This project has moved. For the latest updates, please go here.

Performance comparison btw VC-1.17 and DiskCryptor 1.1.846.118

Topics: Technical Issues
Jul 4, 2016 at 9:29 AM
Edited Jul 4, 2016 at 9:35 AM
Hello,
I give thought to changing my disk encryption (SSD Samsung 850pro's, Win7pro, i7-3930K with AES-NI, 64Gb Ram) from DiskCryptor to Veracrypt. After doing some implementation tests I did also run CrystalDiskMark to see how VC performs under several file size conditions:
Image

There seems to be performance drops especially in 4K test!?
How to explain these values in 4K..-tests?

Thx for any feedback, regards, p.
Jul 5, 2016 at 6:08 AM
Hello,
an additional test with ATTO benchmark:
Image

Trend seems to be same, less performance with VC, especially in writing small files.
Any hint about this behaviour?

Thx + regards!
Jul 19, 2016 at 9:17 AM
Really no hint about this behaviour?
Regards P.
Jul 19, 2016 at 10:31 AM
Edited Jul 19, 2016 at 10:38 AM
I'd suggest to test with much bigger data, like 8GB. The more expensive drive you have, the bigger cache it has. And I'd test only with one pass to prevent caching of data. These are my results - unencrypted vs veracrypted full drive, but on the cheapest adata 550 ssd, it has capacity about 250GB and I have two veracrypted partitions on it (system part, data part).

Unencrypted:
Image
Fully encrypted
Image

This benchmark was made, when I left about half of the drive free (the second partition was not encrypted yet). I have expected improvement, but the is not any. So leaving some free space unencrypted will help probably only with more expensive samsung drive. If I'll have my hands on samsung, I can make some tests.

Encrypted half of the drive, other left blank:
Image

I cannot answer your question, but I can tell that the "look and feel" when working with the harddrive is the same with vc as without. I think the safety with VC worths the loss of some points measureable by the benchmark only ;-) I get noticeably different results each benchmark run and when I make multiple passes, the results are better than those displayed, probably because of caching and since they are fake, I have not displayed them - my adata manages to write large data at 50MB/s (uncached), which I have measured by real usage, so I have dispaleyd the results which were close to this number.
Jul 19, 2016 at 2:20 PM
Thank you so much for testing!
There seems to be a significant drop in performance with small files too. I did make some tests with a picture archiv, about 10 Gb files, size between 1 and 2,5 Mb each. Result is similar as shown above: Drop in 4k-benchmarks if using VC, but not if using DC.
Maybe some adjustments are necessary for SSDs?
Trim feature even for non-system disks?
Cache management?

Maybe idrassi can write a statement about this behaviour?

Thx + regards, p.