I have had a very settled spell in Cubase 8.5.20 where the real time peak meter was almost inactive. In C9.1 it behaves like a hyper-active child. Leaping all over the place. It is quite disturbing that this issue has re-surfaced. Certainly not worth £80. I will revert to 8.5 until this issue is sorted out. A bit of a disappointment.
That’s a bummer - sorry to hear you’re having issues. For me I’m getting a tiny bit better CPU use overall than 8.5 at all buffer settings, which is nice but not nearly as much as needed yet to compete with some other DAWs out there that I’ve used or tried on the same computer.
Hi silhouette, at which latency are you getting the peaks? I’m asking because I have that same CPU (5960x), and my experience with Cubase 9 is quite similar as yours when trying to play virtual instruments at very low latencies (64 samples). In my particular case, if I let hyper-threading active in the BIOS, I get severe drop outs just tinkering with Monark in a new empty project. However, disabling hyper-threading makes the real-time meter behave much steadier at those same low latencies. We have been talking about the possible causes of this issue here: www.steinberg.net/forums/viewtopic.php?f=252&t=107214
I have been running my projects in 8.5.20 at 128. Actually I don’t need to because I use the Kemper Profiling Amp for all my guitar stuff and could run it a bit higher. However I did not have any issue with spiking until today when I moved to C9.1. I have had Hyperthreading enabled without issue thus far. So what has changed in C9?
I would be a bit nervous of changing stuff in the BIOS as that would be above my pay grade.
Try changing Audio Priority setting under Audio Device settings to boost. When I changed this traltime asio peaks was reduced considerably. I don’t know what other problems this change could lead to though. I guess something else will suffer because it is set to normal by default.
C9 has a better, more reaslistic, graphical representation of load in the ASIO meter.
In 8 it was like 2-3% and then you could suddenly redline without a reason, or even worse not redline at all, but get sneaky subtle drop-oups without warning, the meter being all jolly at 5%.
So the load is is more accurate, realistic, which is a good thing. On top of that the real headroom actually improved.
Thanks for your reply. Already on boost - I could try normal when I get back from Christmas shopping.
I am afraid that you are talking about your experience with C9. I am having the opposite experience. What you are talking about with C8 is what I am currently seeing with C9 and quite frankly makes it unusable for me with large projects. This behaviour does not happen for me in C8.5.
My computer is exactly the same - no new software other than C9. It looks like I have wasted £80 as I will go back to C8.5. It does not seem that there is anything new in C9 to tempt me to persevere with this problem as I want to spend my time making music, It would be nice if one of the Steinberg techs could give me some advice.
I will be contacting Support after I have tried creating a track from scratch to make sure that the issues are not exacerbated by the import process from C8.
Apologies, Maybe I missunderstood, you talked about the performance meter, I don’t read anything about actual issues? Dropouts, glitches etc? Are you experiencing those too? Or is it just the meter you are worried about?
That is what I have not determined yet. The real Time Peak meter is all over the place - often in the final third. I need to try a new project to see how it is, however it does not feel right on first run. I certainly would not feel comfortable working with that level of activity unless I knew for sure that it did not mean what it is saying!!!
That’s what I tried to explain in my earlier post. The meter is now far more realistic.
The 8.xx was overoptimistic/broken: it never went past 25% and could even peak and cause dropouts with 2-5% load.
The 9 meter gives a much better view of realistic load and real headroom. Which should be a good thing.
Your actual DAW performance probably even increased a bit.
Don’t get worked up about the meter, it’s just more acurate, not an actual higher load for your CPU 
That could explain things. Anyway, it would be really nice if a moderator could give us a deepest insight about this improved behavior of the real-time peak meter in C9, as it is not listed in the chart of new features.
Jorge,
I’ve noticed in your signature that you’ve chosen to disable HT on your 5960x.
With your 5960x you’ll have better performance (double+!!) if you incease your memory to 64GB and turn hyperthreading ON. You are starving that processor with only 16GB of RAM (16GB / 8 cores = 2GB per core and 1GB per thread in HT) . You might find HT off better for your gaming, but definitely not the best choice for audio and video production! (and on the MSI x99 sli you should update the BIOS and drivers to the latest versions if you haven’t already).
At 4.2GHz overclock you are pushing that processor HARD and are likely to see thermal throttling during extended use without some serious and extensive cooling in place. Better to scale back to between 3.75 and 4 for long use stability and performance.
Fully agree with below.
I do hope you are right. However the proof might take a little time to become apparent. I guess I will hold of whingeing until I can back up my argument. Although as a matter of interest who says the more hyper meter is more accurate? Steinberg? I have just managed some recording that hasn’t gone too badly. Although there was an exception when I tried a mixdown first time round. I guess this is a learning game. I will give it a chance.
I had projects that did not peak above 5% though I had random spikes and red light out of nowhere
These projects now run fine and hover between 25% - 50% load. I have not yet loaded any projects that I could print on 8.5 but not on 9
Just want to weigh in here with similar problems with CBP 9.1. My real time peak meter is jumping all over the place, often pegging the red line while project is in pause mode. I’ve got a modest amount of plugins going, but at least half of those are UAD with a dedicated processor. At this point, I can’t even work in the project. The meter pegs, the project stops or the sound drops out entirely. I’ve got my audio card buffer set to max at 1024. I’ve got multi processing activated and ASIO-guard off. The audio priority is set to boost. Any ideas?
I noticed u defend anything around C9 in multiple threads.
Are you a paid promoter?
The performance of GUI and CPU in C9 for OSX got worse than in C8.
We have 15 Mac systems and tested it on all of them.
12 Imacs and 3 mac pro system running apogees sympk MKII.
This version of cubase is the worst in performance Ive seen yet.
No I’m not a paid promoter. 
I use Cubase on Windows, couldn’t tell about MacOs
How do you know this? Is it documented anywhere?