MQL5 Cloud Network - PR drastically changed.
Hi, I think PR values is likely caped. Few years ago I was running 2500k@5ghz and I was hitting over 200PR scores aswell.
2500k@5ghz single core speeds are much more comparable with 10900k@3.5-4ghz speeds.
Most likely something happened while your i9-10900k machine was processing a job that dramatically slowed it down. Most jobs are easy, and the agents can buzz through a couple hundred in a minute. Some jobs are harder, taking several minutes to complete, but not using a ton of RAM (e.g. only 1 GB per agent) or hard CPU instructions (e.g. 100% CPU usage but only pulling 80w instead of 220w). However, every once in awhile, a job will come in that is very RAM hungry, will use hard CPU instructions, and take hours to complete. Oftentimes these jobs are batched to your computer, so you might have all ten cores running an instance of this monster of a robot. Next thing you know, your system is out of RAM and the CPU is thermal throttling—sometimes seeing thermals higher than stress testing can produce! I have seen per-agent RAM usage over 5 GB—which balloons to 50 GB when happening on 10 agents for 10 cores. As you approach full RAM utilization, Windows will start paging data to the disk, swapping older memory for data that is being accessed now. Combine thermal throttling with paging RAM to the disk, and your PC's processing rate will be slowed down to an absolute crawl! Imagine it doing this all night on a large project. Your real PR would be closer to 5-30 in those moments, but when averaged out, just resulted in a drop to 112. As the days since the "incident" go on, the MQL5 Cloud Computing system realizes that your computer is finishing tasks a lot sooner than expected, and your PR gradually comes back up to match.
Thank you Shalem!
It actually really great explanation! I was wondering about those ram intensive tasks could be at fault.
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
You agree to website policy and terms of use
I run more than a few devices on the cloud network to be sold, but in the last couple days one device on my list dropped in PR by half. From 225 to 112, an i9 10900k@5.1GHz all core. There's also a mobile i7 10800h on it that's getting 187 for PR however it did not change at all.
I'm trying to figure out exactly what might have happened, it occurred a couple days ago when they did the change in PR determination, showing an even averaged number from all agents on a single CPU, instead of an independent PR per agent on one.I've also noticed that the PR now changes once a day on established agents, while I don't mind that nor do I mind the averaging for each agent, in fact the averaging can benefit in ways to some people using their hardware while leasing, I'm just hoping this is a fluke with the new changes. Because if it's not and that's what we can expect the i9 10900k's to score around when other 10th gen CPU's with lower clocks are getting the same or better PR score, it just wouldn't make sense.
Just trying to raise some awareness to what happened in the event something wasn't applied as expected with the new PR determination changes. Normally I wouldn't bring it up but with all other devices both in the the same hardware tier and what not, none of the others had a cut like that, they just all have the even number averaged PR for what they scored around.
EDIT: So for the last day I've been watching and it seems like the score for my 10900k is slowly recovering, I assume this must have been a side effect of the balancing on how the PR for devices has been changed on the backend, I suppose at this rate it will be fixed around this time next week.