Radeon RX 480 Linux Testing Is Happening Right Now

[Michael]

About the tests, I'd be interesting in noise level too, but I've read that you don't have the equipment for it Michael (why are there 10 "Michael"s in the completion list??). Also, I guess you would need a silent room to do your tests. Testing the noise level in the middle of a room full of servers running at their maximum might not be the perfect conditions

How much money would you need to test the noise level in right conditions? That could be some kind of stretch goal (like in Patreon).

I have no idea what a good USB-based sound meter costs... Maybe someone more into audio can comment on the price of a good one that works under Linux. Or even if a really good USB microphone could be adapted for those purposes, etc.

[Dr. Righteous]

I'm really happy to see these Polaris based cards coming out and at a very good price point. This is why I've stuck with AMD.
When ever I feeling that "I've had it" with wrestling with AMD drivers, etc trying to get good performance in linux; I just browse the price of Nvidia cards and shut my mouth.
AMD Cpus my not be the fastest; but for performance per dollar you can't bet them. Same with the GPUs, not the fastest, but one heck of a deal for the money.
For those who must remain budget minded when it comes to computer upgrades; AMD has my back...........

[artivision]

Simple facts:
1) Amd has the greatest open source code. Not only the Linux driver but open FX for developers to, regardless of the OS.
2) Amd on Linux has native support (2x speed) for D3D9 and eventually for 11 to, somewhere after the completion of WineD3D11.
3) Amd can have strong 4-6TFlops Apus with 16GB Hbm2. If they can do it like a small Gpu without motherboard and for Laptops to, you can imagine.

[fld-]

I wonder if the RX 480 will work in a Qemu/KVM Windows 7 VM via vfio

[Oguz286]

I wonder if the RX 480 will work in a Qemu/KVM Windows 7 VM via vfio

I'm pretty sure it will. I'm using a 7750 with Qemu/KVM with a windows 7 vm and it works fine. I do get the occasional lockup but I suppose that's because of KVM.

Unlike nvidia cards, unless you buy their grossly expensive 'professional' cards

[bridgman]

Some irritating remarks here. There won't be a proprietary driver anymore, it's mainlined AMDGPU in the Linux Kernel and the optional proprietary AMDGPU-PRO in userland. Don't name it Catalyst.

Just a quick note - the amdgpu hybrid/pro driver currently requires different kernel code from what is upstream, for the simple reason that getting kernel code upstream requires corresponding open source userspace code. In cases where the only userspace code requiring certain functionality is closed source, we can't get the kernel code upstream yet.

At the moment I believe OpenCL and Vulkan require some out-of-tree kernel code, while closed source OpenGL does not.

[habilain]

I have no idea what a good USB-based sound meter costs... Maybe someone more into audio can comment on the price of a good one that works under Linux. Or even if a really good USB microphone could be adapted for those purposes, etc.

In this circumstance, the difference between a sound meter and a microphone is tiny. A sound meter is a microphone attached to something which measures the max amplitude of what it listens to. Given that you're attaching the microphone to a computer, it'd be trivial to do the sound meter's job in software and just read off the maximum amplitude from the recording. This is basically how a phone-app based sound meter would work, but they're not so accurate because the microphone in a phone isn't really designed for anything above or below the volume of a normal conversation - a slightly higher quality microphone would fix this easily. You probably don't need one which is too high quality though, as you're only caring about amplitude rather than what things sound like.

I'd say the bigger issues are a) Ambient noise from other machines in your test area and b) Proper positioning of the microphone from the noise source (which may be tricky to automate!).

[curaga]

I can recommend the Audio-Technica ATR2500-USB mic. Great reviews and works well in Linux.

Given that you're attaching the microphone to a computer, it'd be trivial to do the sound meter's job in software and just read off the maximum amplitude from the recording.

How do you convert it to decibels accurately? It's just a relative reading. You can tell it's 3x higher than background noise, but without an absolute db scale it's hard to tell how bad it is.

[habilain]

I can recommend the Audio-Technica ATR2500-USB mic. Great reviews and works well in Linux. How do you convert it to decibels accurately? It's just a relative reading. You can tell it's 3x higher than background noise, but without an absolute db scale it's hard to tell how bad it is.

Normally you would calibrate on some known noise sources at a fixed distance, allowing you to know the actual noise at the microphone, and then plot a calibration graph that maps readings to sound level in decibels. It should be fairly easy to find such sources as there's quite a lot of hardware which will tell you how noisy it is. Alternatively, you can create your own if you have a speaker which you know the sound power output of - it should be fairly simple maths to determine how to scale it's power output. There will be some error in this, but that's to be expected and is pretty much unavoidable. Even a proper sound meter will have some error!

[andrei_me]

Michael bridgman anything about ROC and/or HSA?