I’ve got a new computer. Long story short, I swapped a Titan GPU for a GTX 1080Ti (MSI, Aero). Running on the Titan the processing in Step 2 was slower than with my old laptop - but it ran.
After replacing the GPU, while the computer appears to function normally doing everything else, while Pix4D crashes almost immediately in step 2 (Step 1 it completed). I switched down to 4.1.15 in hopes of the older version functioning with the replaced (downgraded) GPU… no luck.
There have been no substantive changes to the computer aside from the GPU replacement. That Pix4D has been unable to work with a higher end GPU is frustrating. That it is unstable now is even more so. I have a large - for me potentially critical - project pending and I need to know I can process data.
If it isn’t throwing a Pix4D error and simply hard crashing in Windows then I would reinstall Windows since you swapped video cards.
It is also possible the last few versions of Pix4D are buggy as most of my projects now hard crash in Step 2 as well. I updated to the preview version as one last step before reinstalling Windows on my machine.
Thanks Adam. It makes some sense. I ran the Titan in the latest version. The 1080ti is nogood in that version, the older version, or the preview version. It’s not flagging a Pix4D error, just saying it stopped working.
And the Preview just hard crashed in Step 1 for me…reformat here I come.
That Pix4D can’t run a Titan GPU would be forgivable if they hadn’t known about it for months already.
That the last 3 versions of the software seem to be this buggy is really difficult to accept. I shouldn’t have to reinstall Windows on a brand new computer for the above.
That I’ve been told to spend more money on the software so we can start to work it out is a bit too much.
Matthew, I feel your pain, having splashed out $4,250 on a dedicated Windows machine for Pix4D. I was getting constant crashes and the computer would simply restart on me. I have to say, the tech support was great from Holden Greene in San Francisco but poor from Switzerland; similar to their general customer service, I find.
Holden pointed out that the drivers that I was downloading from the GPU card manufacturer’s site (Gigabyte) were not as current as the NVIDIA drivers from their GeForce site. This was for a 1080 (not Ti). That was something I fixed, along with turning off any ‘power saving’ settings on my display and PC. So far, everything is working fine and I also knocked down my core processing capability from 20 to 18 as per Holden’s suggestion too.
I’m sure the Pix4D team are focused on making the machine learning capabilities of their platform the best it can be. They should realise that without their customers’ funds there would be no company or product to develop, however, so a little more service and a little less arrogance would go a long way.
Jaime, Thanks for the commiseration. To say I spent 2x that, by way of expressing frustration and bench marking how disappointing the results were when it continually crashed, risks sounding like a childish comparison.
I learned the AMD 1080ti (aero) had a boost program running in the background. The computer’s builder, Digital Storm - who have absolutely excellent customer service by the way - found this and disabled the program. The computer then ran my reference dataset (1375 20MP images) on normal settings (1/2 image, optimal density) in under 2 hours, compared to my high end laptop at 17 hours. The titan ran the dataset in 17.5 hours. That made the $ well-spent. That’s hours saved on each project not waiting for the computer.
Oddly, the computer is unable to process that same dataset on high resolution and density, though the laptop did (42 hours). Suggests to me there remains an issue.
Along the same lines, the GPU is draggy and slow when manipulating the point cloud in the ray viewer. This is as surprising as the failure on high-res.
I’m sure Pix4D are trying to make their software work better. It would be nice to know they managed to work with newer hardware, and nicer to know that they were willing to help a customer even if currently out of license (I rent the software on a monthly basis). My pushing this methodology to my company/'s clients and my own clients benefits Pix4Dbeyond my rental fee.
As we see here and most other people as well, conflict or drivers beyond Pix4D are often the cause of issues. I don’t get paid by Pix4D and I am not here to defend them but we all have to come together and make this stuff work through difficult circumstances.
I build my own machines so a native Win10 install is easy but even the motherboard and video card suite programs cause issues. Honestly, only have the absolute bare minimum installed and only use max performance settings without any power saving.
My Core i9 with 128GB of RAM, M.2 RAID, and dual GTX 1080Ti Extreme machine runs super fast and near flawlessly. But I was doing some Ether mining and had some Windows crashes so when I went back to Pix4D it had issues that only a reformat and clean install fixed. This wasn’t directly a Pix4D issue as it was Windows files but the concept is clear that Pix4D is VERY sensitive to EVERYTHING. I literally only have Win10 and Pix4D installed right now and I went from 10 out of 11 crashes last week to zero now.