Intel Broadwell HD Graphics 5500: Windows 8.1 vs. Linux

[Revelator]

It's all automated parsed and generated... So that's what it was reported from the laptop's installation.

ok but need a fix

[opensource]

Did you actually use the phrase "raw power" in reference to an Intel GPU? How odd.

Of course, we have Beignet and we can use it It's not much but theoretically more than the CPU part.

[duby229]

Of course, we have Beignet and we can use it It's not much but theoretically more than the CPU part.

Functional and "raw power" are 2 very different things.

[winnitro]

SteamOS can't compete with Windows.

In Windows 10 we have absolutely every games from Steam, Origin and Xbox.

In the SteamOS, only "some" Steam games.

[duby229]

Not if MS has their way. Good thing it's not likely to happen.

[SystemCrasher]

In Windows 10 we have absolutely every games from Steam, Origin and Xbox.

In the SteamOS, only "some" Steam games.

This is exactly why Valve needs to support different platforms. If they only rely on windows... they will be doomed in long term. MS never cares about what would happen to OEMs, ISVs, devs and so on. Needless to say, it maybe ok for MS but makes it exceptionally troublesome for anyone who is stupid enough to rely on MS and their techs exclusively. And if someone thinks aggravating most of industry against self is not a problem, you got it wrong. You see, on mobile devices market, in the web, etc - MS is simply nobody. Now it is time to aggravate gamedevs and distributors by posing extra competitions to them at platform level? Hmm, cool idea! They will surely have to consider other platforms .

[Eero]

Changes in performance

The absurd bloom filter used in Phoronix OpenArena test makes the test memory bandwidth bound (and ~halves its perf). Reason why OpenArena performance has dropped, is X intel driver defaulting back to DRI2 (DRI3 doesn't have the memory bandwidth overhead of DRI2, but it's still buggy).

Unigine Tropics improvement is an older improvement in Mesa shader compiler I think.

Unigine Valley drop could be this: https://bugs.freedesktop.org/show_bug.cgi?id=86242

Xonotic high quality improvement is an older improvement in Mesa shader compiler I think.

Plot3D has 0.6 million primitives and nothing else, recently there was a kernel change that improved BDW performance a lot in that kind of extreme micro-benchmark cases and a bit also in more realistic cases.

Triangle test low numbers are because DRI3 in Ubuntu 14.10 Vsyncs it. Newer X intel driver defaults back to DRI2 and there are some Mesa fixes also to the X DRI3 / present extension handling.

Volplosion is shader bound and the change is improvement in the Mesa shader compiler from fall, the test doesn't anymore cause register spills.

I hope Phoronix could do some testing with different X (=2D) drivers so that comments in articles wouldn't always assume issues with the 3D driver...