Our Experience with UVD and R600
It's easy to point to the presentations and say, "look, this is slightly misleading." This doesn't get across the even more convincing interactions we had with AMD personnel over the course of reviewing the HD 2900 XT. Hopefully this will help to illuminate the issue further. I'll have to put a disclaimer on this and say that I don't have a photographic memory and I forgot to bring my diary to the press briefing. This isn't normally the kind of stuff I take notes on for product launches.
At the AMD press event in Tunis, Tunisia (we still can't figure out why we went to Africa to talk about a graphics card), myself and a couple other reviewers had the good fortune of being able to see HD Decode first hand. We watched a side by side comparison of an R6xx based part and a G8x based part with interesting results that mostly pointed out the long way Intervideo and CyberLink have to go in refining their player software. We were under the impression that this was R600, though clarification was not provided.
We talked about UVD and PureVideo, and we were asked if a video decode comparison would be made against NVIDIA hardware for our HD 2900 XT launch article. Upon replying in the affirmative, we got the impression that we should expect the R600 to outperform its competition (the 8800 GTS) in video decode tests. This doesn't seem as certain without UVD, but our marketing speak filters failed to pick up on anything out of the ordinary.
Moving on, we were initially told to expect a driver to enable video decode acceleration shortly before launch. AMD also provided us with a copy of PowerDVD which was said to support AVIVO HD. We did get a newer driver near launch, but in testing it we saw very bad video decode performance. It turns out that we couldn't test video decode at all, as hardware acceleration was not enabled in our press driver. This indicates to us that, even though we saw decode acceleration working in Tunis, AMD was having more trouble than expected in the QA department.
We would expect QA issue with new UVD hardware. If we've got the same video decode hardware as a previous generation part, it seems a little odd to think that we wouldn't have a driver or software to enable the feature.
In going back through our slides for this piece, we noticed that it is explicitly stated that UVD is included on the Radeon Mobility HD 2300. This is an R5xx based part retrofitted with UVD. Of all the parts not to have this new feature, the DX9 class Mobility HD 2300 would have been the one we thought of first. Surely if this part has the new feature, the HD 2900 XT should also include it. The argument for not is, again, that higher end GPUs tend to be paired with CPUs that can handle the workload.
The reason we offer time and again for the fact that G86 and G84 include full video decode while G80 does not is that the slower parts had 6 months more development time. In NVIDIA's case, it makes complete sense that G80 would lack the more refined video decode hardware. For AMD to launch a top to bottom product line and leave a feature like this out of their highest end part is very strange.
So strange, in fact, that combined with all the other experiences we've had with this rather surreal product launch we still just can't help but doubt what we're hearing. To us, the most sensible explanation for all this is that the hardware must be there but is somehow broken - perhaps broken to the point that it will never be fixed on current silicon (i.e. via drivers) and a new spin would be required to address the problems.
In spite of what makes sense to us, AMD has indicated that UVD hardware is not at all present on R600. Their "official" statement has yet to appear, but we'll update this article when/if we get it.
It's easy to point to the presentations and say, "look, this is slightly misleading." This doesn't get across the even more convincing interactions we had with AMD personnel over the course of reviewing the HD 2900 XT. Hopefully this will help to illuminate the issue further. I'll have to put a disclaimer on this and say that I don't have a photographic memory and I forgot to bring my diary to the press briefing. This isn't normally the kind of stuff I take notes on for product launches.
At the AMD press event in Tunis, Tunisia (we still can't figure out why we went to Africa to talk about a graphics card), myself and a couple other reviewers had the good fortune of being able to see HD Decode first hand. We watched a side by side comparison of an R6xx based part and a G8x based part with interesting results that mostly pointed out the long way Intervideo and CyberLink have to go in refining their player software. We were under the impression that this was R600, though clarification was not provided.
We talked about UVD and PureVideo, and we were asked if a video decode comparison would be made against NVIDIA hardware for our HD 2900 XT launch article. Upon replying in the affirmative, we got the impression that we should expect the R600 to outperform its competition (the 8800 GTS) in video decode tests. This doesn't seem as certain without UVD, but our marketing speak filters failed to pick up on anything out of the ordinary.
Moving on, we were initially told to expect a driver to enable video decode acceleration shortly before launch. AMD also provided us with a copy of PowerDVD which was said to support AVIVO HD. We did get a newer driver near launch, but in testing it we saw very bad video decode performance. It turns out that we couldn't test video decode at all, as hardware acceleration was not enabled in our press driver. This indicates to us that, even though we saw decode acceleration working in Tunis, AMD was having more trouble than expected in the QA department.
We would expect QA issue with new UVD hardware. If we've got the same video decode hardware as a previous generation part, it seems a little odd to think that we wouldn't have a driver or software to enable the feature.
In going back through our slides for this piece, we noticed that it is explicitly stated that UVD is included on the Radeon Mobility HD 2300. This is an R5xx based part retrofitted with UVD. Of all the parts not to have this new feature, the DX9 class Mobility HD 2300 would have been the one we thought of first. Surely if this part has the new feature, the HD 2900 XT should also include it. The argument for not is, again, that higher end GPUs tend to be paired with CPUs that can handle the workload.
The reason we offer time and again for the fact that G86 and G84 include full video decode while G80 does not is that the slower parts had 6 months more development time. In NVIDIA's case, it makes complete sense that G80 would lack the more refined video decode hardware. For AMD to launch a top to bottom product line and leave a feature like this out of their highest end part is very strange.
So strange, in fact, that combined with all the other experiences we've had with this rather surreal product launch we still just can't help but doubt what we're hearing. To us, the most sensible explanation for all this is that the hardware must be there but is somehow broken - perhaps broken to the point that it will never be fixed on current silicon (i.e. via drivers) and a new spin would be required to address the problems.
In spite of what makes sense to us, AMD has indicated that UVD hardware is not at all present on R600. Their "official" statement has yet to appear, but we'll update this article when/if we get it.
Facebook
Twitter
RSS
53 Comments
View All Comments
TheOtherRizzo - Wednesday, June 6, 2007 - link
Thanks a lot for looking into it!I hope that you something comes out of this and that you get to do some tests of your own. Obviously I haven't had the chance to test every card with every driver so it would be possible that the "All features on All AGP cards" claim turns out to be innacurate but several forums indicate it's quite probable.
The card that was personally giving me grief was a 6600 GT with all the drivers that came out until february 2007 at which point I switched to ATI.
Cheers and get well soon!
DerekWilson - Tuesday, June 12, 2007 - link
I got a response from NVIDIA and asked for permission to reproduce it here.Here's what we've got on the AGP issue from them ...
-----
Hi Derek,
It is not true that NVIDIA's current drivers disable support of PureVideo in GeForce 6 and 7 series AGP cards.
Our Windows XP drivers for those AGP cards support hardware accelerated decoding of H.264 and MPEG2 HD content and hardware accelerated decoding of H.264, MPEG2 and WMV9/VC-1 SDcontent.
Our Vista drivers drivers for AGP cards support hardware accelerated decoding of H.264, MPEG2 and WMV9/VC-1 HD content and hardware accelerated decoding of H.264, MPEG2 and WMV9/VC-1 SD content.
It is true that over a year ago we had to disable Windows XP support of HD content in the VC-1/WMV9 format on AGP cards due to AGP memory latency. Too much residual data to read from AGP buffers. At the time we made this change to the drivers, we posted a new purevideo_support web page to show this change. Unfortunately at some point since then, the old web page has been re-posted and we are incorrectly showing that we support HD content in the VC-1/WMV9 format on AGP cards. This will be corrected with the posting of a new spreadsheet on Tuesday, June 11.
By the way, have you heard of a web site called www.waybackmachine.org ? It is an internet archive site that keeps track of old versions of web site pages. You can use it to verify that we did indeed change our web pages back on April 23, 2006 to delete WMV9/VC-1 from the AGP card section of the purevideo support table. Just paste the link http://www.nvidia.com/page/purevideo_support.htmli...">http://www.nvidia.com/page/purevideo_support.htmli... their search bar and press "Take Me Back"
I hope this helps and thanks for letting us know about our inaccurate web page.
Best Regards,
Scott V.
TheOtherRizzo - Wednesday, June 13, 2007 - link
They're trying to make it sound less severe than it is. Hardware acceleration is for HD and that's what doesn't work. And of course they didn't know about the mistake, there are only several threads at their own official forums...Sadly I no longer have an appropriate AGP card so I can't test with their newest drivers, but the real show stopper for me was that HW deinterlacing (called "Spatial-Temporal De-Interlacing" in their spreadsheet) didn't work from 9x.xx onwards, SD or HD. This made the card absolutely useless for watching DVB streams. If you are interested in testing have a look here: http://www.avsforum.com/avs-vb/showpost.php?p=8358...">http://www.avsforum.com/avs-vb/showpost.php?p=8358...
Well anyway, nice to see your email had an effect, thx for that.
If you are further interested in the subject of video decoding another thing to look into is why only the 8500/8600 cards perform deblocking on h264 files. Sample clip here: http://x264.nl/h.264.samples/force.php?file=./hd.f...">http://x264.nl/h.264.samples/force.php?file=./hd.f...
and screenshots with dxva on and off here:
http://www1.filehost.to/files/2007-06-14_01/031049...">http://www1.filehost.to/files/2007-06-14_01/031049...
http://www1.filehost.to/files/2007-06-14_01/031059...">http://www1.filehost.to/files/2007-06-14_01/031059...
Cheers
Chunga29 - Monday, June 4, 2007 - link
Probably because they learned their lesson with PureVideo? That was the first time such a feature was heavily touted, and the original version failed to materialize on 6800 cards. This is recent, not something from 2+ years ago. As much as it sucks for AGP users and the 6800 users to not have PureVideo, I can't say I'm too concerned about an in-depth look at old news. This is news for today, and I'm pretty sure if NVIDIA launched a new product today and failed to deliver on a new feature (like PureVideo HD on 8500/8600 cards) that we'd see similar complaints.I'm still wonder WTF ATI/AMD demoed in Tunisia last month. Did they have H.264 on 2400/2600 (which still aren't shipping)? Did they have H.264 on 2900 but it only worked some of the time, and they eventually gave up? Certainly they showed SOMETHING to AnandTech, Tech Report, etc. and that's the real problem. If they hadn't made a point of promoting a feature that appears to be missing, no one would have cared as much.
TheOtherRizzo - Monday, June 4, 2007 - link
I wasn't talking about the broken 6800 AGP cards. This would be old news and Nvidia no longer claims that they support all Purevideo features. I'm talking about current Nvidia advertising (http://www.nvidia.com/page/purevideo_support.html)">http://www.nvidia.com/page/purevideo_support.html) that claims that all of the newer AGP cards have working Purevideo. In reality all Purevideo features were disabled from 9x.xx drivers onwards, even on cards where it used to work. That means no MPEG-2, no h264, no wmv9, no vc-1 and no hardware deinterlacing for all AGP cards.That's not old news, that's lying. Many products on that list are currently still in retail.
Chunga29 - Tuesday, June 5, 2007 - link
Interesting.. I haven't used an AGP card myself for nearly two years now, and I can't remember the last serious look Anandtech gave to the technology. On the one hand, the solution is simple: don't update AGP drivers. On the other hand, I'm not sure why NVIDIA would even bother to remove PV support from AGP cards. Was it buggy? Maybe just an oversight (not likely)? You might try emailing the Anandtech people directly - sometimes they just don't see these comments I'd wager. Derek seems to be the GPU guy with the means to tlak to NVIDIA or AMD/ATI or whoever.TheOtherRizzo - Wednesday, June 6, 2007 - link
Yes, it was buggy in several ways. For example wmv9 support was removed for 6600 GT in 8x.xx drivers because there were audio synch issues with some files (the driver release notes said the synch problem had been "fixed"). h264 decoding on 6600 GT had heavy artifacting with most drivers.And it isn't as simple as not upgrading the drivers: VC-1 decoding wasn't introduced by Nvidia (on any card) until one of the later 9x.xx drivers. So going back to 8x.xx won't help you get VC-1 decoding. It simply never existed for AGP.
I've followed your advice to contact AT per email and have sent Derek Wilson a copy of this thread.
JKing76 - Monday, June 4, 2007 - link
C'mon guys, basic grammar here. A couple other reviewers and I had the good fortune...
Acert93 - Monday, June 4, 2007 - link
I am curious the NV40 debacle with Purevideo encoding/decoding was never really followed up on by the press. That was a pretty major issue of an advertised feature that didn't work.Frumious1 - Monday, June 4, 2007 - link
Probably because it was going to be fixed in a driver update for so long that by the time anyone finally figured out it was broken on 6800 hardware it was almost too late to say much. I remember AT's 6600 articles pointed out that PureVideo was finally working and did complain about the way the situation was handled. AMD would have been better off following that pattern... except UVD apparently does work on 2400 and 2600 cards.Still can't figure out how they screwed the pooch on this one. Is the hardware just malfunctioning on R600 (but not on the other chips), or did they intentionally mislead here? If the hardware is broken, it's a bit odd that it works at all on the other 65nm cards; if the hardware wasn't ever there on R600, why in the hell did AMD even start with their vague marketing? (Don't tell me that manufacturers put UVD on 2900 boxes without actually thinking it was there!)