Intel Haswell HD Graphics 4600 Performance On Ubuntu Linux

[mendieta]

I suspect the Windows-to-Linux comparison is picking up some DX-to-OpenGL differences as well. The games with the biggest differences seem to be the ones which use DX by default on Windows. If you compare with OpenGL renderer on both OSes the delta is usually much smaller, and sometimes goes the other way as others have mentioned.

That said, I suspect the difference in GPU clocks (open source driver defaults to low clocks on Trinity IIRC) makes a bigger difference than everything else.

Thanks, Bridgman. Things have gotten so much better in the last year!

[smitty3268]

Why does the A10-5800K performs so poorly?

I checked some directx benchmarks on internet and saw that it should perform better than the hd4000 and the hd4600.

EDIT:

I found why: the benchmark here is done with the open source driver.

I think adding a line for the A10 and the catalyst driver would have been appropriate.

Not only that, the default clocks are very low and michael obviously didn't bother setting the pm to high profile.

What would be useful on any further AMD comparisons with Haswell:

mesa-git - what michael tested
mesa-git optimized - use the r600sb optimized backend, and make sure clocks are as high as possible (high profile pm). unfortunately, the kernel actively limits clocks on these integrated amd parts and fixing that can double performance, but that's too much work to be fair to use in a benchmark. changing a couple settings/environment variables isn't.
catalyst - see what the official driver gives.

oh, and actually test it with the same tests (unigine) - you know how to fix this.

And while i'm dreaming about how michael could actually improve his tests, i'll just point out he should include some wine and valve/steam tests as well. i can dream...

[Kivada]

Not only that, the default clocks are very low and michael obviously didn't bother setting the pm to high profile.

What would be useful on any further AMD comparisons with Haswell:

mesa-git - what michael tested
mesa-git optimized - use the r600sb optimized backend, and make sure clocks are as high as possible (high profile pm). unfortunately, the kernel actively limits clocks on these integrated amd parts and fixing that can double performance, but that's too much work to be fair to use in a benchmark. changing a couple settings/environment variables isn't.
catalyst - see what the official driver gives.

oh, and actually test it with the same tests (unigine) - you know how to fix this.

And while i'm dreaming about how michael could actually improve his tests, i'll just point out he should include some wine and valve/steam tests as well. i can dream...

But that would require actual effort and thought be put into the review instead of just throwing PTS on it and running to the presses for that sweet sweet advertisement gold. Larabel is already far too lazy to include anything that can't be easily automated by PTS, he refuses to ut forth the minimum effort of the Windows review sites. He's never done an overclocking test, since stable overclocks on Linux can be different from stable overclocks on Windows. Never does noise and heat differences that you will see on Linux since due to driver optimization or lack there of would have different results on Linux.

And Larabel wonders why he gets so little hardware to test from the manufacturers... He's the worst reviewer after super biased and inept testing sites like Tom's Hardware Guide. And at least they get hardware from manufacturers on a regular basis.

[pinguinpc]

Originally Posted by smitty3268

And while i'm dreaming about how michael could actually improve his tests, i'll just point out he should include some wine and valve/steam tests as well. i can dream...

Wine Test is good idea for MAD and especially for NVIDIA but for INTEL is a BAD BAD idea, INTEL IGP performance and compatibility on wine is very very poorly, is worst than MAD and this is say much

All fault is for INTEL drivers and poorly support for recent OpenGL versions

[Kivada]

Wine Test is good idea for MAD and especially for NVIDIA but for INTEL is a BAD BAD idea, INTEL IGP performance and compatibility on wine is very very poorly, all fault is for INTEL drivers and poorly support for recent OpenGL versions

And whats the problem with pointing that out in the review? It's like not running the HL2:Lost Coast benchmark, it's free, it's native, it's not a shitty Q3 engine demo.

[mike4]

Could you add in the upcoming tests a GTX770 (Nvidia drivers) compared to default Haswell. I wonder if I should buy a GTX770 or even could save that money and live with Haswell graphics on X-Plane and FSX.
Many thanks

[AnonymousCoward]

Could you add in the upcoming tests a GTX770 (Nvidia drivers) compared to default Haswell. I wonder if I should buy a GTX770 or even could save that money and live with Haswell graphics on X-Plane and FSX.
Many thanks

EDIT: Never mind, didn't read your post closely enough. The 770 will be a lot faster, though.

[brosis]

As for item 1, look here, the geometric mean of windows catalyst is 23% faster than linux catalyst:
http://openbenchmarking.org/result/1...ro=y&obr_nor=y

There could be result of famous on-demand CPU bottleneck.
I am *not* a catalyst user, more catalyst hater; but only nvidia blob performs (a bit) worse in linux 100% of time; catalyst in linux usually behaves differently from windows version and by no means slower.

And for item 2: linux opensource can be up to 5 times slower than linux catalyst (35% in the median of the chosen tests)

http://openbenchmarking.org/result/1304178-UT-AMDNVIDIA78&obr_sgm=y&obr_sro=y&obr_cmw=y&obr_imw=y

Just adding the two medians, windows is 50% faster than linux opensource. Anyways, I can't believe I put some much time hunting all this info, lol

Ok, offering to you - how about you start reading my posts, instead of blindly trying to justify your own errata?

Now look, all I have done is attach Vadim result to that of Michael and you see how bad current defaults for the driver are and what the performance is.
Just compare the opensource & catalyst 5750 from Michael with that of Vadim (5750-sb):

http://openbenchmarking.org/result/1304178-UT-AMDNVIDIA78,1304309-FO-57500792103#r-14

[s_j_newbury]

The article doesn't mention what x86_energy_perf_policy is set to. It would seem reasonable to ensure it's set to "performance", I'm pretty sure the kernel default is "powersave" and it's quite possible Haswell has more aggressive internal powersaving.

[curaga]

OTOH, AMD is using a chip similar to the A10-5800k, but with 8 CPU cores, for the new Playstation. Why the F*** are they not releasing that to the public? That would be a no brainer for a fantastic value, for an all around balanced APU for general computing and graphics.

That's Jaguar (Bobcat successor) cores, not the big ones in Trinity/Bulldozer/Piledriver. I wouldn't call having 8 Atom-class cores a balanced CPU, since single-thread performance still matters.