@microapp I just finished installing my orchestral libraries (EW Opus Hollywood * Platinum, NI Symphony Series). Using a complete non-scientific test (aka I selected the entire brass and string sections - each is an EW Opus instance on a separate track - and leaned on my keyboard for a couple of octaves) → with 9ms in 11ms out latency (UR44 driver in standard mode with 256 Buffer size - the system didn’t have any issues “instantly” triggering and sustaining upwards of 1600 voices as displayed in the Opus Engine. There’s a couple of EW Spaces reverbs running as well.
I tried buffer of 128 (6ms latency on the UR44) - it triggered the performance red light of doom with the same test though playing normally with all the instruments playing back it was fine.
If had a more scientific test I’d do one but I don’t have one to hand.
Settings wise there were only three things done that are not default to stop windows 11 being a pest:
Steinberg Power Scheme is ACTIVE (you could use windows High Performance scheme - same effect, just this way Cubase activates it on launch)
I have an NVIDIA Graphics card. In the Nvidia control panel under Manage 3D settings I added Cubase12.exe as a program and then changed the Power Settings to Maximum Performance. That way Cubase switches it automatically when it starts
I have a Gsync monitor - not to sure this one impacted Cubase but it did impact another program I had. In the Gsync settings I left it on the Default of being active only for Exclusive Fullscreen applications. In window mode it was interfering with the other app. I figured Cubase didn’t need any such interference.
Normally I setup an entirely different Win11 OS execlusively for Cubase. This time round since I was experimenting I have set Cubase up in my every day OS - which includes Docker services, running video conference apps, and a whole lot of background processes running. Cubase doesn’t seem to care about these at all.
I ran latencymon for 15 minutes as well (with Cubase active in the loading screen but not doing anything so the power plans are active ) wihtout any issues. With the NVIDIA power management turned on (aka not maximum performance) latencymon will show the nvidia kernel driver causing issues.
Only thing now is to keep using the beast and see if it holds up over the test of time (for my other work involving data analytics the machine is “sheesh that was fast” timing. Some activities are taking well under half the time and I am exhausted trying to keep up!).
robw,
It seems like a beast.
Sounds like the core/thread scheduling is no longer an issue and CB plays well with Win 11 in general. The voice count is in the ballpark as well.
Thanks much for the info. It is exactly what I was looking for.
Enjoy!
The big question is, will it withstand the Omnisphere Phoenix Rising patch test! That thing can hang 16 threads in an older i9 with just that ONE instance of ONE patch!
I only have a 12900k, but on my system that patch uses 13% CPU (Taskmanager) and about the same in the ASIO-Meter in Cubase. Interestingly this is regardless of the number of voices played.
FYI - interesting little test… I just loaded 64 different instances of Omnisphere with Phoenix Rising patch, each playing 8 notes simultaneously and it handled it with just 128 buffer (2.77ms / 3.5ms - RME HDSPe AIO PRO) and only started a couple of glitches after more than 50+ instances. Interesting to see how the cores handle it with only 1, 8 and 9-24 maxed out but still plenty of power left in the other cores… looks like bottleneck here is audio buffer .
FYI - at 256 samples audio buffer (5.44ms / 6.16ms) no glitching at all and Audio Performance peaking around 75-85%
Awesome! And it definitely shows the difference between different editions of Intel processors. The Phoenix Rising test has always been one of those unofficial load tests for me. It easily glitched a $3,000 i9 system with NVME drive patch streaming and 16 GB RAM, a $400 DAW and $1,200 Thunderbolt audio interface… With nothing else running on the system. Yes, I know price is not directly related to quality or performance, but it did make me wonder what exactly Omnisphere required for a smooth operation of their product. I did notice the same buffer bottleneck.
It is a never-ending cycle between the latest and greatest VSTs coming out and musicians having to keep acquiring new hardware to keep up… Sometimes we just want to load up something and enjoy playing music and not get shaken out of the “zone” by glitches, pops, crackles, bitcrushing and tempo drops…
@Dobbo79 Was that test based on recorded and playing back or played live?
It looks like recorded but I didn’t want to assume. Performance is great but ASIO Guard would have been involved and might be why the E-cores are able to tackle some of the load.
In my test I played live with Record enabled tracks - still did great but different load pattern on the cores (no opportunity for ASIO Guard). I suspect a P-Core started to putter out on the realtime part first in my case but it’s hard to tell cause you can’t monitor it at all (to quick in this case) with standard tools.
I’m curious to know how the Thread Director and Cubase are playing together in this. Cubase clearly sets its Process Priority to High but I wondered if Win11 or Cubase is smart enough to put the ASIO Guard threads onto e-cores leaving more realtime stuff on the P-Cores. I think Process Lasso might be able to tell me a bit about this but haven’t been able to look at it yet.
My goal was to have a very low latency project from start to finish…to break free from laggy playback when mixing and although the latency compensation button is of course handy, it can really throw the mix out when a load of plugins are deactivated. I think I’ve basically achieved that now…
I had a quad core UAD-2 PCI-E but I’ve decided to ditch UAD now in favour of greater responsiveness… Essentially because of the added latency using their plugins/DSP - not an easy decision but I feel there are many alternatives that sound just as or nearly as good with little or no latency.
So with this setup and carefully chosen plugins, it’s now possible to keep my OVERALL project latency inc. audio buffer at around 10-20ms and that’s with maybe 40-50 tracks and Softube Console 1 on a lot of those, fabfilter plugins and dozens of others, and I can keep this responsiveness right up to final mix stage, to 95% completion… Then I can loosen the buffer a little and/or activate ASIO Guard and easily activate my (more CPU INTENSIVE) mastering chain whilst running everything else in the same project - a nice option to have… I’m living the dream!!
Couldn’t recommend the i9 13900k and Z790 / DDR5 (and of course the RME HDSPe AIO Pro) highly enough as an upgrade, esp. from 9th gen.
FYI in some benchmarks (I did a lot of before and afters) the multi core performance was 6x(!) i9 9900k. I can whole heartedly confirm this is evident in real world performance too.
Outside my music related hobby, I’m seeing similar 6x performance figures compared with my Overclocked 9900k.
I have my 13900k OC’d up to 6.3 GHz on 3 cores, 5.9 on 4 and 58 on all core load with 4.6 on the eCores. For light loads this works great but for heavy loads with much threading, like Cubase under load, the P cores tend closer to 5.7 and the eCores 4.6. I say tend because everything is heat related so under heaviest loads (which my work produces but not Cubase) I get 5.6 p and 4.3 e - still screaming along nicely.
Oddly enough I a looking forward to winter here because being so cooling reactive lowering the ambient temperature results in greater performance!
Whilst this is all good news for us there are some users reporting issues with clicks and pops due to high ISR and DPC latency. It looks to me like some motherboards make a significant difference… Unfortunately there isn’t enough information to confirm as people aren’t reporting their motherboard choices enough to make the picture clear (to much focus on only the cpu). I’m using an ASUS ROG Strix z790-e gaining wifi - that might help some folks find the right combo.
