This is all very detailed and super interesting, I'm learning a lot, so thank you!
Since I've decided anyway that I want to test this master/slave thing with Vienna, I'm gonna buy the slave...Hopefully, that will solve most of my issues, together with SSD and samples being loaded/processed on the slave.
Pictus chimed in in this thread as well, saying that a ASUS® PRIME X299-A motherboard would not work allow the i9-7960X cpu to work properly, because it doesn't have enough VRM power.
Other choices I'm left with from the store are:
- Asus Rog Strix X299-E - Asus Rog Rampage VI Extreme - Gigabyte X299 UD4 PRO
Any of these would work/give justice to the i9 processor?
Thanks for all the details on the psu, now it all makes sense to me!
Kaine wrote: ↑Thu Oct 25, 2018 10:43 amYou could get away with a 500w on paper although higher wattage tends to mean the fan will spin up only at higher loads (most switch off under 50% load) and it means you shouldn't have it spinning most of the time, so nice and silent and possibly why your friend suggested it.Grape_Brotherhood wrote: ↑Wed Oct 24, 2018 8:10 pm A friend of mine advised getting at least a 750W power supply instead of 500W. Is it more a matter of Wattage or pins/plugs? Power supplies find me as a total noob in the subject....
You also don't tend to find many PSU's under 750W coming with 8+4's however, so in this instance, it's inadvertently wattage + pins.
3 ranges I tend to favour.
Cheapish = Seasonic Focus Plus Platinum
Mid-range = Dark Power Pro
All out = Seasonic Prime Ultra Platinum.
The higher the efficiency, the less wasted power and heat loss. The Seasonics have a switch to jump between hybrid and full mode (hybrid is as I discussed up top), whereas the Bequiets are using their own fans on aggressively low fan curves which are just decent low noise solutions anyhow and all three should offer twin ATX connections.
If you're looking to stick with Corsair, I think the RMx 750 should have it covered, as well as units from the AX range.
It doesn't slow down the system, the complaint is that they'll spike the driver every few mins briefly, although the spike they throw out the is about 35% of the total DPC load capability. The reason it doesn't overly concern me with DPC in this instance is that It takes a 101% load to run the risk of anything happening and it's more critical over time whereas quick sharp spikes tend to not really impact it, unless there are other problem drivers already in there.Grape_Brotherhood wrote: ↑Wed Oct 24, 2018 8:10 pm He also advised an AMD graphic card, because it causes less driver problems which slow down the DAW (Cubase in my case).
Do you know any good AMD usable in this environment, to run up to 2 monitors on the slave?
Still, my Nvidia preference largely comes down to CUDA support for video editing duties or largely work outside of audio editing. If you've no need for it, then grab the ATI regardless. Although, that leads me to the other problem I have with ATI which is the fact they seem to have given up on passive cards. The last ones that appeared widely were the 230/250's and they no longer have current driver support they are that old and I've been asking about new passives from team Red for ages now.
They've started to adopt fan profiling and creating hybrid cards (spin up only under heavy load), although they are a few generations behind Nvidia with this and their fan scaling hasn't been as smooth in previous testing, although this could have changed as I haven't been paying attention as much recently to current cards. I'd love to see a fully passive or cheap semi passive option appear from them again through.
Any card out there currently can do 2 screens though, so that at least is the easy answer.
Nope, that's a problem.I'll look into some software to monitor cores separately, thanks. The Activity Monitor told me that CPU reached something between 300-400% peak (100 = one core used fully, 200 = two cores used fully, etc...) so in theory, I had still 1-2 cores of headroom, right?
You can't split a channel chain, it's just too inefficient and time-consuming to do within the restricted audio buffer cycles you're working with. So, say you have 10 cores, it'll fire 10 channels at 10 cores in one shot, process them and as each one finishes in turn, it'll then call for more data and another track is submitted to the CPU for processing.
If all the channels are processed before the cycle completes and the next cycle begins then all is right in the world.
If one of the channels can't finish in time, then it crackles... and as we know, there is no such thing as "a little crackle" in audio. As such, that means if one channel overloads a single core, then the end result is one that you wouldn't be wanting.