once that wall is hit, there's no "alternative", "horizontal route" that SSDs can take - more capacity requires more transistors, and that's that. The wall exists because at some point the gate-oxide layer will reach one atom thick, and you can't get thinner, and traditional transistors rely on the bulk statistical behaviour of many, many electrons. The smaller a transistor, the smaller the number of electrons there are involved in the operation of the device. As the transistor shrinks, eventually you'll reach the point where there are too few electrons and the transistor no longer works in a predictable, deterministic manner.
Fortunately, there's spintronics or something else equally mind-boggling to save the day when the wall is hit with traditional transistors.
Unfortunately, this analogy doesn't work. "The wall" in CPUs was clock speed - the number of transistors in a given physical area had been doubling approximately every 18 months as Moore had predicted - what Intel had been doing was relying on the smaller size of an individual transistor allowing it to operate at a faster rate, and using a few extra transistors in a CPU to extend functional unit pipelines. The problem was that making pipelines longer and longer meant they took larger and larger hits for mis-predicting code branches, and they happen rather a lot on multi-tasking general computing platforms. Another problem was higher clock speeds increasing power consumption and creating hot spots in the CPU - areas so small that it was hard to extract the heat fast enough to allow reliable operation. So now, instead of making CPUs with a die that gets ever physically smaller and higher clocked (same number of transistors from one generation to the next, just each transistor is smaller), the physical size of the die stays about the same, but the smaller size of each transistor allows for more cores to fit in the same area. The multi-core approach certainly gives the consumer better value for money!
But the key here is to appreciate that whichever path the CPU industry had taken (higher clock speed or more cores), the size of the transistors would be getting smaller.
With SSDs (and CPUs, for that matter), there will be a minimum size that it's possible to make a traditional transistor. Once that wall is hit, there's no "alternative", "horizontal route" that SSDs can take - more capacity requires more transistors, and that's that. The wall exists because at some point the gate-oxide layer will reach one atom thick, and you can't get thinner, and traditional transistors rely on the bulk statistical behaviour of many, many electrons. The smaller a transistor, the smaller the number of electrons there are involved in the operation of the device. As the transistor shrinks, eventually you'll reach the point where there are too few electrons and the transistor no longer works in a predictable, deterministic manner.
Fortunately, there's spintronics or something else equally mind-boggling to save the day when the wall is hit with traditional transistors.
I was going to edit your post down before responding but decided not to.
But I'll respond to one thing, transistor size.
The thing you're forgetting is that there is a "new" thing in chip manufacture that's becoming possible, and that's talked about for use particularly with memory devices, and that's 3D architecture.
We're already seeing it in some, and it will become used more often, and will become more important as time goes on. This will allow a greater number of cells in the same space. Cooling requirements go up, but they have solutions for that as well.
Once that "wall" in size is reached, and until newer technologies are available, we'll continue to see greater amounts of memory per chip for at least a while.
The thing you're forgetting is that there is a "new" thing in chip manufacture that's becoming possible, and that's talked about for use particularly with memory devices, and that's 3D architecture.
We're already seeing it in some, and it will become used more often, and will become more important as time goes on. This will allow a greater number of cells in the same space. Cooling requirements go up, but they have solutions for that as well.
Once that "wall" in size is reached, and until newer technologies are available, we'll continue to see greater amounts of memory per chip for at least a while.
Good point. I've not read much about it but must assume that the heat problem is not at all trivial. Certainly you're not going to be able to stack 20 layers of transistors on top of each other. Hopefully 3D will bridge the gap from electronics to spintronics or whatever.
Yes I keep hearing about "chip stacking" that may alleviate some of the issues.
I think Melgross is talking about something else.
A chip consists of a die inside a plastic package, with hair-thin wires connecting the comparatively giant pins that stick out of the package, to the pads around the edge of the die. Chip stacking is where you take several dice and package them together in a single package - for example I believe the Samsung SOCs in the iPhone have the RAM on one die and the CPU on another. The two dice are stacked on top of each other, and then packaged into the single plastic package you see on the iPhone motherboard.
The 3D that Melgross is talking about is, I believe, about making several layers of transistors on a single die.
Good point. I've not read much about it but must assume that the heat problem is not at all trivial. Certainly you're not going to be able to stack 20 layers of transistors on top of each other. Hopefully 3D will bridge the gap from electronics to spintronics or whatever.
Hopefully, they WILL be able to stack 20 layers of transistors on top of each other. If they can, they've solved the heat problems.
But I've seen several methods for solving this. Very sophisticated. One from IBM involves micro channels between the transistors and levels that pump a liquid through the chip with tiny pumps that are themselves nano technology somewhat similar to TI's DLP technology. Amazing to see the lab demo!
Intel has two different concepts as well.
This is all being done within the chips themselves. It's easier with memory chips because of the regular layouts, as opposed to the complex layouts of CPU's and GPU's, though that's what this is primarily meant for.
But I've seen several methods for solving this. Very sophisticated. One from IBM involves micro channels between the transistors and levels that pump a liquid through the chip with tiny pumps that are themselves nano technology somewhat similar to TI's DLP technology. Amazing to see the lab demo!
No way! Nano-tech like DLP blows my mind. Sounds very cool (oops, pun!) but sadly also like a lot of potential for unreliability and/or very low yield.
I think Melgross is talking about something else.
The 3D that Melgross is talking about is, I believe, about making several layers of transistors on a single die.
Correct! They've been working on this for decades, but it's only become a major attraction now.
No way! Nano-tech like DLP blows my mind. Sounds very cool (oops, pun!) but sadly also like a lot of potential for unreliability and/or very low yield.
DLPs are incredibly reliable. Think of how many times a second they must move in a Tv chip, then multiply that by hours, months, years. It becomes an astounding number.
All the time, it's being bathed in a very hot light from a projection bulb, even with the filtering between. These are reliable because there is no actual "bending" going on that causes cracks or change in the structure. That's the advantage of working at the molecular level rather than at the "bulk" level.
DLPs are incredibly reliable. Think of how many times a second they must move in a Tv chip, then multiply that by hours, months, years. It becomes an astounding number.
All the time, it's being bathed in a very hot light from a projection bulb, even with the filtering between. These are reliable because there is no actual "bending" going on that causes cracks or change in the structure. That's the advantage of working at the molecular level rather than at the "bulk" level.
DLPs are reliable but there's no liquid involved. It's difficult to envisage how such a structure as complex as the one you described would get the same yields as traditional 2D approaches. How do they get the cooling liquid in the right place? Like I said, it's mind-blowing.
DLPs are reliable but there's no liquid involved. It's difficult to envisage how such a structure as complex as the one you described would get the same yields as traditional 2D approaches. How do they get the cooling liquid in the right place? Like I said, it's mind-blowing.
The channels are made from a metal that is deposited as a strand. Then silicon is deposited around that. Then the metal is acid etched out, which leaves a channel. Actually, hundreds of them.
This is old manufacturing methodology for chips.
The "pump" parts are made the same way, on the same chips as the rest of the stuff. TI gets pretty high yields.
Realize that all chips are actually 3D structures. There are as many as eight layers deposited and etched to make a chip now.
Here is a popularized account of some of it. If I can find more detail, I'll post it.
http://www.usatoday.com/tech/news/te...ng-chips_N.htm
This is basically showing cooling of chip stacks, which is the older iteration of the designs, which are now being designed for the cooling of the interior of the chips themselves. This uses an external, but tiny pump (not shown).
http://www.sciencedaily.com/releases...0606152512.htm
I tierd to get the article from my online Science journal, but no go.
Here is a popularized account of some of it. If I can find more detail, I'll post it.
http://www.usatoday.com/tech/news/te...ng-chips_N.htm
This is basically showing cooling of chip stacks, which is the older iteration of the designs, which are now being designed for the cooling of the interior of the chips themselves. This uses an external, but tiny pump (not shown).
http://www.sciencedaily.com/releases...0606152512.htm
Cheers Melgross.
This story and your comment partially answer a question I have. I'm planning to get a new 15" MBP 2.8 shortly after Snow comes pre-installed. Clearly, if HDD's can't tax the 1.5 Sata level and SSD's can, they have some advantages. But I'm wondering how telling they are.
So my question was comparing Apple's SSD's with their 7200 rpm 500 GB drive in terms of various aspects of performance. This will be my next main machine for likely 5 years or so, and I want it to be as fast as possible within my budget, with plenty of onboard storage. I've compared the price of upgrading from 5400 to 7200 and thus getting the space I really need compared to spending far more on a much smaller SSD - and if the perf diff from 5400 to 7200 is noticeable, it seems like kind of a sweet spot for me.
Especially as I think in about 2 or maybe 3 years 512MB SSD's should be readily available and more affordable (probably better too) than 256's today - so a few years down the road I'm eyeballing going from 4 to 8 GB RAM (maybe sooner on the RAM), and one of those future SSD babies to juice me up for the second half of my planned use.
Reviews usually stick to Apple default configs, so this is a question I've never heard answered.
So if anyone can enlighten me, how much faster on which tasks ARE today's Apple-supplied SSD's compared to 7200 HDD? And 7200 to 5400?
I was pretty much like you when I got my first PowerBook 17" G4 years and years ago. That workhorse lasted me more than 6 years, and in that time, I made minor upgrades as prices fell. 7200-rpm hard drives, maxed out memory, upgraded dual-layer, dual-format DVD-burner...
And I'm also an amateur photographer. I have a desktop and a NAS server running at home for storage and backup, but I still like to have all my media with me on the go; better to have something and not need it then to need and not have it, right? Unfortunately, the largest SSD with good random read/write performance (that's the most critical aspect of SSD as has been noted a number of times here now), was the new line of OCZ Vertex 256GB SSDs. But those things are expensive, fetching anywhere from $650-$800 depending on where you're buying from.
If all I wanted was performance, I think I could live with a small 80GB Intel X25-M and get unreal speed from it, but without the ability to hold all the things I need. And traveling with an external hard drive just negates the point of easy, light portability. So like you, I am picking up a comparatively cheap Seagate 7200.4 drive for now, which will fit everything I need, while I wait for SSDs to come down in price and increase in capacity over the next 2 or 3 years until it's economically feasible to plunk down money for it. As much as I would love to have the speed now, I just can't give up the storage space in return.
If all I wanted was performance, I think I could live with a small 80GB Intel X25-M and get unreal speed from it, but without the ability to hold all the things I need. And traveling with an external hard drive just negates the point of easy, light portability. So like you, I am picking up a comparatively cheap Seagate 7200.4 drive for now, which will fit everything I need, while I wait for SSDs to come down in price and increase in capacity over the next 2 or 3 years until it's economically feasible to plunk down money for it. As much as I would love to have the speed now, I just can't give up the storage space in return.
I'm going for the hybrid approach. Gonna jettison the optical drive (rarely use it anyway) and move my 500 GB drive to it's place - there are companies making mounting brackets to do just that. Then I will use the Intel 80 GB SSD as my primary drive with my apps.
I'm going for the hybrid approach. Gonna jettison the optical drive (rarely use it anyway) and move my 500 GB drive to it's place - there are companies making mounting brackets to do just that. Then I will use the Intel 80 GB SSD as my primary drive with my apps.
Just be aware of the fact that SSD's slow down considerably as they fill up, much more so than do HDDs, which slow down from rotational issues. The small SSDs slow down quickly if you fill them to a large percentage, and they also slow down from frequent use, which is something thats not an issue with HDDs.
There are several solutions for this slowdown, but right now, none are implemented for regular users.
SSDs are not yet a panacea. I would wait until the new generation comes out in 2010.
Just be aware of the fact that SSD's slow down considerably as they fill up, much more so than do HDDs, which slow down from rotational issues. The small SSDs slow down quickly if you fill them to a large percentage, and they also slow down from frequent use, which is something thats not an issue with HDDs.
There are several solutions for this slowdown, but right now, none are implemented for regular users.
SSDs are not yet a panacea. I would wait until the new generation comes out in 2010.
I would take issue to your statement "SSDs slow down considerably as they fill up, much more than do HDDs"
I'm aware that any storage device that begins to become full will slow down as the controller has to hunt more for empty storage areas but I've seen nothing that shows SSD slow down comparable to HDD.
http://www.computerworld.com/action/...icleId=9132668
The good news is that after an initial dip in performance, SSDs tend to level off, according to Eden Kim, chairman of the Solid State Storage Initative's Consumer SSD Market Development Task Force. Even if they do drop in performance over time -- undercutting a manufacturer's claims -- consumer flash drives are still vastly faster than traditional hard drives, because they can perform two to five times the input/ouput operations (I/Os) per second of a hard drive, he said.
The larger the SSD the less likely we'll run into slow down issues (the same could be said for HDD)
an another good note
http://www.theinquirer.net/inquirer/...s-launch-weeks
RUMOURS ARE growing louder that Intel will be launching new SSDs based on the firm's 34nm NAND chips within just a couple of weeks. Previous reports had said Chipzilla planned to come out with its new 34nm flash memory drives in Q4, but that timetable has been nudged up.
We've been told that with these new 34nm NAND SSDs, users can expect higher performance, higher capacities, and most importantly, lower prices.
Amongst the 34nm offerings hot off the production line will be a 320GB drive, plus 160GB and 80GB capacities too. But there very well could be more. Our sources tell us there will be drives big enough to replace the HDDs in most, if not all
I welcome this move. We need to push down the current pricing and deliver a 320GB SSD drive which is enough space for most people (at least in the notebook realm)