why does fglrx suck so hard?

[mugginz]

I'm doing the same. I hope it's not the same old, same old.

Yep. All we can do is wait as long as possible. I've been putting off the new card since November though so depending on what happens at the end of next month I'll buy something then.

I also think you are both wrong in a way. I think Linux users are resigned to the fact they have to do a few things, the 'manual' way if they want things optimized.

I agree that most users are resigned to the fact that they have to install the drivers, it's just that people would rather they just magically work.

Some push the advantage of the OSS drivers in relation to automatic out of the box hardware support, but that support is sub-standard when compared to the proprietary drivers. Things are getting better, but they're still a long way off for some use cases. Those that can live with what the OSS drivers provide have it easier than those who need what the blobs give.

Although, fortunately, there are a few "shortcuts" for installing, for e.g., the Nvidia driver. There's a few script methods, I think, so that is as good as it gets for an easier install. I've wrestled with the Nvidia install but as long as there is a good script option, I'll take that inconvenience over a buggy driver and driver issues preventing me from using my card.

Ubuntu provides a really good driver install experience out of the box which works for most people as long as you're happy with the versions they provide. Then there are the PPA's which make access to the latest drivers also pretty easy. I've found the standard installer to be pretty fool proof though for the last few years but I've been using one of their preffered distros so this should be the case.

In other words, I'm constantly trying to re-install, update or downgrade drivers or packages to avoid a video problem or various issues impacting me using the full extent of my OS, then it's hard to justify going that route. I read that, simply, ATI > NVidia for hardware and Nvidia > ATI for drivers and that's a bad sitution for Linux right now, I think. You can't get the best of either world in any way.

I'm in the same boat. Even though I'd like to support AMD with my wallet in addition to the fact that I think they probably have the best hardware at the moment I'm just not prepared to go backward with regards to stability and features. No matter what the hardware provides, if the driver can't make it available to the user in a consistent and reliable way then it might as well not exist in the first place.

I've been using nVidia with Linux for about 8 years now and have always found them to be the best from a driver stand point. Not always perfect, just better than the others. If AMD don't come through in April at least there's a reliable alternative that'll provide most of what I want.

[cutterjohn]

...all I can say is that 10.3 better be DAMNED good since they opted to NOT release it this week...

Hoping to see MAJOR crap cleaned up in both linux AND windoze versions of this driver, but obviously not very optimistic given AMD's [s]poor[/]s] non-existent legacy support...

[mugginz]

...all I can say is that 10.3 better be DAMNED good since they opted to NOT release it this week...

Hoping to see MAJOR crap cleaned up in both linux AND windoze versions of this driver, but obviously not very optimistic given AMD's [s]poor[/]s] non-existent legacy support...

Indeed.

I suppose AMD are aware that there is a point in time where people will loose patience in the end-user space as apposed to the workstation space. I just hope they can put the right amount of resources in place in order to come up with the goods before the patience runs out.

Given AMD's position that they consider their Linux market opportunity to be the workstation market I'm not sure I think they see a major imperative in getting consumer level support for fglrx up to scratch any time soon. Without access to their balance sheet for estimated Linux sales I guess I'm not in a position to judge just how much manpower they should be throwing at fglrx, just that it doesn't seem to be enough for me. Luckily I don't need to know the details of that in order to select a hardware vendor. I only need to know what's available now and in the very near future.

I've decided to take a wait and see approach, and I'm not prepared to purchase hardware before reliable software support is actually available because I've seen too many others watch their modern ATI hardware go stale and become less relevant before proper drivers are delivered.

[bridgman]

I think you would want the income statement rather than the balance sheet, but the core issue here is that determining which sales go into Linux is very difficult, with the exception of OEM preload scenarios (where the OEMs can tell us how many systems were ordered with Linux vs other OSes).

Like everyone in the business, we had to make estimates based on a number of data sources including end user feedback, web site hits, distro downloads, driver downloads and various industry estimates. They all tended to come out in about the same range - just over 1% - but we have been allocating development resources assuming a bit over twice that (more like 3x recently) in the belief that Linux client PC market share will grow over time.

If we were to allocate manpower based on Linux sales as you suggest, we would need to move between 1/2 and 2/3 of our Linux development team to work on other OSes. I don't think that's what you should be asking for.

[wagner17]

I think part of the problem may be that's it hard to estimate just Linux sales. I had to upgrade my hardware recently and one of the choices I had to make was between Dell's Linux laptop that came with an Nvdia card or getting a Windows laptop with an ATI card. I ended up getting the one with the ATI card 'cause that was more important to me but I felt like I was traitor.

[mugginz]

Like everyone in the business, we had to make estimates based on a number of data sources including end user feedback, web site hits, distro downloads, driver downloads and various industry estimates.

They all tended to come out in about the same range - just over 1% -
but we have been allocating development resources assuming a bit over twice that (more like 3x recently) in the belief that Linux client PC market share will grow over time.

I think there is growing awareness that AMD are throwing money at the problem and deserve some extra consideration for this. From a potential customer position though the hardware/software combo still has to deliver. If there was simply a frames per second difference and AMD were a little behind I could certainly overlook that.

If we were to allocate manpower based on Linux sales as you suggest, we would need to move between 1/2 and 2/3 of our Linux development team to work on other OSes. I don't think that's what you should be asking for.

No I'm not suggesting they follow the formula you present there. What I meant when I said

Without access to their balance sheet for estimated Linux sales I guess I'm not in a position to judge just how much manpower they should be throwing at fglrx, just that it doesn't seem to be enough for me

what I was getting at is that neither I nor anyone else is in any position to demand more of AMD, but that any purchasing decision I may make doesn't require that level of knowledge of the internal processes. The only types of fact I need to include in the decision making process are the fitness for purpose, performance per dollar, value for money type ones and given that fglrx is as important to the outcome as the hardware itself, if it has significant issues for a particular use case it takes the whole item off the table.

In summary:

• I want to buy AMD
• I can certainly overlook minor speed issues
• I'm not entitled to demand more of AMD
• Fglrx needs to make the hardware available such that it's suitable for use for a modern Linux desktop with all that entails.

My quip, "just that it doesn't seem to be enough for me" is really just an acknowledgement that no matter what they're doing, the driver isn't currently in a form that provides what I need, but I'm hoping that it will soon, and if it does before the end of April then AMD will at least have my money.

[Cuppa-Chino]

I second the last writer, I have just ordered to laptops with hd5650 mobility cards - I was very very tempted to go with nvidia for ease of use, but decided to ride my luck that hd5650 and newer xorg's will get support soon.

Please have it ready before my wife gets here laptop -- explaining to her that xyz that worked on her old desktop does not work is going to dent WAF for linux again...

[Kano]

It seems you know more about ati oss drivers than bridgman

[bridgman]

I actually agree with Q this time, in the sense that the next generation of parts should come out about the same time that the open source driver development has "caught up" with new hardware generations so the devs won't be splitting time between supporting new parts and adding core functionality to existing parts.

On the other hand, I don't expect the next generation to be any more "open source friendly" than the current ones overall.

[wagner17]

I second the last writer, I have just ordered to laptops with hd5650 mobility cards - I was very very tempted to go with nvidia for ease of use, but decided to ride my luck that hd5650 and newer xorg's will get support soon.

Please have it ready before my wife gets here laptop -- explaining to her that xyz that worked on her old desktop does not work is going to dent WAF for linux again...

Non-technical spouses, those poor guinea pigs in our Linux experiments. I am not gonna lie, the first time I made my wife switch to Linux I got her an Nvidia card because I didn't want to hear her complain about the ATI drivers. It wasn't until this year that I finally felt comfortable enough to get her a machine with an ATI card.