QuickAssist Driver for FreeBSD is Here and pfSense Support Coming

Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.

Patrick

Administrator
Staff member
Dec 21, 2010
12,511
5,792
113
All I have to say on this one is finally. There was an "underground" driver in Q3/Q4 last year, but this is finally happening.

I do know the pfSense team is doing a lot with QAT.

On the networking side, this is one of the biggest homelab developments all year.
 
  • Like
Reactions: gigatexal

mstone

Active Member
Mar 11, 2015
505
118
43
46
This is going to be the biggest disappointment rather than the biggest development.
 

Evan

Well-Known Member
Jan 6, 2016
3,346
598
113
This is going to be the biggest disappointment rather than the biggest development.
Could well be as I understand now 3 different QAT 'versions' and each needs a different driver right ? (Sure I read that somewhere)
 

mstone

Active Member
Mar 11, 2015
505
118
43
46
Because people seem to generally not understand the benefits and limitations of the technology and mistakenly expect that it will be a magic wand to deliver dramatic performance improvements which won't pan out in real life. I don't understand why Patrick has been such a cheerleader for the technology, instead of rigorously comparing it to non-QAT solutions for real-world workloads.
 

Patrick

Administrator
Staff member
Dec 21, 2010
12,511
5,792
113
pfSense has the ability to do some of the things necessary for QAT to work well. For example, it can automate setup from driver through to VPN service definition. That is really important with QAT.

We did a lot of work with Coleto Creek that never got published with the new generation coming out. The use cases are a bit limited, and the setup sucks/ sucked. On the other hand, when it worked it was like get XYZ performance with QAT using a fraction of the core count. Often you can hit higher throughput using CPU cores, but then you are using the CPU cores solely for that task.

Not a magic wand, but far from meritless.
 

mstone

Active Member
Mar 11, 2015
505
118
43
46
I agree it has uses, but the coverage this far hasn't done a good job of calling those out. The perception is that it's a widget that makes crypto faster, without much appreciation that not only can it be slower, it's useful mainly insofar as you value not adding cores. It has some possible application on the low power end (but not with Intel's retail pricing) but more so if you have to add crypto to a box that's maxed out with 50 or 100 high power cores where the cost of adding is really steep. For a small scale application where it's the difference between 2, 4, or 8 cores, it generally didn't make sense. (In most cases you'll get more use out of a couple of additional general purpose cores.) A lot of people bought into rangeley because they thought it would speed up their openvpn rooter, and most of those people will end up throwing that equipment out without ever having used the QAT. I hope the coverage of the new parts makes the use cases a lot more clear so people don't waste their time.
 
  • Like
Reactions: sth and eva2000