1 Topic by Potscrubber (edited by Potscrubber 2014-06-04 02:30:29)

  • From: Aotearoa
  • Registered: 2007-10-23
  • Posts: 255

Topic: XT E-PCIe express card driver performance - Resolved

Skip to post 11 to see the resolution on this.

Hi All

I've just today received a Matrox E-PCIe card and done some test recordings with my Thinkpad.  It's working well, though initially I can't get the buffer size down that low without lost asio buffers (LAB's) in Sequoia.  So far, 64 ch @48k or 32 ch @96K cannot go below 512 buffer size without LAB's. 

DPC's are much lower with E-PCIe as opposed to USB 3.  I am using a Molex 3m E-PCIe cable (that should be ok right?).  Recording to a USB 3.0 drive.

I'm just trying to get a feel for the throughput / buffer size capability and initially it's not as low as I expected on E-PCIe.  Anyone else care to chime in and share their experiences?

Thanks
Simon

Madiface XT, Madiface, 3x Micstasy, ADI8QS
Sequoia 17, W10 x64
https://bsound.co.nz/tools-nix

2 Reply by Randyman... (edited by Randyman... 2014-05-22 00:42:23)

  • From: Houston, TX
  • Registered: 2007-05-11
  • Posts: 1,192

Re: XT E-PCIe express card driver performance - Resolved

I'm running mine on a W7/64 Sandy-Bridge-E X79-Sabertooth Desktop system @ 4.7GHz with a Sonnet E-PCIe card and a 2m Molex Cable.  Samsung 840-Pro SSD System Drive, and a fast hardware RAID-5 array for Audio.  The busses (excluding PCIe bus) are also running much faster than stock (I love Sandy Bridge E!!!).

I run the driver at 32 Samples without dropouts.  I do notice quick pops/crackles when stopping my DAW's transport with tons of channels armed, but no issues while recording or playing.

I stay slaved @ 44.1K.  I'm using 2 optical MADI connections and have Nuendo setup to "see" around 88 Ins and 48 Outs (rest are inactive in Nuendo).  I can't recall if I tried with all 196/198 channels visible in Nuendo, but I assume I did before tweaking everything - don't recall any Audio Buffer Issues on my end (and that would not affect the PCIe traffic or system level ASIO buffer, just Nuendo's own handling of the driver's many channels).

Seems that this baby needs a pretty serious system for absolute best performance at lowest latencies - but I'd hope for better than 512 with an i7 system.

That's all I got for now - LMK if you have any questions on my specific configuration.

cool

MADIface-XT+ARC / 3x HDSP MADI / ADI648
2x SSL Alphalink MADI AX
2x Multiface / 2x Digiface /2x ADI8
  • From: Aotearoa
  • Registered: 2007-10-23
  • Posts: 255

Re: XT E-PCIe express card driver performance - Resolved

Thanks for chiming in Randy.  Wow that's a grunty system you've got there, glad you can achieve such low latency! 

Maybe I'll give it a go on my P67 i7 2600 desktop and see how it compares.

Madiface XT, Madiface, 3x Micstasy, ADI8QS
Sequoia 17, W10 x64
https://bsound.co.nz/tools-nix
  • From: Houston, TX
  • Registered: 2007-05-11
  • Posts: 1,192

Re: XT E-PCIe express card driver performance - Resolved

Indeed - That 2600 might have a better chance for success - desktops in general tend to run faster "clock-per-clock" with their full-sized/full-power chipsets and ample power regulation opposed to a laptop's space-and-power compromised components.  What are the specs on your laptop?  I assume you tracked down an E-PCIe Express Card for laptop use?  Pretty cool!

cool

MADIface-XT+ARC / 3x HDSP MADI / ADI648
2x SSL Alphalink MADI AX
2x Multiface / 2x Digiface /2x ADI8

5 Reply by Potscrubber (edited by Potscrubber 2014-05-22 02:49:00)

  • From: Aotearoa
  • Registered: 2007-10-23
  • Posts: 255

Re: XT E-PCIe express card driver performance - Resolved

Yeah.. the desktop will just be an experiment though: I actually want the XT / Thinkpad combo for location recording.  The same TP runs very well with the old Madiface, I should do some comparison testing there (I'm not saying it runs bad with the XT, just that I expect lower buffer possibilities when running E-PCIe).

The TP is pretty high spec (and with a partition optimised for audio work):
Intel Core i7 3520M @ 2.90GHz, W7x64, 8G.  Recording drive is external USB-3 7200 rpm.
NVIDIA NVS 5400M / Intel HD Graphics 4000 w/ Optimus  :: actually - there's a thought.  I should try disabling Optimus and test with either/or graphic adaptor.  Although the E-PCIe card / port isn't sharing resources in the device manager.

Yeah, I got the Matrox express card listed here:
http://www.rme-audio.de/forum/viewtopic … 431#p90431

I will try it in my studio desktop, that has a PCIe slot to Express Card adaptor.  Major adaptions!

Madiface XT, Madiface, 3x Micstasy, ADI8QS
Sequoia 17, W10 x64
https://bsound.co.nz/tools-nix

6 Reply by Randyman... (edited by Randyman... 2014-05-22 03:22:52)

  • From: Houston, TX
  • Registered: 2007-05-11
  • Posts: 1,192

Re: XT E-PCIe express card driver performance - Resolved

Experimenting with the video config sounds like a good place to start.  I'm not a fan of the Optimus type OnBoard>Discrete Video Swapping on something like a dedicated DAW.  A Gaming laptop or General Purpose laptop - Sure smile  Best of battery life and performance when needed.

From my very limited Laptop experience, the thing is in order to use the 5400M Discrete Graphics I think you still have to have the onboard HD4000 Graphics enabled - so it might make more sense to try disabling the 5400M (if possible) and going with the HD4000 by itself.  I believe the HD4000 is what physically connects to the screen.

Maybe one of the awesome DAW Builders that hang around here will pop in with their experiences using Optimus on Laptop DAW's...

Back to the Desktop - I'd assume the Express Card to PCIe adapter is much less intrusive than the older PCI to PCMCIA adapters (those things sucked for ASIO!).  Hopefully it will not impede or influence your desktop comparison results.

Sounds like you are gonna have a busy few days smile  Best luck! cool

MADIface-XT+ARC / 3x HDSP MADI / ADI648
2x SSL Alphalink MADI AX
2x Multiface / 2x Digiface /2x ADI8

7 Reply by Potscrubber (edited by Potscrubber 2014-06-04 02:10:08)

  • From: Aotearoa
  • Registered: 2007-10-23
  • Posts: 255

Re: XT E-PCIe express card driver performance - Resolved

Some interesting results so far.  To wrap up the video question, it didn't make any difference.  However I will set bios to be either the discrete or integrated graphics and not Optimus - just not sure which one yet.  Would be happy to hear any opinions.  But as I said, the bios video card setting didn't make any difference to LABs (Lost Asio Buffers) recording in Sequoia.

My old Madiface (expresscard version) running on the same machine and boot partition, runs like a total champ.  Full channel count at 32 samples buffer and not a single LAB.  Even when I 'abuse' it during recording like enabling and disabling devices in the device manager.  I agree this is ridiculous, but it demonstrates that I was not able to get it to lose a buffer.

Madiface XT, Madiface, 3x Micstasy, ADI8QS
Sequoia 17, W10 x64
https://bsound.co.nz/tools-nix
  • Ulrich
  • S/Mux specialist
  • Offline
  • Registered: 2008-05-31
  • Posts: 316

Re: XT E-PCIe express card driver performance - Resolved

You write you are recording to an external drive? Does it change anything when using the internal one?

I have TP T530 i7 Intel graphics. I am using the old MADIface, which is running perfectly, as you are stating, too. However, I always record to the internal drive.

I wonder how the "simple" MADIface USB compares.

9 Reply by Potscrubber (edited by Potscrubber 2014-06-04 02:11:12)

  • From: Aotearoa
  • Registered: 2007-10-23
  • Posts: 255

Re: XT E-PCIe express card driver performance - Resolved

Ulrich wrote:

You write you are recording to an external drive? Does it change anything when using the internal one?

I have TP T530 i7 Intel graphics. I am using the old MADIface, which is running perfectly, as you are stating, too. However, I always record to the internal drive.

I wonder how the "simple" MADIface USB compares.

Hi Ulrich.  I was going to just reply to you, "that couldn't make a difference, USB3 bandwidth is so high ... blah blah" but I tested it, and yes, it does make a difference.  Recording to a partition on the internal system drive, I can get buffers down to 256 (64ch @48k) without LAB's in Sequoia (but no lower).  So even though, USB3 drive has plenty of bandwidth, it is a factor with CPU use obviously.  That's a bit of a bummer. 

regards
Simon

Madiface XT, Madiface, 3x Micstasy, ADI8QS
Sequoia 17, W10 x64
https://bsound.co.nz/tools-nix

10 Reply by MC

  • MC
  • MC
  • Administrator
  • Offline
  • From: Germany/Thailand
  • Registered: 2007-04-06
  • Posts: 25,673

Re: XT E-PCIe express card driver performance - Resolved

Recording to external drives is not under our control, so there is no way to 'improve' the driver for that. It looks more like the drivers (maybe BIOS as well) of your Thinkpad need improvement if using USB 3 manages to disturb PCIe...

Regards
Matthias Carstens
RME

MC's Website

11 Reply by Potscrubber (edited by Potscrubber 2014-06-04 02:19:08)

  • From: Aotearoa
  • Registered: 2007-10-23
  • Posts: 255

Re: XT E-PCIe express card driver performance - Resolved

SO, it seems the solution was staring me in the face the whole time.  I got improved low latency performance in Sequoia by enabling MMCSS for ASIO in the driver global options tab.  Further improvement was made by disabling multicore cpu utilisation in Sequoia.  My latest LAB-free test was 64 channels @96K, 64 buffers (the minimum at 96K), to USB-3 external drive for an hour, whilst messing around on the machine trying to make it drop a buffer.  Very happy!

Thanks everyone who made suggestions.  I'll note this on the Magix forum, and also in the MMCSS thread:
http://www.rme-audio.de/forum/viewtopic.php?id=17575

Simon

Madiface XT, Madiface, 3x Micstasy, ADI8QS
Sequoia 17, W10 x64
https://bsound.co.nz/tools-nix