64bit support for nvidia and creative

The locking in NVidia's driver shouldn't matter too much. I don't think there is too much kernel activity other than the video card while you make heavy use 3D. It's not that disk or network are in heavy use.
 
nakal said:
I reached a bit "too high" and bought an RV630 based one. I'm still waiting for Xv support, but desktop works very well and as I read this thread, it seems it even works far better than with nv, which apparently cannot do full screen video playback. I have no problems with unaccelerated video playback, even at 1920x1200. :)
I'm in a similar situation. I picked an RV620 card. Fortunately full screen video playback works for me too, even h.264 encoded HD videos. I think if I didn't have an E8500 CPU things might be different. :)

It will be good when Xv and 3D support is available. I've read suggestions that March 2009 will be the month. I'm missing quake 3!
 
The DRM and DDX functionality will probably be present in the next week. The DRM may be linux specific at first but that hopefully won't last long. That would give you Xv and EXA support. March/April is the current rough estimate for 3D parity with the r500 driver.

Adam
 
nakal said:
I also have an ATI, because I'm almost convinced that NVidia does not really care about FreeBSD/amd64. And who wants an amd64 system crippled to i386?
NVidia does care.. Why would they even build an fbsd driver in the first place? As has been said over and over again, the fbsd kernel is missing some functionality NVidia needs to make a decent amd64 driver. Heck, even the ATI driver would benefit from that.

And why would my system be crippled? I've only got 2GB of RAM (plenty for fbsd), so what would I gain by running amd64 instead of i386? Besides bragging rights?
 
SirDice said:
NVidia does care.. Why would they even build an fbsd driver in the first place? As has been said over and over again, the fbsd kernel is missing some functionality NVidia needs to make a decent amd64 driver. Heck, even the ATI driver would benefit from that.

Interesting... Can you please explain that statement a little better? What functionality and how would the currently functioning FreeBSD/amd64 Mesa driver benefit from this currently missing functionality?

Adam
 
I'm familiar with that document. I'm curious what functionality listed there would help improve the open source Mesa drivers, and how it would help.

Adam
 
I thought I'd post about my recent success in getting DRI working with my onboard G33 graphics. Xv works, high resolution video playback no longer kills CPU and 3D is fast enough for Quake 3! I've disabled my ATI in the BIOS for now. You might consider doing the same!
 
cracauer@ said:
The locking in NVidia's driver shouldn't matter too much. I don't think there is too much kernel activity other than the video card while you make heavy use 3D. It's not that disk or network are in heavy use.
WTF? Almost any application displays something. Browsers, mail clients, terminals... that's done by the video card, too.
 
aragon said:
I think if I didn't have an E8500 CPU things might be different. :)

No, it's not! :) I can tell you that my E2200 CPU performs very decently. I have no complaints at all with video playback.

It will be good when Xv and 3D support is available. I've read suggestions that March 2009 will be the month. I'm missing quake 3!

I'm missing Unreal Tournament. :e

SirDice said:
As has been said over and over again, the fbsd kernel is missing some functionality NVidia

I wonder why it is possible to make an AMD/ATI driver then. Apparently, the radeonhd developers don't have any complaints about amd64.

And why would my system be crippled?

Why not use the full amd64 CPU instruction set? amd64 is working better with ZFS and if you have plenty of memory (which is pretty cheap at the moment btw). I have made some CPU speed tests and it seems in certain situations an amd64 system is almost twice as fast as i386 on the same hardware.
 
>I have made some CPU speed tests and it seems in certain situations an amd64 system is almost twice as fast as i386 on the same hardware.

Do you have some real-world tests too? I didn't see any of such improvments - maybe apart from a big database server with heavy load.
 
When I test the speed of CPUs, I try to test the CPU only. So it's not possible to compare it to real world scenarios. I remember, I've seen some drastic improvement while running openssl speed using certain crypto algorithms. I'm using encryption a lot.
 
nakal said:
I also have an ATI, because I'm almost convinced that NVidia does not really care about FreeBSD/amd64. And who wants an amd64 system crippled to i386?

What VGA card did you choose? I reached a bit "too high" and bought an RV630 based one. I'm still waiting for Xv support, but desktop works very well and as I read this thread, it seems it even works far better than with nv, which apparently cannot do full screen video playback. I have no problems with unaccelerated video playback, even at 1920x1200. :)

just RV630 family ?
I was planning to buy a HD 4850 (from newer family) for using in games (windows) and FreeBSD amd64.

As I read the main page of the radeonhd site, this ain't a good path., right ?

where can I find their roadmap (if any) ?

and what about the ati driver (read there about a driver-ati x driver-radeonhd issue) ?

thanks,

none
 
What I said above pretty much holds true for r600 and r700 cards:

The DRM and DDX functionality will probably be present in the next week. The DRM may be linux specific at first but that hopefully won't last long. That would give you Xv and EXA support. March/April is the current rough estimate for 3D parity with the r500 driver.

This is true for both the xf86-video-ati and xf86-video-radeonhd driver. At the present moment, 2D acceleration (without Xv) is present via shadowfb. Xv and 2D acceleration via EXA will be available as soon as the DRM and DDX are updated... The code for this is done, AMD is just waiting for approval which they hoped to have done by Christmas (and obvious missed by at least a few days).

Adam
 
Why not use the full amd64 CPU instruction set? amd64 is working better with ZFS and if you have plenty of memory (which is pretty cheap at the moment btw).
Memory may be cheap but that doesn't mean I will just go out and buy something I do not need. Besides, the mainboard I'm using can't handle more then 2GB. So should I rush out and buy a new mainboard too? Oh.. that would mean getting a new CPU and graphics card too.. Sure.. I'd love to but bills need to be payed too, you know.
 
Since you appear to be around now, SirDice, I'm still interested in knowing specifically which features that nvidia has requested would benefit the open source Mesa drivers and how they would benefit from them.

Adam
 
adamk said:
Since you appear to be around now, SirDice, I'm still interested in knowing specifically which features that nvidia has requested would benefit the open source Mesa drivers and how they would benefit from them.
I am by no means a kernel hacker, so I go on what someone else says. But I did read the original request for those features some time ago. Reading the reasons behind the need made sense to me. And I'm sure you've seen the 25 page thread on the nvnews forum :e

http://marc.info/?l=freebsd-hackers&m=115157983106569&w=2

http://www.nvnews.net/vbulletin/showthread.php?p=1620669#post1620669
 
SirDice said:
I am by no means a kernel hacker, so I go on what someone else says. But I did read the original request for those features some time ago. Reading the reasons behind the need made sense to me. And I'm sure you've seen the 25 page thread on the nvnews forum :e

http://marc.info/?l=freebsd-hackers&m=115157983106569&w=2

http://www.nvnews.net/vbulletin/showthread.php?p=1620669#post1620669

I'm certainly familiar with all the requests, and the various discussions about the requests. And while the requests seem to make sense, you actually said:

As has been said over and over again, the fbsd kernel is missing some functionality NVidia needs to make a decent amd64 driver. Heck, even the ATI driver would benefit from that.

That's what I'm looking for clarification on. I've had numerous discussions with the current DRI maintainer for FreeBSD, and he's never mentioned any of those items at all, much less in any way to suggest that the drivers would benefit from that functionality. So I'm curious how you, or this "someone else", came to that conclusion.

Adam
 
Your lack of a response makes me think that you really did not know what you were talking about when you said those changes would help the ATI driver. Is that the case?

Adam
 
vermaden said:
ATI has just released 3D documentation for R600/R700 chips (yes, the newest series also), so in longer term RadeonHD driver will support full 3D accelration: http://www.phoronix.com/scan.php?page=article&item=amd_r600_oss_3d

Errr. Please re-read the article. They did not release any r600/r700 documentation at this time. It clearly states:

There was about 180 pages of 3D register specifications set to be released, but all of the AMD officials didn't come to a consensus before leaving for the holidays.

What they did release was functioning code for the DRM (the kernel module) along with updates to the 2D driver to support EXA acceleration and Xv, along with an application designed to test 2D and 3D operations by communicating directly with the GPU.

The DRM code is linux specific at this point, though I saw rnoland talking with the ATI guys on IRC yesterday, discussing what changes would need to happen to the BSD DRM code.

Documentation will hopefully be following in the next month, with basic 3D functionality in Mesa in the near future, too. The last estimate I heard for parity with the r500 driver was March/April.

Adam
 
adamk said:
Errr. Please re-read the article. They did not release any r600/r700 documentation at this time.

... great :/

I was not interested in all reading so I read only "3D Code" and assumed that its whole 3D documentation, sorry for misinformation.
 
none said:
where can I find their roadmap (if any) ?

They don't publish any roadmaps, because they cannot guarantee anything.

and what about the ati driver (read there about a driver-ati x driver-radeonhd issue) ?

Simply said: there is none. The authors of radeon and ati drivers always argued with radeonhd developers. As far as I understand, this fight had its peak when radeonhd implemented the generic AtomBIOS support (AMD has always pushed the radeonhd developers to do so). This has made the radeon or ati developers angry because radeonhd began to support older cards out of the box and reached in their territory.

As a kind of revenge, one radeon/ati developer desperately deleted a line from a testing/integration script (that noone really uses), in hope radeonhd will not be considered as the official driver for Xorg. In my opinion, this is a really childish behavior, especially his argumentation and justification. On the other hand the radeonhd developers handled it very easy (they simply asked something like "what the hell are you doing?" and that's all). It makes them a lot more likeable somehow.

When you ask me, I use radeonhd, because it simply works better on my PCs. I need dual-screen pretty often and I like a functioning VT-switch, as I mentioned before.
 
nakal said:
They don't publish any roadmaps, because they cannot guarantee anything.



Simply said: there is none. The authors of radeon and ati drivers always argued with radeonhd developers. As far as I understand, this fight had its peak when radeonhd implemented the generic AtomBIOS support (AMD has always pushed the radeonhd developers to do so). This has made the radeon or ati developers angry because radeonhd began to support older cards out of the box and reached in their territory.

Errr.. You're a little misinformed here.

xf86-video-ati (aka, the 'radeon' driver in this case) supported AtomBIOS first. While 'radeonhd' was the first to support 2D in r500 cards (*not* through AtomBIOS) the same functionality was quickly added to the 'radeon' driver. Many folks still can not understand why the 'radeon' driver wasn't just modified by the Novell developers in the first place to support r500 cards instead of creating a whole new driver.

The developers of the 'radeon' driver then extended this support to use AtomBIOS where available. The radeonhd developers then did the same with their driver. In fact, the 'radeon' driver was the first one to support both 2D and 3D acceleration at the same time on r500 cards.

There was never any anger over radeonhd suddenly supporting older cards since, frankly, radeonhd didn't suddenly start supporting older cards. It is not, and always has been, limited to r500 and newer cards. It does not work on older radeons, at all, and there are no plans on having it do so.

As for me, I prefer the 'radeon' driver simply because I've been using it for years on older GPUs and it works fine. I use dual screen all the time, and regularly switch VTs without issue.

Adam
 
Back
Top