Best setup for multi-display stereoscopic experiments

1
Dear all,

sorry for crossposting with the PsychoPy mailing list, but I'd like to
hear from both communities.

We're replacing our system for stereoscopic stimulation, which uses
two displays for left/right images. I was wondering if anybody can
recommend a good configuration to have exact timing and sync across
displays. I'd greatly appreciate recommendations on graphic cards.

Thanks for your help.
Matteo

--
Matteo Visconti di Oleggio Castello
Ph.D. Student in Cognitive Neuroscience

Dartmouth College, Psychological and Brain Sciences dept.
6207 Moore Hall, Hanover, NH 03755
office: +1 (603) 646-8665
www.linkedin.com/in/matteovisconti
2
Depends on operating system and specific other needs. On Linux AMD graphics cards with the open-source graphics drivers work very well with Psychtoolbox, and PTB has various builtin tricks to make such cards especially useful for stereoscopic dual-display presentation, high-precision color display, and for precise stimulation timing. The qualification here is that all my development and test cards are from the X1000/HD-2000/3000/4000/5000 series, HD-6000 should work similarly, the hardware design has changed quite a bit for HD-7000 and later, and i never had the opportunity to test with such cards, so there may or may not be limitations on such cards. In any case i'm somewhat involved in graphics driver development for AMD cards on Linux, so support-wise they are a good choice for PTB. Although OSX is a horrible system for dual-display presentation, some of PTB's builtin tricks can ease the pain with AMD cards there as well. Windows Vista and earlier is no longer supported by PTB and Windows-7 and later should be generally avoided for multi-display work. But if you absolutely must use Windows-7 then AMD cards with Eyefinity Technology should be ok'ish. NVidia cards with the proprietary NVidia drivers on Linux generally worked well in the past, again the latest models >= GeForce 400 series are not really tested by me due to lack of frequent access to such cards, probably work well, but around GeForce 400-600 a new display engine was introduced into the hardware, which makes it difficult to extrapolate from past hw to new hw, so who knows?

In any case, using Linux as OS for such tasks should be a safer choice (on average).
-mario
3
Hi Mario,

I think we're going to get an amd HD 6450. However, we have 3 monitors with DVI connections. To make them work, I was thinking to get an active displayport hub to split the displayport into three and then use three displayport to dvi converters. Did you test such a setup and do you think this would work in linux and psychtoolbox?

thanks!
Matteo

On Thu, Sep 25, 2014 at 5:43 PM, mario.kleiner@... [PSYCHTOOLBOX] <PSYCHTOOLBOX@yahoogroups.com> wrote:

Depends on operating system and specific other needs. On Linux AMD graphics cards with the open-source graphics drivers work very well with Psychtoolbox, and PTB has various builtin tricks to make such cards especially useful for stereoscopic dual-display presentation, high-precision color display, and for precise stimulation timing. The qualification here is that all my development and test cards are from the X1000/HD-2000/3000/4000/5000 series, HD-6000 should work similarly, the hardware design has changed quite a bit for HD-7000 and later, and i never had the opportunity to test with such cards, so there may or may not be limitations on such cards. In any case i'm somewhat involved in graphics driver development for AMD cards on Linux, so support-wise they are a good choice for PTB. Although OSX is a horrible system for dual-display presentation, some of PTB's builtin tricks can ease the pain with AMD cards there as well. Windows Vista and earlier is no l! onger supported by PTB and Windows-7 and later should be generally avoided for multi-display work. But if you absolutely must use Windows-7 then AMD cards with Eyefinity Technology should be ok'ish. NVidia cards with the proprietary NVidia drivers on Linux generally worked well in the past, again the latest models >= GeForce 400 series are not really tested by me due to lack of frequent access to such cards, probably work well, but around GeForce 400-600 a new display engine was introduced into the hardware, which makes it difficult to extrapolate from past hw to new hw, so who knows?

In any case, using Linux as OS for such tasks should be a safer choice (on average).
-mario




--
Matteo Visconti di Oleggio Castello
Ph.D. Student in Cognitive Neuroscience

Dartmouth College, Psychological and Brain Sciences dept.
6207 Moore Hall, Hanover, NH 03755
office: +1 (603) 646-8665
4
I assume two displays are for the stereo setup, and the 3rd for monitoring the experiment? If so, you could always export your X-session and view the matlab window from another linux machine (I assume, perhaps wrongly, that PTB is fine with this). More pedestrian, just buy a cheap 2nd video card to host the matlab window for monitoring the experiment. That seems safer than adding an active converter system.



On Sun, Oct 5, 2014 at 1:46 PM, Matteo Visconti di Oleggio Castello Matteo.Visconti.Di.Oleggio.Castello.GR@... [PSYCHTOOLBOX] <PSYCHTOOLBOX@yahoogroups.com> wrote:

Hi Mario,

I think we're going to get an amd HD 6450. However, we have 3 monitors with DVI connections. To make them work, I was thinking to get an active displayport hub to split the displayport into three and then use three displayport to dvi converters. Did you test such a setup and do you think this would work in linux and psychtoolbox?

thanks!
Matteo

On Thu, Sep 25, 2014 at 5:43 PM, mario.kleiner@... [PSYCHTOOLBOX] <PSYCHTOOLBOX@yahoogroups.com> wrote:

Depends on operating system and specific other needs. On Linux AMD graphics cards with the open-source graphics drivers work very well with Psychtoolbox, and PTB has various builtin tricks to make such cards especially useful for stereoscopic dual-display presentation, high-precision color display, and for precise stimulation timing. The qualification here is that all my development and test cards are from the X1000/HD-2000/3000/4000/5000 series, HD-6000 should work similarly, the hardware design has changed quite a bit for HD-7000 and later, and i never had the opportunity to test with such cards, so there may or may not be limitations on such cards. In any case i'm somewhat involved in graphics driver development for AMD cards on Linux, so support-wise they are a good choice for PTB. Although OSX is a horrible system for dual-display presentation, some of PTB's builtin tricks can ease the pain with AMD cards there as well. Windows Vista and earlier is no l! onger supported by PTB and Windows-7 and later should be generally avoided for multi-display work. But if you absolutely must use Windows-7 then AMD cards with Eyefinity Technology should be ok'ish. NVidia cards with the proprietary NVidia drivers on Linux generally worked well in the past, again the latest models >= GeForce 400 series are not really tested by me due to lack of frequent access to such cards, probably work well, but around GeForce 400-600 a new display engine was introduced into the hardware, which makes it difficult to extrapolate from past hw to new hw, so who knows?

In any case, using Linux as OS for such tasks should be a safer choice (on average).
-mario




--
Matteo Visconti di Oleggio Castello
Ph.D. Student in Cognitive Neuroscience

Dartmouth College, Psychological and Brain Sciences dept.
6207 Moore Hall, Hanover, NH 03755