# Intel sapphire-rapids processors

**URL:** <https://pyfr.discourse.group/t/intel-sapphire-rapids-processors/1052>\
**Category:** General\
**Tags:** openmp\
**Created:** [8 March 2024 15:20 UTC](https://pyfr.discourse.group/t/intel-sapphire-rapids-processors/1052 "2024-03-08T15:20:27Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![subutai](https://avatars.discourse-cdn.com/v4/letter/s/dbc845/32.png) [@subutai](https://pyfr.discourse.group/u/subutai)\
**Post date:** [8 March 2024 15:20 UTC](https://pyfr.discourse.group/t/intel-sapphire-rapids-processors/1052/1 "2024-03-08T15:20:27Z")

</div>

Dear PyFR community,

Has anyone had experience with the brand new intel sapphire rapids processors (a.k.a. Intel(R) Xeon Platinum 8468) using the oneapi compilers (compiler command: icx)? Especially using the new PyFR version v2.0.0, which comes with cache-blocking speed-up? I want to explore this area since i have a free access to this new cluster for some time. Alas, i had no prior experience with the openmp backend. In fact, I have always used the cuda backend.

Would the icx compiler work faster than gcc? should i compile the libxsmm using icx also?

I would really appreciate any comments on this.  
Regards  
Cengiz

---

<div class="post-metadata">

**Author:** ![fdw](https://avatars.discourse-cdn.com/v4/letter/f/a587f6/32.png) [@fdw](https://pyfr.discourse.group/u/fdw)\
**Post date:** [8 March 2024 16:33 UTC](https://pyfr.discourse.group/t/intel-sapphire-rapids-processors/1052/2 "2024-03-08T16:33:01Z")

</div>

Performance should be reasonable. Although note that the memory bandwidth is still well below that of recent GPUs.

It does not matter what compiler you use for libxsmm. You may want to try different compilers for PyFR, however, and see what works best.

Regards, Freddie.