I am a professional still photographer about to build a Win10 editing station around a socket 1151 Intel 7700K processor (Kaby Lake) to replace an IvyBridge 3770 I put together in early 2012. I see a few decisions for which I'd appreciate others' experience.
Like all of us, I want fast response by Lightroom in Import, development tool response, and Export. I have been working for a few years with 32Gig RAM and will do so on the Win10 system; it's clear that this isn't the only key to speed. I have read a lot about "balanced" systems that best deliver this outcome. I usually import no more than 500 images at a time. Often as few as 100. I won't be covering events that generate 2000-3000 images this next 12 months. I have started avoiding importation of JPEGs and doing a little culling before Import.
First question for the community as I start ordering the components for my new build: Will I get a significant, worthwhile, boost in speed by spending the extra $$ for PCI-channel memory as opposed to larger SATA-connected SSD?
One detailed case/discussion here
...suggested splitting storage of recent images into a dedicated SSD as the best way to speed scrolling in either grid or loupe mode (whether previews were minimal, standard, or Smart). Interesting idea.
NVMe PCI-based flash memory vs. SATA SSDs:
I am experimenting with a new work step: copy all files from a card into an SSD, do some culling or organizing (like not bothering to Import JPEGs) and bring files into Lightroom from their new location. Of course, my catalog and previews are on an SSD (SATA). Further, I assign Photoshop scratch disk to my largest SSD.
Future-proofing the motherboard choice calls for a chipset that can work well with the new memory channel. I am researching the M2 connection. I guess I'd buy at least a 512GB second generation NVMe unit like the Samsung 960. I currently have a Samsung 850 1TB SSD on SATA and will move it to the new machine. It's been in use for a year at most, and not every day. But I did see a reviewer say the improvement of NVMe over SATA in real-world use was minimal. The specs for transfer speed make that surprising.
What is your experience?
The current processors seem to gain only 10% throughput if thoughtful overclocking is applied. 10% doesn't seem like a big benefit if I am shortening the life of any of the system elements. I'll use only air cooling to avoid extra expense and maintenance time devoted to fancier cooling. I buy big airy cases.
I am a stills guy who might experiment with video during the next couple of years but do not anticipate becoming a videographer or have assignments require a lot of rendering time. I believe the extra $ for the "K" is worth it in case I do need the 10% in the future. I know the mobo will cost more but usually a better mobo like that has other features that are worth it.
My current mobo (Intel 3770-series "H" board) turned out the have too-few SATA connectors but that's because for a long time I depended on spinning hard drives. Smaller capacity 3.5 internal HDDs are more reliable than giant 8TB monsters internal or external but the whole storage scene has changed. I use an NAS elsewhere in the house for storage of close-in backups. Perhaps I can avoid needing a LOT of SATA ports.
Quadro or FirePro still not worth it for me, it seems. Nvidia 1060, I think. My clients are NOT super-fussy about color - I am way more particular than they are. Adobe still not providing enough routines sped up by GPU to matter to me. I also do not see any important ways the current Adobe code uses the GPU well enough to have the "Use GPU" preference checked.
Thanks for any feed back people would like to provide, positive or negative.