Best hardware upgrade to boost performance for reconstruction phase?
AnsweredI have a fairly good PC but doing a lot of high quality meshes which will take weeks or months, so wondering how best to upgrade to speed things up, maybe a budget of under 1000 dollars. As far as I can tell from the RC notes, RAM is not the bottleneck, maybe it's CPU so perhaps there is not much to be done for this system short of replacing it but would love to hear other's thoughts.
Video Card (Gigabyte GeForce GTX 1080 WindForce OC 8GB)
CPU (Intel Core i7 6800K LGA 2011-3)
Mainboard (Gigabyte GA-X99-Designare EX Motherboard)
Memory (Corsair Vengeance LPX CMK32GX4M2A2666C16W 32GB (2x16GB) 2666MHz DDR4 in white)
Water Cooling (Corsair Hydro Series H90 140mm CPU Cooler)
SSD (500Gb Samsumg 750 evo)
Power supply is Corsair HX1200i 80 PlusPlatinum 1200W.
-
It seems CPU is what matters most. Once you get past the depth mapping the video cards don't do much more in the process. You then have the depth grouping etc after that which just max out CPU utilisation.
If you were to bump up from your i7 to an i9 core you would churn through the data quicker. More than $100 option though!
Other than that I've tried my cache and photo source folder on a high end M2 drive. Doesn't seem to make too much difference over the standard SSD I was using before.
-
Hi Andrew,
My rig has only ever had SSDs in it. I've seen my SSD that contains the photos max out 100% utilisation so I'd say a mechanical drive would be a bottleneck there. I ran a project with the photos feeding off my NAS and the image feature recognition took approx 4x longer.
So far I haven't seen my M2 with the cache max out utilisation but in saying that the read write speed of the M2 might be helping out rather than the throughput it can provide.
edit: To answer your question though, the jump from mechanical to SSD seems to give a decent boost. The jump up from SSD to M2 appears to not give as much increase. In your case maybe a high end SSD might be better than a new mobo and M2.
Please sign in to leave a comment.
Comments
4 comments