Open ATI R600/700 3D Graphics For Christmas?

[Louise]

I am just getting a crazy idea here.

Would it make life easier if a hyper visor was implemented in the GPU, so all API calls goes through the hypervisor first? I.e. only the closed source driver can access the DRM stuff.

On the Playstation3 only a subset of the hardware is available for Linux.

I don't expect anyone from AMD to answer this, as that could probably get them in trouble. But for everyone else. Couldn't that be done?

[Ex-Cyber]

I am just getting a crazy idea here.

Would it make life easier if a hyper visor was implemented in the GPU, so all API calls goes through the hypervisor first? I.e. only the closed source driver can access the DRM stuff.

How does a hypervisor inside the GPU identify the closed-source driver in a way that can't be faked?

On the Playstation3 only a subset of the hardware is available for Linux.

The difference on PS3 isn't open vs. closed-source drivers, it's "Game OS" mode (which permits no user-controlled code) vs. "Other OS" mode (which permits no GPU access).

[mityukov]

Worst case is that we lose the ability to sell our products into the Windows market as a result of releasing info which results in our DRM implementation no longer being considered sufficiently robust.

Please help me to get what you're saying here (English is not my native language; but, maybe, I'm just "not in the boat")..

How can Windows users care about "to whom else(except of AMD) it's known, how your hardware works technically"?

[Gimzo]

How can Windows users care about "to whom else(except of AMD) it's known, how your hardware works technically"?

It's not that windows users care (most of them doesn't care at all). It is this: If Ati releases any information that could help somebody break HDCP (or at least media companies come to think that), they revoke Ati's HDCP key. That means no Ati GPU can be used to watch "protected content" in windows, which means those windows users (who don't care) simply won't buy Ati hardware because of that.

[Louise]

How does a hypervisor inside the GPU identify the closed-source driver in a way that can't be faked?

Until now the PS3's security system is not well understood by the community, but the XBox 360's is, and it have been proven secure.

So the solution would be to the following:

+---------+        +---------------------------+
|         |        |    closed source driver   |
|  GPU    |        |                           |
|         |        | +-----------+             |
+---------+        | | signature |             |
|  hash   | -----> | | checker   |             |
| checker |        | +-----------+             |
+---------+        +---------------------------+

Inside the GPU, AMD puts a small mask rom, 4k is enough for everyone , which have a real SHA or RC5 hash check routine. This checks the hash of a code block in the driver. Because it is a hash, this code block can never be changed.

The code block contains a real signature checker, that can verify the rest of the driver, that it have not been altered. If it passes, it sends a challenge to the GPU, which it returns with a new challenge to the driver. If the driver passes, opens the GPU for the DRM API.

The difference on PS3 isn't open vs. closed-source drivers, it's "Game OS" mode (which permits no user-controlled code) vs. "Other OS" mode (which permits no GPU access).

That's correct, but it does so through a hyper visor. All instructions are passed though the hyper visor, which is thin layer between the hardware and kernel mode.

The hyper visor controls what instructions can be executed, and in the case of the PS3, the 2D/3D acceleration instructions can't be accessed, because the executeable that starts Linux (Other OS) isn't signed. The hyper visor checks the signature on the executeable.

Besides disabling the 2D/3D stuff, it also disables the last SPE and access to the game partition on the hard drive.

[Chewi]

Ach, I do blame MS for this mess. Like this infamous document says, they could have told Hollywood to get stuffed but they needed Hollywood more than Hollywood needed them.

To AMD, thanks for sticking your neck out, guys.

[mityukov]

It's not that windows users care (most of them doesn't care at all). It is this: If Ati releases any information that could help somebody break HDCP (or at least media companies come to think that), they revoke Ati's HDCP key. That means no Ati GPU can be used to watch "protected content" in windows, which means those windows users (who don't care) simply won't buy Ati hardware because of that.

This makes some sense, thank you.

[bulletxt]

"We can't imagine it being delayed much longer than that"

well after all these months, 1 month more or 1 month less, doesn't change anything. If they release it at december or march 2009, my amd card got old of 1 year but when I payed the card, the seller wanted the money immediatly and not 1 year after so lets try to be realistic yeah?
I don't like these articles. you put "finally" with a "!" for amd/ati articles, instead when NVIDIA does something really good with their drivers, you write it as if it was normal, the tone is completly different.just look at previous articles for reference or go even back for more proof. I'm sure it's not needed to do a copy paste, if you're smart you'll understand the tone by reading the articles.
This isn't correct and clearly shows how you prefer AMD over Nvidia.

Just imagine what kind of article could phoronix write if, instead of NVIDIA doing that thing with video bringing cpu to less than 5%, it was AMD. they'll do a party through streaming.

I'm fed up of these articles that remind me that amd isn't sleeping. it's worse than a blog that tells me every two months that the new version of windows will be more secure.

and for any thouchy person that feels bad reading my post, he must know that I'm an ATI/AMD fan and own and owned most of the times ATI cards. But patience has a limit, even for a fanboy, or at least a smart one. I hope you read this AMD.

[rvdboom]

You realize, of course, that we're talking about a **free** driver, not a binary one. In the later case, well you're not in a worst situation with AMD than with NVIDIA, as both provide a proprietary driver which supports all cards and works reasonably well.
But as far as an open, free driver is concerned (and I'm personnaly interested in just that), AMD has a much better policy than Nvidia and should allow us to have a correctly performing, accelerated free driver in a close future. Intel has also a good support, but its 3D chips are not big performers and run only on Intel platforms.
That is worth an exclamation mark and that's why, in the current state of affaires, there's a good reason to "prefer" AMD (or Intel for what is worth).
You may not agree but you can understand.

[Pfanne]

wouldnt it be possible to cut all the drm shit off the gpu by using only software acceleration via opencl!? much like the accelreation written for gallium3D.
you'll have more or less the same speed but no extra stuff in the chip itself...
well i think i am talking about something new here and the engineers working for amd should be (hopefully ) more clever than me!