1

I have a Kingston SSDNOW on sata 2. Doing a benchmark i get these results:

enter image description here

As can be seen, the sequential write performance is terrible. Below even 100 MB/s. Also why do the tests labelled 4k show such terrible performance. Is my drive dying?

I am on windows7 with trim enabled.

Update
The exact model of the SSD is:

Kingston SSDNow V100 256GB
It claims 250MB/sec read and 230MB/sec write speeds
http://www.amazon.com/Kingston-SSDNow-Notebook-SV100S2N-256GZ/dp/B004BDORLU

5
  • 3
    What's the exact model of your SSD? It makes a huge difference. Apr 30, 2012 at 23:31
  • What kind of a machine do you have the SSD connected to? Certain SATA controllers tend not to play well with SSDs.
    – Mr Alpha
    May 1, 2012 at 9:59
  • @MrAlpha its the intel x58 chipset. connected to the onboard intel sata 2.0 controller.
    – pdeva
    May 3, 2012 at 7:22
  • @DavidSchwartz done.
    – pdeva
    May 3, 2012 at 7:22
  • +1 for giving me a good benchmark tool. But it should be noted that mbps is not the same same MB/s... May 4, 2012 at 16:43

3 Answers 3

4

Typical lower-end Kingston SSDNOW SSDs are rated for 6,000 read IOPS. Yours is rated for 3,300 read IOPS. A 4KB read is 4,096 bytes. So 12.33MB/s is 12.33*1024*1024/(4*1024) IOPS, or 3,156. So that seems reasonable.

Evidently the Kingston SSDNow V+ 100 256GB and SSDNow V 100 256GB provide quite poor NCQ performance, as they suffered when running the CrystalDiskMark random read 4K-QD32 test. This was also a weakness of the original SSDNow V 128GB drive. Whereas the old OCZ Vertex 120GB provided a throughput of 65MB/s, the SSDNow V+ 100 256GB was limited to just 19.9MB/s.

The random write 4K-QD32 performance was slightly better as the Kingston SSDNow V+ 100 256GB achieved a throughput of 37.1MB/s while the SSDNow V 100 256GB managed 28.4MB/s, which was a considerable improvement over the 11.4MB/s of the SSDNow V 128GB drive. - Review

1
  • I have updated the question with the exact model now
    – pdeva
    May 3, 2012 at 7:33
5

Two things immediately leap to mind.

The first is as psusi already mentioned that the drive is really full. Besides the impact on write performance you suffer from the lack free blocks psusi talks about, there are two other effects as well. With a large drive like that, full to the brim, the FTL table will be huge so doing look-ups in it will be slow. This can be part of the explanation for the low 4k score. Also, with that little space Windows will start having problems with fitting in temporary files and such. You really need to free up space one way or another.

The other this is that I suspect your SATA controller is in IDE emulation mode. This means you don't get NCQ which will hurt performance, especially the 4k QD32 numbers. A telltale sign of this is that the 4k and 4k QD32 numbers are essentially the same.

0

The drive is nearly full so there are too few free erase blocks to maintain performance. Delete some files to free up some space and try again.

3
  • His random read performance is the worst number there. Do you think the number of erase blocks affects read performance? May 1, 2012 at 1:04
  • @DavidSchwartz, it is odd that the read performance is worse than write.
    – psusi
    May 1, 2012 at 2:10
  • One big difference is that with a write operation, you don't have to wait for the medium, since the drive can write to cache, complete the operation, and finish writing to the medium later. A read operation cannot complete until the drive has finished reading the data. That suggests that the problem is the raw number of I/O and the rate at which the flash itself can complete them. May 1, 2012 at 5:51

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .