Not for me...
457MB RAM for each task, not so much for me.

It looks like they require 2 GB to download, but much less to run. It is common for the jobs to reserve a lot more memory than they typically use; there could be larger ones also.

Problem Report #1: Five (5) AMD Threadripper CPU’s lockup running Fight Aids at Home (FAH2) > 22 threads.

Severity: Must avoid 100% FAH2 World Community Grid (WCG) project on BOINC using AMD Threadripper CPU’s.
100% Zika projects are okay. Mix of projects are okay including a few FAH2 work units in the mix.

Environment:
OS: Ubuntu 16.04.6 LTS (Unity) on all twenty (20) machines. Linux Version 4.15.0-46 generic.
Boinc: Version 7.6.31 from Ubuntu repository.
CPU’s 5 - AMD Threadrippers 16 Cores 32 Threads Machine fails, Locks up > 22 threads of FAH2
9 - AMD Ryzens 8 Cores 16 Threads Machine is good, all 16 FAH2 threads running.
6 - Intel i7-4790K 4 Cores 8 Threads Machine is good, all 8 FAH2 threads running.

Symptons:
1.) After about a minute following reset, the machine locks up with no mouse or keyboard input accepted.
2.) Low heat output seems to indicate number crunching has also stopped after the lockup.
3.) The only exit is a hard power off by holding the power-on button depressed several seconds. Keyboard CTRL-ALT-DEL does nothing.
4.) Fortunately, the machine operates normally for about a minute after completing power-on startup. Otherwise, a disk OS re-installation would be necessary.
5.) In that first minute of normal operation, one must quickly open BOINC manager and “suspend” World Community Grid (running 32 FAH2 threads). Now the machine will run normally indefinitely without locking up.
6.) FAH2 is then deselected at WCG web site, “My Projects”. All FAH2 work units on five (5) Threadripper machines must be manually “Aborted” by user. However, each of the five (5) Threadripper machines successfully resume number crunching doing 32 queued (“waiting”) Zika work units without locking up.
7.) After having “suspended” WCG in the BOINC console, FAH2 is seen to be running on 22 of 32 threads. #1 Longest running FAH2 thread has run over 4 hours. #22 shortest running good thread has run 1 hour 22 minutes. The last 10 threads out of 32 have run equally just 14 Seconds.
8.) Unfinished Zika threads are “waiting” because the FAH2 threads have a shorter deadline (4 April FAH2 vs 12 April Zika). The five (5) Threadripper machines are each manually recovered as above by aborting all FAH2 work units. All fifteen (15) other machines are untouched, and operate normally throughout the Threadripper recovery. Those good 15 machines have run eight (8) and sixteen (16) FAH2 threads at 100% loading, like Threadrippers normally do.
9.) I have run all 20 machines where all projects are selected at WCG web site, including FAH2 in the mix. FAH2 work units processed good on Threadripper, except when the FAH2 load is 100%.

Comments:
1.) There is a lot if unused memory on all machines, including the Threadripper 1950X which has 16 GB. I wonder if the memory allocation should be allowed to grow larger than is set in BOINC?
2.) The delay before lockup at startup is puzzling (though fortunate)? There is a similar delay period, a window of time in which the BOINC Manager must be opened after startup (and left open), else it will never open afterward.
3.) I wonder if FAH2 software or BOINC could detect this problem and avoid it?

I noticed some of these work units took 8% or a little more of RAM on my 4-core processor. This was not running me out of RAM. I had 8 GBytes RAM in my machine with another 8 GBytes on order. When the RAM came, I installed it, and changed the CR2032 clock battery that was a little over 7 years old.

So now those work units each take only 4% of my RAM (not surprising).