HD Video Decode Quality and Performance Summer '07
by Derek Wilson on July 23, 2007 5:30 AM EST- Posted in
- GPUs
HD HQV Image Quality Analysis
We have already explored Silicon Optix HD HQV in detail. The tests and what we are looking for in them have not changed since our first round. Fortunately, the ability of NVIDIA and AMD hardware to actually perform the tasks required of HD HQV has changed quite a bit.
Both AMD and NVIDIA told us to expect scores of 100 out of 100 using their latest drivers and hardware. We spent quite a bit of time and effort in fully evaluating this test. We feel that we have judged the performance of these solutions fairly and accurately despite the fact that some subjectivity is involved. Here's what we've come up with.
The bottom line is that NVIDIA comes out on top in terms of quality. We've seen arguments for scoring these cards differently, but we feel that this is the most accurate representation of the capabilities offered by each camp.
On the low end, both AMD and NVIDIA hardware begin to stumble in terms of quality. The HD 2400 XT posts quite a lack luster performance, failing in noise reduction and HD deinterlacing (jaggies). But at least it poorly deinterlaces video at full resolution. We excluded tests of NVIDIA's 8500 series, as their video drivers have not yet been optimized for their low end hardware. Even so, we have been given indications not to expect the level of performance we see from the 8600 series. We would guess that the 8500 series will perform on par with the AMD HD 2400 series, though we will really have to wait and see when NVIDIA releases a driver for this.
With video decode hardware built in as a separate block of logic and post processing being handled by the shader hardware, it's clear that the horrendous 3D performance of low end parts has bled through to their video processing capability as well. This is quite disturbing, as it removes quite a bit of potential value from low cost cards that include video decode hardware.
Both AMD and NVIDIA perform flawlessly and identically in every test but the noise reduction test. AMD uses an adaptive noise reduction algorithm that the user is unable to disable or even adjust in any way. NVIDIA, on the other hand, provides an adjustable noise reduction filter. In general, we prefer having the ability to adjust and tweak our settings, but simply having this ability is irrelevant in HQV scores.
The major issue that resulted in our scoring AMD down in noise reduction was that noise was not reduced significantly enough to match what we expected. In addition to the tests, Silicon Optix provides a visual explanation of the features tested, including noise reduction. They show a side by side video of a yellow flower (a different flower than the one presented in the actual noise reduction test). The comparison shows a noisy video on the left and a video with proper noise reduction applied on the right. The bottom line is that there is almost no noise at all in the video on the right.
During the test, although noise is reduced using AMD hardware, it is not reduced to the level of expectation set by the visual explanation of the test. Based on this assessment, we feel that AMD noise reduction deserves a score of 15 out of 25. Silicon Optix explains a score of 15 as: "The level of noise is reduced somewhat and detail is preserved." In order to achieve a higher score, we expect the noise to be reduced to the point where we do not notice any "sparkling" effect in the background of the image at all.
By contrast, with NVIDIA, setting the noise reduction slider anywhere between 51% and 75% gave us a higher degree of noise reduction than AMD with zero quality loss. At 75% and higher we noticed zero noise in the image with no detail loss until noise reduction was set very high. Tests done with the noise reduction slider at 100% show some detail loss, but there is no reason to crank it up that high unless your HD source is incredibly noisy (which will not likely be the case). In addition, at such high levels of noise reduction, we noticed banding and artifacts in some cases. This was especially apparent in the giant space battle near the end of Serenity. It seems to us that computer generated special effects seemed to suffer from this issue more than other aspects of the video.
While, ideally, we would like to see artifacts avoided at all cost, NVIDIA has provided a solution that offers much more flexibility than their competition. With a little experimentation, a higher quality experience can be delivered on NVIDIA hardware than on AMD hardware. In fact, because NVIDIA sets noise reduction to default off, we feel that the overall experience provided to consumers will be higher.
We have already explored Silicon Optix HD HQV in detail. The tests and what we are looking for in them have not changed since our first round. Fortunately, the ability of NVIDIA and AMD hardware to actually perform the tasks required of HD HQV has changed quite a bit.
Both AMD and NVIDIA told us to expect scores of 100 out of 100 using their latest drivers and hardware. We spent quite a bit of time and effort in fully evaluating this test. We feel that we have judged the performance of these solutions fairly and accurately despite the fact that some subjectivity is involved. Here's what we've come up with.
| Silicon Optix HD HQV Scores | ||||||
| Noise Reduction | Video Res Loss | Jaggies | Film Res Loss | Stadium | Total | |
| AMD Radeon HD 2900 XT | 15 | 20 | 20 | 25 | 10 | 90 |
| AMD Radeon HD 2600 XT | 15 | 20 | 20 | 25 | 10 | 90 |
| AMD Radeon HD 2600 Pro | 15 | 20 | 20 | 25 | 10 | 90 |
| AMD Radeon HD 2400 XT | 0 | 20 | 0 | 25 | 10 | 55 |
| NVIDIA GeForce 8800 GTX | 25 | 20 | 20 | 25 | 10 | 100 |
| NVIDIA GeForce 8600 GTS | 25 | 20 | 20 | 25 | 10 | 100 |
| NVIDIA GeForce 8600 GT | 25 | 20 | 20 | 25 | 10 | 100 |
The bottom line is that NVIDIA comes out on top in terms of quality. We've seen arguments for scoring these cards differently, but we feel that this is the most accurate representation of the capabilities offered by each camp.
On the low end, both AMD and NVIDIA hardware begin to stumble in terms of quality. The HD 2400 XT posts quite a lack luster performance, failing in noise reduction and HD deinterlacing (jaggies). But at least it poorly deinterlaces video at full resolution. We excluded tests of NVIDIA's 8500 series, as their video drivers have not yet been optimized for their low end hardware. Even so, we have been given indications not to expect the level of performance we see from the 8600 series. We would guess that the 8500 series will perform on par with the AMD HD 2400 series, though we will really have to wait and see when NVIDIA releases a driver for this.
With video decode hardware built in as a separate block of logic and post processing being handled by the shader hardware, it's clear that the horrendous 3D performance of low end parts has bled through to their video processing capability as well. This is quite disturbing, as it removes quite a bit of potential value from low cost cards that include video decode hardware.
Both AMD and NVIDIA perform flawlessly and identically in every test but the noise reduction test. AMD uses an adaptive noise reduction algorithm that the user is unable to disable or even adjust in any way. NVIDIA, on the other hand, provides an adjustable noise reduction filter. In general, we prefer having the ability to adjust and tweak our settings, but simply having this ability is irrelevant in HQV scores.
The major issue that resulted in our scoring AMD down in noise reduction was that noise was not reduced significantly enough to match what we expected. In addition to the tests, Silicon Optix provides a visual explanation of the features tested, including noise reduction. They show a side by side video of a yellow flower (a different flower than the one presented in the actual noise reduction test). The comparison shows a noisy video on the left and a video with proper noise reduction applied on the right. The bottom line is that there is almost no noise at all in the video on the right.
During the test, although noise is reduced using AMD hardware, it is not reduced to the level of expectation set by the visual explanation of the test. Based on this assessment, we feel that AMD noise reduction deserves a score of 15 out of 25. Silicon Optix explains a score of 15 as: "The level of noise is reduced somewhat and detail is preserved." In order to achieve a higher score, we expect the noise to be reduced to the point where we do not notice any "sparkling" effect in the background of the image at all.
By contrast, with NVIDIA, setting the noise reduction slider anywhere between 51% and 75% gave us a higher degree of noise reduction than AMD with zero quality loss. At 75% and higher we noticed zero noise in the image with no detail loss until noise reduction was set very high. Tests done with the noise reduction slider at 100% show some detail loss, but there is no reason to crank it up that high unless your HD source is incredibly noisy (which will not likely be the case). In addition, at such high levels of noise reduction, we noticed banding and artifacts in some cases. This was especially apparent in the giant space battle near the end of Serenity. It seems to us that computer generated special effects seemed to suffer from this issue more than other aspects of the video.
While, ideally, we would like to see artifacts avoided at all cost, NVIDIA has provided a solution that offers much more flexibility than their competition. With a little experimentation, a higher quality experience can be delivered on NVIDIA hardware than on AMD hardware. In fact, because NVIDIA sets noise reduction to default off, we feel that the overall experience provided to consumers will be higher.
Facebook
Twitter
RSS
63 Comments
View All Comments
DigitalFreak - Monday, July 23, 2007 - link
Based on Derek's results, I ordered the parts for my new HTPC.C2D E6850 (Newegg are bastards for pricing this at $325, but I didn't want to wait)
Intel P35 board
2GB PC2-800
Gigabyte 8600GTS (passive cooling)
Wished there was an 8600 board with HDMI out, but oh well...
SunAngel - Monday, July 23, 2007 - link
You confirmed my suspicions all along. I always wondered if the true motive for SLI and Crossfire was to double the benefits of GPU processing rather than separate the graphics performance of 3D and video acceleration. In my eyes, I see SLI and Crossfire being a "bridge" for 3D graphics and Video accleration cards. What I am referring to is the PCIex16(16 lane) slot been for high powered 3D GPUs and the PCIex16(8 lane) slot being for video accleration GPUs.It is obvious between the HD2900XT and the HD2600XT that one is great at rendering 3D game graphics while the other is great at acceleration motion picture movies.
Personally, this is an okay tactic by the card manufacturers. It segments the performance a little bit better. I do not game the least bit, so the high end cards are something I don't want. But, my taste are different than others that do want it. But those that desire both, can have their cake and eat it too, but using a dual PCIex16 motherboard and installing each type of card.
Overall, good article. You enlightened my purchasing decision. With all the talk about futureproofing that was going around for a while, buying a dual PCIex16 motherboard makes a lot of sense now.
TA152H - Monday, July 23, 2007 - link
I don't think you understand the point of the cards.If you buy the 2900 and a high end processor, you will not have any problems with HD playback, that's the whole point. You don't need a 2600 to go with it. The number of people that buy something as expensive as the 2900XT and a low end processor that is incapable of playing back HD is very, very low to the point where ATI decided it was a mistake to buy it.
So, no, you wouldn't get a 2600 to go with it, you'd get a good processor and the 2900 and that's all you'd need to have the best of both worlds.
Chunga29 - Monday, July 23, 2007 - link
Yes, if by "best" you mean:- Higher CPU utilization when viewing any HD video content, compared to 8800
- Generally lower price/performance in games compared to 8800
- More flaky (in my experience) drivers than 8800 (though I believe AMD might actually be better on Vista - not that I really care at this point)
Don't pat AMD on the back for skipping UVD on R600. NVIDIA didn't bother to work on VP2 for G80, and yet no one is congratulating them on the decision. I agree that the omission is not the end of the world, mostly because I don't think people running 8800/X2900 cards are really all that concerned with H.264 video. If I were looking to go Blu-ray or HD-DVD, I'd be looking at a set-top box to hook up to my HDTV.
My PC is connected to a 24" LCD that I use for work, not watching movies, and trying to put it next to the TV is more effort than it's worth. Unless H.264 suddenly makes a difference for YouTube and the like (hey - I'd love to see higher quality videos online), I can't say I'm all that concerned. Seems to me there's just a vocal minority whining about the latest features that are used by less than 10% of people.
UVD, PureVideo HD, and a partridge in a pear tree: it's all moot to me!
TA152H - Tuesday, July 24, 2007 - link
OK, do you understand the meaning of the word "context"?I'm not going into the merits of Nvidia and ATI. I have used both, I consider Nvidia junk, and I do not buy them. If you have had better luck, then go with them. That's not the point, but anyone with any reading comprehension should have figured that out.
He was talking about putting a 2600 and a 2900 on the same motherboard to get the best of both worlds, meaning having all the performance of the 2900 yet getting the HD capabilities of the 2900. Do you understand that?
My point is you don't need the 2600 to get "the best of both worlds", you just need a good processor and you will not miss that feature. I think Nvidia made the right choice too. Most people are morons, and they want just because they want, and they fail to realize nothing is free. Including useless features at a cost is a bad idea, and ATI did the right thing not to, even though you'll have the idiots that think they are missing out on something. Yes, you are, you're missing out on additional cost, additional electricity use, and additional heat dissipation. You don't need it if you buy a reasonable processor for the item. That's the point. Try to understand context better, and realize what he meant by the best of both worlds.
strikeback03 - Wednesday, July 25, 2007 - link
Assuming your "good processor" falls somewhere between the two tested C2D processors, dropping UVD boosts average processor usage around 42% in Transporter2, 44% in Yozakura, and 24% in Serenity. So which uses more electricity and generates more heat - the additional transistors needed for UVD on the 2900, or moving your CPU off idle to do the work?TA152H - Tuesday, July 24, 2007 - link
Previous post should have said "HD capability of the 2600".Chunga29 - Tuesday, July 24, 2007 - link
For someone trying to act superior, you need to take a look in the mirror (and the dictionary) for a moment. I agree it's silly to use something like a 2600 and 2900 in the same system. However, if you want the "best of both worlds", let's consider for a minute what that means:Best (courtesy of Mirriam Webster):
1 : excelling all others
2 : most productive of good: offering or producing the greatest advantage, utility, or satisfaction
So, if you truly want the best of both worlds, what you really want is:
UVD from ATI RV630
3D from NVIDIA G80
Anything less than that is not the "best" anymore (though I'm sure some ATI fans would put R600 3D above G80 for various reasons).
Try ditching the superlatives instead of copping an attitude and constantly defending ever post you make. If you want to say that R600 with a fast CPU is more than sufficient for H.264 playback as well as providing good 3D performance, you're right. The same goes for G80. If you want to argue that it may have been difficult and not entirely necessary to cram UVD into R600, you can do that, but others will disagree.
Since they were at something like 700 million transistors, they may have been out of room. That seems very bloated (especially considering the final performance), but how many transistors are required for UVD? I'd say it was certainly possible to get UVD in there, but would the benefit be worth the cost? Given the delays, it probably was best to scrap UVD. However, the resulting product certainly isn't able to offer the best possible feature set in every area. In fact, I'd say it's second in practically every area to other GPUs (G80/86 and RV630, depending on the feature). As others have pointed out in the past, that's a lot like the NV30 launch.
autoboy - Monday, July 23, 2007 - link
What??scosta - Monday, July 23, 2007 - link
I think this sentence in page 1 is wrong!Dont you mean ...
Regards