Hi all,

I am noticing on my machine that tasks in my Linux OS consistently take a little over 4000 seconds, but when i am in Windows, they take just under 3000 seconds.

Are other people seeing the same thing?

I remember that at Universe@home it was beneficial to run Linux, so could be that is is just faster to run tasks here in Windows..

CPU: Ryzen 7 3800XT, i think my Computers are visible on this website, so more specs can be seen in profile if interested.

Kind Regards,
YoToP

I looked at my results and I am around 3000 for time. Windows 10 with a Ryzen 1700, 8 cores since I have SMT off. I don't know if I process a little bit faster per work unit but it is a bit funny matching your Windows computer with a more current Ryzen chip, 3800XT

I would guess the time difference between Windows and Linux on your computer would be down to Dennis application, the compilers used and if they use a more efficient library with one of them, Windows?

Are you running just DENNIS on both computer? Do you run any other projects at the same time?

Maybe someone with the project will look at the data per processor, Linux vs Windows. You may have caught something that can be improved

"I would guess the time difference between Windows and Linux on your computer would be down to Dennis application, the compilers used and if they use a more efficient library with one of them, Windows?"
That would be my guess too.

Yeah, i am just running DENIS, So there might be room for improvement on Linux tasks.

Some projects have a huge memory difference dependent upon CPU and OS. Speed is very difficult to collate as not many project have constant length tasks also credits awarded don't follow length of task across different CPU architectures. We are more interested in credit rate rather than outright speed.

I couldn't get very returns out of AMD's on Linux so reluctantly put them on Windows. I haven't tried Windows on my Intel's, the return and speed looks roughly what I'd expect. Only a few projects appear to take advantage of the blistering integer speed of ARM's (possibly using NEON?).

The New human ventricular cell model application on Linux was as fast as Windows, but the latest 'optimised' version is at least 20% slower.

Hi,
What do you consider the 'optimized' app?

Best,
Jesús

---

Jesús Carro
Universidad San Jorge
@InSilicoHeart

Human ventricular cell models optimization

Original poster YoToP's computers were a good example since two machines are identical, except for os, windows & linux
Unfortunately, the linux machine hasn't had contact since March 7th.


https://denis.usj.es/denisathome/hosts_user.php?userid=136213

The one you have labelled as optimized....

0.02 Human ventricular cell models optimization HuVeMOp

The one you have labelled as optimized....

Optimization of the scientific simulation, not optimization of the app....

NHuVe and HuVeMOp are the same application (same sources), we have two different programs to separate the simulations. The only difference is that in the NHuVe we used our model, and in the HuVeMOp we are running O'Hara's model, but the compiled app is the same. Both can run both models, we have two separated apps in DENIS for organization.

Best,
Jesús.

---

Jesús Carro
Universidad San Jorge
@InSilicoHeart