104 replies to this topic

[alfi]

#mount /hdd
mount -t cifs -o user=jack,pass=pw,rsize=8192,wsize=8192,nobrl,forcedirectio //192.168.1.6/hdd /hdd

thanks for the tip, now trying on helenite and seems to work better :-)

previously I was using cifs options rsize=12288,wsize=12888,rw,nolock,soft,udp on garnet, which was giving "watcheable" results even for high bitrate channels (5-6 Mbit avg), lower bitrates were much better. I have tried the same with helenite now and it was not usable (any reason why it does not work with helenite and works with garnet?). also rsize=12288,wsize=12888,nobrl,forcedirectio for some reason does not work..

[dweeb4]

The stb25xx has very limited ethernet performance. But I've heard some rumours that you could improve the network performance of the 500 by replacing come components. So no, it's not all the stb's fault. (as a matter of fact, the 600 performs significantly better with the same chipset).

That's what I'm getting at Pieterg, what is it about the 600, with the same chipset, that allows better performance? Can anyone answer this?

If you mean replacing the capacitors around the ethernet chip in the 500, I've done that & it didn't seem to improve the direct recording results

[Sjaaky]

But I've heard some rumours that you could improve the network performance of the 500 by replacing come components.

This is only true if you've got very very bad network perfomance, like max 200KByte/s. When you reach ±3MB/s directly and ±1.5MB/s for the ftp-cifs roundtrip, the 3 extra capacitors will not help you. Or do you mean other components?

any reason why it does not work with helenite and works with garnet?

The cifs module in Garnet ignores the rsize and wsize setting, it always uses 4096. The new cifs in Helenite doesn't ignore these settings. The forcedirectio and wsize=8192, rsize=8192 give better performance. The reason many people had problems with Helenite is they used their old setting of wsize=16384 or heigher, because it should be better and gives better performance on other boxes, but not on the dm500. Cifs ignored it for previous images, but not anymore.

The new cifs module in Helenite should give better performance than the old one in Garnet, but only when you use the right settings. I've experimented a lot of with different settings. The settings in my previous post work best for me.
If not, you can set wsize=4096 and rsize=4096 and leave out the forcedirectio to mimic the old Garnet behaviour.

With higher wsize it seems the box takes too long to send a packet over the network and is too late to process the incoming transportstream. With higher wsize I get better raw network performance, but recording gets worse.

[pieterg]

But I've heard some rumours that you could improve the network performance of the 500 by replacing come components.

This is only true if you've got very very bad network perfomance, like max 200KByte/s. When you reach ±3MB/s directly and ±1.5MB/s for the ftp-cifs roundtrip, the 3 extra capacitors will not help you. Or do you mean other components?

This must be the fix I heard about. So there's nothing to gain there, in that case.

[dweeb4]

But I've heard some rumours that you could improve the network performance of the 500 by replacing come components.

This is only true if you've got very very bad network perfomance, like max 200KByte/s. When you reach ±3MB/s directly and ±1.5MB/s for the ftp-cifs roundtrip, the 3 extra capacitors will not help you. Or do you mean other components?

This must be the fix I heard about. So there's nothing to gain there, in that case.

And I did the capacitor mod on my 500 sometime ago without any improvement in recording quality.

So can anybody look into their DM600 and tell us the chips you find in there? How does this box perform better than 500 if it has the same chips? Ancilliary components? Would love to be able to put my finger on this conundrum - I don't want to buy a 600 to see.

[knud66]

Dweep it is recommended that you buy a DM600, it is really a wonderfull machine like the DM500.
The DM500 has poor networkstability which makes it almost impossible to make nice recordings to the network. In a DM600 you can build a harddisk to overcome the problems with the network interface and the DM600 is in the new system (OpenEmbedded and i heard also Enigma2) which gives far more possebility's in the future. The DM600 is also build on the STB25xxx. If you want more there you should considder a DM7020. Or may be the DM7025 which is build with a different processor (ATI instead of IBM).

Good Luck.

[dweeb4]

Dweep it is recommended that you buy a DM600, it is really a wonderfull machine like the DM500.
The DM500 has poor networkstability which makes it almost impossible to make nice recordings to the network. In a DM600 you can build a harddisk to overcome the problems with the network interface and the DM600 is in the new system (OpenEmbedded and i heard also Enigma2) which gives far more possebility's in the future. The DM600 is also build on the STB25xxx. If you want more there you should considder a DM7020. Or may be the DM7025 which is build with a different processor (ATI instead of IBM).

Good Luck.

I know knud66, that all other DM models are better than the 500 but it's the 500 I have & I want to find out why it's networking is weak!

Can you open your DM600 & take some pics of the internals to post here? This might give some help!

[dweeb4]

Hey, I found some DM600 internals photos here: http://satellitemani...net/?t=11118123
Looks like the same Asix networking chip - still studying the photos

[knud66]

The DM600 i had was for programming purposes, is gone to the owner, but for sure a good machine. The difference in the price is also small.
But we have to work with the hardware that is in our hands.
I tryed to make pictures in every angle there was possible of the internal of de DM600 but sadly my electronic media left me alone with nothing. (Memory broken)
The modification mentioned early in the thread is also the only one i know to improve the network in the DM500.

[dweeb4]

The DM600 i had was for programming purposes, is gone to the owner, but for sure a good machine. The difference in the price is also small.
But we have to work with the hardware that is in our hands.
I tryed to make pictures in every angle there was possible of the internal of de DM600 but sadly my electronic media left me alone with nothing. (Memory broken)
The modification mentioned early in the thread is also the only one i know to improve the network in the DM500.

Ah yes, I know the mod (see above posts), but I don't give up that easily. If we could identify how the 600, with the same chipset, is superior to 500 for direct recording (not to internal HDD) then we could have some idea where to focus efforts in order to try & improve 500, if possible.

Of course all this would be easier if Dream would release a schematic for the box but this is unlikely

[knud66]

Of course all this would be easier if Dream would release a schematic for the box but this is unlikely

This would be the easy way. :D/

It should be possible to find something like that but i am afraid that only the people how like to copy the good work of Dream Multimedia have the schematic for the box. i really don't have a clue why the performance on the network interface of the DM500 is slow preforming. Perhaps there is a difference in the 'closed' driver source of Dream Multimedia that is making the performance bad.

Sorry for my bad english, it is still not getting better.

[RedFury]

It surely looks like the same network chip (the ASIX) as the DM500 has.
This doesn't have to be a problem, because there's nothing wrong with this chip.
The ASIX can handle up to 100MBit Full Duplex.

You have to find the weakest shackle in the chain, and this can be the processor, driver,
databus between CPU and ASIX etc.

Indeed there IS a hardware problem with the DM500, wich can be modified using the instructions
in this document:

http://www.pli-image...33&getfile=2914

[quilmore]

Indeed there IS a hardware problem with the DM500, wich can be modified using the instructions
in this document:

http://www.pli-image...33&getfile=2914

have you tried it?
has anyone tried it?
I think Dweeb4 said he's tried the capacitor thing but to no improvement

[RedFury]

Yeah I have tried it, and I gained a few KB/sec. Not much, but a little improvement nonetheless.

The success of this mod all depends of what is behind the network.
If you have good quality Cat5e cabling, and short lenghts (like I do) everything will work (mostly).

It's not that DMM made a huge mistake and 'forgot' components in their design,
they just slightly undersized a few. This wasn't found in the prototyping phase because network performance was probably OK
on the test bench (with short cables).

I think that can be a reason why some people have no problems at all, and some have.

But first always make sure you got your network right. Decent Hubs/switches/routers/cables/servers etc.

I recorded a movie last night from premiere (large stream) and it worked flawless.
This was with PLi Helenite Final (new framebuffer) on my modded DM500, over 15m Cat5e cable to a 3com switch, over 10m Cat5e cable
to my DM7025 (over a CIFS share with 8192 buffers with forcedirectio and nobrl.

[Sjaaky]

Some dm500's have a throughput of 200KByte/s. After the capacitor mod they get the 'normal' throughput of 3-4 MByte/s.

[dweeb4]

Redfury,
That's an interesting document - I looked for something along these lines when I first started trying to use my 500 for recording & found the caps mod which I tried but found no improvement - my network throughput is 3-4 MBytes/sec so I don't think this is the cause of my recorded glitches but I will revisit the hardware mod & check it along with the other mods mentioned.

Why I (Edit:now) believe that software is the problem/solution is that when I use the new streamts module I find that streaming is now almost glitch free (a big improvement). AFAIK, the recording function is very similar to the streaming function.

Edit: The other factor is that the 500 & 600 are pretty identical in the hardware but the 600 can record over the network without error

[Roodkapke]

Well, yes and no.

You need to use the read and write-buffers for recording.
Assuming that you use CIFS.
Errorfree recording with a 600 isn't always errorfree.
If you try and record from for instance ZDF or ORF, the 600 produces errors too.
These channels produce up to 8 Mbit/s raw material.
That often results in errors, due to bandwith problems.

[RedFury]

When you have a troughput of say 4 MBytes/sec, thats 32MBit/sec. This is a fairly respectable troughput.
You are correct, there is in this case no need to look for hardware issues further. Although the mod will probable do no harm, it's better not to open the box for the risk involved making this mod.

Recording on the DM500 still pushes the box to it s limits.
Below a list with a few point of attention in recording/playback:

- Short cables (when possible possible)
- Decent network hardware (better take a switch then a hub on your network (a 100MBit full duplex autosensing one).
- DON'T use the (4-port) internal switch of your adsl router/modem. Although this works nice for internet, it won't deliver max performance for the network.
- Keep the numer of plugins/processes on your DM500 as low as possible. (i.e. no subtitle plugins, cron daemons etc)
- Record on decent hardware. Your old 486 won't quite cut the cheese.
- When you record on a PC, make sure no other background processes are disturbing the harddisk with datatransfer.
- When you use a NAS, make sure it can handle the traffic! Buy a decent one and not Brand-X el cheapo.
- CIFS works better thet NFS (why? no idea! in theory it should be the other way around.)
- Try playing with the CIFS read and write buffer sizes and other options.
- When you record to a windows PC, don't use Vista!!!
- When you record to a windows PC, disable all other network protocols except TCP/IP.

To narrow the bottleneck down, connect your recording device (NAS,PC etc) directly to your dm500 with a cross-over network cable.
Try to see if that makes a difference.

When you own both a DM500 and a DM600, and the 600 records perfectly on the same cabling and share as the DM500,
then maybe (and I hope not) DMM changed something in their drivers. We have no sources of them, so we can't see this.

But, trust me, I will not rest until your dM500 also records to your expectation!

[dweeb4]

OK Rood,
It's different (but does it not use read & write to STDout in streams) BUT old streamts = lots of glitches; new streamts = very few glitches. I think this says something about the software helping the cause. So 600 can produce error recordings (I don't own one so didn't know) but these are the exception rather than the rule as is the case with the 500. If you analyse the recorded file using Mpeg2Repair it will show all errors ( even though the file might playback falwlessly)

Redfury, I'm happy you wont rest until my 500 is recording properly. I will be sure to let you know. I already made the caps mod but will look at the others just for the sake of thoroughness. I can tick off all the issues you mention in your list, I find no better performance when I use a short crossover cable between box & modern laptop running XP.

[RedFury]

dweeb4, can you post the output of the following commands?

-> ifconfig
-> mount
-> ps -A
-> lsmod

If the previous list didn't apply to you, you have a very nasy little sattelite tuner!
The reason there is so little rumour about the DM600, is the fact that is has a HDD of it's own.
So Why would you record over the network with it?