My first build with an AMD FX-8350 I am hoping it will be ok for encryption

[Essence_of_War]

I am judging by the time it takes to encrypt a 4TB hard drive
AES took over 7 hours
Twofish cascaded with Sepent took 7 hours but seemed to finish first.
These were tested on 7200RPM drives NAS drives. I did time it. they said 6 hours but took 7 both of them.

Ok. And if you do a bit of math, 4TB at 7 hrs ~130-140 MB/s, which happens to be just aobut the maximum speed you can write data to a 4TB hard drive.

The reason you don't see any difference between AES and Twofish is that
1) You were IO limited - both Twofish and AES were fast enough to saturate your bandwidth to disk.
2) You didn't monitor CPU usage - if you had, you would have likely seen your CPU idling under AES and fully loading a core while doing Twofish.

 

[Headfoot]

OP: The point is you could have gotten something really cheap and low power like a Jaguar chip or a Atom Rangley/Avoton 8-core both of which have AESNI built in and are very low power and still cap out your HDD speed with encryption bandwidth. Both are good NAS chips

 

[BirdDad]

Ok. And if you do a bit of math, 4TB at 7 hrs ~130-140 MB/s, which happens to be just aobut the maximum speed you can write data to a 4TB hard drive.

The reason you don't see any difference between AES and Twofish is that
1) You were IO limited - both Twofish and AES were fast enough to saturate your bandwidth to disk.
2) You didn't monitor CPU usage - if you had, you would have likely seen your CPU idling under AES and fully loading a core while doing Twofish.

Normally the hard drives transfer at around 140- 170MBps

 

[Essence_of_War]

Normally the hard drives transfer at around 140- 170MBps

I'm sorry, I honestly can't tell what you're getting at here.

 

[Jovec]

Normally the hard drives transfer at around 140- 170MBps

Essence_of_War's points were correct. You are I/O limited - the time it takes to read the unencrypted data from the hard drive and write the encrypted data back to the hard drive far outweighs the time is takes to do the encryption, whether in software or with hardware AES instructions.

If all the CPU is doing is encrypting then hardware AES-NI support won't make too big of a difference. Combining that encryption with other CPU heavy tasks is where hardware AES-NI will help.

 

[BirdDad]

Very nice speedup on boot with AES

 

[BirdDad]

It can't be entirely I/O bound, my laptop which is an i7 quad encrypts at only 120MBps that's using the USB 3.0 port so something else is going on there.

 

[Essence_of_War]

It can't be entirely I/O bound, my laptop which is an i7 quad encrypts at only 120MBps that's using the USB 3.0 port so something else is going on there.

...

How can we possibly evaluate this without knowing what application you're using for the encryption, what encryption method you're using, what the IO systems look like, etc?

 

[MiddleOfTheRoad]

Despite it cost 390 dollars, it has 50% more threads than it!

$1000+ chip and $400 chip being compared to one that costs around $99 to $175 (depending on actual model -- FX-8310, etc.)

Talk about a totally apples to oranges -- totally pointless comparison.

 

[MiddleOfTheRoad]

What we need is a frames per dollar or hashes per dollar type benchmark.

Passmark is probably as close as it gets (although I wish they'd take out all the stupid mobile CPU's off the list for simplicity):
https://www.cpubenchmark.net/cpu_value_available.html