HTPC - TV Tuner Reviews
by Jarred Walton on December 7, 2005 12:05 AM EST- Posted in
- Smartphones
- Mobile
Performance Considerations
We've talked a lot about the cards, but we haven't said much about the actual performance or CPU usage. It varies by card, as well as what content you're watching, so we captured some graphs showing CPU usage in a variety of situations. For the most part, anything faster than around 1.6 GHz P4 or 1600+ Athlon should suffice, though just to be safe, you might want closer to 2.4 GHz/2400+. Using the Sempron 64 running at 2.50 GHz was more than sufficient for everything but the MyHD analog recording. Let's start first by looking at how each card performs in analog mode. The MyHD and Theater 550 are running on the Athlon 64 4000+ HP system, while the Fusion5 is in the Sempron system. Don't worry - we're only looking for a rough estimate of CPU power requirements, and the Fusion5 worked fine with the Sempron as well as the slightly faster HP.
For watching analog content, not only is the Theater 550 a better image, but it also uses far less CPU power. In fact, the image for watching and recording with the Theater 550 is the same. Can you spot the point where we enabled recording in the graph? (The higher CPU usage at the start is actually PowerCinema loading, so don't confuse that with the latter portion of the chart.) We began recording the station on the 550 around the spot marked by the two "peaks" where CPU use spiked to 40% or so. Otherwise, CPU use remains around 10%.
The Fusion does quite well when only viewing analog channels, averaging around 20% CPU use on the Sempron system and slightly less (15%) on the 4000+ (which we didn't capture). Recording with the Fusion5 kicks CPU usage up to about 60%. That's a pretty big jump, but we were running with a maximized viewport (720p content), and we didn't experience any stalls or other problems. For watching an analog channel, the Fusion5 is a close second to the Theater 550, though it's a distant second for recording the content.
Finally, we have the MyHD card. Similar to the Fusion, watching an analog channel uses less CPU time than recording the channel. However, less in this case means about 40% CPU usage. Enabling even the fastest DivX encoding (with MP3 audio) kicks the CPU up to almost 80% usage. The video capture doesn't cut out, but there does seem to be a slight delay every second or so. Even though there is supposedly 20% idle CPU time, the capture isn't really clean, and it's about equal to watching TV with only mediocre reception. We don't have a graph, but moving to the 2.6 GHz X2 chip with Normal quality DivX encoding still used a whopping 65% of the CPU time - remember that there are two cores, so that would be over 130% CPU time for the 4000+. The faster X2 also removed the slight lag in the video, all while producing a much better encode.
Here's another version of the data, with a few additional reference points. The graphs are colored according to performance: green indicates that there were no difficulties, yellow means that there may be periodic issues, and red means that the result was unacceptable.
Moving to digital channels removes the Theater 550 from the list, but we have both the DxVA accelerated and non-DxVA modes in FusionHDTV. Most modern GPUs can use the DxVA acceleration, so consider the non-DxVA charts to be a worst case scenario. All of the results were taken while viewing a 1080i football game broadcast.
With DxVA enabled, watching 1080i with the Fusion5 requires about 20% of your CPU time. Recoding the same content appears to add about 10% more CPU use, and we get close to 30%. The non-DxVA mode requires quite a bit more power to handle the decoding of the digital signal, using 45% of the CPU power. As before, recording the same content appears to add another 10%. Remember that unlike analog signals, digital TV is already compressed. Decompressing is quite a bit easier than compressing a signal, which is why a Sempron can handle 1920x1080 video at 30 FPS without trouble, whereas the same CPU chokes when attempting to encode even 640x480 video at 30 FPS.
Viewing digital content with the MyHD is completely different from using the card in analog mode. CPU usage on the 4000+ was around 25% when viewing a 1080i signal, though disabling the overlay screen knocks that down around 15%. The record mode oddly enough has consistently lower CPU usage, at around 12 or 13%. The hardware acceleration of DTV decoding that the MyHD card provides is clearly put to good use.
Again, here's the simplified graph. All of the results are green, as there were no difficulties. Judging by the CPU usage on the Sempron for the MyHD, most of the CPU usage is used by the overlay window.
We've talked a lot about the cards, but we haven't said much about the actual performance or CPU usage. It varies by card, as well as what content you're watching, so we captured some graphs showing CPU usage in a variety of situations. For the most part, anything faster than around 1.6 GHz P4 or 1600+ Athlon should suffice, though just to be safe, you might want closer to 2.4 GHz/2400+. Using the Sempron 64 running at 2.50 GHz was more than sufficient for everything but the MyHD analog recording. Let's start first by looking at how each card performs in analog mode. The MyHD and Theater 550 are running on the Athlon 64 4000+ HP system, while the Fusion5 is in the Sempron system. Don't worry - we're only looking for a rough estimate of CPU power requirements, and the Fusion5 worked fine with the Sempron as well as the slightly faster HP.
Analog Channels (watching on left, recording on right)
| FusionHDTV | |
 |
 |
| Click to enlarge. | |
| MyHD | |
|  |
| Click to enlarge. | |
| Theater 550 | |
 |
|
| Click to enlarge. | |
For watching analog content, not only is the Theater 550 a better image, but it also uses far less CPU power. In fact, the image for watching and recording with the Theater 550 is the same. Can you spot the point where we enabled recording in the graph? (The higher CPU usage at the start is actually PowerCinema loading, so don't confuse that with the latter portion of the chart.) We began recording the station on the 550 around the spot marked by the two "peaks" where CPU use spiked to 40% or so. Otherwise, CPU use remains around 10%.
The Fusion does quite well when only viewing analog channels, averaging around 20% CPU use on the Sempron system and slightly less (15%) on the 4000+ (which we didn't capture). Recording with the Fusion5 kicks CPU usage up to about 60%. That's a pretty big jump, but we were running with a maximized viewport (720p content), and we didn't experience any stalls or other problems. For watching an analog channel, the Fusion5 is a close second to the Theater 550, though it's a distant second for recording the content.
Finally, we have the MyHD card. Similar to the Fusion, watching an analog channel uses less CPU time than recording the channel. However, less in this case means about 40% CPU usage. Enabling even the fastest DivX encoding (with MP3 audio) kicks the CPU up to almost 80% usage. The video capture doesn't cut out, but there does seem to be a slight delay every second or so. Even though there is supposedly 20% idle CPU time, the capture isn't really clean, and it's about equal to watching TV with only mediocre reception. We don't have a graph, but moving to the 2.6 GHz X2 chip with Normal quality DivX encoding still used a whopping 65% of the CPU time - remember that there are two cores, so that would be over 130% CPU time for the 4000+. The faster X2 also removed the slight lag in the video, all while producing a much better encode.
Here's another version of the data, with a few additional reference points. The graphs are colored according to performance: green indicates that there were no difficulties, yellow means that there may be periodic issues, and red means that the result was unacceptable.
Digital Channels (watching on left, recording on right)
| FusionHDTV DxVA | |
 |
 |
| Click to enlarge. | |
| FusionHDTV non-DxVA | |
 |
 |
| Click to enlarge. | |
| MyHD | |
 |
 |
| Click to enlarge. | |
Moving to digital channels removes the Theater 550 from the list, but we have both the DxVA accelerated and non-DxVA modes in FusionHDTV. Most modern GPUs can use the DxVA acceleration, so consider the non-DxVA charts to be a worst case scenario. All of the results were taken while viewing a 1080i football game broadcast.
With DxVA enabled, watching 1080i with the Fusion5 requires about 20% of your CPU time. Recoding the same content appears to add about 10% more CPU use, and we get close to 30%. The non-DxVA mode requires quite a bit more power to handle the decoding of the digital signal, using 45% of the CPU power. As before, recording the same content appears to add another 10%. Remember that unlike analog signals, digital TV is already compressed. Decompressing is quite a bit easier than compressing a signal, which is why a Sempron can handle 1920x1080 video at 30 FPS without trouble, whereas the same CPU chokes when attempting to encode even 640x480 video at 30 FPS.
Viewing digital content with the MyHD is completely different from using the card in analog mode. CPU usage on the 4000+ was around 25% when viewing a 1080i signal, though disabling the overlay screen knocks that down around 15%. The record mode oddly enough has consistently lower CPU usage, at around 12 or 13%. The hardware acceleration of DTV decoding that the MyHD card provides is clearly put to good use.
Again, here's the simplified graph. All of the results are green, as there were no difficulties. Judging by the CPU usage on the Sempron for the MyHD, most of the CPU usage is used by the overlay window.
Facebook
Twitter
RSS
77 Comments
View All Comments
Brian23 - Thursday, December 8, 2005 - link
I'm supprised more people aren't talking about this. I haven't downloaded them, but I assume they are recordings of OTA broadcasts. If that's the case, then I'm suprised Anandtech posted them. In the news section they're always posting stuff about how people are getting sued for sharing movies. Most of the movies out there on torrents have been broadcast OTA at some point which would make them public domain. Everyone is so quick to judge someone when they post a torrent of something, but mysteriously no-one has said anything about this yet.JarredWalton - Thursday, December 8, 2005 - link
They're basically 30 second clips demonstrating analog/HDTV videos. If anyone wants to come after me for posting sample content... wow, they have way too much time on their hands. I guess I didn't bother to get "express, written consent" though....Brian23 - Thursday, December 8, 2005 - link
I don't fault you for posting this, and I don't think you need express written consent. I think the whole lawsuit thing over p2p file sharing is a bunch of crap. My point I was trying to make in my previous post was that I was suprised that no-one was was posting anything about copywrite violations. Normally the people here at Anandtech scream and yell if there's something being distributed that's copywritten, no matter what it is. I think people should be able to share stuff that they recorded from OTA.valnar - Wednesday, December 7, 2005 - link
Perhaps I missed something in the review, but what's the point of posting screen shot comparisons of a transport stream? They would all be identical across all the cards, unless there were bit errors in the stream. Of course they all look good! They should also look the same too. The only way to show a difference would be through a video card, such as a particular ATI or nVidia in either overlay or VMR9 mode. Or with the hardware output of the MyHD series of cards, which is vastly superior to the software based cards. If you have an HDTV and have used the DVI or Component out of the MDP-1xx, they are fantastic. Of course, a screenshot wouldn't show that. Maybe a digital photo or something(?).It should also be noted that watching HDTV on overlay is a big no-no. The overlay was never meant to register more than 540 lines of resolution. While deciding between overlay or VMR9 for DVD (740x480/NTSC) watching is a matter of religious debate, overlay can never resolve HD properly. I would submit that the reviewer possibly didn't see any of those cards in their full glory.
-Robert
xtknight - Wednesday, December 7, 2005 - link
Where is this overlay limitation stated and what happens when you go above said lines? I'm sorry but I find that hard to believe. SIL overlay vs. SIL VMR9 looks the exact same to me as I remember it. What does 540 lines mean? What resolution is the max (x by y)?JarredWalton - Wednesday, December 7, 2005 - link
I'm a little confused by this overlay limitation as well. I mean, obviously the Fusion5 only works through overlay, but the video stream is transferred digitally from the TV Tuner to the graphics card. There should be no limitation other than bandwidth, I thought? Maybe I'm using the term "overlay" incorrectly here?As for the HD images, I mentioned that they are all identical for TP captures. They were included to show people exactly how much data HDTV contains (and how much better looking HDTV is) relative to analog/SDTV. The upsampled SDTV image is also there to show how poor that looks in comparison to a real HDTV stream - and I get far too much upsampled content on most HD channels.
valnar - Thursday, December 8, 2005 - link
These limits are discussed in all the usual forums, like avsforum.com.JarredWalton - Thursday, December 8, 2005 - link
Link please? I just don't see how the hardware overlay interface on your video card could possibly be limited to only 720 x 480. Windows Media Player uses the overlay mode for playing video. Last I checked, it's fully capable of playing 1080I resolution. NTSC signals have nothing to do with overlay mode inside computers.valnar - Thursday, December 8, 2005 - link
It not limited per se, but you won't see much of an improvement at a rez above 540 lines horizontal. Everything (like WMP) can use the Overlay, but if you were to truly show a video in 1080P mode, for example, it wouldn't be as sharp - because of most video card overlay limitations. It's hard to explain, as the resolution of overlay is more an "analog" style of limitation, to use that metaphor inaccurately. It simply doesn't resolve the detail of HDTV, period. Get a 100" screen and a projector and it becomes obvious. Forgive me for not being able to articulate it, hence my recommendation to visit Home Theater/HTPC specific forums.xtknight - Thursday, December 8, 2005 - link
I can't find anything about this limitation with google or avsforum search. Are you mixing up overlay with interlacing? It would be nice if you could provide us with a link because we effectively have no clue what you're talking about here...