Yes over the weekend I will review the results of the 2 experiments that just finished and start new ones. The idea is to continue like until now. With two populations of 1000 agents (task) each.

---

And thanks from me, too. That went very smoothly, and allocation of v4.03 hasn't been disturbed. Another resend has arrived for processing when this one finishes, without manual intervention.

==> WARNING: A newer version of conda exists. <==
  current version: 4.8.3
  latest version: 23.1.0

Please update conda by running

    $ conda update -n base -c defaults conda

Does anyone know if I need to install Miniconda and/or Anaconda to satisfy this error message?
E.g.: https://conda.io/projects/conda/en/latest/user-guide/install/linux.html
My Linux Mint Synaptic Package Manager can't find any program containing "conda."
Maybe this is just something for the server-side staff but then why post an error message to confuse crunchers?

Maybe this is just something for the server-side staff but then why post an error message to confuse crunchers?

It's not an error, it's simply a warning - information, if you like.

The project supply the conda package (which is why Mint doesn't know about it), and they're obviously happy with the version they're using. You don't need to do anything.

==> WARNING: A newer version of conda exists. <==
  current version: 4.8.3
  latest version: 23.1.0

Please update conda by running

    $ conda update -n base -c defaults conda

Does anyone know if I need to install Miniconda and/or Anaconda to satisfy this error message?
E.g.: https://conda.io/projects/conda/en/latest/user-guide/install/linux.html
My Linux Mint Synaptic Package Manager can't find any program containing "conda."
Maybe this is just something for the server-side staff but then why post an error message to confuse crunchers?

even if you installed it, it wouldnt change anything and you'd get the same warning message. as Richard wrote, these tasks use its own environment. they do not use your locally installed conda at all. which is why they work on systems that do not have conda installed at all. this is all by design to avoid any version conflicts or dependencies on the local system. it has been this way from the beginning.

additionally, this message was only present when trying to run the old/incompatible 4.01 app. you do not get that message from the correct 4.03 app. 4.01 was re-published by accident and is an app version about 1.5 years old. it is not compatible with the design/structure/requirements of how these tasks function today. the project admins have removed this version so you wont see this problem again.

---

what catches my eye:

the Pythons which I got downloaded within the past 2 days seem to use a lot less system memory than the ones before.
Has Abouh made any changes to this effect?

❸ Name Abouh Meaning
Exceptional qualities that make this name special are negotiation skills and an amazing sense of tact. When developed these two assets will help you get what you want and achieve all goals.

Cooperation is a key aspect of your life!

Since you are far less successful in life if you do not find a level of unity with others. Most problems that people find too tricky to solve are often no match against your ingenuity.

Harmony in your surrounding is a key to happiness and feelings of relaxation. Having friends or family fight affects you greatly in a negative sense. That is why you have the reputation of being a peacemaker(only out of necessity).

At your best you become very kind hearted, charming and full of positive energy. An amazing person to spend time with.

hi all,
...i used to contribute to GPUgrid with 1 gtx1080ti since 2020, 4core nonHT cpu
since the mostly GPU based acemd tasks i see a lot has changed regarding the effective HW requirements,
so i m wondering what is a somewhat optimal HW setup with the current python tasks, or what is potentialy the biggest current bottleneck?
-CPU core/thread #?
-RAM size?
-RAM speed vs latency preference?
-SSD speed
...i concur ECC RAM is very needed for long runtimes/nonstop usage.
___________
Frank from Slovakia, EU

For Python, you don't need to worry about anything other than having 8-10 cpu cores to support the task.

Enough system memory, at least 16GB.

Enough virtual swap space, at least 50GB.

Enough VRAM on the gpu, at least 4-6GB.

feri, If I might add to Keith Myers' excellent synopsis, the speed at which these tasks run appears more dependent upon CPU ability than GPU ability. You might want to consider that if you are thinking about assembling a host dedicated to running pythons and you maybe have an old GTX 1060 6GB or something else with sufficient VRAM (GTX1650) laying around.

..a friend of mine actualy has a gtx1060 6GB laying around
thanks for the insights

Look at Richard Haselgrove's results with 6GB GTX 1060's

Look at Richard Haselgrove's results with 6GB GTX 1060's

They're GTX 1660s, but 6 GB is right. They run fine on a setting of 3 CPUs + 1 GPU - a bit over 8 hours for the current jobs.

I've noticed that on the same computer (with dual boot), tasks finish almost twice as fast on Ubuntu compared to Windows. I've tried running tasks on Linux only a few days ago and did so on Windows before.
Has there been any recent change or do tasks just compute more efficiently on Linux?

They have always been faster on Linux

---

... tasks finish almost twice as fast on Ubuntu compared to Windows. I've tried running tasks on Linux only a few days ago and did so on Windows before.

They have always been faster on Linux

that's correct. What surprises me though is that tasks finish almost twice as fast. I don't think that this was true before, was it?

Close to 10 hours are needed on Windows and almost six on Linux. I also find the difference striking, that's why I asked

Anyone having problems getting the ATM tasks to upload? I have 4 completed jobs on 3 machines trying to upload and have not been able to make contact for nearly a day now. Two tasks on one machine making that device unable to get any more work.

Got several ATM stuck in upload there is now 2 days left to deadline.

I've been watching the ATMs on the linux hosts (since they won't run on my windoze machines) to find a siderr file of a finished WU to study the linux (while I try to learn it).
I haven't found one. Only 'in process'; most show previous failures which vary from host to host. I'd be interested to see one completed if anybody can post a link.
Thanks.