Menu
▾
▴
linuxconsole-dev
|
From: Michael P. <mic...@op...> - 2006-07-09 21:27:42
|
>Does anyone have success in running a dual head setup with two PCI-Express graphics cards ? If you want excellent PCI-E/PCI-E performance and cost isn't a concern, get yourself an nvidia nforce based SLI-capable motherboard, like the Asus A8N-SLI or the socket AM2 Asus M2N-SLI. We have used those two extensively for 2-user high end gaming systems. Paired with a dual-core CPU you can play games like Quake4 head to head with good frame rates depending on the video cards. However, if all you need is good 2D performance, AGP/PCI or PCI-E/PCI video should be just fine as long as you use all nvidia cards with the nvidia driver. Just pick up an MX4000 PCI card, the video overlay works just fine on all cards. -- Michael Pardee Open Sense Solutions LLC http://open-sense.com 888-323-1742 920-494-3222 |
|
From: Michael P. <mic...@op...> - 2006-07-25 12:46:37
|
Hi Ludovic, Sorry for the undersight on my part, I just checked the ogle AND Xorg cpu usage: PCI: 22-26% ogle_mpeg_vs 13-16% Xorg AGP: 29-30% ogle_mpeg_vs 0-1% Xorg So indeed the performance is worse on the pci cards. I will try this on a dual pci-e system later. XV does seem to be working on both screens, but why is the player cpu usage lower but Xorg usage higher on PCI? Why would it be so much more variable on PCI? Thanks, Michael Pardee Open Sense Solutions LLC http://open-sense.com 888-323-1742 920-494-3222 > > The interresting part is cpu usage of X11 process. In all case, with xv, > mplayer only has to copy data to a Xshm. It doesn't directly hit the > hardware... X11 then has to copy these data into the memory of the > graphics > cards and then instruct the GPU to blit/overlay it on screen... > > So what is X11 cpu usage in you case ? > > Regards, > Ludovic > > |
|
From: Jean-Daniel P. <jd...@di...> - 2006-07-25 20:24:43
|
On Tue, Jul 25, 2006 at 07:39:52AM -0500, Michael Pardee wrote:
> So indeed the performance is worse on the pci cards. I will try this o=
n=20
> a dual pci-e system later.
I performed the same test as in my previous post : playing the very=20
same video sample on both cards and the gloabl cpu load is roughly th=
e=20
same, with more or less 8%/9% for mplayer and 10% for XFree86 (4.3.0.=
1=20
debian) on both cards.
now you can of course reach better performances on AGP, like 80 fps=20
with fgfs which is impossible with the PCI card, at least because the=
=20
gpu is much weaker.
AGP is nvidia 4200 ti
PCI is nvidia MX 440
--
Jean-Daniel Pauget - http://www.nekodune.com/
T=E9l: +33 (0)2 33 17 20 16
2, rue Andr=E9 PELCA
50580 Denneville-Plage
France
|
|
From: ludovic p. <pl...@nn...> - 2006-07-20 18:17:24
|
Hello, I followed your advice and bought a MX4000 PCI... In a single head setup = (an=20 amd XP 2700) , mplayer playing a video with XV eats only about 10%... (us= ing=20 latest nvidia drivers) But using the same board as secondary head (on an amd64 3000) didn't impr= ove=20 much my situation : In this case, CPU usage is around ~32% (this is 5% le= ss=20 than the MGA G200 I had). I can't believe that the AGP board eats so much bandwith event when unuse= d ! Does someone have the same problem ? Regards, Ludovic Le dimanche 9 juillet 2006 23:27, Michael Pardee a =C3=A9crit : > >Does anyone have success in running a dual head setup with two PCI-Exp= ress > > graphics cards ? > > If you want excellent PCI-E/PCI-E performance and cost isn't a concern,= get > yourself an nvidia nforce based SLI-capable motherboard, like the Asus > A8N-SLI or the socket AM2 Asus M2N-SLI. We have used those two extensi= vely > for 2-user high end gaming systems. Paired with a dual-core CPU you ca= n > play games like Quake4 head to head with good frame rates depending on = the > video cards. > > However, if all you need is good 2D performance, AGP/PCI or PCI-E/PCI v= ideo > should be just fine as long as you use all nvidia cards with the nvidia > driver. Just pick up an MX4000 PCI card, the video overlay works just = fine > on all cards. --=20 The best way to accelerate a Macintoy is at 9.8 meters per second per sec= ond. |
|
From: Michael P. <lin...@gr...> - 2006-07-20 19:03:35
|
I wonder if this has something to do with mplayer. When you first asked the question I went over to a 4 user system with MX4000 pci cards and played a video with realplayer using XV and saw < 10% cpu usage. Even without XV, high resolution realplayer videos still don't use up 35% of the cpu. Are you sure XV is operational when used as a secondary card? You can test with gstreamer-properties if you are using gnome. If you post a link to the mplayer video and question I'll try to play it and see what happens here. Thanks, Mike -- Michael Pardee Open Sense Solutions LLC http://open-sense.com 888-323-1742 920-494-3222 |
|
From: ludovic p. <pl...@nn...> - 2006-07-23 17:04:42
|
No, I have the same with mplayer and xine. XV seems to works fine on seco= nd=20 head (xvinfo), but it just uses more cpu than when the card is alone.=20 I think it could be related to DMA... I suspect the nvidia drivers to do = DMA=20 correctly only on first head and not on second... I had a look at the xv code of the mga driver (the code which send data t= o=20 video memory) : there is no dma here and it gives arround 35% cpu usage w= hen=20 playing video. Since i have the same perfs on the pci nvidia, it is possi= ble=20 that it uses the same method (copying data using cpu, without dma)... The video I used to test was a capture from dvb-t (mpeg2). So you can get= the=20 same by playing a dvb for example... I'd really like to know if you have = XV=20 working on all you heads with nice cpu usage... Regards, Ludovic Le jeudi 20 juillet 2006 21:03, Michael Pardee a =C3=A9crit : > I wonder if this has something to do with mplayer. When you first aske= d > the question I went over to a 4 user system with MX4000 pci cards and > played a video with realplayer using XV and saw < 10% cpu usage. > Even without XV, high resolution realplayer videos still don't use up 3= 5% > of the cpu. > Are you sure XV is operational when used as a secondary card? You can = test > with gstreamer-properties if you are using gnome. > If you post a link to the mplayer video and question I'll try to play i= t > and see what happens here. > Thanks, > Mike --=20 The "cutting edge" is getting rather dull. -- Andy Purshottam |
|
From: Jean-Daniel P. <jd...@di...> - 2006-07-23 20:56:17
|
On Sun, Jul 23, 2006 at 07:04:35PM +0200, ludovic pollet wrote:
> I'd really like to know if you have XV working on all you heads with ni=
ce=20
> cpu usage...
playing the very same video sample (divx, but does it matter ?) 25 fps on=
=20
both my card gives almost the same result.
both are using xv, mplayer said :
=AB VO: [xv] 720x576 =3D> 768x576 Planar YV12 =BB
GeForce4 Ti 4200 (AGP 4x):
real 2m11.211s
user 0m13.145s
sys 0m0.344s
GeForce4 MX 440 (plain old PCI)
real 2m9.529s
user 0m18.829s
sys 0m0.636s
lurking around with ps or top gives 8% 9% max for mplayer on both cards,=20
but most of the time it's around 5%
Still, I think measuring with "time" is more precise.
using full screen by resizing the video to 1280x1024 doesn't gives much
difference :
GeForce4 Ti 4200 (AGP 4x):
real 0m26.300s
user 0m3.480s
sys 0m0.240s
GeForce4 MX 440 (plain old PCI)
real 0m26.960s
user 0m3.000s
sys 0m0.880s
=09
Now if I play fgfs, one is really much slower than the other...
--
Jean-Daniel Pauget - http://nekodune.com/
T=E9l: +33 (0)2 33 17 20 16
2, rue Andr=E9 PELCA
50580 Denneville-Plage
France
|
|
From: Ludovic P. <pl...@fr...> - 2006-07-24 05:14:25
|
The interresting part is cpu usage of X11 process. In all case, with xv,=20 mplayer only has to copy data to a Xshm. It doesn't directly hit the=20 hardware... X11 then has to copy these data into the memory of the graphi= cs=20 cards and then instruct the GPU to blit/overlay it on screen... So what is X11 cpu usage in you case ? Regards, Ludovic Le dimanche 23 juillet 2006 22:56, Jean-Daniel Pauget a =E9crit : > On Sun, Jul 23, 2006 at 07:04:35PM +0200, ludovic pollet wrote: > > I'd really like to know if you have XV working on all you heads with = nice > > cpu usage... > > playing the very same video sample (divx, but does it matter ?) 25 fps = on > both my card gives almost the same result. > both are using xv, mplayer said : > =AB VO: [xv] 720x576 =3D> 768x576 Planar YV12 =BB > > GeForce4 Ti 4200 (AGP 4x): > real 2m11.211s > user 0m13.145s > sys 0m0.344s > GeForce4 MX 440 (plain old PCI) > real 2m9.529s > user 0m18.829s > sys 0m0.636s > > lurking around with ps or top gives 8% 9% max for mplayer on both cards= , > but most of the time it's around 5% > Still, I think measuring with "time" is more precise. > > using full screen by resizing the video to 1280x1024 doesn't gives much > difference : > > GeForce4 Ti 4200 (AGP 4x): > real 0m26.300s > user 0m3.480s > sys 0m0.240s > GeForce4 MX 440 (plain old PCI) > real 0m26.960s > user 0m3.000s > sys 0m0.880s > > Now if I play fgfs, one is really much slower than the other... > > > -- > Jean-Daniel Pauget - http://nekodune.com/ > T=E9l: +33 (0)2 33 17 20 16 > 2, rue Andr=E9 PELCA > 50580 Denneville-Plage > France > > -----------------------------------------------------------------------= -- > Take Surveys. Earn Cash. Influence the Future of IT > Join SourceForge.net's Techsay panel and you'll get the chance to share > your opinions on IT & business topics through brief surveys -- and earn > cash > http://www.techsay.com/default.php?page=3Djoin.php&p=3Dsourceforge&CID=3D= DEVDEV > _______________________________________________ > Linuxconsole-dev mailing list > Lin...@li... > https://lists.sourceforge.net/lists/listinfo/linuxconsole-dev --=20 Le vice toujours sombre aime l'obscurite. -- Nicolas Boileau, Satires |
×
Want the latest updates on software, tech news, and AI?
Get latest updates about software, tech news, and AI from SourceForge directly in your inbox once a month.
Thanks for helping keep SourceForge clean.
X