Intel Publishes More Haswell Graphics Driver Code
Phoronix: Intel Publishes More Haswell Graphics Driver Code
While Intel's Ivy Bridge launch is imminent, and I'm still digging through information concerning today's Intel Valleyview code drop that brings Ivy Bridge graphics to their next-generation Atom as they do away with PowerVR graphics for their SoCs, more graphics driver code to enable Haswell support has landed this evening...
Awesome!! Thank you, Intel!!
Originally Posted by phoronix
While I'm a bit uncertain what those "power wells" are, given what someone else mentioned in another thread (namely that GT3 has 40EU), being able to power down unused EUs would be really useful since power usage is, again, a big reason to use their IGP.
If AMD were landing the first Radeon HD8000 series support a year or more before the card launch, I'd buy their products, too. Dang, Intel, you guys are really good at this "planning" thing!
Just imagine if AMD could do something similar for their discrete cards... WOW.... that would have meant that the code drop which came to us this week for HD7000 support would've landed back in, oh I don't know, mid-2011? And we'd have the Gallium3d driver already committed to mesa master by January 2012 or so? And we'd have color tiling and a shader compiler and OpenGL 3 by the end of 2012 for a chip that launched that same year?
Oh, who am I kidding. AMD's open source guys are so overworked, and they have such a huge backlog of generations to catch up on, that I can't see this ever happening unless they quadruple (or more) their current manpower.
So, my next laptop is going to have an Intel CPU and an Intel IGP. This is the deciding factor for me. Are you listening, AMD? Your "let's play catch-up several months after our GPUs are released" open source strategy is costing you money. And I tend to spend at least 3-5 thousand dollars on a laptop, so you're really missing out! If I could buy a Trinity laptop with confidence that I could at least have hardware-accelerated compositing when I bring it home and install a current distro (compiling from git is also okay), I would totally do that. Instead, Intel gets my money.
I'll still buy your discrete GPUs, because I'm a sucker and because I boot Windows to play games. But I ONLY ever run Linux on my laptops, and I DON'T use it to play games, but I want a low-power GPU that can do basic 3d stuff and play flash in a browser without lagging. Right now I have no reason to believe that Trinity will be that far along in the open source drivers by the time I'm ready to buy a new laptop.
«AMD has finally released the open-source driver code to support the Radeon HD 7000 "Southern Islands" GPUs and next-generation Fusion "Trinity" APUs under Linux with their open-source driver.» (Source)
Originally Posted by allquixotic
Trinity is basically a Cayman part, as i understand it. I'm sure there will be some specific differences it has, but I'd guess support for it will be pretty good by the time it comes out.
Originally Posted by Nedanfor
From the article:
Originally Posted by Nedanfor
Having the kernel bits without the Gallium3d bits is like having loose hydrogen atoms without oxygen. No oxygen, no water. No Gallium3d, no driver. Having half the pieces is better than none from a theoretical perspective, but from a practical perspective of actually using the hardware to do interesting things, it's exactly equal to having zero support. Unless you've found a way to survive by imbibing liquid hydrogen...?
Originally Posted by Phoronix
The other interesting thing to note is that it took the Evergreen ASICs over two years (or if you want to be extremely charitable, a year and a half) from the time their Gallium3d driver was first introduced, until the time when they brought up a majority of the hardware support features and got their OpenGL 2.1 acceleration working well.
Assuming that AMD pulls a miracle and manages to halve the time it takes to go from "bare bones G3D driver that crashes running glxgears" to "something we can ship by default in distros", then the wait time is reduced to between 0.75 years and 1 year depending on your optimism factor.
0.75 years from Trinity release would be about January 2013 if Trinity is released within the next 30-60 days. And that's an extremely optimistic projection. Assuming that they release it just in time for it to get picked up by 6-month distros, that means it would show up no earlier than Ubuntu 13.04 for end-users who don't compile from git (which, by the Phoronix surveys, is something like 94% of Phoronix viewers). Ubuntu 13.04 will be released about one full year after the anticipated release date of Trinity.
When you consider that end-users and enthusiasts upgrade between every 2 and 4 years, you are basically cutting the usable Linux lifespan of the hardware between 25 and 50% for people who want open drivers for their (eventual) stability, performance and lack of proprietary goo. The only people who will feel the wait noticeably less are the "late adopters", such as businesses, who are just now starting to upgrade their hardware to mid-grade HD3000 series chips.
It's basically a disincentive to live on the edge. And yet, if AMD's marketing department gets their way, they'll have you buy the chip as soon as it's out, because it's the latest and greatest thing. Slight problem: you can't use it unless you want to bow down to Bill Gates or the proprietary alligators in your kernel.
Last edited by allquixotic; 03-22-2012 at 06:54 PM.
The userspace bits for Trinity were published the same afternoon, as Michael hinted. Maybe look in the commit log before badmouthing us ?
That's because the foundation work required for GL 2.1 was being added at the same time, and the Gallium3D paths were brand new as well. That only has to happen once.
Originally Posted by allquixotic
Last edited by bridgman; 03-22-2012 at 07:07 PM.
So Trinity will be supported at launch?
Tempest in a teapot?
Anyway, I'm looking forward to a fanless HD 7000-based card (7550?)
That is the expectation. I'm sure we'll find some issues with specific SKUs that didn't appear on the engineering boards, but other than testing on a broad range of hardware all the support should be in place now.