10.3: Stereo Driver Hooks & Final Thoughts

The last feature making its appearance in next month’s Catalyst 10.3 drivers will be the inclusion of some underlying hooks in the drivers for 3rd party 3D display hardware. AMD hasn’t been completely ignoring NVIDIA’s success with 3DVision, and while they’re not getting directly in to the 3D arena like NVIDIA has, they’re going to be providing the tools for 3rd parties to get in if they want to.

The big change here is that they’re going to provide driver hooks for 3rd party products to use to improve and simplify the operation of those products. One example AMD is throwing out is that their drivers will now be able to do quad buffering so that 3D products can double-buffer each eye separately. These hooks will also allow the hardware to output stereo images at 120Hz similar to how NVIDIA implements 3DVision, so that each eye can be offered images at 60Hz without needing to use a more esoteric solution such as iZ3D’s double-DVI setup.

Our expectation here is that with these new hooks a 3rd party will offer a 3DVision-like kit utilizing shutter glasses and a 120Hz monitor, although the quad buffer changes in particular are rather generic and can be used (not to mention necessary ) for any other form of 3D technology that takes off. Nothing has been announced yet, but it’s likely only a matter of time.

Final Thoughts

Although AMD is always working on the Catalyst drivers, major updates tend to come in spurts and this is a prime example of that case. With the 10.2 and 10.3 releases we are seeing the first post-launch driver drop for the Radeon 5000 series. Some of the things we’re seeing today such as Catalyst profiles and fully-functional Ultra Low Power State are things we would have liked to see at the 5000 series launch, while other things such as the new Eyefinity tweaks are going to be a nice addition to the existing capabilities of the hardware.

For the Catalyst profiles and Mobility driver support in particular, it’s going to be worth keeping an eye on how well AMD does in implementing these things. Profile support is simply a matter of being timely with new game releases, although we’re still going to ask for the ability to write our own profiles anyhow. As for the Mobility driver program, moving from an opt-in to an opt-out model should prove to make the program much more successful than AMD’s previous effort, but the one remaining wildcard is what OEMs have opted-out of the program, and with what products. AMD is only expecting business products to be opted-out, but as far as we know, any OEM can opt-out for any reason, so there’s still a remote chance of snatching defeat from the jaws of victory.

There’s also one thing we were hoping to see in the 10.2/10.3 drivers which has not come to pass, and that’s OpenCL support. AMD is continuing to only provide OpenCL runtime functionality through the installation of the Stream SDK, which means even though AMD’s 4000 and 5000 series hardware is capable of OpenCL, none of those cards can run those programs out of the box. AMD’s current reason is that they don’t want to expand the size of their drivers any further (they’re currently 123MB for the 10.3 betas) which is a valid concern looking at the size of their OpenCL runtime, but at the same time we can’t imagine this is good for OpenCL adoption in the long-run.

Who is going to develop applications using OpenCL if half your user base (not counting Intel IGP users) can’t run your application out of the box? Even worse, you currently need to sign up for an AMD Developer Central account before you can download the Stream SDK in order to get the runtime - and what user is going to do that? NVIDIA is way ahead of AMD here, having shipped OpenCL support in their drivers for several months now, and they’ve been able to do so while keeping their drivers at about the same size as AMD’s (let’s not forget the PhysX runtime either). Unless AMD expects everyone to go the DirectCompute route (in which case we can kiss cross-platform GPGPU usage goodbye) AMD’s GPGPU efforts are currently stuck in place.

Finally, from a testing perspective the 10.3 drivers are still in beta, but much like the Catalyst 9.12 hotfix was to the 10.2 driver launching today, the 10.3 driver set we’re working with has been shaping up rather well. AMD hasn’t told us when they’ll be launching besides the fact that it will be in March, but based on the drivers we’re seeing we wouldn’t be surprised if it was an early launch rather than a mid-month or later launch as AMD is common for AMD.

10.3: AMD’s New Mobility Driver Program
Comments Locked


View All Comments

  • bhaberle - Monday, March 8, 2010 - link

    It looks like ATI Catalyst 10.3 Beta leaked out. I am pretty happy with it. You can download it here ( http://www.geekmontage.com/ati-catalyst-10-3-beta-...">http://www.geekmontage.com/ati-catalyst-10-3-beta-... ) I see it on a few other sites too, just do a google search.
  • dvdreplication - Thursday, February 25, 2010 - link

    Well dana i have placed my order for it. Hope that it 'll be a very nice product. Thanks for sharing.

  • Scali - Monday, February 22, 2010 - link

    Is it just me, or did ATi not talk about OpenCL support at all?
    As far as I can tell, it's still not included in the 10.2 drivers, I still need to install the Stream SDK to get the actual OpenCL runtimes.
    This is not acceptable for distribution to end-users. They will just need to include the OpenCL runtimes in the driver package.
  • DanaGoyette - Saturday, February 20, 2010 - link

    Mobility drivers for HD-series graphics cards sounds great... but what about laptops that have the FireGL-branded cards, such as mine?
    For example, HP's latest Win7 driver for my laptop is 8.632 -- which is really old.
    My ideal case: mobility drivers that offer the professional features (such as "10-bit pixel format", though I don't have a display capable of using that) -- even if they're not ISV-certified.

    Also, ATI has had some form of quad-buffered stereo on the workstation cards for a while; it's cool to see it now expanded to cover consumer cards (and hopefully laptops, as well).
  • LordanSS - Friday, February 19, 2010 - link

    I had to, one more friggin time, revert back to 9.12 hotfix because the darned mouse cursor bug is back.

    Why the hell can't they get this straight? Geez.
  • Quidam67 - Thursday, February 18, 2010 - link

    Sorry if this is off-topic, but it seems like a place I might get soem help on this:

    Having recently returned to AMD with the purchase of a HD5770 I'm having problems with the latest drivers, specifically they just don't seem to (properly) recognise my Panasonic V series LCD. The result is there is no DVI or HDMI (have tried both) Item added to the root tree menu in CCC. This means that amoungst other things, I can't alter the scaling -which by default results in a desktop screen that is significantly shrunk in the middle of my display.

    Luckily for me I also own a Samsung LCD. CCC recognises the Samsung and I get the DVI tree menu, which then allows me to adjust the scaling to zero. This setting is "remembered" when I plug back into my Panasonic -but what a headache, and I assume I'll need to do this everytime I update to a new ccc/driverset.

    So is this a fault with the Panny and the EDID data it sends, or is this an issue with ATI's software? The ting is I tried a much older rig running a HD3850 AGP and 8.9 drivers, and they seemed to detect the TV fine.

  • Quidam67 - Friday, February 19, 2010 - link

    For what it is worth, the 10.2 drivers seem to have fixed the problem I referred to above. I know get a menu item titled +DVT(DVI)4 in ccc giving me the scaling options back again.
    Thank god for that. I can only assume that between driver releases 9 thru 10.1 they broke something
  • Bolas - Thursday, February 18, 2010 - link

    Last year I came up with a joke about this subject. Hope you like it.

    How is ATI like a VW commercial?

    They're both "drivers wanted".
  • Roland00 - Wednesday, February 17, 2010 - link

    Does this mean you can use 3 monitors without buying a Displayport monitor and an active adapter? For if you have 4 cards you will have 4 TMDS signal generators.
  • Roland00 - Wednesday, February 17, 2010 - link

    Meant to say 2 cards with 4 TMDS signal generators

Log in

Don't have an account? Sign up now