gpu

After a long journey getting AMD graphics cards working on the Raspberry Pi 5, we finally have a stable patch for the amdgpu Linux kernel driver, and it works on AMD RX 400, 500, 6000, and (current-generation) 7000-series GPUs.

With that, we also have stable Vulkan graphics and compute API support.

When I wrote about getting a Radeon Pro W7700 running on the Pi, I also mentioned AMD is not planning on supporting Arm with their ROCm GPU acceleration framework. At least not anytime soon.

Luckily, the Vulkan SDK can be used in its place, and in some cases even outperforms ROCm—especially on consumer cards where ROCm isn't even supported on x86!

After years of work among a bunch of people in the Pi community (special callout to Coreforge!), we finally have multiple generations of AMD graphics cards working on the Raspberry Pi 5.

We recently got Polaris-era GPUs working (like the RX460), but in the past month we've gotten 6000 and 7000-series GPUs up and running. And many parts of the driver work at full performance—well, as much as can be had on the Raspberry Pi's single PCIe Gen 3 lane (8 GT/sec)!

I've been testing tons of modern AAA games, like Doom Eternal and Crysis Remastered, and can get 10-15 fps at 4K with Ray Tracing on, or 15-20 fps at 4K. Dropping down to 1080p is not enough to overcome the Pi's CPU bottleneck—only at resolutions under 720p does the Pi's CPU and the single PCIe lane not seem to get in the way quite as much.

After I saw Pineboards 4K Pi 5 external GPU gaming demo at Maker Faire Hanover, I decided it was time to set up my GPU test rig and see how the Pi OS amdgpu Linux kernel patch is going.

I tested it out on a livestream over the weekend, but I thought I'd document the current state of the patch, how to apply it, and what else is left to do to get full external GPU support on the Raspberry Pi.

I also have a full video up with more demonstrations of the GPU in use, you can watch it below:

My journey testing various graphics cards on the Raspberry Pi began soon after the Compute Module 4 was launched in 2020. Since then I've tested almost 20 graphics cards—with a little success.

But there were two roadblocks to getting drivers for even older AMD radeon drivers working well:

I'm working on a project—a very dumb project, mind you—and I was trying to acquire the two current-gen flagship GPUs: an Nvidia RTX 4090, and an AMD Radeon RX 7900 XTX.

In some weird stroke of luck (it has been difficult to find either in stock), I was able to get one of each this week.

(lol at the size difference...)

Besides exorbitant price gouging, Nvidia's ownership of the crown in terms of GPU performance remains in this generation, as the 4090 blows past any competing card so far. But AMD's 7900 XTX was poised to be the best value in terms of price, performance, and efficiency (at least compared to any Nvidia offering).

In October 2020, after Raspberry Pi introduced the Compute Module 4, I started out on a journey to get an external graphics card working on the Pi.

At the time, it'd been over a decade since the last time I'd built a PC, and I had a lot to learn about PCI Express, the state of graphics card drivers in Linux, and PCI Express support on various ARM SoCs.

Last year, due to some extreme luck and help from a viewer, I picked up an AMD RX 6700 XT for MSRP (around $500).

My initial goal was to see if I could get the card working on a Raspberry Pi. Though my initial effort was fruitless, I've since hacked the driver to work through at least a few 'rings' of AMD's doorbell init process.

Earlier today I did a livestream on my YouTube channel to attempt using an Nvidia GeForce GTX 1080 on a Raspberry Pi Compute Module 4.

As with all my testing, I'm documenting everything I learn in this GitHub issue, which is part of the Raspberry Pi PCI Express Card Database website.

It's only been a few hours, but I've already gotten good suggestions for better debugging than I was able to do on the stream. And someone pointed out it might be the case, due to 32-bit memory limitations on the BCM2711's PCIe bus, that no GPU with more than 4 GB of onboard RAM could work. Though it's hard to confirm there'd be no software workaround—even 1 and 2 GB graphics cards (AMD and Nvidia) are crashing the kernel in similar ways.

The full livestream is available on replay and is embedded below:

Last year I tested two older graphics cards—a Radeon 5450 and a GeForce GT710—on a Raspberry Pi Compute Module 4.

This year, I've been testing three more graphics cards—a GeForce GTX 750 Ti, a Radeon RX 550, and the diminutive ASRock Rack M2_VGA.

The Compute Module 4, if you didn't know already, exposes the BCM2711's single PCI express lane, and the official IO Board has a nice, standard, 1x PCIe slot into which you can plug any PCI express device.

The Raspberry Pi Compute Module 4 eschews a built-in USB 3.0 controller and exposes a 1x PCI Express lane.

The slightly older Raspberry Pi 4 model B could be hacked to get access to the PCIe lane (sacrificing the VL805 USB 3.0 controller chip in the process), but it was a bit of a delicate operation and only a few daring souls tried it.

Watch this video for more detail about my experience using these GPUs on the CM4:
GPUs on a Raspberry Pi Compute Module 4!