Serious Performance Hit on SSD

[Jim Doty]

I just replaced the HDD in my Sony VAIO with a new Samsung SSD 860 EVO 1TB. I used Samsung Magician to run performance benchmarks, and before encryption I got the following:

Sequential Read: 3,109 MB/s
Sequential Write: 3,344 MB/s
Random (IOPS) Read: 58,349
Random (IOPS) Write: 30,273

After full encryption of the System Partition, I reran the benchmarks and got:

Sequential Read: 812 MB/s
Sequential Write: 646 MB/s
Random (IOPS) Read: 19,287
Random (IOPS) Write: 16,845

That's a 4:1 reduction in Sequential and 2:1 to 3:1 reduction in random.

Windows 10 Pro, 64 bit
Intel i5-3210M CPU @ 2.5GHz
8 GB RAM
VeraCrypt 1.23-Hotfix-2 (64 bit)

Is there any fix for this?

[theincogtion]

I switched from VeraCrypt to Bitlocker due to this performance issues. I am using an M.2 SSD
Check the Screenshots attached

[DDD]

good to know. I also found a big performance hit on a regular hard drive. Why don't you upgrade to 1.24U6?

[Jim Doty]

Thanks for the tip theincogtion. I rejected Bitlocker before because it had all sorts of security issues, but they have now corrected those, so I gave it a try:

6/3/2020
With full encryption of the System Partition--VeraCrypt 1.23-Hotfix-2 (64 bit)--I reran the benchmarks and got:
Sequential Read: 709 MB/s
Sequential Write: 770 MB/s
Random (IOPS) Read: 32,470
Random (IOPS) Write: 21,484
After removing encryption of the System Partition--no encryption at all--I reran the benchmarks and got:
Sequential Read: 3,986 MB/s
Sequential Write: 4,613 MB/s
Random (IOPS) Read: 83,251
Random (IOPS) Write: 37,353
After encrypting the System Partition with Bitlocker and a long Password, I reran the benchmarks and got:
Sequential Read: 2,438 MB/s
Sequential Write: 2,707 MB/s
Random (IOPS) Read: 65,429
Random (IOPS) Write: 36,132

That's a big improvement over VeraCrypt, approximately 4:1.

DDD, I'm not willing to go to the trouble of trying 1.24U6 unless I know it offers some significant benefit over what Bitlocker is doing for me now.

[theincogtion]

Diskcryptor is also worth looking at. It has also some performance issues but far better then vera crypt
https://forum.diskcryptor.org/viewtopic.php?f=18&t=302

[Dave]

DiskCryptor uses a lot more assembly language and this could make it much faster to use? DiskCryptor is written in C and Assembly (according to Wikipedia). C++ is apparently slower than C? So maybe DiskCryptor could be faster overall?

[xxx]

i recently moved one system to ssd, and it really takes a huge performance hit
the ecrypted partition vs another, unencrypted performance hit is very, very big
sequential writes/writes are ok, they are about 1/3 about what they are unencrypted, but that's expected after all
but rnd4k q32t16 went down from over 2000MB/S to 54!
and rnd4k q1t1 went down from over 60MB/S to 20
this issue has been reported years ago from i can find on the forums, and i know what the devs said, that it is a time consuming fix, but shouldn't it be eventually implemented?
i mean, from 2000Mb to 54...
thanks