For the same or less cost, you’d probably be better off just getting a used server with multiple Intel CPU’s loaded with RAM and toss your cost savings toward a server upgrade in say 3 years… Just my 2 cents
Server CPU’s being limited to low bandwidth is a misnomer, the clock speeds are normally dependent on the number of cores as they all have about the same power it’s just how it gets distributed… If you get lower core count CPU’s, you get higher bandwidth per core and older servers running 2 4-6core v2-v3CPU’s can still net you at 4GHZ per core at relatively low cost, you just get way more cores given that you have 2 CPU’s
but the initial price tag is much higher…the problem could be the mainboard as well…
preconfigured systems help a lot in this market segment, but you pay features not needed in a DAW
a workstation machine with comparable specs is a good choice as well, but again,
it costs more than an equal powered desktop system but provides additional vendor support…
and workstations are available with Intel and AMD… and I heard about a new one with ARM CPU… something with fruits 
I literally have a rack of servers that I’ve paid no more than acouple hundred bucks apiece for, all would give a brand spankin new single CPU a run for it’s money at a 10th the cost… I beg to differ
Not to mention with the extra space, you could toss something like XCP-ng on the server and install multiple VM’s on it and use something like Audio Gridder to fly your plugins into your session
I doubt that…
for a DAW it is also important which graphic card is usable… how many USB ports you have and what kind of PCIe sockets are available
There is no “One fits all” system
Are you telling me that I have to put a loud ass pc into my recording room because my little fanless pc that does nothing but record the audio isn’t going to cut it ha ha? Never had any issues myself, or recorded any nasty fan noises in my mics either
Obviously, I only run Cubase on the pc in my recording room and fly in all my plugins from a server stored in another room
What is your problem?
I didn’t told you to do something…
Who’s got a problem? I thought I was posting solutions… Only took me losing one session because a plugin decided to go nutso before I learned my lesson
The only value is in those final audio tracks, I got wise enough to keep those on a low overhead box so if there was ever a crash… It wouldn’t be on that box!
Would just be nice if Steinberg put out their own native version of something like Audio Gridder so I could hold a higher confidence in it, preferably based off an open OS like FreeBSD so a Windows license for every different VM of plugins or VSTi’s wasn’t a requirement
That’s probably true if:
a) You’re sticking to just recording without running heavy DSP/VSTi simultaneously, in which case the load on cores is low… or
b) You’re comparing your server to a slow/old single CPU.
But if you want to run CPU-heavy plugins, and more than just one, while recording, clock speed is going to be important. And it’s highly unlikely that your old server is going to outperform a current generation Intel or AMD CPU.
There’s a reason for why the highest scores in the benchmarks that does exist almost always go to either “prosumer” or workstation CPUs, and virtually never ever server chips.
If the option is either a current generation AMD Ryzen CPU or a couple of old Intel server chips then I think it’s terrible advice to recommend the latter.
Right up until you lose your last hour of stellar takes because your new pc crashed… Older low core count CPU’s can get up to 4-4.5GHZ, at this point in time that’s plenty, but in the event it’s not and the pc hangs you just lose the flown in plugin on said track and not the entire track… After losing some great takes, I NEVER run plugins or VSTi’s on the same box I entrust to recording those takes to tape… But hey, trust your own experience I guess
Not to mention all the RAM you can stuff in a server for VSTi’s…
@Mouga
Hi all
The old times of just relying on clock speed like with your 5hgz 7700k has been replaced by IPC and general architecture for latency.
Cubase uses one thread per track until all threads are used and the doubles them up, generally.
AMD 5000 series chips like risen 5950x and 5900x may provide both low latency and capacity for higher track counts. 5800x would also be an excellent choice to replace a 7700k.
I am on gen 1 threadripper with the worst possible latency and I still get 128samples with my ff800. On my Intel 6700k on z270 I get 96samples with the same project. The new zen three 5000 series chips have significantly less latency than the the Zen 1 and Zen 2 Amd CPUs with slightly better latency and far greater capacity in standard desktop than the Intel counterparts.
The bottom line is AMD is getting faster and Intel isn’t. Ultimately it’s what you comfortable with and how confident you are in a platform change. For me going from duel xeon to threadripper gen 1 was rocky 5 years on I have no hesitation in recommending AMD for high reliability installs.
Going from a 7700k to anything Intel or AMD that is current will get you a significant speed increase, like 2x - 3x depending on your track counts. Higher core count more capable machines generally give more stability in larger project that have higher CPU loads.
Enjoy your new machine shopping.
Edit:
I don’t think you’ll be able to get down to 32samples but you will be able to have more plugins and tracks at a similar latency.
@BungalowBill
If I was running a recording studio with lots of live recording activity, I would very likely follow your approach. And what you’re saying about extending the useful life of older server systems that have been retired from data centers is one of the great tricks to get significant computing power for relatively little money.
And for working with sample based instruments having lots of RAM is totally sweet.
However all of that sweet RAM doesn’t help so much when trying to run a physical modelling plugin that eats CPU cycles for breakfast and needs hardly any RAM. But importantly, not everyone uses or needs such plugins on a regular basis. Those that do, tend to lust for CPU cycles like a crypto miner lusts for graphics cards. 
I have quite a bit of data center experience in my dark past, but my music making has been a mostly solitary exercise with relatively little need for microphones or recording of other musicians (I mostly work with tracks they have previously recorded somewhere else). If I had more of a live recording environment, I would totally follow your approach. Get all of those cooling fans far away from microphones and human ears listening to studio monitors!
If you don’t mind sharing, I have a couple of questions of curiosity for you:
Is low latency generally a priority in your studio? If yes, how low have you been able to get in such a networked environment?
Are you shipping audio between machines (and rooms) as
- digital packets (e.g. Ethernet/USB/Firewire/Thunderbolt)
or as - digital streams (TOSlink/ADAT)
or as - analog audio?
Thanks again for sharing your experience - in addition to music making myself, I’m an enthusiastic student of the myriads of different situations, objectives, environments, approaches and experiences of of other music makers!
There’s really no non-ugly solution to live recording latency involving plugins whether on the same box or flown in… For something like guitars I’ll usually run the guitar through a real amp so they have that direct monitor from the amp and then split the signal so I can grab a DI as well, vocals I’ll normally run through a cue send and drop an all in one type plugin which has low overhead just to flavor the input for vocalist confidence (the direct monitor on your interface if it has one is your friend as well, don’t be afraid to twist that knob)…
As for CPU cycles, using something like XCP-ng allows you to also dedicate multiple cores to a specific VM and have them show up as but 1 fast core to the OS, plus AMD servers are starting to hit the used server market so you aren’t lock into just Intel anymore…
As for transferring audio, I still use the trusty old snakes if I’m moving between analog gear but I make my best attempt to get things into the digital domain as quickly as possible… I used to use pci-e attached boxes but those got too messy, so I switched first to firewire but those were messy as well so now I find myself on USB3 which seems to hold up fairly well but far easier to replace bad cables and such… ADAT used to be that ugly fiend I moved my digital around with but I find myself moving my gear more and more toward AVB, and while it still hiccups every so often it only seems to be maturing
1 Like
@BungalowBill
Thank you for your answers. Sounds like it’s still a game of compromises and continuing evolution - just like most things in life
Good to hear that AVB is getting better!
IPC has always been a factor and so has frequency. But I get your point in that we probably see much larger gains in IPC than frequency bumps. I should stop just typing “frequency”.
I agree with you is what I’m saying.
1 Like
This is a useful comparison from Dom Sigalas & Scan computers using Cubase:
I’ve done a lot of research into this & my AMD system (custom build from Scan - similar to Dom’s but with more storage & slightly upgraded PSU & cooling) will arrive in the next few weeks when the AMD 5950X CPUs finally arrive (I ordered it 2 months ago!).
In summary, AMD wipes the floor with Intel. Even the newest 11th gen I9-11900 is way behind the AMD in multi-core performance. Single core is better than the video above (which uses 2019 9th gen Intel I9) but early tests seem to suggest that this is at the expense of thermal issues.
3 Likes
The times for Intel dominance concerning Cubase use are simply ending in 2021.
It’s as simple as that.
Went from 9900k to 5950x and it is an amazing upgrade. I also have VE PRO machines with a lot of KOntakts that I also built the 5950x machine for. Kontakt unfortunately is clearly not optimized for AMD and performs much better with overall voice count on INTEl than AMD. Cubase runs amazing on AMD.
When it comes to audio processing you are better of with the best single core processing and 4 or 6 cores then lower single core processing and for instance 12 cores.
This is inherent to quality audio processing and even more so for high quality even with multi-thread optimizations. This is about the most taxing task you can throw at any sort of CPU. I could write book about this but i will keep it short and end it with a search tip:
Google : audio processing single core performance vs multicore performance
Or something along those lines and you will find what you are looking for.
Glenn, the thing is that we don’t have many tests and comparisons that are apples-to-apples, and the ones we do have (DAWbench, by Sound On Sound and Scanproaudio) show consistently that CPUs with over 8 cores perform better than those with fewer cores.
It doesn’t mean that more speed and better IPC isn’t better, but certainly if there was an option for a CPU with fewer cores to consistently outperform the other ones it’d show up in those tests. And it doesn’t.