I don't think that bigger buffer size will got u more WUs for FAH2. You can leave it on 1 day or 10 days, but with limited amount of work, yours PC will only got one stock of WUs most probably similar to number of cores it has. At least that is in my case. FAH2 is slight more CPU demanding than MCM. As I can recall, time for completing old and new FAH2 WUs is almost the same.

Agreed, @kapetan_proton, I did not see a daily increase in FAH2 work units when I set the cache to higher number for hours.

FAH2 is little more time than MCM for me 3 hours dayly i had 31 FAH2 on One day
now sinds the change 0 WU for FAH

But it is one of the only project which depends on work returned to generate more work. If someone has 10 + 10 days stored and cached, that slows down work generation tremendously with even this 7 day deadline. Worse case there but you get my meaning.
In addition project preferences allow one to set how many WUs are sent to each machine.
OT, but I hope CPDN's ridiculously long deadlines get shortened in a similar manner, too.

What you may be missing is that each FAH WU is one stage out of about 150 to 200 that make up a total run; if each stage might take as long as 7 days to turn over, the total time to complete an experiment could be several years (at which point, why bother running the experiments at all?!?)

There are, of course. two ways one can moderate the amount of FAH2 work that runs on a machine. One approach is to use an app_config.xml file to limit the number of FAH2 tasks that can run at the same time. The other is to use a WCG device profile to limit FAH2 to some appropriate small number of tasks downloaded at any one time. The best approach may be to combine the two techniques - for example, on an 8-core machine ask for up to three FAH2 tasks at a time but only run two at a time...

(Applying some sort of limits using the above technique is even more important for MIP1 because of the way it thrashes L3 cache! But that's another topic for another forum.)

And if you think that sounds like unnecessary effort to achieve an effect, I would point out that, before they implemented the "send at most X" mechanism, I sometimes had to stop accepting MCM1 or SCC1 work because I wasn't being sent anything but - if you have a mixed workload, it should balance out unless you only select the long-running (and less available) tasks such as ARP1 and HST1 to keep FAH2 company!

By the way, the first FAH2 work units had 10 times as many iterations per unit, a deadline of four days and a trickle mechanism that made a lot of people very unhappy - browse early threads in this forum for evidence of that!

I hope the above helps...

Cheers - Al.