# LFortran runs POT3D on H200 faster than NVFORTRAN **URL:** \ **Category:** LFortran\ **Created:** [November 12, 2026, 8:03am UTC](https://fortran-lang.discourse.group/t/lfortran-runs-pot3d-on-h200-faster-than-nvfortran/11294 "2026-11-12T08:03:14Z")\ **Posts on this page:** 6\ **Page:** 1
**Author:** ![certik](https://yyz2.discourse-cdn.com/free1/user_avatar/fortran-lang.discourse.group/certik/32/4_2.png) [@certik](https://fortran-lang.discourse.group/u/certik)\ **Post date:** [November 12, 2026, 8:03am UTC](https://fortran-lang.discourse.group/t/lfortran-runs-pot3d-on-h200-faster-than-nvfortran/11294/1 "2026-11-12T08:03:14Z")
I just merged the last PR into main, so LFortran now runs the @sumseq’s POT3D code with `do concurrent` (no other directives or pragmas) on NVIDIA GPU and is faster than NVFORTRAN (both with `do concurrent` and also with OpenMP target pragmas), all details are at: [Tracking: POT3D on CPU and NVIDIA, AMD and Apple GPUs · Issue #17015 · lfortran/lfortran · GitHub](https://github.com/lfortran/lfortran/issues/17015#issuecomment-6817846344). Here is a little table with a few tests from POT3D: | test | nvfortran `main`: wall / Total / POTFLD | LFortran: wall / Total / POTFLD | | --- | --- | --- | | potential\_field\_source\_surface | 5.56 / 1.768 / 1.659 | **4.93 / 1.559 / 1.417** | | potential\_field\_current\_sheet | 6.61 / 3.096 / 2.902 | **6.39 / 2.795 / 2.555** | | open\_field | 7.38 / 3.840 / 3.588 | 7.38 / **3.492 / 3.211** | It also runs the same code on AMD GPUs and also Apple GPU (but you need to use a custom single precision version of it, since Apple GPU doesn’t support 64-bit floating point). This is obviously very fresh, so LFortran’s performance likely doesn’t translate to all other codes yet, but I would say it’s not a bad start. I am sure that NVIDIA is now hard at work to speedup their Fortran. It took a while, but I think I am finally ready to compete. 🙂 ---
**Author:** ![FortranFan](https://avatars.discourse-cdn.com/v4/letter/f/96bed5/32.png) [@FortranFan](https://fortran-lang.discourse.group/u/FortranFan)\ **Post date:** [November 12, 2026, 11:31am UTC](https://fortran-lang.discourse.group/t/lfortran-runs-pot3d-on-h200-faster-than-nvfortran/11294/2 "2026-11-12T11:31:37Z")
> [@certik](#): > > ..LFortran now runs the @sumseq’s POT3D code with `do concurrent` (no other directives or pragmas) on NVIDIA GPU and is faster than NVFORTRAN .. Brilliant, absolutely brilliant! My “ai assistant”, when queried about codes the LFortran team can immediately try and likely achieve rapid success and complete the march towards their own “official” BETA status, listed the following: 1. [OpenIFS](https://www.google.com/url?sa=i&source=web&rct=j&url=https://github.com/ecmwf-ifs/openifs&ved=2ahUKEwi75ef-qIKYAxXllokEHb-EOwQQy_kOegoIAggACAEIFBAF&opi=89978449&cd&psig=AOvVaw3VZxjFxcdT4Qocu3k7kFqX&ust=1794569099367000) from ECMWF 2. [NWChem](https://www.google.com/url?sa=i&source=web&rct=j&url=https://github.com/nwchemgit/nwchem&ved=2ahUKEwi75ef-qIKYAxXllokEHb-EOwQQy_kOegoIAggACAEIFBAH&opi=89978449&cd&psig=AOvVaw3VZxjFxcdT4Qocu3k7kFqX&ust=1794569099367000) 3. [Quantum ESPRESSO](https://www.google.com/url?sa=i&source=web&rct=j&url=https://gitlab.com/QEF/q-e&ved=2ahUKEwi75ef-qIKYAxXllokEHb-EOwQQy_kOegoIAggACAEIFBAK&opi=89978449&cd&psig=AOvVaw3VZxjFxcdT4Qocu3k7kFqX&ust=1794569099367000) ---
**Author:** ![RJaBi](https://yyz2.discourse-cdn.com/free1/user_avatar/fortran-lang.discourse.group/rjabi/32/5161_2.png) [@RJaBi](https://fortran-lang.discourse.group/u/RJaBi)\ **Post date:** [November 13, 2026, 1:54am UTC](https://fortran-lang.discourse.group/t/lfortran-runs-pot3d-on-h200-faster-than-nvfortran/11294/3 "2026-11-13T01:54:18Z")
Is there any thoughts towards Lfortran and Intel GPUs? ---
**Author:** ![certik](https://yyz2.discourse-cdn.com/free1/user_avatar/fortran-lang.discourse.group/certik/32/4_2.png) [@certik](https://fortran-lang.discourse.group/u/certik)\ **Post date:** [November 13, 2026, 2:19am UTC](https://fortran-lang.discourse.group/t/lfortran-runs-pot3d-on-h200-faster-than-nvfortran/11294/4 "2026-11-13T02:19:54Z")
Planned. If I add a backend, will you use it? ---
**Author:** ![sumseq](https://yyz2.discourse-cdn.com/free1/user_avatar/fortran-lang.discourse.group/sumseq/32/3083_2.png) [@sumseq](https://fortran-lang.discourse.group/u/sumseq)\ **Post date:** [November 13, 2026, 6:12am UTC](https://fortran-lang.discourse.group/t/lfortran-runs-pot3d-on-h200-faster-than-nvfortran/11294/5 "2026-11-13T06:12:52Z")
As soon as I start using LFortran, I would use the Intel GPU backend - we run one of our codes (HipFT) on a B580 GPU now using IFX. ---
**Author:** ![certik](https://yyz2.discourse-cdn.com/free1/user_avatar/fortran-lang.discourse.group/certik/32/4_2.png) [@certik](https://fortran-lang.discourse.group/u/certik)\ **Post date:** [November 13, 2026, 8:36am UTC](https://fortran-lang.discourse.group/t/lfortran-runs-pot3d-on-h200-faster-than-nvfortran/11294/6 "2026-11-13T08:36:23Z")
Ok, I’ll get HipFT working and also add the Intel backend.