Hdtv

[phenixdragon]

When HDTV is fully out, will SageTV be able to have the same tuner card to show and record analog and digital w/o having to change between settings? I know this is a big problem with most cards out there now. I think the Access card you have to change if you use their software to watch it.

[stanger89]

None of the current HDTV cards will be able to record analog with Sage since none of them have hardware MPEG encoders on them. Recording HD is easier since broadcast HDTV is already a compressed MPEG stream that only has to be written to the hard drive, Sage can already do this with the AccessDTV. Not quite sure what you mean about changing.

While still experimental, the AccessDTV works pretty much like a 250 in Sage, meaning that Sage records OTA HD with the ADTV and plays it back with the decoder you have selected. The playback seems to the part causing problems right now. However the Sasem Onair HD Editor sounds like it can play just about any HDTV file fine.

BTW, I'm really close to pulling the trigger on an ADTV, especially now that the Broadcast Flag has been passed.

[phenixdragon]

Well what I mean is that while reading about a lot of the diffrent cards and looking for what software I would run to record all the cards will not display digital and analog at the same time. Such has lets say ch. 4 where I live comes in HDTV, ch. 5 doesn't. I would not be able to change from ch. 4 to ch. 5 without going into the software and changing it from digital to analog. I think AccessDTV is the only card that you are able to do that. They pretty much all have analog and digital coax inputs but they just don't play nicely.

Now if I have the ADTV and 250 cards installed would they know which to record with?

[stanger89]

AFAIK, each input has its own channel lineup so you could delete the analog versions of the digital channels from the 250's lineup so the ADTV would be the only source. Hopefully I'll get a chance to try this out someday.

[phenixdragon]

Thats pretty cool then. So would the OSD show all the channels?

[salsbst]

The OSD shows the combination of all channels from all sources. Duplicates are not duplicated, however. It is up to the application to choose which source it will use if you view/record a channel that is available from more than one source. Obviously it is important to allow the user to make choices about whether or not to watch/record digital or not, so hopefully that feature will be available in v2.

[phenixdragon]

Sounds good.

[edmc]

It occurs to me that my single-drive HTPC may be a bit taxed if I were to attempt to timeshift HD content. I recall that 720p/1080i compressed is something like 45Mb/s (that's mega-BITS-per-second). To do timeshifting, I'm going to be Writing one stream at that rate while Reading another. And those two streams will definately be to/from different locations on the disk.

Am I going to require RAID-0 (striping), or can any of the available drives sustain 2x45Mb/s (psuedo random) bandwidth?

For reference, my HTPC has no problem streaming 2 480i MPEG-2 streams from my 2 Hauppauge PVR-250 Tuner Cards while decoding a third stream doing (timeshifted) playback. But those streams are more like 5Mb/s...

[phenixdragon]

Personally I plan on running 1 card per unit, and if I need more then I would put up a server, such as something as this Pundit (if I use more in the future) I would have one card and then network the rest from the server or other units.

[stanger89]

Actually OTA HD is max 19.2Mbps, that's 2Mbytes/sec. My 200Gb Wester Digital can sustain reads of over 50Mbytes/sec that's 400Mbps. Just about any hard drive large enough for PVR duties should be fast enough to timeshift HD.

[CHING2201]

Personally, If cost isnt an issue you could get a bunch of 100-200mb scsi (cheetah's come to mind) and stripe them.

or a few WD Raptors (sata) only problem is these currently are only 36gb big. I've got a couple of them in raid 0 on my machine here, they really scream! They also run hot too, them running at 10000 rpm. Im using a couple 120 GB Seagate SATA drives for my SAGE box, in raid 0. I think they'll handle just about anything throughput wise so they'll likely handle HDTV, cant wait until I get my card!!

JL

[edmc]

stanger89> Actually OTA HD is max 19.2Mbps,...

Indeed welcome news - that'll be quite a bit easier to support :-)

stanger89> My 200Gb Wester Digital can sustain reads of over 50Mbytes/sec that's 400Mbps.

Careful... I believe 50MB/s (a.k.a. 400Mb/s) is the "media transfer rate" for that drive - not a particularly stellar number actually.

But that is not the scenario we expect. Such a "sustained" xfer rate would be possible only for sequential access. The performance must be significantly less for "random" accesses. Using a cluster size of 64KB helps, but it can't get the drive to have random bandwidth equivalent to sequential bandwidth.

I looked up the WD Caviar 200MB EIDE Drive - perhaps that is the drive you have. The pertinent specs are:

Avg. Read Seek Time: 8.9ms
Avg. Write Seek Time: 10.9ms
Buffer to Disk: 736Mb/s (max media xfer rate)

So lets assume the best combo of these: 8.9ms & 736Mb/s

Consider the time it takes to do a Random 64KB Read:

8.9ms to get to the 1st sector to be read (from above)

64KB / 736Mb/s = 0.7ms to xfer the data from media to buffer

ignore the EIDE xfer (which, at Mode 5 Ultra ATA [100MB/s] is much faster than the media xfer rate).

Random Read takes a minimum of 8.9ms+0.7ms = 9.6ms

Now transports that can support overlapped commands can certainly help here. In the least, they can mask the 0.7ms xfer time by overlapping that latency with the 8.9ms seek of a subsequent Read. As far as I know, EIDE drives don't support this - but lets just assume yours does (perhaps it's SCSI).

Hence, each Random Read could complete every 8.9ms. But this means that 0.7ms worth of data takes 8.9ms to fetch for a link utilization of 0.7/8.9 = 8%.

So that optimal 736Mb/s media transfer rate drops to only around 59Mb/s. Again, this is in the optimal case for a "fully random" Read sequence.

In the scenario I questioned (i.e. where SageTV is both Recording and Playing back a stream simultaneously), both Reads and Writes are going on. And recall Writes are even less efficient than reads for this WD Caviar drive. Further, those Reads and Writes are definately to different portions of the disk - thus ensuring that the I/O sequence the drive sees will appear quite random (as opposed to sequential).

Hence, it is great news indeed that current OTA HD streams top out at 19.2Mb/s. Sounds like a single drive might just be able to make it...

[stanger89]

Actually 50MBytes/sec is what I get off my drive for a sustained transfer rate. I tested it with hd_speed. If you don't believe me check out the storagereview.com review of my drive here:
http://www.storagereview.com/article...D2000xB_2.html

Granted that is sequential transfer rate, but most transfers is Sage should be largely sequential.

[RBTConsultants]

I can record 2 high quality Sage recordings (5.5 gb per hour) and 1 HD recording (9 gig per hour) while playing back another HD recording. This is from one XP logical drive which is 2 spanned IDE 120 gb drives on the same IDE channel.

I/O bandwidth is just not a problem.

Bob

[mikesm]

I/O bandwidth is just not a problem.

I think that confirms that we aren't talking about fully random I/O requests.

RBTConsultants, what filesystem are you running on that rig? NTFS? Do you do anything to keep the drive defragmented? How full is the spanned array?

Thanks,
Mike

[edmc]

stanger89> Actually 50MBytes/sec is what I get off my drive for
stanger89> a sustained transfer rate....Granted that is
stanger89> sequential transfer rate, but most transfers is
stanger89> Sage should be largely sequential.

"sequential" or "random" is the question. As I said earlier, I have no doubt that your WD drive can sustain 50MB/s for sequential xfers. Indeed, if your drive is the WD Caviar, your max Media Xfer Rate is over 90MB/s (for those "outer tracks" of course).

RBTConsultants> This is from one XP logical drive which is 2 spanned IDE 120 gb drives on the same IDE channel.

RBT, you are "spanned", not "striped"?

Of course spanning (i.e. concatenation) would help as well provided that the multiple streams/files were placed on different physical drives/spindles. Striping (particularly with a stripe size == 64KB Cluster size) would ensure this.

mikesm> I think that confirms that we aren't talking about fully random I/O requests.

Sure would be nice to know for sure... Or, rather, what we are seeing between best-case sequential and worst-case fully-random. Over in Unix-land, I know there are tools to show this, but I've not been able to locate a tool to report actual fragmentation of a particular file for NTFS.

I guess one simple test would be to just read a file and see what sort of b/w one could achieve - whatever it's fragmentation...

[stanger89]

Originally posted by edmc

Sure would be nice to know for sure... Or, rather, what we are seeing between best-case sequential and worst-case fully-random. Over in Unix-land, I know there are tools to show this, but I've not been able to locate a tool to report actual fragmentation of a particular file for NTFS.

In WinXP, if you run the defrag utility, analyze a drive and view the report it will tell you which files are fragmented and how many fragments make up each file.

[RBTConsultants]

Sorry I didn't get back to this thread sooner. My 2 drives are spanned, not striped, using plain old NTFS. Both are on the primary channel of an ASUS p4pe motherboard built in IDE. I have on several occasions recorded 2 things on SAGETV and something on ACCESSDTV while watching a previously recorded HDTV program using ONAIR HD Editor.

I don't defrag or anything. The disk is primarily used for PVR and shows get deleted as they are watched. I've never seen a problem.

Bob

[djg]

I recently got a second accessDTV and have it in a networked pc running SageRecorder. It is connected as a network encoder to a pc running SageTV with an accessDTV and PVR250. The Sage drive in the server is a dedicated WD 250 GB on a Promise ATA100 PCI card with WinXP Pro on a separate drive on the mb controller (Asus P4P800 with P4 2.66 non-HT). I'm running a 100Mb wired network thru an router/switch. I've also got the Sage Client on the pc with the ADTV network encoder. Experimenting with this setup I've gotten these results:

- The disk has no trouble recording two HD streams and playing back either of the streams on the server.
- The disk has no trouble recording two HD streams and playing back a third HD stream on the server.
- With the server in sleep mode, I can watch either stream on the client with only an occasional stutter.
- If both the server and the client are playing the live streams, I get frequent stutters on the client but the server is fine, but it depends on the resolution of the material playing on the server.

With both client and server displaying, the server cpu is pegged at 100% when the material being played on the server is true 1920x1080i HDTV so the stuttering on the client seems due to the server being unable to keep the network fed. The network runs at about 20% utilization with only recording from the encoder and about 50% with the client playing plus encoding. With the client closed and true HDTV displayed on the server, the server cpu is running at 85%, with SD (640x480p) programming the cpu is at 55%, all while recording two streams. So to make best use of Sage and HDTV, the fastest cpu is beneficial.

Dave

[mikesm]

Are you doing software decoding for playback on the client and server, or using the Access DTV's native videoout for that? If using software decoding, which MPEG-2 codec are you using?

Thanks,
Mike