AMD Radeon RX 480 On Linux

[LasseKongo]

Thanks for a nice review, finally signed up for Phoronix premium after visiting the site daily for a while.
The performance of the RX480 was around where I expected (also in Windows), the power consumption was
a bit higher than I thought, but overall it looks like I might upgrade my 280X to a RX480. Really pleased with how
the driver situation for AMD GPUs are evolving, that is an incitement to buy in it self.

The only problem now is the price, where I live (Sweden) the 8 GB version sells for ~290USD ex. sales tax which is quite
far from the 240USD recommended price. Also I probably want a better cooling solution, and that will add even
more to the price. Lets hope the price drops some when the dust settles in a few weeks.

[puleglot]

Two questions came to my mind:
1. What about using a RX 480 on a mainboard that only offers PCI Express 2.0? Will it work and if it does how much impact on performance if any will the PCIe 2.0-interface have?

PCI-E interfaces are backward and forward compatible. As for the performance impact look at the following tests for example:

http://www.techpowerup.com/reviews/Intel/Ivy_Bridge_PCI-Express_Scaling/

My TONGA is connected to PCI-E 2.0 x8 bus and it's pretty happy.

[Michael]

I really wish there were higher quality and more representative compute/OpenCL benchmarks. That's a pretty solid chunk of the Linux audience, I'd say and the current set of benchmarks are little and ill conceived IMHO.

- This GpuTest is pretty outdated OpenGL compute shader stuff, I doubt it's very representative for compute.
- Mixbench is, well, as its author describes: "The purpose of this benchmark tool is to evaluate performance bounds of GPUs on mixed operational intensity kernels. The executed kernel is customized on a range of different operational intensity values."
Without seeing a _curve_ of flops/byte vs performance, not even the flops/byte at which the tests are run, those numbers are meaningless. Even a simple peak flop rate (with the arithmetic intensity at which this is reached) would be more meaningful than just a single number.
- Last, the SHOC MD5 is frankly a niche benchmark that's heavy on integer and bitwise ops so it's special and that load is frankly not too interesting for most compute use-cases.

I'm baffled why aren't any of the standard/classic computational/HPC benchmarks included that are the de-facto tools to characterize the performance of a processor/compute architecture. These are in fact a the foundation of SHOC, included in the "Level 0" (device memory bw, max flops, etc.) and "Level 1" tests (FFT, scan, reduction, gemm, sort, triad, etc.): https://github.com/vetter/shoc/wiki

Michael, can you please take this feedback and consider it seriously. These GPUs are not anymore gamers-only anymore an I'm sure there are plenty of people in the technical computing/HPC community interested (I know at least a dozen ) in more thorough and representative compute benchmarks.

I have more OpenCL tests coming.

[bridgman]

Two questions came to my mind:
1. What about using a RX 480 on a mainboard that only offers PCI Express 2.0? Will it work and if it does how much impact on performance if any will the PCIe 2.0-interface have?
2. Is there any AMD APU available yet that profits from the same open source driver model?

I wouldn't expect to see much difference between PCIE 2.0 and 3.0 unless you are running a narrow bus (x4 would probably show a noticeable difference, x8 not much, x16 essentially nada).

Kaveri and Carrizo both use the same open source stack, although upstream kernel code uses the radeon kernel driver by default rather than amdgpu. The new Bristol Ridge and Stoney Ridge parts will also use the same stack.

At the moment Kaveri is the only APU being sold as a socketed part for desktop/DIY systems, eg A10-7860K, while Carrizo/BR/SR are soldered onto the mobo for use in laptops, all-in-ones and small-footprint boxes (whatever the current name for them is).

[nuetzel]

Dear R9 280X, R9 380X (Antigua XT, full Toga) or any other RX480 upgrader,

would you be so kind and donate your card to Michel?
He could add these missing parts to his farm.
I'll do with my poor 6670 (Turks XT) if I got mine RX480 (costum one) after our vacation.
Maybe some more (older) if he wants.

Thank you very much for all your support!

[Dedale]

I have more OpenCL tests coming.

Cool, thank you.

Tridam from the French HFR made an interesting remark. The card pumps more than what a PCIE slot plus a 6 pin connector can provide. So it goes beyond what a PCIE should provide. So they say to buy them only if you trust enough your motherboard.

http://www.hardware.fr/articles/951-31/conclusion.html (In French)

[puleglot]

Dear R9 280X, R9 380X (Antigua XT, full Toga) or any other RX480 upgrader,

R9 280X is a TAHITI XT, almost the same as HD 7970.

[kaprikawn]

Excellent article, this looks to be the GPU I've been waiting for, OSS drivers with competitive performance. I shall be picking one up when the partner boards start appearing. Goodbye 9790.

[sverris]

Cool, thank you.

Tridam from the French HFR made an interesting remark. The card pumps more than what a PCIE slot plus a 6 pin connector can provide. So it goes beyond what a PCIE should provide. So they say to buy them only if you trust enough your motherboard.

http://www.hardware.fr/articles/951-31/conclusion.html (In French)

Thats an interesting question, which occurred also in some German tests (heise.de, golem.de). The PCIe-Standard has 'some' tolerance, for safety reasons, but who knows how much? Will f.i. very cheap motherboards sooner or later run into issues, then?

[mibo]

Thank you Michael for the extensive testing.

Do I read the power graph correctly, that the PC+RX480 consumes 54 Watts when idle?
Some Nvidia cards need 10 Watts less...
In the two other tests I read (anandtech.com and computerbase.de) they mentioned a (windows) driver bug that keeps the idle power consumption higher than needed...
Is there also such "bug" in the Linux drivers (as OSS and PRO driver show the same 54W)? Is there a chance, this idle power usage gets lowered with newer driver?

Nevertheless, I'm waiting now for the more silent AIB versions. And then is upgrade time :-) (from HD7950)