Hi, I have a scientific visualisation app (cgui from http://www.bioeng.auckland.ac.nz ) that I have tested on 3 different machines and I have found that the graphics performance is opposite to what I expected. I was wondering if anyone could give me some pointers about how to figure out what might be causing the slowness on 2 of the machines. The machines are: 1. Compaq Presario 2100 series laptop. Athlon XP 2400, 512Mb Ram, Ati Radeon IGP320M. The accelerated features of this graphics chipset are not supported by XFree86. 2. PC with AMD Duron 900 + Matrox GA400 graphics adpator. 3. PC with Athon XP 1800, running Xvnc 3.3.6. Viewed across 100 Mbit ethernet from vncviewer on Windows. The thing that is confusing me is that when viewing a shaded bezier surface the best performance is on machine 3 - using vnc. I would have expected machine 1 at least to have equivalent performance to machine 3 given that the rendering is being done in software in both cases. But it is noticably a lot slower on machine 1, any ideas why that might be? I haven't done any framerate measurements, the "test" I used was to load a human torso model and rotate it with the mouse. Machine 2 was the slowest at this task, so apparently there is no hardware support for shading in the mga driver or the driver doesn't like my hardware. Machine 2 does however run glxgears the fastest. Any suggestions appreciated. g -- Glenn Ramsey <glenn(a)componic.co.nz> 07 8627077 http://www.componic.co.nz
On Tue, 23 Sep 2003, Glenn Ramsey wrote:
I have a scientific visualisation app (cgui from http://www.bioeng.auckland.ac.nz ) that I have tested on 3 different machines and I have found that the graphics performance is opposite to what I expected. I was wondering if anyone could give me some pointers about how to figure out what might be causing the slowness on 2 of the machines.
The machines are:
1. Compaq Presario 2100 series laptop. Athlon XP 2400, 512Mb Ram, Ati Radeon IGP320M. The accelerated features of this graphics chipset are not supported by XFree86.
And you'll probably find the AGP bridge isn't either, which may be another cause of slowdown. Oh, and the XFree86 DRI stuff won't help at all either - the mainline X tree is *old* in this regard. If you're feeling brave, check out the current XFree86 snapshots - ATI have been contributing a *lot* of code, and I think they've got 2D acceleration for the ATI IGP chipsets in there now (SWMBO has a laptop with one of these, and she's asking me to do this...). If you're *really* lucky, the latest X snapshot will actually build!
2. PC with AMD Duron 900 + Matrox GA400 graphics adpator.
Hmm... don't Matrox have their own binary driver somewhere?
3. PC with Athon XP 1800, running Xvnc 3.3.6. Viewed across 100 Mbit ethernet from vncviewer on Windows.
Machine 2 was the slowest at this task, so apparently there is no hardware support for shading in the mga driver or the driver doesn't like my hardware. Machine 2 does however run glxgears the fastest.
This tells me that your app doesn't use GLX, so it's probably doing all the rendering using the machine's CPU, RAM, $deity knows what else. As the Duron 900 is the slowest machine you've got. I'm still boggled at anything over VNC being the fastest, though. That's just bizarre. Cheers Richard -- Richard Stevenson If you want to know what God thinks of money, just look at the people He gave it to -- Dorothy Parker
On Tue, Sep 23, 2003 at 10:00:29AM +1200, Richard Stevenson wrote:
On Tue, 23 Sep 2003, Glenn Ramsey wrote:
Hmm... don't Matrox have their own binary driver somewhere?
3. PC with Athon XP 1800, running Xvnc 3.3.6. Viewed across 100 Mbit ethernet from vncviewer on Windows.
Machine 2 was the slowest at this task, so apparently there is no hardware support for shading in the mga driver or the driver doesn't like my hardware. Machine 2 does however run glxgears the fastest.
This tells me that your app doesn't use GLX, so it's probably doing all the rendering using the machine's CPU, RAM, $deity knows what else. As the Duron 900 is the slowest machine you've got.
Also check the GLU libraries. For nvidia cards the binary driver includes libGL and libGLU files - on my debian machine I had to make fake packages to satisfy dependencies, otherwise a mesa (software) version of libGLU (and libGL) got installed. John McPherson
And you'll probably find the AGP bridge isn't either, which may be another cause of slowdown. Oh, and the XFree86 DRI stuff won't help at all either - the mainline X tree is *old* in this regard. If you're feeling brave, check out the current XFree86 snapshots - ATI have been contributing a *lot* of code, and I think they've got 2D acceleration for the ATI IGP chipsets in there now (SWMBO has a laptop with one of these, and she's asking me to do this...). If you're *really* lucky, the latest X snapshot will actually build!
I can't speak for the ATI support in XFree86. But we have a bunch of Intel 865 boards here that have integrated i830 video chips. Their 3D performance (for simple-ish stuff) is quite good using DRI/DRM. Though some of the GL screensavers tend to crash X. And with the i8x0 chipsets you're limited to 16bit colour if you want 3D.
This tells me that your app doesn't use GLX, so it's probably doing all the rendering using the machine's CPU, RAM, $deity knows what else. As the Duron 900 is the slowest machine you've got.
I'm still boggled at anything over VNC being the fastest, though. That's just bizarre.
The biggest problem historically with graphics has always been the bandwidth between video memory and main memory. Graphics are just plain big and eat the data bandwidth something chronic. So it is unsurprising that VNC is faster than using a video card. For 2 reasons: 1. Xvnc doesn't use the graphics hardware. It renders to an in main memory buffer. No data is travelling on the AGP or PCI buses. 2. The VNC protocol is a polling protocol that only refreshes the screen when necessary and only transmits what has changed. It is also asynchronous to the video updates. This means that only a fraction of the video updates are being sent over the ethernet wire. Regards -- Oliver Jones » Director » oliver.jones(a)deeperdesign.com » +64 (21) 41 2238 Deeper Design Limited » +64 (7) 377 3328 » www.deeperdesign.com
On Tue, 23 Sep 2003, Oliver Jones wrote:
I'm still boggled at anything over VNC being the fastest, though. That's just bizarre.
The biggest problem historically with graphics has always been the bandwidth between video memory and main memory. Graphics are just plain big and eat the data bandwidth something chronic. So it is unsurprising that VNC is faster than using a video card. For 2 reasons:
Good points both - I'd forgotten that VNC worked that way. In a previous life, I had to administer an NT-based backup system using VNC (over a Saturn cable modem link). You'd move the mouse and then go out to dinner while you waited for the update to come back. It was probably the backup server, though. Legato Networker works (as much as it ever could) on a big Alpha running Unix. It doesn't cope well with an NT box, even if that NT box has a CPU a million times faster. Cheers Richard -- Richard Stevenson Smith & Wesson: The original point and click interface
Richard Stevenson wrote:
Good points both - I'd forgotten that VNC worked that way. In a previous life, I had to administer an NT-based backup system using VNC (over a Saturn cable modem link). You'd move the mouse and then go out to dinner while you waited for the update to come back. It was probably the backup server, though. Legato Networker works (as much as it ever could) on a big Alpha running Unix. It doesn't cope well with an NT box, even if that NT box has a CPU a million times faster.
I've found that vnc works the best when the server is Xvnc and the viewer is on windows. Which is not surprising since that is what it was originally designed for. While the windows vncserver allows the machine to be usable it seems to not redraw properly a lot of the time for things like menus. This is mentioned in the docs. On the machine where I use Xvnc over the 100M lan it is just like being there. g -- Glenn Ramsey <glenn(a)componic.co.nz> 07 8627077 http://www.componic.co.nz
As an aside, TightVNC is much better than old vanilla vnc. You can do things like compress the tcp stream using zlib and use jpeg image compression. Quite nippy over ADSL. Regards On Tue, 2003-09-23 at 10:45, Richard Stevenson wrote:
On Tue, 23 Sep 2003, Oliver Jones wrote:
I'm still boggled at anything over VNC being the fastest, though. That's just bizarre.
The biggest problem historically with graphics has always been the bandwidth between video memory and main memory. Graphics are just plain big and eat the data bandwidth something chronic. So it is unsurprising that VNC is faster than using a video card. For 2 reasons:
Good points both - I'd forgotten that VNC worked that way. In a previous life, I had to administer an NT-based backup system using VNC (over a Saturn cable modem link). You'd move the mouse and then go out to dinner while you waited for the update to come back. It was probably the backup server, though. Legato Networker works (as much as it ever could) on a big Alpha running Unix. It doesn't cope well with an NT box, even if that NT box has a CPU a million times faster.
Cheers
Richard
-- Oliver Jones » Director » oliver.jones(a)deeperdesign.com » +64 (21) 41 2238 Deeper Design Limited » +64 (7) 377 3328 » www.deeperdesign.com
Oliver Jones wrote:
As an aside, TightVNC is much better than old vanilla vnc. You can do things like compress the tcp stream using zlib and use jpeg image compression. Quite nippy over ADSL.
The thing I like about the windows vnc viewer is that it does fullscreen and (this is important) allows you to easily swap back to your native desktop using a keypress (ctrl-esc). TightVNC has a fullscreen mode and it works under KDE3.1 but I couldn't figure out how to swap back to windowed mode once fullscreen mode was entered.(1) Anyway in X since you can run X apps remotely and most modern wm's allow multiple desktop then vnc is kind of redundant for X to X. g (1) whoops... the keypress is F8 ... just found out why I missed that, it's "man xtightvncviewer" not "man tightvnc" and on my Mandrake 9.1 man -k was broken at the time, something to do with the installer thinking I was in Korea. -- Glenn Ramsey <glenn(a)componic.co.nz> 07 8627077 http://www.componic.co.nz
On Tue, 2003-09-23 at 13:33, Glenn Ramsey wrote:
The thing I like about the windows vnc viewer is that it does fullscreen and (this is important) allows you to easily swap back to your native desktop using a keypress (ctrl-esc).
I hope you have tried Xfree inside cygwin.. its the greatest thing since sliced bread.. more stable than any other x server inside windows in my experience.. i use my linux server exported over wireless to my laptop all the time.. its just like using it locally but i can be anywhere in the neighbourhood.
kyle carter wrote:
On Tue, 2003-09-23 at 13:33, Glenn Ramsey wrote:
The thing I like about the windows vnc viewer is that it does fullscreen and (this is important) allows you to easily swap back to your native desktop using a keypress (ctrl-esc).
I hope you have tried Xfree inside cygwin.. its the greatest thing since sliced bread.. more stable than any other x server inside windows in my experience..
No I haven't tried it. When I get some time (RSN) I'll have a look. What would be the advantage to using that over vnc, given that currently I have no complaints about vnc and it does everything I need it to? g -- Glenn Ramsey <glenn(a)componic.co.nz> 07 8627077 http://www.componic.co.nz
Oliver Jones wrote:
I'm still boggled at anything over VNC being the fastest, though. That's just bizarre./
The biggest problem historically with graphics has always been the bandwidth between video memory and main memory. Graphics are just plain big and eat the data bandwidth something chronic. So it is unsurprising that VNC is faster than using a video card. For 2 reasons:
1. Xvnc doesn't use the graphics hardware. It renders to an in main memory buffer. No data is travelling on the AGP or PCI buses. 2. The VNC protocol is a polling protocol that only refreshes the screen when necessary and only transmits what has changed. It is also asynchronous to the video updates. This means that only a fraction of the video updates are being sent over the ethernet wire.
Running the app on the local xserver (xfree86 4.2.1) it appears to go just little bit faster than over vnc. The chipset in this machine (the one you sold to me!) is S3 ProSavage and it uses the the vesa driver. g -- Glenn Ramsey <glenn(a)componic.co.nz> 07 8627077 http://www.componic.co.nz
Running the app on the local xserver (xfree86 4.2.1) it appears to go just little bit faster than over vnc. The chipset in this machine (the one you sold to me!) is S3 ProSavage and it uses the the vesa driver.
Hmm. Vesa driver eh. That will be slow. No accel support I don't think. Best get a better video card. ;) Regards -- Oliver Jones » Director » oliver.jones(a)deeperdesign.com » +64 (21) 41 2238 Deeper Design Limited » +64 (7) 377 3328 » www.deeperdesign.com
Oliver Jones wrote:
/Running the app on the local xserver (xfree86 4.2.1) it appears to go just little bit faster than over vnc. The chipset in this machine (the one you sold to me!) is S3 ProSavage and it uses the the vesa driver./
Hmm. Vesa driver eh. That will be slow. No accel support I don't think. Best get a better video card. ;)
He he. But seriously that is the point, I have another machine that does have a better video card. Taking vnc out of the picture, this machine still runs the app the fastest even though it seems to have the least capable spec. g -- Glenn Ramsey <glenn(a)componic.co.nz> 07 8627077 http://www.componic.co.nz
participants (5)
-
Glenn Ramsey -
John R. McPherson -
kyle carter -
Oliver Jones -
Richard Stevenson