SATA vs USB 3 performance

[denarced]

Hi,

 

When you're buying an internal HDD (non-SSD), people keep asking me: "Why didn't you just get a good USB 3 external one?". My answer has always been that it's easier to have an internal one because otherwise my server would be surrounded by external USB drives.

 

But the question does raise another one: are the two solutions equal? And I'm talking equal in practical terms, 10% difference isn't really meaningful. I've seen some comments here in the forum that USB 3 drive's random access speed is horrible and the interface itself takes at least 10% of the speed. But I haven't actually seen any recent performance tests on this one. I tried to search through Linus' stuff and couldn't find a video on this. I found some sites that had some performance comparisons but they were from 2009 and therefore probably irrelevant since USB 3 had some performance issues in the early days.

 

Is there a major difference? I know that external drives are handier but that's another topic ...

[Ssoele]

USB 3.0 will always be worse in this situation then SATA, because the drive in the enclosure also uses SATA, so you effectively add an interface on top of it, which only can make it slower.

[denarced]

USB 3.0 will always be worse in this situation then SATA, because the drive in the enclosure also uses SATA, so you effectively add an interface on top of it, which only can make it slower.

True. A clarification: I'm interested in how much slower.

[looney]

I tested this once by booting a VM on A external USB 3 HDD (WD Scorpio blue) and then connecting the same HDD via sata.

It resulted in a 9% speed difference, but when I came to transferring files there was barely a noticeable difference.

 

So it depends on you scenario, lots off small files (booting) will have a much bigger impact then large transfers. 

 

 

Please keep in mind that this is only based on a test I did a few months back and there are a lot of USB 3 chips out there just to list one of the variables that will have a impact on the test.

[denarced]

I tested this once by booting a VM on A external USB 3 HDD (WD Scorpio blue) and then connecting the same HDD via sata.

It resulted in a 9% speed difference, but when I came to transferring files there was barely a noticeable difference.

 

So it depends on you scenario, lots off small files (booting) will have a much bigger impact then large transfers. 

 

 

Please keep in mind that this is only based on a test I did a few months back and there are a lot of USB 3 chips out there just to list one of the variables that will have a impact on the test.

Thanks for your response. Even thou this is just one USB 3 chip with specific HDD, it's nice to see that the difference isn't that great. As mentioned previously in this thread, the random file performance was supposed to be bad but 9% difference isn't all that bad. I'm presuming that lots of small files include some random access reading.

[LogicalDrm]

With hdds difference between SATA2 and USB3 in theoretical speeds is really minimal. I would go for getting internal one since they are cheaper and have more options in size and manufacturer. You can always get external enclosure with USB3 connector if there's need for it.

[cpuxtech]

It all depends on how portable you want your data to be, but typically as it was said earlier, you are just adding another interface overhead to SATA if you use a USB 3 External drive so yeah you will take somewhat of a performance hit.