From: Emil Velikov <emil.l.velikov-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
To: Carlos Garces <carlos.garces-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
Cc: nouveau-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW@public.gmane.org
Subject: Re: Bad performace for NV18 driver
Date: Sat, 08 Jun 2013 21:31:15 +0100 [thread overview]
Message-ID: <51B39493.2070207@gmail.com> (raw)
In-Reply-To: <CAD-add=DhPzgFR-73x54rFc7KtKSSTXAz-WKA7=tKm3MhSHbJQ-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org>
On 08/06/13 19:06, Carlos Garces wrote:
> After update the system now I use:
>
> Kernel 3.9.2
> x server 1.12.4
> nouveau 1.0.7
>
Good call
> the xorg config are at:
>
> http://pastebin.com/2LUxLiVs
>
Bad call, your resolution issues are due to messed up config
If you really insist on having such an "interesting" setup, start from
scratch and build on top of it - one section at a time. I would
recommend getting the "minimal" xorg.conf [1] as a base
> It's the one installed by my distro, without any modification
>
> I have made 2 test.
>
> Normal boot
> Xorg at http://pastebin.com/reZdAXEF
> GLX Renderer is detected as GLX "Mesa DRI nv18 x86/MMX+/3DNow!+/SSE GLX"
>
Note that nouveau correctly detects that the monitor is 1024x768 60Hz
capable.
"(II) NOUVEAU(0): Modeline "1024x768"x60.0"
Something is the config you're feeding it causes the issue
>
> Same configuration with nomodeset
> Xorg log at http://pastebin.com/ZbXBG8sz
> GLX Renderer: Gallium 0.4 on softpipe
>
> Results for Cairo test with and without nomodeset
> http://openbenchmarking.org/result/1306082-CGAR-CAIROTE23
>
> Full results from Friday (before update Mesa)
> http://openbenchmarking.org/result/1306072-UT-2DGPUTEST95
>
>
> The result with Cairo are better in the updated system, 0.63 FPS vs
> 0.54 FPS but still worse than boot with nomodeset (7,84 FPS)
>
> Any ideas to improve the results with Cairo?
>
If you do a bit of search[2] you'll notice that in quite a few cases
cairo/sw is quite fast
Quite a few of the benchmarks in your original post were faster using
nouveau
On that note, I believe that a question in terms of "Program XXX is
rendering very slow, is there any way to improve that ?" would be more
beneficial
Regards
Emil
P.S. Please avoid top-posting
[1]
Section "Device"
Identifier "my_fancy_identifier"
Driver "nouveau"
EndSection
[2] http://ickle.wordpress.com/
> On Sat, Jun 8, 2013 at 12:18 AM, Ilia Mirkin <imirkin-FrUbXkNCsVf2fBVCVOL8/A@public.gmane.org> wrote:
>> On Fri, Jun 7, 2013 at 5:56 PM, Carlos Garces <carlos.garces-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org> wrote:
>>> Hi!
>>>
>>> I have some strange results on my test with nouveau drivers
>>>
>>> I'm using a old GeForce4 MX 4000 (NV18) with 128 MB.
>>>
>>> I have made 2 test.
>>>
>>> -The default configuration, using nomodeset without nouveau_vieux_dri.so
>>>
>>> $inxi -G
>>> Graphics: Card: NVIDIA NV18 [GeForce4 MX 4000]
>>> X.Org: 1.12.4 drivers: vesa,nouveau (unloaded: fbdev)
>>> Resolution: 1024x768@0.0hz
>>> GLX Renderer: N/A GLX Version: N/A
>>>
>>> -The "optimized" configuration, after removing nomodeset, installing
>>> the 3d driver (nouveau_vieux_dri.so, package libgl1-mesa-dri) and
>>> configure the xorg to use only nouveau
>>>
>>> $inxi -G
>>> Graphics: Card: NVIDIA NV18 [GeForce4 MX 4000] X.Org: 1.12.4 driver:
>>> nouveau Resolution: 800x600-XyBiMD9AD2I@public.gmane.org
>>> GLX Renderer: Mesa DRI nv18 x86/MMX+/3DNow!+/SSE GLX
>>> Version: 1.2 Mesa 8.0.5
>>>
>>> The resolution is 800x600 because was impossible for me set up 1024x768
>>>
>>> There is the full xorg log
>>> http://pastebin.com/2ZbMwyRK
>>>
>>> The results are strange for me. I expect better performance on
>>> "optimized" configuration but some test using Cairo are very bad (0,54
>>> vs 5,38).
>>>
>>> The test are available at
>>> http://openbenchmarking.org/result/1306076-UT-2DGPUTEST63
>>>
>>> Any idea to improve the performace of NV18?
>>> Any know issue with nouveau_vieux_dri.so and NV18?
>>
>> It's a little difficult to tell anything without seeing your xorg.conf
>> -- it seems like 1024x768 should work fine. You also appear to be in
>> 16-bit mode, and using fairly old software (xorg 1.12, mesa 8.0,
>> xf86-video-nouveau 1.0.1). What kernel version are you on? Although
>> it's a fairly old card so perhaps there hasn't been much improvement
>> on it. In recent X versions, things get auto-detected for the most
>> part and you can forgo most explicit configuration like
>> modelines/etc... not sure if that's the case for xorg 1.12 or not.
> _______________________________________________
> Nouveau mailing list
> Nouveau-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW@public.gmane.org
> http://lists.freedesktop.org/mailman/listinfo/nouveau
>
prev parent reply other threads:[~2013-06-08 20:31 UTC|newest]
Thread overview: 4+ messages / expand[flat|nested] mbox.gz Atom feed top
2013-06-07 21:56 Bad performace for NV18 driver Carlos Garces
[not found] ` <CAD-addk1m19g_onyyM1Lx4Rnji32sPDpAffHaJ=39=33aZGnHg-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org>
2013-06-07 22:18 ` Ilia Mirkin
[not found] ` <CAKb7UvjVE52o5ZCFg43b-yRGGPgWkRkn7ctEu8EscySfskHX2Q-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org>
2013-06-08 18:06 ` Carlos Garces
[not found] ` <CAD-add=DhPzgFR-73x54rFc7KtKSSTXAz-WKA7=tKm3MhSHbJQ-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org>
2013-06-08 20:31 ` Emil Velikov [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=51B39493.2070207@gmail.com \
--to=emil.l.velikov-re5jqeeqqe8avxtiumwx3w@public.gmane.org \
--cc=carlos.garces-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org \
--cc=nouveau-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW@public.gmane.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.