I’d like to get a solid opinion on this matter. I’ve seen in a few places in the psychopy documentation a warning against using integrated graphics card. I would like to know know big of a problem it really is?
I usually run psychopy on an i7 8gen or i9 computer with at least 16 and 32 Gigs of RAM respectively. My machine comes with an Intel® UHD Graphics 630.
Most of my experiments consist of presenting images or video stimuli, running eye tracking and interfacing with psychophysiology recording (e.g. Biopac MP160) synced using blackboxtoolkit or parallel ports), and I routinely use 2 or 3 displays.
Hi, the only answer to this is that you will need to test the performance of your particular set-up. Your computer itself is less likely to be a bottleneck there than its graphics card, particularly if you are using 2 or 3 displays.
will show that there can be a lot of variability across operating systems and other factors, even when using good equipment. So it is crucial to test performance on your own system, rather than rely on general recommendations.
Hi, they way I understand the guidelines you should rather “necro” an old thread rather than start a new one.
And using integrated GPUs probably always comes back up again.
The issue my lab has, is that we need to buy new PCs. But due to university regulations we have to buy PCs from Dell. Now Dell allows us to either buy a PC with the integrated graphics or buy a graphics card for professional computer designers which only has DisplayPorts. Meaning we would have to buy a new set of monitors as well, as they only have VGA and DVI ports.
Since we don’t have the PCs in advance for testing purposes I wanted to ask what the current experience is with integrated GPUs. Particularly the Intel UHD 770. It does support OpenGL up to 4.5.
We mainly do reaction time experiments where we display simple pictures on a grey background. So nothing too fancy. Reckon we would be fine with using the iGPUs?
Well, unless you test it or find somebody who uses this computer you won’t know for sure. Is there nobody around which has a comparable PC that you could use for testing? What will you use for button presses? The keyboard has quite a delay that adds to your RT-variance.
Would the current Dell motherboards even have VGA/DVI ports? We just bought some of the Precision 3680, which only has DisplayPorts. Flipping through their other offerings, I only saw DisplayPort/HDMI…
If you have old GPUs, you could always move them to the new PCs? Or buy the GPUs you want separately? For one datapoint, my institution cares very much how we purchase complete computers, but is more relaxed about purchasing of individual components.
Older GPUs can just be moved into new computers due how motherboards work. But, depending on the GPU and Motherboard, it might not function the same and potentially have driver issues (like needing drivers for windows 8, but you are running windows 10).
@JensBoelte We can’t test them in the sense of installing our own software first to run experiments. We use standard Dell Keyboards. I am not sure if my boss cares about super precise timings or just that the experiments run at all. I don’t analyse the date, I just gather it, so I never thought about that.
@aforren1 They don’t have the ports natively, but they can be added. Although our IT-Center has told me that DisplayPort and DVI-Digital just need a passive adapter to work. Supposedly there are cables that have DP plug on one end and DVI-D on the other. The signal is mostly the same.
If that won’t work we got the OK from the IT-Center to add our own GraphicsCard later on.
Our university is usually very strict with getting Dell PCs and Laptops, because the contract allows us to buy them very cheap, but requires that we only ever buy from Dell.
@stanley1O1 You’re right, we may be able to just reuse the older GPUs. They are a few years old, but if they ran under Win 10 they should run under Win 11, too.
Testing means to get an oscilloscope, present a visual stimulus at a certain rate and measure its display rate with the oscilloscope. For testing the hardware, you do not need your IT department but some electrician or electrical engineer. Software tests are not worth your time IMHO.
Oh, I see. Well, we don’t have equipment like that and I don’t think we have electical engineers at this university.
I don’t think we need things to be this super specific anyway. And we only need the RT to compare which group is faster or slower. Absolute values are not that important and neither are delays, as long as any they are equal among all groups.
The problem with running windows 11 wouldn’t necessarily be about the GPU itself, it would more so be about the CPU and RAM. If the CPU and RAM are too slow/weak, running windows 11 is going to be a problem.
Also we have the same problems as you with university guidelines on computer purchasing (security concerns). A couple things you can do is buy adapters rather than buy new monitors. And discuss with your IT about a special requisition for a testing computer, then assure them that the computer itself will stay off the network, and any software/updates will be done via usb.
If participant input RT is a concern, I would recommend something like a Cedrus button box. It’s pricey, but psychopy already has built-in Cedrus integration. If using a keyboard I would make sure a) it’s a wired connection and b) it’s USB 3.0, plugged into a USB 3.0 port on the computer, and then limit the amount of USB devices connected.