Intel graphics benchmarked by Phoronix

[Beso]

2008/12/23 Michael Larabel <michael.larabel at phoronix.com>:
> Nikos Chantziaras wrote:
>> Devin Heitmueller wrote:
>>
>>> On Tue, Dec 23, 2008 at 11:22 AM, Nikos Chantziaras <realnc at arcor.de> wrote:
>>>
>>>>> http://www.phoronix.com/scan.php?page=article&item=intel_graphics_q408&num=1
>>>>>
>>>> That test is wrong.  You don't compare different drivers on different
>>>> distros.  You compare them on the same distro.  Now I get to say that
>>>> it's not the driver that is slower, but it's the newer Ubuntu version
>>>> that slows it down.
>>>>
>>>> They should learn to benchmark :P  And they should learn to also name
>>>> the benchmarks appropriately (that it compares performance of two Ubuntu
>>>> versions rather than Intel driver versions.)
>>>>
>>> [...]
>>> Is it the Xorg team's perspective that, "we'll only look into huge
>>> performance regressions if you first do all the work to prove that the
>>> huge slowdown cannot be anything but the Xorg stack?"
>>>
>>> I'm not trying to be confrontational, but I think burying your head in
>>> the sand until you are given 100% proof that it's the Intel driver may
>>> not be the best approach here.
>>>
>>
>> Well, it's certainly possible that the driver became that much slower.
>> But doing the benchmark right would be so easy, yet Phoronix didn't do
>> it.  All I need to do here, is simply remove the old driver and install
>> the new one and re-run the benchmark.  Everything else stays the same.
>> Dead easy.  Yet Phoronix didn't do it.  Hopefully someone here with an
>> Intel GPU can do it better.
>>
>> _______________________________________________
>> xorg mailing list
>> xorg at lists.freedesktop.org
>> http://lists.freedesktop.org/mailman/listinfo/xorg
>>
>>
>
> The original intent of this testing was going to be for an Ubuntu 8.10
> vs. Ubuntu 9.04 development article, however, after seeing these Intel
> performance regressions, those graphics results were then published.
> It's clearly mentioned in that article that these were the results from
> the stock packages of Ubuntu 8.10 and the Jaunty repo as of Dec 18.
>
> As is stated in the article, once the various packages (kernel, X Server
> 1.6, et al) near a point of stability, they'll all be tested properly
> and built from source in a vanilla configuration with all other
> variables remaining the same.
>
> Best Regards,
> Michael Larabel
>
i agree with michael. at the end thre's the following:

Intel's current selection of integrated graphics processors (IGPs) are
already slow when it comes to just the hardware, but with the graphics
stack currently found in the Ubuntu 9.04 development branch the
performance is simply miserable. The performance drops were nearly
across the board with both 2D and 3D regressions being very
noticeable.


it's said that the actual stack found in jaunty is miserable, which is
different from
saying that intel stack is miserable. on the other hand, jaunty hasn't
been yet released
in beta so it's clear that there would be some major issues.
these tests prove that jaunty isn't there for the use of normal users
but not that intel's
new stack is not ok. i'd suggest people to read more carefully before
spitting out
something not true, that the original part hasn't been saying.

also, i think that these git tests are better run on a gentoo (or a
gentoo based distro) box,
since it has a much better way of handling different package versions,
especially the ones from git.

-- 
dott. ing. beso