A 13 Line Patch That Boosts Intel Sandy Bridge Performance
Phoronix: A 13 Line Patch That Boosts Intel Sandy Bridge Performance
After some initial Linux troubles, last month we finally got Intel Sandy Bridge graphics working under Linux. The latest Intel CPUs (such as the Core i5 2500K) with integrated graphics are blazingly fast, and the classic Intel Mesa driver was fast compared to other open-source Mesa / Gallium3D drivers, but it still was a ways behind the low-end discrete graphics cards with the proprietary AMD / NVIDIA drivers for Linux. It was also shown that the Intel Linux Mesa driver is much slower than the Intel Windows driver for Sandy Bridge, as we had also found was the case for previous generations of Intel graphics. Committed to the Mesa mainline Git repository this week though was a very important Sandy Bridge change. While the commit only touched 13 lines of code (11 lines of new code, 2 lines of changed code), it has dramatically improved the Sandy Bridge Linux performance as our results show in this article.
Good News for Open Source Display Drivers
Between this Sandy Bridge improvement and the recent benchmarking of Nouveau on a GeForce GTX 580, I'm optimistic about the performance that we'll be getting out of the box from the distributions being released in autumn.
This is just amazing! An opensource linux graphics driver that is near windows driver performance. Obviously the tremulous test seems capped by vsync.
well this is some good news. I'm keen to see know how the mobile SB GPU does. Btw. what about vaapi video decoding? does this actually work? what about some initial benchmarks on that?
I'm not convinced that vsync would cap at 30fps unless Michael is using a very old monitor or has selected a strange mode on a HDTV
If the framerate is low vsync caps at half of the refresh rate.
Great, now if Intel where to improve their drivers for the previous generation of i5 GPUs to Windows standards I'd be happy.
Or at least to GMA 915 standards and by that I mean pushing the performance even close to what the hardware is actually capable of.
My old laptop with a Celeron M 1.3GHz and GMA 915 could do most 720p content but the CPU was too slow for 1080p. On my new laptop with an i5-450M and GMA HD I still can't do 1080p even with VA-API which leaves me kind of puzzled.
On top of that SB can now reach close to Windows performance by simply increasing the thread count?
Not saying I could do better but is everything between GMA 915 and SB really that hard to write fast drivers for?
Wow wow wow! That looks incredible.
But what about Intel HD (1st gen) graphics? What about pre Intel HD graphics? Their drivers still lag immensely behind Windows drivers.
Ok, this is wonderful!
But it leaves me officially confused. I always thought the Intel Linux driver is slower than the Windows one because it is using Mesa, which is inferior to Intel's proprietary OpenGL implementation. This patch seems to prove that wrong. Now this can have several explanations:
1. Intel's proprietary OpenGL implementation is on the same level as Mesa. But that would mean there is something else slowing down the older generations of their graphic cards, since they are so slower than on Windows. Is it something similarly trivial? Surely if its not in OpenGL implementation, it shouldn't be hard to get the same performance as the Windows driver?
2. There is actually a similar issue in the Windows driver, meaning the card can actually perform much faster on Windows too. I find this unlikely, that would mean the card is actually much faster than we thought
3. I'm missing something really important here...
What you are missing here is that Linux driver teams tend to be A LOT smaller than their Windows counterteams. It is a simple fact in life that you put most of the money where it is most important which is as of right now Windows users.
Originally Posted by nite