Page 1 of 2 12 LastLast
Results 1 to 15 of 18

Thread: tivonet to turbonet - upgrade worthwhile?

  1. #1
    Join Date
    Aug 2002
    Posts
    65

    tivonet to turbonet - upgrade worthwhile?

    subject says it all really

    my extraction speeds are about 1:1 at the moment (i.e. 30 mins of video takes just under 30 mins)

    what sort of improvement would turbonet see??


    thanks
    david

  2. #2
    Join Date
    Jun 2002
    Location
    Was Frozen North now Sunny South
    Posts
    351
    I also have a tivonet and am interested in what others who have upgraded have to say. It seems to me that when the Turbonet card first came out folks were talking about a minimal increase in thruput i.e. less that 30% or something. But maybe this has changed w/new drivers?
    Philips Standalone v3.01 w/2-80G drives and Tivonet.

  3. #3
    Join Date
    Jul 2001
    Posts
    657
    I have turbonet on a sa and a dsr6k and have given my old tivonet to my sister for her sa.

    averages using jdiner's newest tytool/tserver_mfs6 in doublesocket mode, direct connection using a crossover cable with no tweaks, all standard tivo services running, running "right outta the box" with "old" drivers:

    dtivo w/turbonet - 1.5mbps

    sa w/turbonet - 1.4mbps

    sa w/tivonet - .9mbps

  4. #4
    Join Date
    Jun 2002
    Location
    Was Frozen North now Sunny South
    Posts
    351
    I can confirm the 900kbps for my SA, so the accept the rest of your numbers Fugg. Certainly a >50% increase is worth looking at!
    Philips Standalone v3.01 w/2-80G drives and Tivonet.

  5. #5
    Join Date
    Feb 2003
    Location
    Northumberland, UK
    Posts
    26
    I am getting very mixed results with my Turbonet card.

    I get 3 very different rates with 3 machines.

    Machine 1 P3 800Mhz Windows 2000 Pro 1.3MB/sec
    Machine 2 Dual P3 500Mhz Windows 2000 server 0.6MB/sec
    Machine 3 P3 800Mhz Win XP Pro 0.3MB/sec

    Tivo tuned to a null channel all machines have 500MB ram but only using a fraction of that. In the case of the Dual processor, only one appeared to handle the transfer.

    Cannot offer any explanation for the variability as file transfers between these machines normally happen at around 6MB/sec

    YMMV.

  6. #6
    Join Date
    Feb 2003
    Location
    Northumberland, UK
    Posts
    26
    [QUOTE]
    I get 3 very different rates with 3 machines.

    Machine 1 P3 800Mhz Windows 2000 Pro 1.3MB/sec
    Machine 2 Dual P3 500Mhz Windows 2000 server 0.6MB/sec
    Machine 3 P3 800Mhz Win XP Pro 0.3MB/sec

    [QUOTE]

    now with another machine same spec as machines 1 & 3

    Machine 4 P3 800Mhz Win NT4.0 1.75MB/sec double socket Tytools r6.2
    Machine 4 P3 800Mhz Win NT4.0 1.67MB/sec single socket Tytools r6.2
    Machine 4 P3 800Mhz Win NT4.0 1.57MB/sec double socket Tytools r6.1
    Machine 4 P3 800Mhz Win NT4.0 1.52MB/sec single socket Tytools r6.1

    Note this is a sample size of 1 so the differences may not mean that much.

    do not have time to investigate why this is atm but I guess I will use NT4 for extraction
    Last edited by akeogh; 02-27-2003 at 04:41 AM.

  7. #7
    Join Date
    Dec 2001
    Posts
    50
    DSR6000 Philips DirecTivo
    Turbonet card
    tserver6_mfs
    2.52 with superzaps Xplusz
    120 gig Hard Drive

    PC - 2.0+ AMD Athlon
    120 gig Hard Drive
    384 Meg DDR Ram
    Tytools6.1r1, double socket mode

    Tivo in kitchen, router and switches in office, 30' away. Connect using cat5 cable on home LAN.

    typical times have been:

    with hub in between (collisions) - .7Mps
    with switch in between (no collisions) - 1.92Mps

    bman

  8. #8
    Join Date
    Aug 2002
    Posts
    65
    I played around with my T60 DirecTivo connecting to 3 different computers and had three different speeds for download. All were close, but fairly consistant.

    #1 - Dual PIII 1 GHz w/512 MB (Intel NIC)
    #2 - PIII 5 MHz w/256 MB (3Com NIC)
    #3 - P4 2.4 Ghz w 1 GB (NIC on board)

    The Dual PIII was slightly faster than the other two. I know that it wasn't because of the dual processors, so I looked at the NIC on the PCs. After disabling the on board NIC on the P4 and installing a Intel NIC the speed increased and it now slightly better than the dual PIII.

    I believe that most of the difference is how the TurboNet communicates with different NICs. Time permitting I will grab a small selection of NICs from the office and test this theory a little more.

    Old7

  9. #9
    Join Date
    Feb 2003
    Location
    Northumberland, UK
    Posts
    26
    Have just looked at transfers with a network analyser and I can now see why NT4 is so much faster than Win2k or XP.

    Data Packet size on NT4 is 1460 bytes whereas it is 536 on the other two machines. Will fix in the morning and retest.

    This could well explain why some people see crap transfer times with Tytools 6

  10. #10
    Join Date
    Oct 2001
    Location
    Out West
    Posts
    3,171
    Originally posted by akeogh
    Have just looked at transfers with a network analyser and I can now see why NT4 is so much faster than Win2k or XP.

    Data Packet size on NT4 is 1460 bytes whereas it is 536 on the other two machines. Will fix in the morning and retest.

    This could well explain why some people see crap transfer times with Tytools 6
    That's not an NT thing. You need to make sure to use Double Sockets option in Tytool to get the full packet sizes.
    SpongeBob is not a contraceptive - Bart S.
    A tabloid, is that one of those really strong mints? - Homer S.

  11. #11
    Join Date
    Feb 2003
    Location
    Northumberland, UK
    Posts
    26

    Bubblelamp

    That's not an NT thing. You need to make sure to use Double Sockets option in Tytool to get the full packet sizes.
    May I respectfully suggest Bubblelamp that you are wrong!

    Perhaps JDiner could comment to this please.

    I have looked at the socket opening request behind the "GET" button in Tytools alpha #6 r2

    NT4 , XP and win2k request a max segment size of 1460

    But XP and win2k
    also have a SACK Permitted option with a length of 2

    So without the SACK option, tserver_mfs6 or 5 will respond with data in 1460 byte packets but with the SACK option the data comes back in 536 byte packets.

    Note the behaviour is exactly the same with Tytools 5r4 so it is nothing to do with double sockets.

    Josh, is this something which tserver_mfs6 can deal with? can you increase the TCP window at the Tivo from 8KB to something greater or does that lie in the turbonet area?

    edit 10:44am why is the forum clock approx 22 mins slow?
    After a bit more research into the TCP rfc , ignore the bit about the 8KB window thats just the limit on what Tivo will accept on the connection.

    The question is why does Tivo drop the packet size to 536 when the sack option is present in the connection request?
    Last edited by akeogh; 03-01-2003 at 06:21 AM.

  12. #12
    Join Date
    Oct 2001
    Location
    Out West
    Posts
    3,171
    Originally posted by akeogh
    May I respectfully suggest Bubblelamp that you are wrong!

    Perhaps JDiner could comment to this please.

    I have looked at the socket opening request behind the "GET" button in Tytools alpha #6 r2

    NT4 , XP and win2k request a max segment size of 1460

    But XP and win2k
    also have a SACK Permitted option with a length of 2

    So without the SACK option, tserver_mfs6 or 5 will respond with data in 1460 byte packets but with the SACK option the data comes back in 536 byte packets.

    Note the behaviour is exactly the same with Tytools 5r4 so it is nothing to do with double sockets.

    Josh, is this something which tserver_mfs6 can deal with? can you increase the TCP window at the Tivo from 8KB to something greater or does that lie in the turbonet area?

    edit 10:44am why is the forum clock approx 22 mins slow?
    After a bit more research into the TCP rfc , ignore the bit about the 8KB window thats just the limit on what Tivo will accept on the connection.

    The question is why does Tivo drop the packet size to 536 when the sack option is present in the connection request?
    That is interesting. Once or twice when I've run Tytool from one of my Win2K boxes, it seemed as if the first connection to the Tivo ran slow. But if I closed the app and started it again, it ran at full speed. I wasn't able to reproduce it consistently, so I never stuck my Sniffer on it. So it seems there's still some handshaking weirdness going on. Thanks for digging into this.
    SpongeBob is not a contraceptive - Bart S.
    A tabloid, is that one of those really strong mints? - Homer S.

  13. #13
    Join Date
    Sep 2002
    Posts
    12

    SACKS setting in Win2K

    Selective Acknowledgement can be controlled via the registry in Win2K.

    ----------
    From http://support.microsoft.com/default...en-us%3B224829

    The SackOpts value in the following registry key can be edited to control the use of selective acknowledgements:
    HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\Tcpip\Parameters

    The valid binary value is 0 or 1, the default value is 1. This parameter controls whether or not Selective ACK (SACK - RFC 2018) support is enabled.


    With SACK enabled (default), a packet or series of packets can be dropped, and the receiver informs the sender which data has been received, and where there may be "holes" in the data. The sender can then selectively retransmit the missing data without a retransmission of blocks of data that have already been received successfully. SACK is controlled by the SackOpts registry parameter.

  14. #14
    Join Date
    Sep 2001
    Location
    West of Bermuda
    Posts
    1,017
    good thread, lot of good info

    i've got the following setup:

    dsr6k dtivo with turbonet running 2.5 s/w
    both tuners on null channel or music choice during xfer
    linksys 8 port 10/100 switch (not hub)
    d-link 10/100 enet card in pc
    900mhz duron win2k pro 384mb ram

    my numbers are all over the map. using tytool6r1, i always get crappy numbers with dbl sockets, single is always faster. if i start with a fresh reboot on the tivo, i can get 1.5mbytes/sec on the first download.

    however, it never gets that good again until i reboot the tivo. i can cycle tytool on the pc, no effect. i can cycle the server on the tivo, no effect. the only thing that speeds it back up is rebooting the tivo. i'm lucky to get 0.8 or 0.9mbytes/sec on subsequent xfers.

    i don't have ths sackopts value in my registry, so i'm assuming it's turned off.

    over on tivocommunity, jafa (turbonet guy) says there's a bug in tivo's tcp stack that (iirc) causes tcp packet sizes to be forced no larger than 536. i don't have packet trace ability on my pc, does anyone have recommendation of free tools for that purpose?

    any other performance related threads around here than anyone could point us to?

    ronny

  15. #15
    Join Date
    Oct 2001
    Location
    Out West
    Posts
    3,171
    Originally posted by ronnythunder
    good thread, lot of good info

    i've got the following setup:

    dsr6k dtivo with turbonet running 2.5 s/w
    both tuners on null channel or music choice during xfer
    linksys 8 port 10/100 switch (not hub)
    d-link 10/100 enet card in pc
    900mhz duron win2k pro 384mb ram

    my numbers are all over the map. using tytool6r1, i always get crappy numbers with dbl sockets, single is always faster. if i start with a fresh reboot on the tivo, i can get 1.5mbytes/sec on the first download.

    however, it never gets that good again until i reboot the tivo. i can cycle tytool on the pc, no effect. i can cycle the server on the tivo, no effect. the only thing that speeds it back up is rebooting the tivo. i'm lucky to get 0.8 or 0.9mbytes/sec on subsequent xfers.

    i don't have ths sackopts value in my registry, so i'm assuming it's turned off.

    over on tivocommunity, jafa (turbonet guy) says there's a bug in tivo's tcp stack that (iirc) causes tcp packet sizes to be forced no larger than 536. i don't have packet trace ability on my pc, does anyone have recommendation of free tools for that purpose?

    any other performance related threads around here than anyone could point us to?

    ronny
    The bug is not in the TCP stack, as an FTP transfer to/from the Tivo uses full packets. This is an issue that jdiner has tried to address with the double-sockets but it didn't solve the problems in all cases. He is aware of it, as I don't let him forget it.
    SpongeBob is not a contraceptive - Bart S.
    A tabloid, is that one of those really strong mints? - Homer S.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •