On my Windows 10 PC the application, using VirtualBox, seems to go faster than that of wingmen using Linux. I just started a Linux box with SuSE Leap 15.0 and joined the project. The Linux box is a HP Laptop with an AMD E-450 CPU of 2014 vintage. The Windows 10 PC has an intel i5 CPU and VirtualBox 6.1.10. I use it also on LHC@home projects which need VirtualBox even on Linux boxen, since LHC@home uses a Scientific Linux which is a RedHat plus some CERN Libraries.
Tullio

---

I normally crunch on Ubuntu 18.04 (now 20.04), but was going to try it out on Win7 64-bit. However, I got too many "VBox unmanageable" errors to give it a good test. I had to about about half of them. It was an i7-4771 with 32 GB memory, so it was not a memory problem. And I tried both VirtualBox 6.1.14 and 5.2.42; same results. Maybe someone else can give it a try.

When I get "VBox unmanageable" message I simply kill the BOINC manager and restart it.Usually on my i5 CPU with six threads I obtain two quantum chemistry tasks running and three "waiting for memory". My PC has a 12 GB system RAM and a 4 GB Video RAM. On another PC with an AMD Ryzen 5 1400 CPU I have 24 GB RAM, rarely used more than 50%, so I think 12 GB RAM is sufficient. I check its usage with GPU-Z when I use my GPU board, a GTX 1060 with 3 GB Video RAM on the AMD PC and a GTX 1650 with 4 GB Video RAM on the Intel PC. I never overclock anything and my fans are silent in both cases. I am running the Linux version on a laptop with an E-450 CPU and 8 GB RAM, with SuSE Linux Leap 15.0 but it is very slow.
Tullio
On the Windows 10 PC with 24 GB RAM I have created a Linux Virtual Machine using SuSE Linux Tumbleweed which is a development version with kernel 5.8.4-1 but it is updated so frequently that is is scarcely usable.

---

The fastest computer on this project seems to be the host ID : 2750.
AuthenticAMD
AMD EPYC Processor (with IBPB)
Ubuntu 20.04.1 LTS
4 processors
It won all the battles against its wingmen.
The windows hosts can go as fast as the linux hosts but they are less reliable because of the use of VirtualBox.(it needs to be handled , cautiously , and doesn't suit with big hosts and multiple VMs running simultaneously.)
Each cruncher has to do a good compromise between speed and reliability , in order to not waste his resources .
Sometimes , it's better to run a bit slower but surely , especially when it is NewChem long application (several days long).
Running with windows, you have to often check the wus to see if the behavior is correct.With Linux it's not necessary,the probability to fail is lower.

I have been using VirtualBox since 2008 in LHC@home on both Linux and Windows hosts. The only problem with Windows was that I had to disable the McAfee antivirus which I found installed on HP Desktops. But it works perfectly for me in this project.
Tullio

---

I would not be so enthusiastic about VirtualBox.This is a great software but i learned that its behavior depends mainly of its version.
I had to downgrade to 5.2.32 version to crunch this project with my computer.The next versions up to 6.1 didn't work for me.I don't know why.
I prefer to stay with its old version as far as i can, to avoid to test other settings.
@Tullio :
If you look carefully your works units , you will see that several ones have a cpu time equal to zero.
It means that there is no scientific output for these ones.This is not normal.
Even if you have a very fast computer , the cpu time should be greater than 0 second (there is 3 parts in the scientific calculation).
This project is very sensitive to the surrounding conditions (It is exploring the frontier between known and unknown science and there is no certitude about calculations and their result)..
To maximize the reliability of your jobs , i would advise you to crunch only one work unit with VirtualBox and assign other threads to non VirtualBox projects.Then you could see if these wus with zero time cpu decrease or disappear.
If it's not sufficient , then try another version of VirtualBox.
If it's too much difficult to set up ,try to crunch in a Linux VM inside VirtualBox.
Curiously , a linux app inside VirtualBox is more resilient and reliable than a windows app using VirtualBox for this project.
Unfortunately for me, I don't have the possibilty to do it because i have a lack of ram memory.
Each one has to manage with the tools at his disposition...
Good luck.

I have installed VirtualBox 6.1.14 and have completed several tasks, only one is unmanageable but I know how to resolve this problem by killing and restarting the BOINC manager. But all uploads are frozen. Why?
Tullio

---

Tried on 4 computers with different VirtualBox versions from 5.2.8, 5.2.44, 6.0.x and 6.1.x and the 5.2.x version it is the only one that is working for this project on Windows 10.
Native app(t1) on Linux it is working very well, it's about the same crunch time while running 7 tasks on windows (not enough memory) and 8 tasks on Linux using 1 GB of memory.

VBox 6.1.14 works well on my Windows 10 PC. Only if I run also Theory@home and Atlas@home from LHC@home there is not sufficient memory. Atlas@home uses 2 CPUs.
Tullio

---

I have completed 11 tasks on a Ryzen 5 1400 CPU running OpenSUSE Linux Tumbleweed, which is a development version with kernel 5.8.7-1. But since that Linux box, which is a Virtual machine on a Windows 10 PC, is enlisted in Science United all its credits will go to Science United, which is managed by David Anderson, the BOINC Director.
Tullio

---

The kernel of the Linux SuSE Tumbleweed is now 5.8.10-1, being updated very frequently, so I have to reboot each time It is now running nwchem, which was sent to it by Science United. It is the only system I have left in Science United.
Tullio

---

I am running nwchem on a SuSE Tumbleweed with kernel 5.10.7-1-default.
Tullio

---

The windows hosts can go as fast as the linux hosts but they are less reliable because of the use of VirtualBox.(it needs to be handled , cautiously , and doesn't suit with big hosts and multiple VMs running simultaneously.)
Each cruncher has to do a good compromise between speed and reliability , in order to not waste his resources .
Sometimes , it's better to run a bit slower but surely , especially when it is NewChem long application (several days long).
Running with windows, you have to often check the wus to see if the behavior is correct.With Linux it's not necessary,the probability to fail is lower.

I was about to post the same thing. I have run QuChemPedIA on Ryzen 3600 machines under Win10 (VBox 5.2.44) and Ubuntu 20.04.3.
It seems that the times are the same as nearly as I can determine. But you get hung work units under Windows, and not under Ubuntu.

I now have it on a Ryzen 3900X (Ubuntu 20.04.3) and will let it go for a while.

I am using VirtualBox on QuChemPedIA@home, with 1900 MB RAM usage, Atllas@homw on LHC@homw, 4899 MB RAM, rosetta python, 6144 RAM, and I don't jhave any problem on Atlas and rosetta python. Onlu on QuChem i get some unmanageable task and I exit BOINC and restart. I am running nwchem on a Linux Virtual Machine using OpenSuSE Tumbleweed, which is a development version with BOINC 7.18.1, kernel 5.15.12 so ifrequently updated that I have to reboot almost daily, with no problem. It als runs Einstein@home.
Tullio

---