Overall, I think the battle for the living room is just starting. Other devices may have already gotten a head-start, but Nvidia’s big leap puts the Shield into the fight in a strong way. Short of being a computer or HTPC, the Nvidia Shield covers several basic, but important, entertainment needs — it’s a gaming system, streams video, plays music, uses apps and may provide a way to finally cut the cable company chord that’s been taking all your money.
Realistically, it’s still a new product, but the options aren’t that few and there’s definitely more room for Shield and similar products to grow. It will be interesting to see what Nvidia and its competitors will come up with next.
Emil Velikov announced the release this morning of Mesa 10.5.7.
The tests I did on an Intel Xeon box were freshly built versions of GCC 4.9.2, GCC 5.1.0, and the latest SVN snapshot of the GCC 6.0 compiler that's under development for release in 2016. The three compilers were built the same way and then they went on to build our many open-source Linux CPU benchmarks via the Phoronix Test Suite automated benchmarking software.
One of the biggest limitations of Nouveau remains and that's the lack of proper re-clocking support. In today's testing, I attempted to re-clock the graphics cards that support it with Nouveau, but most of the time they couldn't be statically re-clocked to their highest performance state. Most Kepler GPUs can be re-clocked to the mid-level (0a) but would lock-up and produce artifacts as illustrated above when attempting to hit the highest performance state for the video memory. Of the GPUs tested, only the GeForce GTX 650 was able to hit the 0f (highest) state and run without failing. Nouveau developers have been working on re-clocking for years but it's an incredibly large task without support or technical documentation from Nouveau. It seems unlikely that this year the feature will be accomplished, especially for seeing any dynamic re-clocking out-of-the-box.
Last year Imagination launched a MIPS development board that went on sale at the end of last year. In not seeing any significant benchmarks or performance coverage from this MIPS Creator CI20 over the past few months, I finally got around to buying one of these MIPS development boards from Imagination Technology. While the CI20 seemed promising at first, so far I'm very unhappy with this board and it's been even less stable than the Imagination PowerVR drivers on Linux going back to the Poulsbo days.
Yesterday I ran a set of interesting CPU-focused benchmarks on 45 of the Linux systems within my custom-built basement server room that represents many of the systems powering the upstream daily open-source testing at LinuxBenchmarking.com. The tests ran yesterday were primarily processor focused as not all of these computers/servers are equipped for handling GPU testing, etc. This article is basically to provide a look at many different old and new, low-end and high-end systems. The software stacks for the different systems vary based upon what they're testing day-to-day within the server farm, so take these results as you wish. Most of the systems are running Ubuntu 14.04/14.10/15.04 with recent Linux kernel versions.
Today we're looking at the performance of the latest proprietary graphics drivers on the Linux desktop at the time of testing: NVIDIA 352.09 Beta and the Catalyst 15.4 Beta as packaged for Ubuntu Vivid -- fglrx 15.20.2 / OpenGL 4.4.13374. Later in the week should be the open-source Intel/AMD/NVIDIA Linux graphics driver results for celebrating the Phoronix birthday. For this article there were 17 graphics cards tested all supported by these latest proprietary drivers -- the graphics cards used were those that were available and in my possession at the time of testing, which sways to the NVIDIA side. There's basically every major NVIDIA graphics card covered given they're frequently sending out samples to Phoronix for Linux testing while in the past few years on the AMD side they have barely sent out any GPUs for Linux testing... All of the AMD GCN GPUs tested in this article were retail GPUs I purchased. Anyhow, the graphics cards able to be tested for this article were: