What is the best speed of sd card to use with a flash memory based laptop? Say, an HP Stream 13. I'm not very familiar with all the classes of sd cards and they are hard to understand. Thank you!
Sd cards
- Thread starter stranger706
- Start date
- Status
The flash memory on the laptop has nothing to do with the SD cards, it's completely independent. Faster SD cards just mean you can save things to them faster, which is important for things like video cameras which have to quickly store lots of data on an SD card.
An SD, SDHC OR SDXC with UHS-I
If you are using it as storage it doesn't matter. If your device lets you run applications from it buy UHS-1 or Class 10.
Really I'd probably only buy class 10 or better since it's a couple dollars more and who knows what you'll use it for in the future.
Really I'd probably only buy class 10 or better since it's a couple dollars more and who knows what you'll use it for in the future.
What are you going to be using this card for? If it's for storing of larger files (like audio/video media), then high speed cards would be good (Class 10, UHS I, etc.).
However, I've heard on here from someone (Panda?), that Class 2/4 cards are faster for random read/write on very small files, so depending on what you'll be storing/accessing from that card, that may be something to consider. But I'm not 100% sure this holds true for every Class 2/4 card out there.
However, I've heard on here from someone (Panda?), that Class 2/4 cards are faster for random read/write on very small files, so depending on what you'll be storing/accessing from that card, that may be something to consider. But I'm not 100% sure this holds true for every Class 2/4 card out there.
It is a bit more "technical" than that.
The rating of speed used to be any random sector access in the old class 2/4 days, so it is hard to design a card that is fast if you write files starting in the middle of one flash block and ending in the middle of another flash block (each block was about 1-2MBytes back then).
Then those camcorder manufacturers complained that SD cards are not fast enough but CF are too expensive and bulky, so they took a compromise that the speed is rated on 4MB or bigger writes, and would be aligned to block boundary. So the newer cards are faster when accessed that way and will not get in trouble crossing block boundary (it has to be at least 8 if not 16MB each by now).
This means they will not worry about small file access from Windows or OSX anymore as they are not writing files that big that fast like camcorders or SLR camera. That also means a 2/4 class card may not be slower than a 6/8/10 class card.
However, flash memory speed depends on all sorts of things from how many chips / dies / plane you access in parallel (interleave), to what kind of chips you use (SLC / 2 bit MLC / 3 bit MLC), to which generation of chips you use (older chips have larger flash memory cell so they can charge / discharge faster for the same lifespan, so you can read / write faster). You really have to benchmark them to see what is the speed.
Does that make sense? or am I making it too complicated?
The rating of speed used to be any random sector access in the old class 2/4 days, so it is hard to design a card that is fast if you write files starting in the middle of one flash block and ending in the middle of another flash block (each block was about 1-2MBytes back then).
Then those camcorder manufacturers complained that SD cards are not fast enough but CF are too expensive and bulky, so they took a compromise that the speed is rated on 4MB or bigger writes, and would be aligned to block boundary. So the newer cards are faster when accessed that way and will not get in trouble crossing block boundary (it has to be at least 8 if not 16MB each by now).
This means they will not worry about small file access from Windows or OSX anymore as they are not writing files that big that fast like camcorders or SLR camera. That also means a 2/4 class card may not be slower than a 6/8/10 class card.
However, flash memory speed depends on all sorts of things from how many chips / dies / plane you access in parallel (interleave), to what kind of chips you use (SLC / 2 bit MLC / 3 bit MLC), to which generation of chips you use (older chips have larger flash memory cell so they can charge / discharge faster for the same lifespan, so you can read / write faster). You really have to benchmark them to see what is the speed.
Does that make sense? or am I making it too complicated?
Last edited:
PandaBear, you make sense, and your information is very useful.
It explains why many Class 10 cards are slower than old 2 GB unrated SD cards at everything except very large writes and reads.
It seems that progress is made on size (larger) but unless there is some breakthrough, newer flash storage is worse for durability and small file write+read speeds.
It explains why many Class 10 cards are slower than old 2 GB unrated SD cards at everything except very large writes and reads.
It seems that progress is made on size (larger) but unless there is some breakthrough, newer flash storage is worse for durability and small file write+read speeds.
Originally Posted By: BearZDefect
It seems that progress is made on size (larger) but unless there is some breakthrough, newer flash storage is worse for durability and small file write+read speeds.
The manufacturers are just following the typical mainstream requirements, and this happens to be storage of larger files, such as video and audio. Even a typical JPEG photo out of a camera is now at least several megabytes in size, and tens of megabytes if you shoot in RAW. And from that perspective, these new high speed cards work well.
Moral of the story is don't try to use one of these cards as your OS drive. For that, you should use SSD.
It seems that progress is made on size (larger) but unless there is some breakthrough, newer flash storage is worse for durability and small file write+read speeds.
The manufacturers are just following the typical mainstream requirements, and this happens to be storage of larger files, such as video and audio. Even a typical JPEG photo out of a camera is now at least several megabytes in size, and tens of megabytes if you shoot in RAW. And from that perspective, these new high speed cards work well.
Moral of the story is don't try to use one of these cards as your OS drive. For that, you should use SSD.
An example of a card I bought last year (64GB SDXC from PNY). It's fast with files 64KB or larger. Not so much with smaller files.
The cards will mainly be used to store photos, music and iphone backups.
So a class 10 would be good?
Thanks for the help everyone.
So a class 10 would be good?
Thanks for the help everyone.
Originally Posted By: eljefino
Originally Posted By: PandaBear
Does that make sense? or am I making it too complicated?
To me that explanation read like this
But thank you just the same.
Originally Posted By: PandaBear
Does that make sense? or am I making it too complicated?
To me that explanation read like this
But thank you just the same.
The Rockwell Retro Encabulator makes perfect sense to me
Like everything else in life, "they don't make them like they used to". Flash memory is the same because you can only put so many atoms on each surface / volume, and surface / volume means cost, so you have to cut them down if you want to make them cheaper year after year.
Engineers find ways to work around the compromise, either by randomize the access to make them last longer (wear leveling), bundling more bits together to reduce overhead (each chunks waste the same amount of transistors for logics that don't store stuff, so bigger chunk is more efficient), just let more errors and defects happen and fix it with statistics (error correction code), slowing things down and use more chips in parallel.
The reason "smaller files" are slower is because you have to read the same 64KB, modify the middle of it for a small file, then write the same 64KB back, instead of just writing 64KB without reading the old data first. (you are going to over write it anyways, so why bother reading the old data?)
Like everything else in life, "they don't make them like they used to". Flash memory is the same because you can only put so many atoms on each surface / volume, and surface / volume means cost, so you have to cut them down if you want to make them cheaper year after year.
Engineers find ways to work around the compromise, either by randomize the access to make them last longer (wear leveling), bundling more bits together to reduce overhead (each chunks waste the same amount of transistors for logics that don't store stuff, so bigger chunk is more efficient), just let more errors and defects happen and fix it with statistics (error correction code), slowing things down and use more chips in parallel.
The reason "smaller files" are slower is because you have to read the same 64KB, modify the middle of it for a small file, then write the same 64KB back, instead of just writing 64KB without reading the old data first. (you are going to over write it anyways, so why bother reading the old data?)
Last edited:
- Status
