I’m currently researching components for my next composing computer build. Ideally, I’d like to get 256gb’s of RAM. I don’t technically need that much and I know 192gb’s will work just fine, but I want to future proof this build as much as possible. I’m also overhauling my template and which sample libraries I’m using so I’d really like to have that extra 64gb’s RAM overhead just in case I find any more libraries I want to add to my template. I’m fully aware of optimizing sample library players by lowering the pre-load buffer size, which I’m absolutely going to do for this next setup. Not sure how that’ll affect CPU performance.
Initially, I was going with an Intel 14900k which, according to Dawbench, will work very very well for using a ton of VST’s, fx, audio tracks, etc. but a colleague recommended Ryzen and (if I want to go overboard he said) Threadripper CPU’s. He also said Sascha Knorr, the in-house demo composer for Orchestral Tools, uses a Threadripper and 256gb’s of RAM. Dawbench posted their test results for the TR7000 series and it came up a bit lacking. Here’s the link for anyone interested: Threadripper 7000 Series Blog | SCAN UK
I wish I could personally ask Sascha how well his rig works and if he has any issues with it, but alas, I cannot. Anybody on here got any experience with a Threadripper setup?
Also, I’ve read that Cubase benefits from having more cores (partially why I’m looking into the 32 core Threadripper), but I can’t find a definitive answer on whether it’s best to have more performance cores or efficiency cores for a C13 setup (might wait til C14 for a full upgrade. I’ve read C13 is a bit buggy). Any input here is greatly appreciated!
Dawbench is the test itself, the tester was Pete from Scan.
Anyway, what buffer size do you need? If you can live with 256 then the TR looks like it’s killer. You said you were thinking about lowering it though so maybe less good.
One thing that has changed recently is that the next Windows version (as well as a current special update you can trigger manually) is supposed to improve performance mainly on the latest 9000-series Ryzen chips, but also on the Zen 4 based CPUs. So with some luck perhaps the TR chips start functioning on the VI test at lower buffer sizes. Who knows though. I guess I’d also wonder about the value of the TR. The 32-core is 2500 just for a CPU. A decent chunk of change.
Not sure about Cubase specifically.
The new 9000-series CPU should have a review out soon-ish too I think (also from Scan) so that might be worth waiting for… and new Intel chips of course.
All fair points. I’m very curious about how the 9000 series will perform for DAW work. I also forgot about the Arrow Lake release rumored to be late next month. Super curious about those too.
So now, one thing I’m still confused about: the 14900k has p- and e-cores. From what I’ve found, Ryzen CPU’s are just p-cores, albeit 8 total cores less than the 14900k. I’ve read you can work just fine with large sessions on 16 cores, but I’ve also read Cubase wants as many cores as you can give it so I’m confused. Add onto that, I’ve read Cubase specifically wants p-cores, so now I’m totally confused.
Not really.
For DAWs still singlecore performance is the one to look at and where usually the bottleneck is.
Those DAWBench measurements are basically the only benchmarks we can look at for DAW use, but they do not represent real world scenarios.
In those tests they duplicate the same track with the same plugins and then count how many of them can be run on a certain CPU at the same time on a certain buffer size. They don’t take groups, mix busses, or master busses into consideration and so they are basically useless.
As soon as you group together those tracks and let them run through a mixbuss and/or master buss with plugins on them the heavy load will jump from multicore to singlecore usage. Mostly CPU0 will be taxed mostly in that situation. So maxed out you will see an overall CPU usage of Cubase of about 40-70% in task manager (all cores together, averaged), but your first core will be maxed out and reach its performance limit causing crackling and artifacts (in the performance viewer). Thats why often people don’t understand, why their machine is at its limit, but there is still 50% CPU usage left to be used, according to task manager.
I personally don’t want any hybrid CPU architectures (or hyperthreading) in a DAW.
I’m currently watching AMD developments, and my next build will likely be around the Ryzen 7 9700X, which will need a new motherboard, and for that I’m waiting to see what appears with the X870/X870E chipset.
This issue has been sorted out.
Also HT will be disabled completely on Intel, so thats not an issue anymore as well (if it has ever been an issue anyway).
I’d wait for the newer 15th gen as @Googly_Smythe wrote and then decide were to go. Actually thats what I am doing.
So far things are looking good.
But this is only true if you actually end up with that one long thread though. For any workload where that doesn’t happen more cores are beneficial. And really all it takes in addition to that is for a large part of the load to be AVX512 for example and then things will likely change, again.
And for what it’s worth Vin/TAFKAT did do an alternative DAWbench test with buses involved. Not sure if Pete from Scan is going to use it though. Vin probably will once he gets to it.
ahh ok that makes a lot of sense. So Threadripper in this case is not great. Now I guess it’s really down to Ryzen vs. whatever Intel can come up with, especially later this year.
Oh lord. I knew 14th gen was having problems but I didn’t know said problems were that bad. Well, good thing this build of mine is slated for next year at the earliest!
Something else I’ve noticed as I continue researching. Many DAW-centered pc builders have Xeon processors as an option for their high dollar machines. Obviously, they make more money off you buying a more expensive machine, but I do wonder if these processors are actually a viable option for we Cubase users. Anybody on here got any experience with Xeon processors?
I’m also looking back into the Ryzen 9000 series, partly because of a developing contingency plan. I’ll occasionally look into Arrow Lake, which sounds very promising, but with Intel recently announcing the switch to their 18a node, some are speculating that Intel may be in a disastrous downward spiral if it all proves to be an over-promise-under-deliver situation. If that’s the case, we may not see Intel for much longer, so we’ll need to look elsewhere for computer processors. My immediate thought is AMD. Sadly, there isn’t much to go off of for the 9000 series, what with it being a new processor.
