Been looking over the bm charts for 3.5" desktop hdd's and am wondering what the difference (in layman terms) is between Thoughput vs. Access Time?
Would one over the other be "more important", or have more of an effect, in making a hdd 'faster'?
The current HDD's I have now are pretty full and am interested in buying a new one.
I assume if it faster then the one I have my o/s on now, I will transfer it to the new one and use the original for storage.
access time is (basically) a measure of how long it takes for the read/write heads to move from one place to another, and start sending you data.
throughput is, once the data starts coming over, how quickly is it coming to you. Is it coming through a garden hose or a fire hose?
Both are important in their own way, depending on the type of data you are using, and work. For example, a transactional database load would be more concerned with access time, and video editing with throughput.
Access time is almost always more important but isn't a perfect measurement of a disk's agility. And anyway, it should be measured over something like 100MB, not over the whole disk, as the files you want to access aren't randomly scattered over the platter.
Suppose you have to make a list of all the names on the tabs of the files in a filing cabinet. So you open the top drawer of the filing cabinet and start thumbing your way through the tabs, jotting down the name of each one as you go. That's sequential access, and when you do this you can go through the tabs very quickly.
Now let's say you have a list of names and you have to check to see if you have files for them. So you look at the first name, decide which drawer the file should be in, pull open the drawer, thumb through the files to see if the file is there, and put checkmark beside the name on your list if you find it. Now you do the same for the next name on the list, which is probably in a different drawer. And so on. This is random access, and as you can imagine it takes a lot longer to handle each file this way.
"Access time" refers to how long it takes to find a file in the filing cabinet. It includes pulling open the drawer and thumbing through the files to find the one you want.
"Throughput" is how many file tabs you can go through each second when you're just looking at one after another.
In terms of a computer system, "access time" controls how quickly your computer boots up and starts applications when you click on them, because those actions require the computer to find hundreds or thousands of little files that can be located all over the disk.
"Throughput" is most important when you copy large files or use a program to read or write a very large file (note that playing videos or music files doesn't require a disk with fast throughput since the computer doesn't have to read them all that quickly in order to play them at normal speed).
For most people, access time is the measure that is most related to the everyday performance of their computer work. SSDs have access times that are about 100X faster than any mechanical hard drive, and that's why a lot of people are willing to pay a premium for them.
"Throughput" is only important when you copy large files or use a program to read or write a very large file (note that playing videos or music files doesn't require a disk with fast throughput since the computer doesn't have to read them all that quickly in order to play them at normal speed).
I would say throughput is -more- (not only) important when copying large files.
A higher throughput will improve performance on every single read or write operation. With small files the improvement will be minimal, with large files the improvement will be substantial.
If throughput was not important when booting or starting applications then it would not make a difference whether you boot from an internal drive or a USB 2.0 drive. Of course a USB 2.0 drive is much slower than an internal drive even though the access times are the same.
OK, so you've got this disk that can transfer 100MB/sec. That means it takes 0.04msec to read 4K bytes. However the disk takes 10msec just to FIND the 4K bytes in the first place, so that means the total time to read is 10.04msec.
If you DOUBLE the transfer rate, the read now takes 10.02msec.
But if you cut the ACCESS TIME in half, the read now takes 5.04msec.
Which do you think is more important for small files?
For LARGE files it works like this: to read a 100MB file it takes 0.01 seconds (10msec) to FIND the file, and 1 second to read it - total time of 1.01 seconds.
Double the transfer rate and you're down to 0.51 seconds.
Cut the access time in half and you're at 1.005 seconds.
Ergo: for small, random reads it's ACCESS TIME that's most important.
For large, sequential files it's TRANSFER RATE that's most important.
Booting from a USB drive is slow for two reasons: 1) USB Flash drives are not particularly fast, and 2) The USB protocol adds a lot of latency which kind of eliminates the no-seek advantage of the drive.
Your USB results surprise me - USB is based on a 1ms "pulse" which synchronizes data packets, I'd have expected something closer to a 2ms penalty for USB for the request and response, which would have been more significant for a typical hard drive with typical access times in the 8-10ms range.
The OP was asking about hard drives, and in that realm the transfer rate benefit for small files is pretty small compared to the access time.
But you are of course completely correct that the relative benefits of access time and transfer rate vary over the spectrum of large to small files. I've edited the post you referred to, changing "throughput is only important..." to "throughput is most important...".