Saluton!
I used to run a 500MHz G4, and now run a G5 PPC 2.3GHz Mac. (BOINC on one core, Folding at Home on the other.)
The old SETI work units were about half the size of the large Einstein wu's on the Mac. Currently (and this is a very rough estimate), the SETI enhanced wu's are abut 75% the “crunchiness” (roughly estimated crunch time) of the large Einstein wu's.
SIMAP still puts out wu's about once a month, so don't give up on them. (However, I recently learned the value of clearing the BOINC long term debt. You may want to look this up.)
I've stopped SZTAKI on my Mac until they release a new version. I'm still getting wu's that crunch for 80+ hours (on the G5) and barely get over 50%. Also, it seems the hard drive idling causes SZTAKI to report errors of loosing connection to the hard drive. The other projects (SIMAP, SETI, Einstein, and WCG) do not have this problem. These wu's complete, but since they report an error, count as invalid, are dropped from the research, and do not get credit. (Quick aside: As I've said there, it's not loosing the credits themselves that annoys me. The credits simply mean the work was good and accepted. I pay for electricity to run my computers and to run the air conditioning to keep them and me cool. My electric bill had doubled since I started BOINC'ing, but I justify it in that I have few hobbies that require much money. When I don't get credits, that means I'm *paying* a little over $1US per day extra for power to run the computers and air conditioner for bad work. I understand if it's a beta project, but if I loose over two or three days of crunch time on *one* wu, that could have been money better spent.) So, anyway, if you crunch for SZTAKI on a Mac, check your credits weekly and drop/suspend it if half or more wu's are invalid.
Actually, now that I think about it... the new SZTAKI wu's require SSE2. The G3 processor has SSE, but I don't think SSE2 existed for any chipset back then. I don't think SZTAKI will work at all on a G3.
WCG units are rather large and processor intensive. Since this is actually a collection 3/2/5 projects (three official, two in up-and-running states, five counting the two “secret” beta projects), it is hard to give good information on this. I can say the FightAIDS@home wu's (FAAH) tend to be 10-12 hours on a high end PPC G5. These would probably not be suitable for a G3. The Help Defeat Cancer wu's (HDC) are under half this size, and so should be crunchy enough for a G3. I have not run any Proteome projects (HPF) on the Mac. One of the “secret” beta projects, a wu labeled “fcg” took about 4 hours, which is just a little under what a HDC unit takes. If you sign up for WCG, you may want to only run HDC wu's. You can try FAAH, but those took about 18-22 hours on a G4 500MHz.
One last thing I have noticed. Since the Mac users are a smaller group, and generally work does not overlap between Mac / Windows clients (work assigned to PPC Mac's will be assigned only to that platform), the PPC work units tend to be larger. The PPC chip can process some types of calculations (SSE-type) much faster than Intel, and there are fewer wu's going around. So, the projects try to get more work per unit by making the units larger. I've noticed that work units are not only physically larger (sometimes significantly larger), but often take longer to crunch on a high-end G5 than an older AMD Athlon XP processor (the XP 3200+ was sometimes faster, and the G5 was sometimes not much faster than the XP 2200+). I've seen the same thing for Folding at Home work.
Gxis Rivedon!