Does that seem right? Using the nVidia GPU app on a GTX 1660 Ti.

---

Click here to see My Detailed BOINC Stats

That's nice , can I have one ;-)

They've been going since Dec 7, only on one host--but the only other tasks I've done recently are CPU on other hosts which all get 72 credit.

---

Click here to see My Detailed BOINC Stats

They've been going since Dec 7, only on one host--but the only other tasks I've done recently are CPU on other hosts which all get 72 credit.

The "tiny" CPU tasks are bundles of 9 blocks from distributed.net, those "huge" 6000+ credits WUs are bundles of usually 768-823 blocks. You can see that in the WU name, dnetc_r72_1672198181_13_770 is for example a bundle of 770 blocks. We get here always 8 credits/block, so yes, 6000+ credits for such WU is right.

EDIT: and since I see this, I have to ask, why are you Apple-users always aborting cuda tasks instead of just disabling them in your project preferences if you don't want to use your GPU in that host for whatever reason? I see this a lot and almost always that are Apple users who do that, I even started a thread about this issue a while ago.

---

Ok, just seems like a lot of credits for a task that completes in 10 minutes or less. Since that host has switched over to MilkyWay tasks, my daily numbers are only about 1/4 of what they were. I know the "standardization" of credit has been a long issue with the various projects.

I'll have to check out that host. Even though the tasks show "Aborted by user" the exit status was "201 (0x000000C9) EXIT_MISSING_COPROC" so I suspect there's something wrong with the interaction of OS/BOINC version, and the old GeForce 320M graphics in that particular MacBook Pro.

---

Click here to see My Detailed BOINC Stats

Ok, just seems like a lot of credits for a task that completes in 10 minutes or less. Since that host has switched over to MilkyWay tasks, my daily numbers are only about 1/4 of what they were. I know the "standardization" of credit has been a long issue with the various projects.

Most Nvidia consumer GPUs have very poor FP64 performance required for Milkyway, your GTX 1660 Ti has a DP:SP ratio of 1:32 resulting in just 169.9 GFLOPS FP64 while having 5.437 TFLOPS FP32/INT32 which is used here. My significantly older GTX 275 has a 1:8 ratio resulting in 84.24 GFLOPS FP64, which is half of what you have, while having only 673.9 GFLOPS FP32, so the difference in credits isn't that huge for me, I get here around 1.29x of what I get from Milkyway, GPUs with 1:5 and better ratios get probably more from Milkyway than from Moo!.

Even though the tasks show "Aborted by user" the exit status was "201 (0x000000C9) EXIT_MISSING_COPROC" so I suspect there's something wrong with the interaction of OS/BOINC version, and the old GeForce 320M graphics in that particular MacBook Pro.

That might explain why there are so many Apple hosts with that behavior.

---