Folding Forum

viewtopic.php?f=61&t=32871

I made a topic about how servers balance WU's between different hardware, but the first few entries are now about someone having multiple GPU's.

Yes the topics were merged. If you look at Bruce's last post in the thread, he indicated such.

"They merged my topic to make help manageable! What can I do? Make even more topics!"

jonault wrote:Yes the topics were merged. If you look at Bruce's last post in the thread, he indicated such.

I can't make much sense of this.

My topic is about if there is any balancing going on, depending on what hardware is requesting work.

That is merged with a topic related to the current situation where servers are overloaded.

If you are refering to this viewtopic.php?p=315478#p315478 reply, that reply is aimed to the overloaded state.

I am very well aware that there is a lot higher load on the servers than usual, there are posts all over the forum about that, I didn't make yet another post about that. My post is inspired by it, because the current situation gives us some new results we haven't seen before, but I didn't ask "why am I not getting any", it is the other way around. I can see one computer getting WU's, and another not. They have very different hardware configurations, and what I am interested in learning about is if there are WU's which need a minimum of threads or hardware architecture.

It is people who are given the answer and refuse to accept that answer that make me happy I am allowed to push the ignore button.

JimboPalmer wrote:"They merged my topic to make help manageable! What can I do? Make even more topics!"

Yes, two unrelated topics got merged, that is why I ask. One is about why only one gpu got used, I am asking about how WU's gets distributed to different hardware types and architectures.

JimboPalmer wrote:It is people who are given the answer and refuse to accept that answer that make me happy I am allowed to push the ignore button.

Feel absolutely free to. I just don't learn much when I accept an answer to something else than what I actually asked.

I didn't make the topic because people HAD to reply. I made it so if someone saw it and knew about how it worked, they could reply. If the topic sat idle for months because no one knew about it, then that would be perfectly fine. But what I can't use are guesses.

But enough about that, your reply in this topic was not related to your reply in the other topic, it was to why it got merged with a thread about something very different.

To really break the question down, I was interested in knowing if some WU's could only get processed on newest hardware architectures, if some needed more threads than others, if some needed instructions sets who could only be found on some cpu's. But "servers are overloaded" are what it ended up as, and apparently that should be enough answer to that question.

TheMikeyDK wrote:To really break the question down, I was interested in knowing if some WU's could only get processed on newest hardware architectures, if some needed more threads than others, if some needed instructions sets who could only be found on some cpu's. But "servers are overloaded" are what it ended up as, and apparently that should be enough answer to that question.

No. WUs run on "cores" (executables that run the science part of FAH). The cores are generic enough to be able to run on a wide range of systems. Most, if not all, are based on the GROMACS "engine". In some cases, GROMACS has optimized code paths to use a special instruction set, if available, over generic instructions. If the processor does not support it, it simply uses the generic processing steps.