Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
iPod I think I'm in love... with a THING
- Thread starter funkychunkz
- Start date
- Sort by reaction score
EGT said:
I heard that too. If I'm not mistaken, it was Samsung that (according to them) had achieved transfer rates high enough to make a flash drive. And it'll be VERY pricey, only for hight end laptops at first... but those would be fantastic for ultra portables (just imagine the thinness of such a drive!). Add the fact that Intel (I don't know about the others) is aggressively optimizing its laptop processors, reducing the amount of heat they dissipate, I think in some years we'll be seeing laptops with no moving parts at all (well, other than the lid and keyboard
), and absolutely silent... I can't wait to get one of those! 
Ok, there is a lot of FUD about flash media going around at the moment, so I am going to have to spend a little time digging around the net for real solid information... however, when it comes to data storage, there are several things to consider concerning speed, and also lots of things to think about other than speed.
First, as it's on everyone tongues, we'll do speed. The newest flash media has sustained read and write times that are GETTING CLOSE to, but not as fast as traditional IDE based harddrives. Here is the first set of tests that I have found, from barefeats. Here we see that copying a 115mb file to and from various devices, including FW HDD shows decent resuts with some of the fasted flash media, but pretty poor from the 'cheap stuff'. So sctually getting the data off the media is significantly (currently at least) faster with traditional HDDs.
Ok, but what about other measures of speed? A HDDs seek time is just as important as it's throughput, perhaps moreso in many cases. HDD seek times (how long it takes to find what you want to read out, or the place to write a file to) is measured in milliseconds, typically around 6-9ms now-a-days. The fastest flash media has seek time measured in nanoseconds. So, finding the data you want is exponentially faster than with a harddrive. So for accessing lots of small files randomly, flash is going to be a good choice, but for moving large amounts of data around you will run into a speed barrier. It's tough to find tests on this, but here is an example of a product press release with some manufacturer specs.
I predict that 2.5" and 3.5" standard sized drives will eventually become self contained flash based disk arrays. Using several seperate flash disks (say 8 8gb modules to start), each with relatively low data throughput, but all with near instant access times. By striping the accross the flashchips (each with their own control module and a single unifying RAID controller) you can combine all of the relatively slow data streams into a single very fast one. Add that to the extremely low/fast seek times of each device and you could have an incredibly fast data storage system, albeit a very expensive one at todays prices.
The other issues to consider with data storage is data density, heat, capacity, cost, and relability. Flash generates very little heat and uses very little energy so it's an excellent choice for either portable machines or compact server situations. The reliability factor is a double edged sword. No moving parts give flash media a very low failure rate compared to traditional HDDs, but the life span of flash blocks is limited to a fixed number of write cycles. This has improved as of late with better manufacturing techniques and also with better control logic that does a good job of balancing write operations.
The advantage with flash is that you can easily predict when it is going to 'fail' and when it does, you can still read the data, just not write more. Harddrives tend to fail more spectacularly, by doing it at random and locking all your data up entirely. Combined with an array method like I described above you could easily design a system with hotswapable replacable modules, which would also lead to easy upgradability (buy a 3.5" drive shell with the raid controller in it which has slots for, say, up to 24 8gb modules. Start with 8 in there and add more as you need them).
Finally, regarding the nano; I suspect they are using some high grade flash media in these, and that the speeds will be respectable; close to pushing the limits of realistic USB2 throughput. I could be wrong, we will have to see what some benchmarking has to say over the coming weeks/months.
First, as it's on everyone tongues, we'll do speed. The newest flash media has sustained read and write times that are GETTING CLOSE to, but not as fast as traditional IDE based harddrives. Here is the first set of tests that I have found, from barefeats. Here we see that copying a 115mb file to and from various devices, including FW HDD shows decent resuts with some of the fasted flash media, but pretty poor from the 'cheap stuff'. So sctually getting the data off the media is significantly (currently at least) faster with traditional HDDs.
Ok, but what about other measures of speed? A HDDs seek time is just as important as it's throughput, perhaps moreso in many cases. HDD seek times (how long it takes to find what you want to read out, or the place to write a file to) is measured in milliseconds, typically around 6-9ms now-a-days. The fastest flash media has seek time measured in nanoseconds. So, finding the data you want is exponentially faster than with a harddrive. So for accessing lots of small files randomly, flash is going to be a good choice, but for moving large amounts of data around you will run into a speed barrier. It's tough to find tests on this, but here is an example of a product press release with some manufacturer specs.
I predict that 2.5" and 3.5" standard sized drives will eventually become self contained flash based disk arrays. Using several seperate flash disks (say 8 8gb modules to start), each with relatively low data throughput, but all with near instant access times. By striping the accross the flashchips (each with their own control module and a single unifying RAID controller) you can combine all of the relatively slow data streams into a single very fast one. Add that to the extremely low/fast seek times of each device and you could have an incredibly fast data storage system, albeit a very expensive one at todays prices.
The other issues to consider with data storage is data density, heat, capacity, cost, and relability. Flash generates very little heat and uses very little energy so it's an excellent choice for either portable machines or compact server situations. The reliability factor is a double edged sword. No moving parts give flash media a very low failure rate compared to traditional HDDs, but the life span of flash blocks is limited to a fixed number of write cycles. This has improved as of late with better manufacturing techniques and also with better control logic that does a good job of balancing write operations.
The advantage with flash is that you can easily predict when it is going to 'fail' and when it does, you can still read the data, just not write more. Harddrives tend to fail more spectacularly, by doing it at random and locking all your data up entirely. Combined with an array method like I described above you could easily design a system with hotswapable replacable modules, which would also lead to easy upgradability (buy a 3.5" drive shell with the raid controller in it which has slots for, say, up to 24 8gb modules. Start with 8 in there and add more as you need them).
Finally, regarding the nano; I suspect they are using some high grade flash media in these, and that the speeds will be respectable; close to pushing the limits of realistic USB2 throughput. I could be wrong, we will have to see what some benchmarking has to say over the coming weeks/months.