False information in the articles: No UVD in R600

[XeRo]

Already posted. Again the issue is some manufacturers state UVD on their box. If it doesn't have it then it's false advertising.

[cadaveca]

y would it be AMDs fault?
AMD isnt responsible for the packaging of its partners.

IF and only IF AMD releases any 2900 cards by itself like build by ATI and mentions on its packaging that UVD is init then AMD can be held responsible for the cards that they sold but not over the cards its partners are selling.

Actually, partner's could produce documents forwarded to them that state UVD should be on the box...i don't think it's likely, but it's almost always possible to "pass the buck". In this instance, then AMD would be at fault. Review sites got thier info from somewhere after all, and under NDA, too. So, this depends on who holds the NDA...reviewers and AMD, or reviews and board partners...irregradless, this is not something that would be questioned in consumer rights...I'd not file suit against AMD anyway, HIS is obviously at fault. I have Sapphire box that does not state this on it, and the first card in my hands was an HIS...Sapphire card was bought due to it's packaging as well, as the "3DMark06 inside" logo. Again, this is because of packaging...

[DeathReborn]

And it does. It has full HD decode. It has good image quality. It has everything that UVD was created for.

The only diference is that it is integrated in the grafic core, and is not in an external chip.
The external chip was created for chips (HD 2600/HD 2400) that doesn´t have suficient raw power to do that kind off work.

You only have to ensure that all the funcionality´s of UVD are present in the product. In reality they are.
The card get 128 points out of 130 point in quality tests and it´s performance in HD decoding is even better then an 8800.

So is the UVD dedicated silicon (on the GPU) or is it just using the shaders to do the work?

[v_rr]

So is the UVD dedicated silicon (on the GPU) or is it just using the shaders to do the work?

The only thing that user as to test is if the performance is as good as it is promised -> Performance is very good
The quality of image is as good as it is promised -> it is
It as all features of that are promised -> it has

So having the external chip or have it integrated on GPU or have the work done by the shaders is exactly the same. The user only have to ensure that everything is OK in performance/quality/features.

[cadaveca]

except for the cpu usage... I myself cannot get over that one. UVD=low cpu usage. 2900 decode=high cpu usage. End result is not the same.

[SnipingWaste]

except for the cpu usage... I myself cannot get over that one. UVD=low cpu usage. 2900 decode=high cpu usage. End result is not the same.

What do you consider high CPU usage. From the few site that tested HD playback at 1080P the CPU usage was around 25% to 30%. I have a felling the HD2900XT has a decoding assist to take some of the work off the CPU because other video cards with out any decoding help, the CPU usage is up in the 80% range.

[v_rr]

except for the cpu usage... I myself cannot get over that one. UVD=low cpu usage. 2900 decode=high cpu usage. End result is not the same.

On those tests I see 22% and 16% of average CPU usage. If that is high CPU usage, the guys with a 8800 are just in the mud. And well... I don´t see anyone with 8800 complayning about their high CPU usage.

[cadaveca]

lol. my tests('cause i got the cards) show much higher usage. Personally, i think drverheaven has a driver that I do not. you are forgetting about all those sites sampsa posted that are not able to get decode working...driverheaven is just one site.

now, while I agree, there is SOME work being done, let's take AMD's own statements about how UVD works...

ATI Avivo HD technology is designed for high-end, high definition video playback in home theater systems. Free your CPU to power other applications with new UVD (Unified Video Decoder) hardware processing of HD video formats.

http://ati.amd.com/technology/avivo/features.html

Notice this falls under the AVIVO features.

But now for the damnation, the very next paragraph:

UVD is a dedicated video decode processing unit on ATI Radeon HD 2000 series graphics processors that offloads both CPU and GPU rendering pipelines. UVD technology reduces power use, decreases system noise and increases notebook battery life during HD video playback.

now, this translates as hardware decode, sure, but UVD is a specific type of hardware decoding.

Dave Beaumann, ATI marketing man, states quite simply, there is no hardware decode on HD2900 at all, and it is all done on cpu.

So who's right? Websites paid for advertizing, prone to mistakes, or an high-level exec from ATI?

My cpu usage numbers say Dave is. Driverheaven says differently, but then again, they are a "bought and paid for" marketing site, and I have the cards right here and now. I believe the numbers my pc puts out, and not some numbers on a website that have been verified by noone.

So...

The UVD is designed to offload the entire decode process for H.264, H.263 and VC-1 motion video, at the maximum bitrates defined by the DVD, Blu-Ray and HD-DVD controlling groups, so that every SKU of this new generation of Radeon could perform (almost) full CPU offload acceleration for video playback for all three current optical disc formats.




The five major stages of the UVD are shown above, with the entropy decode supporting full sped CABAC and CAVLC for H.264, and VC-1's entropy encode scheme, entirely in silicon. Compressed video data is fed in to the unit at the front end of the block, and then that's passed out for post processing by other hardware solutions on-chip, which differ depending on the GPU in the family.

this stipulates that the work is done in a dedicated hardware unit, but on R600, it's done in shader, which we can hardly call dedicated. Notice that "offload the entire decode process"...25-35% is hardly offloading the entire workload

Problem at hand is that the current driver does not support ANY work offload, so the 70-80% cpu usage I'm getting clearly shows no hardware decoding AT ALL, no UVD, no AVIVO, jsut the cpu doing it's work.

So, although the decode IS happening, AND it's happening quite well, this is not something that me, the end consumer, can replicate. This "shader decode" falls under AVIVO, but it does not fall under UVD...and this is where our trains of thought diverge.

You call it UVD...AMD calls it AVIVO. AMD also has a specific definition, if you will, for UVD, that does not jive with what you are stating...you show no UVD..you show "shader decode", which is a whole differnt technology under the AVIVO umbrella!

You describe AVIVO, I say sure, AVIVO is there, but a specific part of AVIVO, called UVD, that the packaging says is in my product, is not.

Now, before you go spouting some more numbers from websites that clearly have the story wrong, please go buy a card, and show me REAL numbers...show me the lowered cpu usage, and not theoretical. Becuae after all, no hardware decoding is happening on my machine, and not because of a lack of trying.

[v_rr]

lol. my tests('cause i got the cards) show much higher usage. Personally, i think drverheaven has a driver that I do not. you are forgetting about all those sites sampsa posted that are not able to get decode working...driverheaven is just one site.

Can be simple driver problems. More review says about the same. Just search.
Those with aceletaion ON are the 8.38, at least is what it looks.

Dave Beaumann, ATI marketing man, states quite simply, there is no hardware decode on HD2900 at all, and it is all done on cpu.

That´s simple impossible. Read again. He says that drivers with decoding are 8.38.

So who's right? Websites paid for advertizing, prone to mistakes, or an high-level exec from ATI?

That´s quote from a forum...... he don´t say that. Read again.

My cpu usage numbers say Dave is. Driverheaven says differently, but then again, they are a "bought and paid for" marketing site, and I have the cards right here and now. I believe the numbers my pc puts out, and not some numbers on a website that have been verified by noone.

At least read the hole thing that the man said and stop with conspiracy:

HD 2900 does not have UVD; it'll be paried with HD video capable CPU's. The drivers used in that test don't yet have decode acceleration enabled either, first revision of that comes with 8.38.

this stipulates that the work is done in a dedicated hardware unit, but on R600, it's done in shader, which we can hardly call dedicated. Notice that "offload the entire decode process"...25-35% is hardly offloading the entire workload

It´s impossible to have the CPU at 0%. The CPU have to interact in the process. Something less than 25% is simply irrelevant specialy on those kind off movies whith heavy IO transfers. The video-card only decodes the video, the rest is made by CPU.

[cadaveca]

Sure, and I agree with you on all of that, but you miss the whole "UVD on the product packaging, but not on the card".

I don't care that hardware decode WILL work(even though, it never will work...)..it doesn't now, not even with 8.38(yes, I have that driver too).

What concerns me is getting UVD to work...and that will never happen. SO, functionality on the box is not on the unit. SPECIFIC FUNCTIONALITY. not just simple decode acceleration...

hardware decode means that the videocard does the decode, and not the cpu. However, cpu work is SOFTWARE DECODE, NOT HARDWARE. Sure, hardware is pushing the data, but the software is doing the work..again, even in the instance Dave mentions, this is not hardware decode...it's decode acceleration..meaning part of the work the software does is done on the gpu, IN SOFTWARE(shader), not hardware.

HARDWARE DECODE IS NOT PRESENT IN 2900XT. Shader(software) ACCELERATED decode IS on 2900XT.

Do you get that? shader code doing the decode. SHADER CODE. Not hardware..software run on the shaders.

So, now that we have come full circle... THERE IS NO HARDWARE DECODE ON R600...MERELY HARDWARE ACCELERATED DECODE USING SHADERS(SOFTWARE).

Sampsa tends to not post something like this without reason, after all. It got stickied, becuase of the merit to his post. You call him a liar too, then?

[v_rr]

Sure, and I agree with you on all of that, but you miss the whole "UVD on the product packaging, but not on the card".

I don't care that hardware decode WILL work(even though, it never will work...)..it doesn't now, not even with 8.38(yes, I have that driver too).

There are 3/4 version of those drivers. You can have the wrong one, or simply you don´t even have the proper software installed. It´s not has simple has put the HD-DVD on the drive and click play.
Search how to get software properly instaled (specialy in XP that needs some software of Microsoft) and an HD player that suport offload to the grafic card.

HARDWARE DEOCDE IS NOT PRESENT IN 2900XT. Shader(software) ACCELERATED decode IS on 2900XT.

Shader is hardware-decode. Software decode is equal to have the CPU do all the work.

Do you get that? shader code doing the decode. SHADER CODE. Not hardware..software run on the shaders.

Ofcourse software run on shader. That thing don´t work alone by it-self. Everything on your PC must be programed software.

So, now that we have come full circle... THERE IS NO HARDWARE DECODE ON R600...MERELY HARDWARE ACCELERATED DECODE USING SHADERS(SOFTWARE).

you are saying that shaders is software?
Shaders are a part off the GPU. That GPU has processing power (a lot) and it can process lots off data. In this case it processes the decoding off the movie. All on the GPU, not CPU.

[BenchZowner]

This is getting frustrating...

[cadaveca]

you are saying that shaders is software?
Shaders are a part off the GPU. That GPU has processing power (a lot) and it can process lots off data. In this case it processes the decoding off the movie. All on the GPU, not CPU.

The cpu usage is from the software translating the raw data from the disc, seperating the streams, choosing the right one,decoding it into shader language, which is then processed by the videocard into final video, "post processing", deblocking, colour correction, etc, and all. As such, the data stream can be manipulated by the end user, as before the cpu can translate this data into shader code it must first be decrypted, and here we are about to get into the real meat of what's afoot here.


In hardware encoding the videocard would get the raw data, decide which bitstream to use, hand it off to it's neccesary parts and process it. At this point, there is no possibility of user-intervention, as the encrypted data gets "decoded" at a level the user cannot access, and thereby protecting the copyright holder for the content.

Becuase the datastream is processed "on cpu" into shader code, there is a chance for this un-encrypted code to be captured before it ever hits the gpu, thereby bypassing any copyright protection the media may have offered. This level of interaction is not possible when the bitsream is hardware decoded...and on the HD2900XT, there is no hardware decoding taking place...merely processing of an already decrpyted bitstream.


Now, how they managed to pull this off without the HDDVD and BluRay camps crying out...well...maybe someone needs to tell them.

[gdogg]

The cpu usage is from the software translating the raw data from the disc, seperating the streams, choosing the right one,decoding it into shader language, which is then processed by the videocard into final video, "post processing", deblocking, colour correction, etc, and all. As such, the data stream can be manipulated by the end user, as before the cpu can translate this data into shader code it must first be decrypted, and here we are about to get into the real meat of what's afoot here.


In hardware encoding the videocard would get the raw data, decide which bitstream to use, hand it off to it's neccesary parts and process it. At this point, there is no possibility of user-intervention, as the encrypted data gets "decoded" at a level the user cannot access, and thereby protecting the copyright holder for the content.

Becuase the datastream is processed "on cpu" into shader code, there is a chance for this un-encrypted code to be captured before it ever hits the gpu, thereby bypassing any copyright protection the media may have offered. This level of interaction is not possible when the bitsream is hardware decoded...and on the HD2900XT, there is no hardware decoding taking place...merely processing of an already decrpyted bitstream.


Now, how they managed to pull this off without the HDDVD and BluRay camps crying out...well...maybe someone needs to tell them.

If you are seeing 80% cpu usage, its not using the video card period. End of story.

No need to make a big show out of this.

[cadaveca]

Sure there is, because my box says HDCP too!...lol but that's just the root of it, becasue now I need an HDCP cpu too! Or did you not understand what Dave meant by that comment?

[v_rr]

Sure there is, because my box says HDCP too!...lol but that's just the root of it, becasue now I need an HDCP cpu too! Or did you not understand what Dave meant by that comment?

Read what he said again, and stop with teory off paying to site reviews that is just ridiculous.

One thing is that you don´t know how to get Hardware decoding from the GPU working, another is teory off paying to site reviews.
Sorry but thats very very ridiculous.

You are using the wrong driver or tool, because even an X1600 mobile can assist in HD movies. And a 2900XT is a more powerfull than a X1600, and lots off sites simply have their cards working and with CPU usage of 15%.

You are just blaming everything that you see. And before posting try to put fings to work and search how to put fings to work or simply wait for oficials Cat 7.5.

Has you can see you are the only one blaming everything. Everyone has already understood that the card assist HD better then 8800, has good image quality and all features.

[Eastcoasthandle]

Transformers using EXCLUSIVE TRAILER @ 1080p, HD 2900XT, Driver 8.37.4.2
CPU usage average around 20% and spiked as high as 29% twice.

AMD's dedicated Universal Video Decoder (UVD), present on both HD 2600 and 2400 series, does the whole job, internally, on a dedicated processor, leaving the CPU and GPU to do other wonderful things. As we cannot publish results for Radeon HD 2600 and HD 2400 cards, we cannot confirm this, but watching a frenetic 1080p demo run off a HD 2600 card, CPU utilisation hovered at around 15 per cent.

The Radeon HD 2900 XT being the GPU powerhouse that AMD thinks it is, video decoding is handled by the voluminous shaders.

Following on from NVIDIA's GeForce 8600 line, the Radeon HD 2000 series has HDCP-support built in, with the number of ciphers matched up to the outputs. All HDCP-compliant ports will offer dual-link compatibility, too. Great for for owners of Dell, Apple and HP 30in panel, all 50 of you.
source

Regardless if it's universal or unified you understand what the article is trying to say.

[cadaveca]

trailer is not content from player...player includes more than one audio and one video stream. Hence the higher usage.


Assist is not what the box says, box says UVD. UVD is not in the box, I'm filing suit.


Lol you think that level of site pays for hardware? Ah, I wish i could be so niave!

[Eastcoasthandle]

trailer is not content from player...player includes more than one audio and one video stream. Hence the higher usage.


Assist is not what the box says, box says UVD. UVD is not in the box, I'm filing suit.


Lol you think that level of site pays for hardware? Ah, I wish i could be so niave!

I am curious to see if your results are still high with just the trailer or not. Go ahead, try it out and report CPU usage.

[cadaveca]

ah, 15% when using lesser card...now that's what Iwas expecting, and if it worked on 2900, it would probably be closer to 10% ona dualcore. I'm currently sitting @ quadcore and r600's in crossfire, i'll try both in corssfire config and regular single rendering, to see if it has any effect, however I have seen upwards of 70% cpu usage from a 3ghz Conroe when playing direct from blu-ray drive.

Now, this is why these websites "hold no water" in my books...they are playing trailers, such as you have, which is not the same as playing from a disk, which features both HD and SD content, as well as multiple audio streams, which noone of these trailers contain!


But i have watched that trailer....quadcore cpu usage from 10%-22%, across all 4 cores. In a dualcore system, this would be about double, of course. that's far too high, imho. Also keep in mind you probably ahve an overclock, and most users of such tech will not.


But then again, maybe this is what ATI was banking on when they made this design choice...most people that buy such a card have systems that can deal with such content decently, but pair these cards with a lesser cpu, and all hell can break loose!

Anyway, here's the law I've been refering to, straight from the justice department:

(c) any description or illustration of the type, quality, performance, function, origin or method of manufacture or production of a prepackaged product that may reasonably be regarded as likely to deceive a consumer with respect to the matter so described or illustrated.

http://laws.justice.gc.ca/en/showdoc...nchorbo-ga:s_4

[Eastcoasthandle]

ah, 15% when using lesser card...now that's what Iwas expecting, and if it worked on 2900, it would probably be closer to 10% ona dualcore. I'm currently sitting @ quadcore and r600's in crossfire, i'll try both in corssfire config and regular single rendering, to see if it has any effect, however I have seen upwards of 70% cpu usage from a 3ghz Conroe when playing direct from blu-ray drive.

Now, this is why these websites "hold no water" in my books...they are playing trailers, such as you have, which is not the same as playing from a disk, which features both HD and SD content, as well as multiple audio streams, which noone of these trailers contain!


But i have watched that trailer....quadcore cpu usage from 10%-22%, across all 4 cores. In a dualcore system, this would be about double, of course. that's far too high, imho. Also keep in mind you probably ahve an overclock, and most users of such tech will not.


But then again, maybe this is what ATI was banking on when they made this design choice...most people that buy such a card have systems that can deal with such content decently, but pair these cards with a lesser cpu, and all hell can break loose!

Anyway, here's the law I've been refering to, straight from the justice department:


http://laws.justice.gc.ca/en/showdoc...nchorbo-ga:s_4

What BR device are you using to play your HD-DVD content on your computer?

[cadaveca]

in the instance of "direct from disk", no. Lite-on BluRay drive, internal.



You can use XBOX HDDVD player over usb? lol post edited!

[Eastcoasthandle]

in the instance of "direct from disk", no. Lite-on BluRay drive, internal.



You can use XBOX HDDVD player over usb? lol post edited!

Correct

It turns out, thanks to the standard USB connection that the Xbox 360 uses, you can use the external HD-DVD drive on your PC; and it's easier than you might think.

But I wonder why you think audio would slow the CPU down when a decent audio card is suppose to off load some (if not all) from the cpu?

[cadaveca]

the audio processing doesn't slow the cpu down, but deciding which stream of audio should be played, whether it be language, or actual codec, and then pulling this out from the mix is what's the issue, or at least that's what I surmise. Of course the audio data stream must be encypted/decrypted just like the video(DRM, HDCP, and all that), so there's that too. Don't forget that this content must be played on HDCP panel, and usually using HDMI connector too, so the vidcard must handle this data too.


I was serious...I really did not expect the xilleon engine to be "missing" from this gpu...

[Eastcoasthandle]

the audio processing doesn't slow the cpu down, but deciding which stream of audio should be played, whether it be language, or actual codec, and then pulling this out from the mix is what's the issue, or at least that's what I surmise. Of course the audio data stream must be encypted/decrypted just like the video(DRM, HDCP, and all that), so there's that too. Don't forget that this content must be played on HDCP panel, and usually using HDMI connector too, so the vidcard must handle this data too.


I was serious...I really did not expect the xilleon engine to be "missing" from this gpu...

Well, it does list it as a feature with the HD 2900XT at ATI's website as a HDTV encoder. The problem is there is no information how it's being used if at all on the HD 2900XT. All of this is really making no sense to me.
Do we believe some poster on another forum?
Do we believe that UVD is included based on what's put on my HIS box and the use of Xilleon found at ATI's website?