Radeon GPUs Are Increasingly Competing With NVIDIA GPUs On Latest RadeonSI/RADV Drivers

[Leopard]

That was a forecast made 10 years ago about what we expected 6-7 years ago.

So we cannot say this surpassed your expectations because this %50-60 thing happened a little bit late?

Sorry , since my native language is not English it is hard for me to understand sometimes.

[bridgman]

Sorry about that , did i get this wrong? Open drivers are already better than Amd GPU Pro ones on OpenGL wise , what is that all %50 - 60 thing? Comparasion with Windows drivers?

That was a forecast made 10 years ago about what we expected 6-7 years ago.

[OneBitUser]

The year of the open source graphics is finally upon us - AMD finally has a first class high end GPU that can go neck and neck with Nvidia in price / performance.

And its out of stock everywhere and when in stock at least twice over its MSRP thanks to miners.

What a buzzkill.

I feel for you/us...

I have bought a GTX 970 last summer, because there were no other options, and now I'm itching to switch back to Radeon with it's ease of use, but the prices are completely out of control.
Nevermind, at least NV is not that bad to setup on *buntus...


What's the current status of switchable graphics on Linux?
I'm considering switching to a notebook, but all the info I've found on the subject is 3-5 years old, which is ancient history by Linux desktop standards.

I have a Precision 7510 at my disposal, it really got me hyped for a switch to notebooks.

[Leopard]

What I said (almost 10 years ago now) was that with a simple shader translator rather than a real shader compiler and without per-app performance investigation and optimization we should be able to reach 60-70% of closed source driver performance. In hindsight that was a bit optimistic - performance actually plateaued at more like 50-60%.

IIRC the first crack in those numbers was when Vadim's "Shader Backend" (SB) optimizer started being enabled by default. The current driver has a "real" LLVM-based shader compiler plus a couple of years of per-app performance investigation and optimization by Marek, Nicolai and others.

BTW I'm pretty sure I actually only said it once but then spent a few years correcting people who had misquoted me.

Sorry about that , did i get this wrong? Open drivers are already better than Amd GPU Pro ones on OpenGL wise , what is that all %50 - 60 thing?

Comparasion with Windows drivers?

[agurenko]

So pleased with the results. It's very impressive how far AMD drivers went in a year+ time frame. I'd love to swap my GTX1060 to RX580 or maybe even Vega56, but I need a new PSU for that since they are much more power hungry. Is there any known issue right now with AMD graphics that would be a regression for me coming from nVidia camp? The monitor output is now merged, right?

[finalzone]

how hybrid graphics are working this days?

Running hybrid laptop using Kaveri and Hainan (Sea Island and South Island respectively) with kernel 4.16 snapshot, Mesa 18.0.0 and enabled amdgpu driver on Gnome 3.28 Wayland , the navigation is smooth. Some works need to be done on RADV which currently failed to render Vulkan scene despite passing Vulkan validation for both CIK and SI.

vulkaninfo | grep device
WARNING: radv is not a conformant vulkan implementation, testing use only.
WARNING: radv is not a conformant vulkan implementation, testing use only.
WARNING: radv is not a conformant vulkan implementation, testing use only.
WARNING: radv is not a conformant vulkan implementation, testing use only.
    VK_KHR_get_physical_device_properties2: extension revision  1
    deviceID       = 0x6665
    deviceType     = DISCRETE_GPU
    deviceName     = AMD RADV HAINAN (LLVM 6.0.0)
    deviceID       = 0x130d
    deviceType     = INTEGRATED_GPU
    deviceName     = AMD RADV KAVERI (LLVM 6.0.0)
    deviceID       = 0x6665
    deviceType     = DISCRETE_GPU
    deviceName     = AMD RADV HAINAN (LLVM 6.0.0)
    deviceID       = 0x130d
    deviceType     = INTEGRATED_GPU
    deviceName     = AMD RADV KAVERI (LLVM 6.0.0)

[Aleksei]

Here is the sad irony.

The year where a foss graphic stack becomes perfectly acceptable and the miminmal performance loss is easily worth it for someone who isn't a serious gamer and not buy top tier expensive cards anyway.

that said, this is entirely irrelevant as the cryptocurrency miners have left none of them available.

From what I gathered during my switch from 1060GTX to RX560 - miners are not that interested in mid-range AMD. Don't know why, they have their own little science/woodoo.

[zboszor]

If they could, they would. They are in the business of selling products after all. The problem is they are getting bought up as fast as they can make them. I've read they are hesitant to invest in increasing their manufacturing output, because if the "using consumer GPU for cryptocurrency mining" bubble bursts, they now have a glut of both product and manufacturing capacity. Makes more business sense to keep producing at current volume knowing they will all sell quickly and at full price.

Unfortunately, it sounds pretty much like the story of the Motorola 680x0 CPUs. Almost all of their CPUs were sucked up by military applications because these were so much better than anything else at the time and they didn't need to compete for a while. Then all of a sudden Intel released the i486 @33MHz or higher and it was on par with or slightly faster than the 68060 @25MHz. The Pentium (after fixing the FPU flaw) was another huge performance jump and Motorola didn't have the R&D budget to compete anymore.

[dungeon]

Interesting article but not sure how it affects production of Vega (unless I am misunderstanding what you are trying to say).

I tought of how prices of memory increased prices on the market, be it DDR4, GDDR5, GDDR6 or HBM2 does not matter.

People wanna fastest memory, plus also of great capacity - that isn't cheap.

[TemplarGR]

If they could, they would. They are in the business of selling products after all. The problem is they are getting bought up as fast as they can make them. I've read they are hesitant to invest in increasing their manufacturing output, because if the "using consumer GPU for cryptocurrency mining" bubble bursts, they now have a glut of both product and manufacturing capacity. Makes more business sense to keep producing at current volume knowing they will all sell quickly and at full price.

Problem with this approach is they are losing mindshare. AMD are selling all the cards they make at full price but those cards don't end up in proper consumer hands instead they end up in mining farms. This means less proper application market share, less bug fixing, less user mindshare, less advertisement, less incentive for game and application developers to actually optimise for AMD hardware.

In the end, they may be selling in the short term but hurt their long term position. Increasing their volume shortterm, if only to help getting more cards into gamer hands, is worth the risk in my opinion. The alternative, just surrendering the whole gaming market to Nvidia, is much worse than the bubble busting and having some inventory unpushed.