Today I have noticed that my Mag in Amicable Numbers project went unexpectedly up by ~20% within a day or two. I was even more surprised about RAC – up by ~50%. I had to start digging up. Amicable Numbers project rewards 6,836.19 credits for each work unit (WU). On my GTX 1060 card it usually takes 1500s to 2000s to process one WU. I’ve never gave it much of a thought. When I checked today, some recent WU were processed in just ~500s, a few in under 200s! But they are still rewarded 6,836.19 credits. Did my card went on some kind of a turbo? Well, no. Processing times for those WU were much shorter. But why?
 | Amicable Numbers |
|---|
Let’s see what Sergei Chernykh, project admin says:
We've had a bunch of very GPU-friendly (highly parallelizable) work units recently, but as far as I can tell, we're back to regular work units again. All work units were calibrated to have approximately the same amount of computations for CPU version.
Run times on GPU are usually stable too, but sometimes they can vary a lot.
As we see, processing times are not easily controllable. There seems to be a problem, though – why they are calibrated for CPUs first, if project is GPU oriented?
Sergei is very helpful and I have received the following answer:
The speed increase for GPU over CPU depends on how well a WU can be parallelized, but it's hard to predict without actually running each WU. It's even different between various GPU models. CPUs are more stable because they don't run that many parallel threads. That's why WUs were normalized for CPUs.
Running each WU for calibration only doesn’t make sense, and it seems that perfect solution doesn’t exist. It doesn’t pose much of a problem, though. Everybody has the same chances and scores equalize over time.
If you have noticed, my Mag grew up less than RAC, but it did, anyway. I think this is due to inactive users. If someone didn’t crunch the project when shorter WU were available, his RAC didn’t receive a boost and his magnitude dived a bit.
| Gridcoin Research |  |
|---|