I've been testing a Samsung EVO 32GB SDHC card and have been seeing some strange benchmark results on write speeds. I'm using a SanDisk All-In-One USB 3.0 reader with a setup that tests other SD cards consistently (up to 85MB/s writes), so I don't think equipment is an issue here.
To begin, the Samsung EVO SD card is new out of the package, never written or formatted.
The relatively slow sequential large writes and faster random 512KB writes was unexpected. However, I noticed similar results in another article here ( 19.0MB/s sequential write / 31.6MB/s random large block write ).
Perhaps this card was doing some initialization that slowed down the first sequential test. So I went ahead and ran the same test again.
Sequential is faster, random is slower. CDM is reporting the best of 5 passes so these results might not tell the whole story. I ran some more single pass tests and they continued about the same as the second one. At this point I did a full overwrite format using SD Formatter 4.0 and then ran some more single tests. Performance was basically unchanged after the format.
I then ran ATTO and this is where things got strange again:
ATTO is using uncompressed data for testing, maybe that had an effect. Anyway, I went on and ran the same test using another reader. Lexar SR1 USB 3.0:
That didn't go as well. So I put it back in the SanDisk All-in-One:
And now I'm starting to wonder what is happening, but I run the same test again:
And back on the Lexar SR1:
Is there something going on with the card controller here? Is it confused and/or adjusting to these benchmark tests?
For some some real world tests I copied a few sets of 24 RAW images over (40-50MB each) from a SSD. I was seeing writes in the 20MB/s to 25MB/s range.
Running CDM with some sequential writes:
I actually had 2 tests report 40MB/s but didn't capture the first result.
QuickBench:
What is going on with this card? Could this be a normal(?) controller behavior, or possibly an issue with the card/error correcting.
To begin, the Samsung EVO SD card is new out of the package, never written or formatted.
Code:
-----------------------------------------------------------------------
CrystalDiskMark 3.0.3 x64 (C) 2007-2013 hiyohiyo
Crystal Dew World : http://crystalmark.info/
-----------------------------------------------------------------------
* MB/s = 1,000,000 byte/s [SATA/300 = 300,000,000 byte/s]
Sequential Read : 46.325 MB/s
[B]Sequential Write : 19.457 MB/s[/B]
Random Read 512KB : 41.880 MB/s
[B]Random Write 512KB : 35.283 MB/s[/B]
Random Read 4KB (QD=1) : 4.623 MB/s [ 1128.6 IOPS]
Random Write 4KB (QD=1) : 1.084 MB/s [ 264.8 IOPS]
Random Read 4KB (QD=32) : 4.691 MB/s [ 1145.2 IOPS]
Random Write 4KB (QD=32) : 1.210 MB/s [ 295.3 IOPS]
Test : 1000 MB [I: 0.0% (0.0/29.3 GB)] (x5)
Date : 2014/07/21 14:26:39
OS : Windows 7 Professional SP1 [6.1 Build 7601] (x64)The relatively slow sequential large writes and faster random 512KB writes was unexpected. However, I noticed similar results in another article here ( 19.0MB/s sequential write / 31.6MB/s random large block write ).
Perhaps this card was doing some initialization that slowed down the first sequential test. So I went ahead and ran the same test again.
Code:
-----------------------------------------------------------------------
CrystalDiskMark 3.0.3 x64 (C) 2007-2013 hiyohiyo
Crystal Dew World : http://crystalmark.info/
-----------------------------------------------------------------------
* MB/s = 1,000,000 byte/s [SATA/300 = 300,000,000 byte/s]
Sequential Read : 46.465 MB/s
[B]Sequential Write : 27.873 MB/s[/B]
Random Read 512KB : 42.091 MB/s
[B]Random Write 512KB : 23.261 MB/s[/B]
Random Read 4KB (QD=1) : 4.596 MB/s [ 1122.2 IOPS]
Random Write 4KB (QD=1) : 1.668 MB/s [ 407.3 IOPS]
Random Read 4KB (QD=32) : 4.675 MB/s [ 1141.3 IOPS]
Random Write 4KB (QD=32) : 1.680 MB/s [ 410.1 IOPS]
Test : 1000 MB [I: 0.0% (0.0/29.3 GB)] (x5)
Date : 2014/07/21 15:11:21
OS : Windows 7 Professional SP1 [6.1 Build 7601] (x64)Sequential is faster, random is slower. CDM is reporting the best of 5 passes so these results might not tell the whole story. I ran some more single pass tests and they continued about the same as the second one. At this point I did a full overwrite format using SD Formatter 4.0 and then ran some more single tests. Performance was basically unchanged after the format.
I then ran ATTO and this is where things got strange again:
ATTO is using uncompressed data for testing, maybe that had an effect. Anyway, I went on and ran the same test using another reader. Lexar SR1 USB 3.0:
That didn't go as well. So I put it back in the SanDisk All-in-One:
And now I'm starting to wonder what is happening, but I run the same test again:
And back on the Lexar SR1:
Is there something going on with the card controller here? Is it confused and/or adjusting to these benchmark tests?
For some some real world tests I copied a few sets of 24 RAW images over (40-50MB each) from a SSD. I was seeing writes in the 20MB/s to 25MB/s range.
Running CDM with some sequential writes:
Code:
50MB
Sequential Write : 40.442 MB/s
Sequential Write : 24.013 MB/s
Sequential Write : 24.057 MB/s
Sequential Write : 24.098 MB/s
100MB
Sequential Write : 26.904 MB/s
Sequential Write : 26.811 MB/s
Sequential Write : 24.025 MB/s
Sequential Write : 26.439 MB/sI actually had 2 tests report 40MB/s but didn't capture the first result.
QuickBench:
Code:
---- Standard Test Details: 4 Kbytes - 1024 Kbytes (1 Megabyte) ----
Transfer Size Sequential Read Sequential Write Random Read Random Write
4 KBytes 020.729 MB/sec 001.644 MB/sec 005.002 MB/sec 002.026 MB/sec
8 KBytes 028.290 MB/sec 002.359 MB/sec 009.367 MB/sec 004.079 MB/sec
16 KBytes 034.782 MB/sec 003.429 MB/sec 015.316 MB/sec 005.515 MB/sec
32 KBytes 039.438 MB/sec 004.038 MB/sec 017.935 MB/sec 005.910 MB/sec
64 KBytes 041.665 MB/sec 007.705 MB/sec 022.122 MB/sec 015.389 MB/sec
128 KBytes 043.196 MB/sec 011.038 MB/sec 029.451 MB/sec 027.101 MB/sec
256 KBytes 043.790 MB/sec 013.657 MB/sec 035.087 MB/sec 029.988 MB/sec
512 KBytes 044.171 MB/sec 017.420 MB/sec 038.876 MB/sec 016.973 MB/sec
1024 KBytes 044.223 MB/sec 016.935 MB/sec 041.440 MB/sec 032.143 MB/sec
Test Average: 037.809 MB/sec 008.692 MB/sec 023.844 MB/sec 015.458 MB/sec
Lowest Mark: 020.729 MB/sec 001.644 MB/sec 005.002 MB/sec 002.026 MB/sec
Highest Mark: 044.223 MB/sec 017.420 MB/sec 041.440 MB/sec 032.143 MB/sec
---- Large Test Details: 2 MB - 10 MB ----
Transfer Size Read Speed Write Speed
2 Megabytes 044.361 MB/sec 035.919 MB/sec
3 Megabytes 044.229 MB/sec 035.626 MB/sec
4 Megabytes 044.420 MB/sec 035.561 MB/sec
5 Megabytes 044.219 MB/sec 035.571 MB/sec
6 Megabytes 044.269 MB/sec 035.418 MB/sec
7 Megabytes 044.085 MB/sec 020.436 MB/sec
8 Megabytes 044.194 MB/sec 016.862 MB/sec
9 Megabytes 044.054 MB/sec 017.775 MB/sec
10 Megabytes 044.186 MB/sec 019.074 MB/sec
Test Average: 044.224 MB/sec 028.027 MB/sec
Lowest Mark: 044.054 MB/sec 016.862 MB/sec
Highest Mark: 044.420 MB/sec 035.919 MB/sec
Code:
---- Large Test Details: 2 MB - 10 MB ----
Transfer Size Read Speed Write Speed
2 Megabytes 044.192 MB/sec 029.758 MB/sec
3 Megabytes 044.366 MB/sec 029.170 MB/sec
4 Megabytes 044.323 MB/sec 035.429 MB/sec
5 Megabytes 044.254 MB/sec 035.470 MB/sec
6 Megabytes 044.360 MB/sec 035.263 MB/sec
7 Megabytes 044.051 MB/sec 018.434 MB/sec
8 Megabytes 044.155 MB/sec 019.499 MB/sec
9 Megabytes 044.092 MB/sec 018.765 MB/sec
10 Megabytes 044.138 MB/sec 019.146 MB/sec
Test Average: 044.215 MB/sec 026.771 MB/sec
Lowest Mark: 044.051 MB/sec 018.434 MB/sec
Highest Mark: 044.366 MB/sec 035.470 MB/secWhat is going on with this card? Could this be a normal(?) controller behavior, or possibly an issue with the card/error correcting.
