LinuxQuestions.org - New SSD Speed Less Than Half Of Expected

- Linux - Hardware (https://www.linuxquestions.org/questions/linux-hardware-18/)

- - New SSD Speed Less Than Half Of Expected (https://www.linuxquestions.org/questions/linux-hardware-18/new-ssd-speed-less-than-half-of-expected-4175519765/)

New SSD Speed Less Than Half Of Expected

I just got a new SSD to put in my laptop. A Compal VAW70, the SSD is a OWC Mercury Electra 120GB 6G. It advertised a read spead of 556 MB/s & Write of 523 MB/s. And all reviews I found of the drive show speeds around that ~520 MB/s or so. But I am only getting 225. I got it up to 257 with some tweaking of the scheduler and read ahead value. But it's still less than half of what it should be and I don't know why.

smartctl shows the link is at 6 Gb/s.

Code:

smartctl -a /dev/sda | grep SATA

SATA Version is: SATA 3.0, 6.0 Gb/s (current: 6.0 Gb/s)

So it shouldn't be limited by the link. I have btrfs on the root partition and the appropriate options in fstab for an SSD. At least I think I have it all set up correctly. But if I did it should be running a lot faster.

I have also stuck the drive in my tower and get the same read speeds. While my other SSD a Kinston HyperX 120GB gets 432 MB/s.

Here is some info on my hardware.

Code:

lspci | grep SATA

00:1f.2 SATA controller: Intel Corporation 8 Series SATA Controller 1 [AHCI mode] (rev 04)

Code:

hdparm -i /dev/sda



/dev/sda:



 Model=OWC Mercury Electra 6G SSD, FwRev=582ABBF0, SerialNo=OW140820AS1510270

 Config={ HardSect NotMFM HdSw>15uSec Fixed DTR>10Mbs RotSpdTol>.5% }

 RawCHS=16383/16/63, TrkSize=0, SectSize=0, ECCbytes=4

 BuffType=unknown, BuffSize=unknown, MaxMultSect=1, MultSect=1

 CurCHS=16383/16/63, CurSects=16514064, LBA=yes, LBAsects=234441648

 IORDY=on/off, tPIO={min:120,w/IORDY:120}, tDMA={min:120,rec:120}

 PIO modes:  pio0 pio1 pio2 pio3 pio4 

 DMA modes:  mdma0 mdma1 mdma2 

 UDMA modes: udma0 udma1 udma2 udma3 udma4 udma5 *udma6 

 AdvancedPM=yes: unknown setting WriteCache=enabled

 Drive conforms to: unknown:  ATA/ATAPI-2,3,4,5,6,7

Code:

hdparm -t /dev/sda



/dev/sda:

 Timing buffered disk reads: 674 MB in  3.01 seconds = 224.10 MB/sec

Code:

cat /etc/fstab 

/dev/sda2      /                btrfs    rw,relatime,ssd,dicard,space_cache  0 0

/dev/sda1      swap            swap    defaults        0 0

nas.local:/shares    /home/tom/nfs    nfs      noauto,users      0 0

My speeds fluctuate - I have seen nearly 500 MB/S from my ssd. today in the same box it's 260-335MB/S. Don't believe theoretical maxima (e.g. 6Gigs) will be realizable in a less than optimal framework (i.e. a pc) with it's own hidden and little advertised bottlenecks, under an OS built to multitask.

225 MB/S is a hell of a lot faster than the platter based disks.

What is the quality of the system? High end or common?

I know nothing about the OP's system, but it's not unheard of for the bottleneck to be system bandwidth rather than a drive's capability.

I was able to get the speed up to 285 by adjusting the read_aheak_kb value to 8192. But never got it higher. I wasn't doing anything else on the machine at the time. On one I booted up to Parted Magic on my flash drive. So the OS wasn't even using the drive.

I've tried it in 3 different computers. In my gaming box I have an ASROCK 970 Extreme 3 Mobo with FX-6300 CPU. In my main computer, an intel DB75EN mobo with i5-3330 CPU. I have an SSD in there too. It gets 480-501 MB/s (now that I've tweaked it too). And yet the new drive only gets 280.

I was originally thinking it was the laptop I had it in that was the cause of the problem. But since the drive gets the same speed in the other boxes I tried it out in. I'm thinking that is the actual speed of the drive. I did some more reading on SSD benchmarking. There can be a big difference in the read/write speed depending on the type of dating being read. Incompressable data can slow down certain controllers a lot. I'm not sure how what hdparm would be. Just random data on the disk?

I know it's a lot faster than the HDD that was in it. It boots up in a flash now. And retoring the backup and making new backs only took about 3-4 minutes. But it really erks me that I can only get half the speed they claimed. If it was say 15% slower I wouldn't have an issue with that. The rig they were using is probably a hell of a lot better than mine. But to get less than half the speed they claim it is cabpable off just seems like there is something wrong. Something slowing it down.

ok, seeing as you are into testing, try
dd if=/dev/zero of=file-on-ssd bs=100M count=5
That will give you an infinitely compressible file. Then something like
dd if=file-on-ssd of=/dev/null
I just tried it. Write 100MB - 1.0G/S; Write 500MB - 410MB/S; read 500MB - 694MB/S.

It shows you content does matter, and testing doesn't :-).

Do all your machines mobos actually support sata3(?).

I tried writing and read the zeroed file and got similar numbers to you. So if all my files of size zero my laptop would be blazing fast hehe.

Quote:

Do all your machines mobos actually support sata3(?).

Yes all 3 machines have SATAIII controllers. And I verified that they are the link was running at SATAIII speeds with smartctl and by checking dmesg.

I just got home from work and I installed gnome-disks. Ran it's benchmark tool and got some interesting results. The first 5% of the samples were around 260-280 MB/s like I was getting with hdparm. but then after that it shot up to around 560. And got an average of 534.3 MB/s.

I guess the drive is slower to start but performs better on longer reads? But yeah there is no configuration problem like I was thinking. It can get the read speads it claimed. So I'm happy with my zippy new SSD :).

Thanks for the help everyone!

Sounds like some cache issue??? Can't imagine why it would "get faster" otherwise.

Maybe a partition alignment issue? The sector alignment value should be 2x the drive NAND erase block size

I had an issue with this and a Samsung evo 840 drive, got buffered disk reads like 200-280 MB/sec (hdparm -t output) then I wiped the drive (with SSD memory cell cleaning), re-did the partitions with sector alignment value of 3072 in gdisk (2x 1536kb drive NAND erase block size). Now hdparm -t outputs 524.19 MB/sec which was the claimed perf by Samsung litterature

Edit, just read last post... Seems another issue if the drive goes full speed after a while