Anandtech 的 GPU 转码对比 Badaboom vs. AMD Avivo Video Converter

Edison

http://anandtech.com/video/showdoc.aspx?i=3475&p=1
And now, the rest of the storyTo say that the output quality is bad would be entirely too generous. Rather than just take our word for it, here are some screenshots of the video. The actual video is worse as the corruption and anomalies change over time as well, but this is enough to show the problems.

This is the 64-bit version of Avivo encoding WMV files:


Our highest quality output video was S02E01 of MacGyver which lacked any deinterlacing.


Bad Boys delivered the second best output with massive corruption frequently occurring.


The Empire Strikes Back with completely unwatchable video that is nothing but completely corrupted and can even crash a player.

The 32-bit version has fewer problems. Codec choice didn't seem to fix corruption on the 64-bit version, but under the 32-bit version we were able to get cleaner output encoding to WMV (but not iPod (H.264) which remained corrupted).


Here's the Empire shot from the video encoded to WMV under Vista 32. Looks better.

MacGyver is unchanged in quality with the interlacing still a major problem, but at least some of the corruption issues are fixed for those running 32-bit Vista.

Our goal was to test the case where an end user may want to encode DVDs for use on their iPod, as this would allow us to more easily compare the software to Badaboom. Comparing quality comes out well in favor of NVIDIA on either the 64-bit or 32-bit version of Avivo when encoding to iPod video, so this makes any performance comparison much more difficult. We can't honestly directly compare the two software packages because of the major difference in the quality of output they generate.

To further illustrate the point I uploaded a segment to Viddler, but the further conversion to a flash video greatly improved quality. Some of the corruption is still there, but lost is the choppiness and other motion issues as well as the fact that every second or two we get a frame that looks like this thrown in:


It ain't easy being green.

Anyway, here's video clip. Yes, it also encoded with no sound, which is apparently also a feature of the Avivo video converter with some video formats.

               
We did also try using other sources to transcode from with Avivo, but our attempt to use H.264, DivX, or WMV sources all failed with different results. Some had video with no sound, some had only sound and no video, and some just gave us an error saying the file wasn't a valid movie and it wouldn't even try to transcode it.

And since when is video transcoding not a deterministic process? While working with our Star Wars clip, we noticed that our files weren't all the same size. Our first thought was that maybe when running on different video cards the transcoder operated differently. But after extended testing, we discovered that we don't always get the same output even if nothing changes. Not only do two different cards not output the same data, but one card isn't guaranteed to give you the same result from run to run.

This, on top of all the other issues, makes any performance comparison moot. But we ran the numbers anyway. And here's what we got:


This is a roughly 18 minute section of The Empire Strikes Back that took between 38 and 43 seconds to encode, and we recorded the first run. Subsequent runs came out between 38 and 43 seconds for all the cards. Regardless of which video card we plug in, performance is the same (and this is despite the output not even being the same). This honestly seems to indicate that not much is happening on the GPU. All the variables change with this hardware and nothing is the same width or clock speed from top to bottom.

We used GPU-Z to test utilization and saw that every 4 seconds or so during the encode process, we would see a blip of about 15% utilization on the GPU then it would immediately fall back down to zero. This means that AMD is using the GPU for very very little and most of the performance of the encoder is coming from the Core i7 processor we've got in there. To test our theory, we ran the processor underclocked at 2.6 GHz to see what happened to performance. Performance dropped from between 38 and 43 seconds per run to 56 to 60 seconds per run. This indicates that the performance of Avivo is incredibly CPU bound.

We can see from looking at the task manager during the transcode that there are about three threads that are pretty hard on the CPU. No matter what CPU speed we were running, one core was always pegged. Two other cores sat at over 50% utilization on the 3.2GHz Core i7. Overall utilization hovered between 30 and 40 percent for the most part. When we decreased the clock speed, all three threads ended up pegging their assigned cores at different times, which could indicate the much larger than 25% performance improvement in moving to higher speeds (the change from 2.6GHz to 3.2GHz is just over 23%).

So, we've got a transcoder that doesn't produce watchable output in many cases, doesn't produce consistent output, and doesn't vary in performance with GPU hardware (meaning it doesn't use the GPU very much at all). It doesn't handle incorrect interlacing flags gracefully and there is no option to force deinterlacing. But that's minor compared to the rest of the problems we are seeing here.

Edison

Elemental's Badaboom 1.0: The RedemptionRemember Elemental? It’s the company that put out Badaboom, the world’s first GPU accelerated H.264 video transcoder built using CUDA. NVIDIA was particularly excited about Badaboom as it finally gave NVIDIA a consumer-level CUDA application to point to when making the argument of why its GPUs were better than both ATI’s GPUs and Intel’s CPUs alone.


Unfortunately, the beta release of Badaboom needed some work. It didn’t do anything well, at all. After that original Badaboom review I met with Sam Blackman, Elemental’s CEO and we went through the list of things that needed to be fixed.

I should give credit to Mr. Blackman, normally whenever we post any sort of a critical review of any product, the company is fiercely upset with us. I argued with Intel PR for years over our Pentium 4 reviews, AMD felt our review of the Radeon HD 3870 was unfair, and if we don’t mention PhysX as a feature advantage of its GPUs then NVIDIA gets a little emotional. As harsh as the original review was however, Sam wasn’t irrationally upset, I believe his exact words were “that was harsh” and then it was straight to “what can we do to make it better?”.

It’s Sam’s attitude that was reflected so greatly in what became Badaboom 1.0.
The changes were sweeping, now gone is the Pro version, which is welcome given that the Pro version was anything but that. Elemental is instead only focusing on the consumer version and will be rolling in features into this version over time.

The initial consumer release was only supposed to support up to 480p output files, while the new 1.0 release can do up to 720p (the old “pro” version supported up to 1080p). The 1.1 release due out in the next few days will add 1080p support. While originally being slated for use in the Badaboom Pro, AVCHD and HDV input formats are now both a part of the $29.99 consumer version.

All in all, killing off the pro version and folding mostly everything into the consumer version made a lot of sense.

There are still some pretty serious limitations: 1) there’s no official support for Blu-ray movies, 2) no official support for DivX, 3) the highest H.264 profile supported is still baseline (although Elemental plans on adding Main support in 1.1 and High profile support in the future).

Elemental did add support for Dolby Digital audio input, although DTS is still being worked on. The only audio output format supported is still AAC-LC.

The total sum of all of this is that Elemental’s first version of Badaboom now has a focus, a very specific one, but it gives us a target to shoot for. This isn’t an application that you’re going to use to backup your Blu-ray collection, it’s not even very useful for backing up your DVDs, but what it can do very well is transcode your DVDs for use on a portable media player like an iPhone or iPod.

Funky Issues? Resolved
The biggest problem with the previous version of Badaboom was that it couldn’t do anything right. I tried transcoding Blu-rays, DivX files, chimpanzees, DVDs, and each input file had some sort of quirk associated with it. Even taking a simple DVD, which Badaboom was supposed to support flawlessly, and transcoding it sometimes left me with an unusable output file of the wrong frame rate.

Focusing Badaboom’s attention, Elemental now made one thing work very well: DVDs. Point Badaboom at an unencrypted VIDEO_TS folder or a DVD disc/image and it will now perfectly rip the DVD to the appropriate resolution.

I should mention that DRM is rearing its ugly head here once more as Badaboom won’t automatically convert an encrypted DVD. Thankfully Slysoft’s AnyDVD simply running in the background is enough for Badaboom to transcode any DVD. If you haven’t used AnyDVD, I highly recommend it - it’s a great way of getting rid of encryption on both DVDs and Blu-ray discs.

Elemental also fixed the weird image quality issues, the output no longer gets scaled out of its correct aspect ratio when downscaled. Hooray.

Badaboom: Quad-Core DesiredBadaboom obviously does very well with a fast GPU, but the CPU requirements are also reasonably high. Keeping the GeForce GTX 280 fed actually ate up 50% of the CPU power of our Core 2 Quad Q9450 in our tests, it seems that Badaboom won’t scale beyond two cores.

The problem is that Elemental and NVIDIA make the argument that using the GPU to transcode video frees up your CPU to do other tasks while you’re doing this. The reality is that this is only true if you’ve got four cores, otherwise your dual-core CPU is just as pegged as it would be if you were doing a CPU-based video transcode. The difference here being that the transcode is going a lot faster.

While NVIDIA wants you to spend less money on the CPU and put the savings towards a faster GPU, the correct approach continues to be buying a decent CPU and a decent GPU, even with GPU accelerated video encoding. If you’re going to be doing a lot of video encoding, a quad-core CPU is still a good idea regardless of whether you’re doing your encoding on the GPU or not.

Edison

Image QualityHere’s where Elemental gets off easy. Since Badaboom is best used as an application for getting DVDs onto your iPhone or other low res format, image quality isn’t as big of a deal as it would be if you were viewing these things on a TV.

Compared to the x264 codec, Badaboom’s output seems just fine:


Elemental's Badaboom 1.0

x264
Again, Badaboom avoids the more difficult image quality comparisons by not being useful for high quality conversions.

AMD shows up to this gun fight with a knife, as Avivo’s image quality isn’t acceptable. While the Avivo video converter is free, it’s not useful.

PerformanceOnce again, I looked at the performance of Badaboom vs. transcoding on a CPU using Handbrake 0.93 (which uses the x264 codec). This time around we have Intel’s Core i7 965, running at 3.2GHz. The comparison stacks up pretty much as it did before:



The issue is that the Core i7 isn’t running with all 8 threads maxed, instead Handbrake appears to be only utilizing 30 - 40% of the available execution resources, which amounts to less than all four physical cores.

I suspect with better CPU utilization we could have a scenario where the Core i7 was able to perhaps match the performance of the GeForce GTX 280. The only problem then becomes the cost difference.

The Avivo Video Converter does complete our conversion task in around half the time of the GeForce GTX 280 running Badaboom, however the output file is unusable so the performance advantage is meaningless in our opinion. If AMD could fix things however...

Edison

Badaboom 1.1 PreviewElemental was nice enough to get us a preview build of Badaboom 1.1, due out in the next week or so.





The 1.1 release is pretty significant, arguably just as significant as 1.0 given the list of improvements:

This version offers additional input files, including DivX, Xvid, VC-1, and MKV. The file formats in 1.0 are still supported.

- H.264 Main profile output.

- 1920x1080 (1080p) output.

- New output profiles: Blackberry Bold, YouTube, and Zune.

- Multi-GPU capability: Badaboom can now let you select which CUDA-enabled GPU to run the transcoding on in the Advanced options. You can open multiple Badaboom applications and run different video files simultaneously on different GPUs, seeing similar high performance as you see with running Badaboom now. For example, two NVIDIA GPUs can transcode two movies in the time it would take Badaboom 1.0 to transcode one. For this initial version, there are a couple caveats:

1. SLI must be disabled (if it is enabled) .
2. Each GPU needs to be connected to a display.
3. Each GPU must have a desktop enabled on it.

I’m not particularly interested in the multi-GPU support that 1.1 offers given the caveats, but the rest of the feature list is excellent. With DivX support you can now take your old DivX shows and movies, re-encode them using Badaboom to save space. While trying to transcode a DivX file in the early version of Badaboom failed miserably, it worked just fine in the 1.1 preview.


oh wow, real settings

Main profile with CABAC support is also enabled, making Badaboom 1.1 closer to a real alternative for high quality rips of DVDs and Blu-rays. The 1.1 beta isn’t ready for prime time but I wanted to see what it could do, so I grabbed my Casino Royale Blu-ray, ripped it to the hard drive (resulting in a 46GB iso). The Blu-ray file structure is pretty straightforward, in the BDMVSTREAM directory you’ll find a bunch of m2ts files, in this case the 34GB 00000.m2ts file is the main 1080p movie.


Please don't throw me in jail

Since I just ran the BD through AnyDVD HD it no longer had any encryption thus Badaboom was ok with transcoding it. To my surprise, it just worked. Using the Custom Media Center profile I was able to select a video bitrate of up to 25Mbps, I stuck with 11Mbps which should be enough for most < 50” HDTVs and normal viewing distances. The maximum audio bitrate is 320kbps, which is what I selected for the transcode. The resulting file was an 11.4GB .mp4 file that took 2 hours and 2 minutes to transcode at an average of 28.3 fps (note that this included the credits as well) on a GeForce GTX 280.


Click to Enlarge

The Blu-ray original was mastered at 24 fps, so with Badaboom on a GTX 280 we can get greater than real-time transcoding.

Delivering as promised, Elemental also enabled 1080p output with Badaboom 1.1. While I suspect that most people using to backup their Blu-ray collection will choose to go down to 720p since you can basically cut file sizes in half and maintain good enough quality for most HD displays.



The biggest limitation I see is that the output file is relatively useless on a HTPC. While Badaboom can provide a quick and easy way to rip a Blu-ray to a smaller, more backup-friendly format, you do lose the ability to preserve the DD/DTS audio tracks. Forget about lossless 7.1 support, I’m just talking about maintaining basic DD/DTS 5.1 that your HTPC/receiver are already setup to play.
Elemental still dodges the bullet of having to be a real Blu-ray backup solution by not addressing the audio side of the equation, but it’s clear where Badaboom is headed. I’m still looking into how well it fares on the image quality side compared to x264, but for now it looks like Badaboom 1.1 has potential in this department.

The real competition will be Intel’s Core i7 which, thanks to its incredible encoding performance can actually do quite well in the HD transcode department. NVIDIA has the advantage of its GPUs being much cheaper (you don’t have to buy a whole new platform) but at least at this point Intel has the quality advantage given that the best audio/video transcoding tools on the market are still x86-only and without a CUDA counterpart.

Travis

结论是AVIVO又一次当了个披着GPGPU外皮的几乎没有GPU加速的低质量转码器

Edison

Final Words
We're happy to see AMD including ATI Stream in their latest driver release. It's great that both NVIDIA and AMD are doing what they can to advance GPU computing right now. We still won't see any truly major strides made in consumer level applications until we have OpenCL and DirectX 11 to bring hardware agnostic general purpose data parallel programming to the masses, but getting tools (even proprietary ones) out there and in the hands of developers will definitely help.

We feel similarly about the marketability of ATI Stream as we do about CUDA. GPU computing is still only a niche and there aren't enough applications out there that really bring the kind of value to the consumer that we want and expect. The decision about which graphics card you are going to pick up shouldn't come down to ATI Stream and CUDA unless you are really into one of the applications out there that runs on one or both of these technologies. For the average gamer, we definitely recommend making your purchasing decisions on how hardware performs in the games you want to play.

All that said, we are very disappointed with AMD's Avivo video converter as a vehicle to show off ATI Stream. It is a poor application that provides little to no value in exchange for the immense frustration end users will have when trying to transcode video. It is not worth the time to it takes to download or the space it takes up on your hard drive.

In the course of evaluating Avivo, our second look at Badaboom showed us a much better product than we previewed that adequately fills a niche and provides good support for getting video on to an iPod or iPhone quickly.  

Badaboom 1.1 shows Elemental's commitment to the cause.  Normally when I'm promised that things will get better, and that features will be added, they don't.  Or if they do, they take a long time.  It is now less than four months since we first previewed Badaboom and with version 1.1, much of what we asked for has been included.  There's still a long way to go and Elemental still has the difficult tasks of matching the quality of established codecs like x264 and MainConcept, but these past two revisions of Badaboom prove one thing: Elemental is serious and willing to listen to feedback.   

No matter how you slice it though, Elemental has a much better product than AMD is offering with the Avivo video converter.

The 8.12 drivers in general do offer some fixes for problems we've had since October. Many of our readers noticed the string of somewhat negative jabs we took at Catalyst over the past few months. We'll spare everyone a redux, but just because this driver is more stable, feature complete, and includes some important outstanding hotfixes doesn't mean the problems AMD has with their approach to driver development have been solved. The train wreck that has been the last few months of Catalyst has happened before and it will happen again as long as AMD puts too many resources into pushing drivers out every month and not enough into making sure those drivers are of high enough quality.

Edison

原帖由 Travis 于 2008-12-16 14:51 发表
结论是AVIVO又一次当了个披着GPGPU外皮的几乎没有GPU加速的低质量转码器

你说的是这个？

"AMD shows up to this gun fight with a knife, as Avivo’s image quality isn’t acceptable. While the Avivo video converter is free, it’s not useful."

" we are very disappointed with AMD's Avivo video converter as a vehicle to show off ATI Stream. It is a poor application that provides little to no value in exchange for the immense frustration end users will have when trying to transcode video. It is not worth the time to it takes to download or the space it takes up on your hard drive."

gz0921

AMD一贯的伎俩，东西放在那里了，至于能不能用就不好说了，反正宣传的时候我有话说，有东西和NV对比就行了~

UVD出来时是这样，这回又是这样~

Travis

关于Badaboom的一小段话

Badaboom obviously does very well with a fast GPU, but the CPU requirements are also reasonably high. Keeping the GeForce GTX 280 fed actually ate up 50% of the CPU power of our Core 2 Quad Q9450 in our tests, it seems that Badaboom won’t scale beyond two cores.

The problem is that Elemental and NVIDIA make the argument that using the GPU to transcode video frees up your CPU to do other tasks while you’re doing this. The reality is that this is only true if you’ve got four cores, otherwise your dual-core CPU is just as pegged as it would be if you were doing a CPU-based video transcode. The difference here being that the transcode is going a lot faster.

While NVIDIA wants you to spend less money on the CPU and put the savings towards a faster GPU, the correct approach continues to be buying a decent CPU and a decent GPU, even with GPU accelerated video encoding. If you’re going to be doing a lot of video encoding, a quad-core CPU is still a good idea regardless of whether you’re doing your encoding on the GPU or not.

tx97

原帖由 Gartour 于 2008-12-16 15:04 发表
Badaboom 好用啊，赶快去买正版吧。可别用破解哦{titter:]

别在那里酸葡萄了，ATI的那个测试如果属实，就几乎没有使用价值。叫别人别用破解？你先换全套正版再来说吧。而且这里讨论的问题和破解有关吗？X些人士是企图转移讨论问题的方向吧{lol:] {titter:]

DonaldGoD

原帖由 Gartour 于 2008-12-16 15:04 发表


Badaboom 好用啊，赶快去买正版吧。可别用破解哦{titter:]

你指望AMD能开发多少免费的通用软件呢？

只为你存在

据说8.12配套的那个转换软件根本就没利用到GPU

rhythmofwolf

最近测试了测试做工好的N卡，的确看不出和A卡的差距来

可惜国内买不到做工好的N卡，真可悲。

做工好的都天价，不适合俺们这些低端消费群体。

Asuka

用4850转了一段MP4在手机上看，还行

可能是手机屏幕太烂了

boris_lee

It is not worth the time to it takes to download or the space it takes up on your hard drive.

枪，一定是枪{huffy:] {huffy:]

tx97

原帖由 Gartour 于 2008-12-16 15:48 发表


何来酸葡萄？

我说过Badaboom不好用了？没有啊

还是说我买不起Badaboom所以认为Badaboom太贵了？也没有啊

某些饭饭果然是杀人不见血的{titter:]

那麻烦不是“饭”的某人解释一下您的原意是什么？{lol:]

Travis

AMD继承了google的风格，先发布，再打补丁{lol:]

lacri

前段时间驱动之家也出个一个对比评测，貌似AVIVO的压片速度比较快，但效果不说惨不忍睹，至少也是破破烂烂，个人意见这很致命！