Hello,
My system is an openSUSE 13.1 updated to the latest patch level.
Motherboard is Asrock H81M with Intel H81 chipset, obviously.
I have a Sundisk Cruzer Extreme USB 3.0 stick (formated as EXT4). My problem is that the write speeds are no better than 35MB/sec when reading is made from the local HDD. The HDD is a Seagate, with at least 120MB/sec read speeds. What is extremely strange, is that I can write to the USB sticks with 100% the nominal write speed (90MB/s+) when I read files from an NFS share (which by the way is an export of a SLOWER HDD )
I have the same problem with all USB 3.0 sticks (I have 2 more Extremes and a Flash Voyager GT). Tthey seem to work as a slightly better USB 2.0).
How can I fix that?
> I have the same problem with all USB 3.0 sticks (I have 2 more Extremes
> and a Flash Voyager GT). Tthey seem to work as a slightly better USB
> 2.0).
Usb flash stick write speed is way lower than what USB 3 provides. It is
even slower than USB 2 speed, so there is no advantage with usb3.
You have to test write speed with a rotating hard disk as the destination.
–
Cheers / Saludos,
Carlos E. R.
(from 13.1 x86_64 “Bottle” at Telcontar)
On 2014-09-11 18:16, gogalthorp wrote:
>
> Actually out there yes. depending on size up to 190 meg write is
> published. But it all depends look at the graph on this page
>
> http://tinyurl.com/kmaugkm
>
> Note the drop off with smaller chunks.
Small files took longer to write? Curious.
Although I don’t really understand the table. The ‘Y’ axis is labeled
“writes”. Top of the chart says “256 MB Length, queue depth 4”. I don’t
understand what they are testing and how.
Does it means that they are writing a single 256 MB block (random data?)
1 time, then 2 times, then 4 times, up to 8192 times? Then the table
speed would be just a result of a big internal cache…
There is a comment at the page bottom that says: “At 7.46 MB/s this
drive has the highest 4K random write speed we have seen by a factor of 10”
So it really writes at 7.46 MB/s, not 200. :-?
–
Cheers / Saludos,
Carlos E. R.
(from 13.1 x86_64 “Bottle” at Telcontar)
The question is not about the USB sticks capabilities though. They are capable or thouse speeds, as the NFS copy proved to me. The question is why copy from LOCAL HDD is that slow.
Can you post the results of **mount -a ** when the disk is inserted?
Try manually mounting it with the async option, it was an issue back in the day.
As for the USB stick speeds, I have a Sandisk Extreme 64 which in OS X writes around 100-150MB/s on average (that’s MB/s, not Mbps) for large contiguous files - however random writing small files is a pain and slows it down to a crawl.
On 2014-09-12 10:36, tpe wrote:
>
> The question is not about the USB sticks capabilities though. They are
> capable or thouse speeds, as the NFS copy proved to me. The question is
> why copy from LOCAL HDD is that slow.
Well… Then you have to test the capabilities of each media separately.
echo "write test, one big file:"
time dd if=/dev/null of=/media/usbfilesystem/file500 bs=10M count=50
time sync
time dd if=/dev/null of=/media/usbfilesystem/file5000 bs=10M count=500
time sync
# perhaps umount/mount media, to clear cache, here.
echo "read test, one big file:"
time dd of=/dev/null if=/media/usbfilesystem/file500 bs=10M count=50
time sync
time dd of=/dev/null if=/media/usbfilesystem/file5000 bs=10M count=500
time sync
Then repeat for the hard disk, testing both the ntfs partition and the
ext4 partition - assuming they are on the same disk. You just need to
change the parameters in the command above - but be very careful, it
will write anywhere you tell it to! Being root, it can overwrite the
partition table, filesystem structures, anything. Better use it as plain
user.
–
Cheers / Saludos,
Carlos E. R.
(from 13.1 x86_64 “Bottle” at Telcontar)
The problem is not the USB. USB CAN write more than 90MB/s.
The problem is that LOCAL HDD has reduced performance. I will re-open the thread as another post.
On 2014-09-13 17:46, tpe wrote:
>
> The problem is not the USB. USB CAN write more than 90MB/s.
> The problem is that LOCAL HDD has reduced performance. I will re-open
> the thread as another post.
Well, the commands I asked you to do would find out if the local hdd has
reduced performance, but you refuse to help yourself…