ATI support poor or just slow?

[Panix]

I have read that the FOSS driver isn't totally covered and it's not true 'open source' or extensive/comprehensive. Obviously, a lot of features aren't covered or supported. Therefore, I don't get why all the die-hards are so up in arms when users have to use FGLRX. I think the majority of people just want to use their card and the features that the card is able to do. In Windows, this is rarely a problem, even for ATI. The only concern might be a few bugs or issues but these are usually cleared up pretty fast. But, there's hardly ever complaints of tearing or not being able to use THREE monitors IN WINDOZE!

But, what is open source? Is it 100% free code or something else?

Just curious.

I'm waiting for Catalyst 10.7 to come out or reports from fellow ATI Evergreen owners to see if the improvements are enough for me.

I think the open source concept is good and ideal but I categorize myself in the group here that just want the features the card is able to do and to use it without glaring issues. I do a lot with video or intend to (so I want 'HTPC-type' features min.) and some minor glitches that might not bother someone else may bother me.

Anyway, I was just wondering about whether ATI can ever catch up to the Windows side and whether their FOSS driver will encompass more features ever or if both will be limited in Linux with constant catch-up status.

[Panix]

'Can't edit but wanted to add, when I say 'what is open source,' I'm wondering if it's limited to what's 'allowed' by the licensing. So, it's open to just an extent? Is the restriction totally tied to licensing?

[bridgman]

'Can't edit but wanted to add, when I say 'what is open source,' I'm wondering if it's limited to what's 'allowed' by the licensing. So, it's open to just an extent? Is the restriction totally tied to licensing?

Panix, I don't really understand your questions any more. What "restrictions" are you talking about here ?

Taking a guess that you mean "current differences in performance and features between open and closed drivers", licensing is completely orthogonal to implementation.

If you are talking about binary-only drivers vs open source as "licensing" then there is a small connection, in that HW vendors can usually make use of certain HW features in a binary driver which can't be used by open source drivers, but I don't think any of those differences contribute to the performance and feature differences you see today, with the obvious exception of UVD.

[gbeauche]

but opensource in the amd way means spec+driver thats because amd pay dev's to help the opensource people

Likewise for Intel. This means specs + drivers + even dedicated labs for Open Source projects and people working full time on it.

Besides, a lot of OSS fanboys cry hard for HW specs but when they are finally here, well, there is nobody more available to actually use them. Things based on ideology always turn bad... What could be useful is access to source code of binary drivers so that certain people don't have to wait for e.g. 8 months simply for a 10-chars fix... ;-)

UVD isn't a feature because hd2900 do not have one and deliver the same 'feature-set'

This is what you think because you don't exactly know or understand the extends of the features and the specifics of how this is implemented and what parts are not...

[Panix]

I'm talking about whether the open source drivers are developed with restrictions as part of the dynamics. For e.g., 3D is said to be lacking with the FOSS drivers. There's no 'official' FOSS driver support at all with Evergreen cards. So, the advantages with the FOSS driver aren't even evident with newer cards. ATI cards ARE NOT usable in Linux. At least, the newer ones. Features are lacking and there is only one driver available which is not optimized or fully-featured even though the card was released about a year ago.

The Workstation market is not my concern, I'm a consumer who wants to use the card I bought. I just find it peculiar that some of these OSS fans praise this side but cannot fully use the video card. That's fine if they're okay with it but don't pretend that situation doesn't exist.

I was asking whether licensing is one part of it why the development is slow. Not all aspects of the code are released? Does that clarify things?

[bridgman]

Panix, you're entitled to your own opinions on useability but *please* try to be more careful it comes to time (which is not really open to interpretation). You started saying months ago that Evergreen had been out for over a year, when in fact the very first card in the family had shipped maybe 6 months prior to your post.

The Evergreen family launched between October 09 and February 10, so we're talking 5-8 months at most.

XvBA has not yet been released, so whether it works on all GPUs equally is not really an issue yet (although we understand that some are making use of the unreleased code today and would like to do the same on Evergreen), so presumably you're talking about the fact that there are different bugs on different generations of hardware ? Is that what you mean by "optimized" ?

Licensing (ie choice of license) has essentially nothing to do with release of code. That's why I didn't understand your question and was asking for clarification earlier.

Do you mean "open source vs binary only" or are you actually talking about licensing ? Are you talking about open or binary drivers here ?

What do you mean by "dynamics" ?

What do you mean by "NOT useable in Linux" ? A lot of people use them every day.

[Panix]

Okay, if I wasn't totally accurate, my apologies. But, one year, six months... does it really matter? Honestly? The point is, many features are left not supported or working after the release of the card. I hate to compare to Windows but if I fire up my XP install, I could, in theory, use all features of the card. I know some sacrifices need to be made going to the Linux side but if I look at what Nvidia covers, it seems like more and less bugs/issues. The hardware is not as good but if I'm going with an ATI card, I want to be able to use the card for what I do.

I don't mind that there are two drivers to choose from. Actually, it's good and ideal to have the open source option. My question refer to whether the licensing hinders development or what extent of code is released or allowed into the driver.

Or is it just so complicated or technical that it takes that amount of time?

By optimized, I mean the feature set and to what extent, it is supported and working. WIP status and bugs/issues with respect to certain features equals 'not working' imho.

I don't like that I can get a weaker Nvidia card (9600GT, 9800GT etc.) and I can use it to the fullest feature set in either Windows or Linux without losing a step. I can use 2D, 3D, I have hardware acceleration and I can use the cute 3D graphics desktop effects without various issues. With comparable ATI cards or the newer Evergreen, this isn't possible. The features are in progress and bugs still remain.

I thought the FOSS driver development depends on what parts of the code are allowed into the mainstream or that it's just a complex and timeconsuming process....that is, if the personnel is sufficient.

As for the binary drivers, the reputation is that the driver development and optimizing support on various features is also slow and buggy. Again, the overall support is always behind on X Server and kernel development.

[bridgman]

Panix, you're really not being clear (or maybe it's just me ) :

My question refer to whether the licensing hinders development or what extent of code is released or allowed into the driver.

...

Or is it just so complicated or technical that it takes that amount of time?

Which driver are you talking about ?

By optimized, I mean the feature set and to what extent, it is supported and working. WIP status and bugs/issues with respect to certain features equals 'not working' imho.

With respect, if "bugs" = "not working" then nobody is shipping anything today

I don't like that I can get a weaker Nvidia card (9600GT, 9800GT etc.) and I can use it to the fullest feature set in either Windows or Linux without losing a step. I can use 2D, 3D, I have hardware acceleration and I can use the cute 3D graphics desktop effects without various issues. With comparable ATI cards or the newer Evergreen, this isn't possible. The features are in progress and bugs still remain.

What, specifically, can you not do with Evergreen or on the older cards ?

I thought the FOSS driver development depends on what parts of the code are allowed into the mainstream or that it's just a complex and timeconsuming process....that is, if the personnel is sufficient.

I don't understand your question. The pacing item for community FOSS driver development is usually availability of developers. Are you talking about features & functionality in general or the specifics of supporting new hardware, since each has a different process ? Supporting new hardware requires that we get drivers running using in house information so that we can generate documentation; adding features to already-supported GPUs is just a question of having enough developers in the community.

As for the binary drivers, the reputation is that the driver development and optimizing support on various features is also slow and buggy. Again, the overall support is always behind on X Server and kernel development.

Are you sure that's still the case ? If you look closely I think you'll see that has been improving gradually over the last couple of years. The installers & scripts are still targeted to the distros & versions we officially support, but in general I think the time between release of a new X / kernel version and fglrx support has been shrinking to the point where it's a non issue for most users.

Seriously, at some point you're actually going to have to try something yourself rather than trying to assemble broad generalities out of internet posts.

[bridgman]

Comparing to Windows doesn't really make a lot of sense, since it has ~50x the market share in the PC space.

Comparing to competitors is fine, as long as it's unbiased, but honestly I'm not seeing that in your posts. What I think you're saying is that you would like to cherry-pick the best aspects from each vendor and buy that product, and the fact you can't do that is somehow our fault

[gbeauche]

Or is it just so complicated or technical that it takes that amount of time?

I once read that fglrx source code was around 15 to 30M lines of code. Besides, most of that is common code shared with Windows. Linux support is "just" the glue. Do you really think that the FOSS driver will match the proprietary driver that fast, if ever, with all its features and performance? If it were, this would mean ATI's Linux and Windows teams were vastly incompetent for years... I don't want to break your hopes but, based on what I have just said, don't expect the FOSS driver to ever match fglrx features and performance.

And yes, this surely is complicated or technical. There were many people crying for HW specs so that they could use them. Well, how many people turned out to actually use those specs to write the driver? Not that many in the end... When you know how complicated it is to port Windows features to Linux by "just" writing glue code to them(*), you can imagine how complicated it is to write those features from scratch. (*) That counts in months to years, or some people just lied to some others.

I don't like that I can get a weaker Nvidia card (9600GT, 9800GT etc.) and I can use it to the fullest feature set in either Windows or Linux without losing a step.

If you know you can use the fullest feature set in either Windows or Linux with NVIDIA, then just buy one. At least, *you* have that choice... A weaker card (in your opinion) with working and efficient drivers is better than better HW with indecent SW stack. Please don't be a masochist if you know that.