My desktop PC is quite old but it does what I ask of it, except it’s very slow when exporting using DeepPRIME because it only has on-board Intel graphics. I have no need to replace the machinet until Win 10 reaches end-of-life. Having said that, having to wait 4-5 mins for a DeepPRIME export is a bit tedious.
Posts here from users like @rrblint and @mwsilvers have reported rapid export times from a GTX 1050 ti based graphics card, even though that’s below PL5’s nominal min spec of GTX 1060. That prompted me to see if I could find a cheap GTX 1050 ti card on eBay. I did, it arrived earlier today.
It took longer to download the latest driver (my broadband is not fibre) than it did to actually fit the card into my machine but now that it is all installed it seems to be working as expected. My first test DeepPRIME export using the GPU on the card took a mere 8 seconds.
That great. Your processing time is much faster than mine. You may have a high spec’d machine than I do. Currently DeepPRIME processing of my Nikon Z fc 21MP raw files takes about 14 second for a single image and 12 seconds per image when batching 10 because of the decreased overhead when batching. DeepPRIME XD processing generally takes 2 to 2.5 times longer.
I doubt it. That first export was of a .CR2 file from my old Canon 400D, i.e. the RAW file was only ~10 MB. A ~ 30 MB RAW file from my Canon 90D has just taken 29 secs.
I can live with that as well, it’s still a huge improvement over no dedicated GPU at all
@mwsilvers I am glad that you were pleased and surprised at @stuck’s figures because I was concerned that figures I posted where higher than those shown by @stuck
My images, BHT and Golf Course, were RW2 or ORF of 20MegaPixels.
I then replaced the GTX 1050Ti with an RTX3070 (but that would have cost somewhat more than a second hand 1050Ti) and the i7 4790K with a Ryzen 5600G and got
The issue with the Nikon images is a problem with higher numbers of simultaneous execution which seems to make DxPL(Win) “stall” and I have created batches of the “Egypt” and “Nikon” performance testing images.
I am sorry that I don’t have to post an apology but your new performance is way better that before and as the 2000 series cards are superseded then some of those will come onto the second hand market at better prices!
Oh, okay. That would account for it. Still, it is a major improvement over what you had at a relatively small price.
Even if you could find a higher end used graphics card at a price you could live with, It might require replacing your current power supply. That was the main reason I chose the GTX 1050 ti. I acquired it new just before the world wide supply chain issues and chip shortage got out of hsnd in 2020. Graphics cards were not nearly as expensive and I could have gotten a significantly better one for perhaps another $175 USD, but my power supply would not have supported it and I would have had to purchase a new one.
@mwsilvers a lot is said about graphics cards and their power requirements and they are absolutely true for sustained graphics loads, e.g. playing games, architectural rendering etc., but DxPL does not (appear to) place a large sustained load on the graphics card!
Tests I did with the i7 and the RTX 3060 showed instantaneous peaks of no more than 350 watts but they were just peaks and were not sustained in any way. Newer graphics cards are using less power, albeit using a “monster” card e.g. 2080, 3080, 4080 and above are certainly going to put a load on the power supply.
My original i7 was running on a 550watt Seasonic Platinum power supply and that has now gone into the new machine (3060 + 5600G) but my old i7 with the 1050Ti now has an old, cheap 550W power supply and appears to be perfectly “happy” (I hope, it is certainly running cool).
It is all relative… When I purchased mine, the GTX 1050 ti design was still available new for somewhere in the.low to mid $200 USD range from several manufacturers, including Nvidia, MSI and others. I have the MSI version. Within a few months The price went up significantly to the mid $400 USD range. The card I would have been able to get for $175 more than what I paid, would have cost a little over $400 USD at the time, and that card within a few months of my original purchase increased to the mid $600 USD range!
Glad you got the 1050 ti. It’s not a speed demon but it’s a whole lot faster than CPU alone and it’s a workhorse and should serve you well for quite a while.
Which is why I decided on this card, it’s a very cost effective upgrade for my photo editing but no gaming requirements.