I would like to ask for WU's for Intel GPU's. Not as plentiful as the big guys but Dell makes a lot of Intel based systems that contain GPU's that crunch other projects quite nicely.

Thanks
Bill F

---

In October 1969 I took an oath to support and defend the Constitution of the United States against all enemies, foreign and domestic;
There was no expiration date.

If this project had people who knew how to adapt it to run on intel GPU's, Im sure they would love for them to do so. Unfortunately, as far as I am aware there are not a lot of people currently on the admin team for this project, and they are already very busy. In short, I would love to be able to run this on an intel GPU or an AMD GPU or any other architecture that ever existed, but the limiting factor is the amount of time available to the developers.

That being said, it WOULD be really cool to see an intel GPU or an AMD GPU capable application.

I also would like to recommend adding Intel GPUs...with the addition of separate Intel video cards, the amount of Intel GPUs available for crunching is on the rise.

the Asteroids code makes heavy use of FP64 double precision and Intel GPUs lack hardware FP64 support.

So it likely won’t happen until Intel starts making GPUs with FP64.

---

Slight addendum.

Intel iGPUs only support FP64 on 6th/7th/8th/9th gen. Devices before or after this won’t work. And not until they get hardware FP64 in future generations of their dedicated GPUs (if they do), current gen Arc cards still lack FP64 at all.

https://community.intel.com/t5/Graphics/What-processors-have-an-internal-Gen11-GPU-with-FP64-double/m-p/691214

There also exists a driver workaround for Linux systems to emulate FP64 on their newer Linux driver for the newer devices that don’t natively support FP64, but the same thing is not supported in Windows, and it’s painfully slow anyway.

---

In Linux the support of Intel is improving at a fairly good rate..... Some reading: https://www.phoronix.com/news/Intel-Xe-Submission-Linux-6.8 https://www.phoronix.com/news/Intel-CR-23.30.26918.9

Lack of hardware FP64 is the biggest issue with Asteroids. FP64 is required here.

---

I have seen a reference to FP64 in the upcoming Xe driver for Linux---I'll try to find the information & post it here.

I have found these references:
1. https://github.com/intel/compute-runtime/blob/master/opencl/doc/FAQ.md#feature-double-precision-emulation-fp64
2. https://community.intel.com/t5/GPU-Compute-Software/Iris-Xe-Graphics-can-t-compute-double-precision-floating-point/td-p/1281976
3. https://foldingforum.org/viewtopic.php?t=38527 (as this is an older posting, I expect improvements on this in the last year or so...)

As the Intel compute-runtime information notes: This feature is only available in Linux.

I have seen other references to FP64 in developer notes but have not been able to pin those references down as of yet....I will continue to look. I'm the main poster for Intel Arc & Xe driver information on the EndeavourOS forum.

I’m not sure the value in reposting information that I’ve already linked to and mentioned.

Intel GPUs at this time largely don’t support FP64. And the workarounds that do exist are very slow and only work in select few environments.

This project relies heavily on FP64 calculations. Intel GPU is not viable at this time. That could change if future generations of GPUs have hardware FP64 support, but for right now it’s basically a non-starter.

---

This project relies heavily on FP64 calculations. Intel GPU is not viable at this time. That could change if future generations of GPUs have hardware FP64 support, but for right now it’s basically a non-starter.

Intel didnt support FP64 ?!?!

That's not entirely true, the Intel ARC series can only calculate FP64 in emulated form but they can do it.

I don't know how the value on Wikipedia came about, but if it is correct, 4-4.9 TerraFlops would be possible with FP64 emulation.

This project relies heavily on FP64 calculations. Intel GPU is not viable at this time. That could change if future generations of GPUs have hardware FP64 support, but for right now it’s basically a non-starter.

Intel didnt support FP64 ?!?!

That's not entirely true, the Intel ARC series can only calculate FP64 in emulated form but they can do it.

I don't know how the value on Wikipedia came about, but if it is correct, 4-4.9 TerraFlops would be possible with FP64 emulation.

no that's not correct at all. any info you find about FP64 TFlops on ARC GPUs was pre-launch speculation that never got corrected. you will never get 4-5 TFlop from emulated FP64. that's faster than a Radeon VII which has strong hardware FP64 performance.

the ARC Alchemist cards do not support FP64 in hardware at all. the emulation only works for Linux in terms of GPGPU compute (and it's very slow), and where it "works" in Windows is to handle the very very few instances in some games that need it (but the slow performance doesn't hinder the overall processing of the game).

sorry, it's not going to happen until Intel includes hardware FP64. maybe they will in the next generation of ARC with Battlemage

---

I appreciate the technical clarity Ian&Steve C. provided regarding the FP64 hardware requirements. It is clear that until Intel integrates robust hardware-level FP64 support, full-scale integration is not a viable path for the current application.

However, as we look toward the upcoming Intel Battlemage and potentially Celestial architectures, the landscape of integrated and discrete compute may shift. Would it be worth the project's time to establish a 'community testing' baseline for these future architectures once hardware-level FP64 becomes available?

Having a clear roadmap for when and how the project might integrate support for new high-performance architectures—as they eventually meet these requirements—could help prevent these threads from looping on the same arguments. It’s not about pushing for current emulated support, but rather about preparing the project for when the hardware finally catches up to the scientific requirements.

> Would it be worth the project's time to establish a 'community testing' baseline for these future architectures once hardware-level FP64 becomes available?
Actually, Intel brought back FP64 support with Battlemage 1,5 year ago, so there's a high chance a B580 might work.

We have an experimental Intel build, but it was created before Battlemage launched. It was only tested on iGPUs, and the few that supported FP64 were too slow to make it worthwhile.

As we’ve said many times before, the app won’t work on Alchemist GPUs, including iGPUs based on that architecture. They don't have the hw support.