>>Do you know why the CPU usage goes through the roof with two crossfired 5870's?

That's usually because the default core 0 is not optimal for your host. You should try to set it to 3 in project preferences (see http://moowrap.net/prefs.php?subset=project) and see what happens. Be careful not to use that core for your nVidia host too (you need to use different venues). For nVidia optimal core is usually 9, 10 or 11.

-w

1)single 5870 on Ubuntu_64: run time =1,766.99s, credit = 6,808.00
2)dual 5870 crossfired: run time =109.08s, credit = 102.00
What's up with that dual run time and credit??

Your dual-card host was getting tiny work units sent to it because scheduler was thinking it was slow. This can happen if there's not enough statistic history to determine it's speed from completion times.

At the moment it looks like that scheduler has learned it's actually a fast host so it's getting sent huge work units. Does things look good to you now?

Hmm, BTW, I can only see one ATI host here in your records. The other host has three nVidia cards. If you actually have two ATI hosts then it's possible that your one host record is used for both host, which can be problematic. Maybe you should try to reset Moo project on one of the hosts to hopefully fix this strangeness?

As a side note, there's two different ways scheduler scales the calculated host speed. It either uses completion times from recently completed tasks or uses a PFC scale. This latter one seems to give incorrect answers most of the time and I might have to tweak it if it constantly keeps giving people problems.

-w

There are different size work units available, the 102 point ones only have 12 stat units in them, the 6,800 point ones have about 768 stat units in them.

So at 8.5 points per stat unit 102 points is correct for a 12 stat unit work unit.

Conan