Productivity PC

forgotmyname

Unless you can think in nanoseconds the latency will make very little difference.

Compare the real life tests with faster and slower RAM and the end results are often the same..  Just because one test that
targets that in particular and gives a different number does not mean it's actually noticeable in real life.

A higher binned CPU will make more difference as could the BIOS on the motherboard. You could probably get a far greater latency improvement
by configuring windows properly and dumping some of the useless elements that autoruns even though you don't use it.

ECC also introduces it's own latency but do you want low latency or reliability?

As mentioned it seems to be a gaming rig your building not a workstation, your building it to save time but with CAS and 3D modelling
surely you want raw power?
It's like saying a lower latency RAM sticks save me 5 seconds over a year if only I had fitted a better CPU that could have saved me 3 weeks
worth of time processing the data..

My previous setup would randomly hang when converting a 3D file, It would be unrepsonsive so I could not open device manager and see
what the error was. Upgrading and the new system could still function whilst the CAD program bogged down, turned out it was using a load
of RAM the machine did not have and it just bottlenecked.    Is 64GB enough???

onomatopoeia99

The thing I'd think about is that gaming cards like the 4090 sacrifice accuracy for speed to get high frame rates.  Workstation cards like the AMD W7000 and W6000 series or Nvidia A6000 don't.

forgotmyname

I do wonder about the accuracy of video cards over the years. I had a set of files used for photogrammetry. I keep the same
files/folder and run new systems through the process to test accuracy and speed.

I found the 2060Super video card to be fairly fast but the results were way off the mark of all my previous tests. So many failed
points the image was nowhere near as complete as all previous and subsequent tests.  The results were slightly variable also.
Not on the speed but the accuracy/completeness of the final output file.

Using the Cuda cores to process the files rather than the CPU gave a large boost in speed but with 32 cores of CPU power
it's not exactly slow to use the CPU alone.  More consistent also.

I have good quality DDR 3600 RAM but run it at 3200 for the extra stability, neglible loss in performance. In use you would
not know the difference.

Am-I-Paranoid

The 12V connector issues are mostly fixed, just make sure it is REALLY well seated and you should be okay.

I wouldn't want one of those i9's due to ridiculous power and heat, but if you'll be doing CPU encoding and can't afford a Threadripper, then it will be faster than the equivalent Ryzen (7950X) for short encodes - but it will thermal throttle fairly fast if you haven't got enough cooling - I think most reviews recommend 420mm AOI or custom water cooling with 420mm rads if you want it to maintain that advantage over the Ryzen9.

You've mentions 64Gb sticks and 64Gb kits - which is it? If you're concerned about latency, then you're probably goign to have to go for 4x32Gb sticks (128Gb RAM)?

This is what I was looking at for my build, but I think CL30 DDR5 for video editing (Cinebench, etc) is probably overkill and gets you no real gains... that's more for gaming (which is what I'm building for).

 https://www.gskill.com/product/165/396/1691400033/F5-6000J3040G32GX2-FX5

Am-I-Paranoid

Corsair just released these 64Gb (2x32Gb) kits too: https://www.techpowerup.com/review/corsair-dominator-titanium-ddr5-6000-2x-32-gb/