How To Benchmark Cubase? Tom & Dom Make a Test to Measure Performance.
Вставка
- Опубліковано 5 жов 2024
- Check out the ZenStation PC here: www.scan.co.uk...
How do you best measure Cubase's performance accurately? Tom and Dom try and make a test project that anyone can run evaluate the performance .
Download the test here.
www.dropbox.co...
Please comment with your results!
Benchmark Result at 32 samples (or as low as you can go) and at 512 samples
Computer or CPU Model
Audio Interface
We don’t need every detail of your computer!
Check Dom's System Review • Should You Upgrade to ...
And the rest of the Shop
www.scan.co.uk...
Here's one to compare to the current top of the range iMac.
Apple iMac - Intel i9 10910 10 core CPU , 32GB RAM, Universal Audio Apollo X8, Cubase 11 Pro
67 Instances at 32 Buffer
78 Instances at 512 Buffer
Hi there - did you see my post? Would be very interesting to hear what your CPU% is reading and your Cubase Average Load meter
i7 6700K @ 4.4 Ghz /
Noctua NH-D15 cooler / Gigabyte Z170x-Ultra Gaming / 16 gb Corsair DDR4 @ 2400mhz / NVida 3gb 1060 - 2 video monitors / Kingston M2 system drive / Samsung 860 evo project drive / Cubase pro 10.5.3 / windows 10 home / Focusrite Clarett thunderbolt 4 pre + Mackie control (original not pro). 26 instances at 32ms buffer 44.1khz 32 instances at 512ms 44.1. The strange thing I found was I couldn't get the system to drop out at all, on 32ms buffer it just choked completely with 27 instances and played without dropouts on 26 instances. Same thing at 512ms buffer. No dropouts with 32 instances and complete choke on 33 instances. All 4 cores were bouncing around between 96 and 99% core temps between 56 and 58 C
Any shot you try the M1?
I got 88 instances at 48 samples (the lowest my Fireface 802 allows)
i5 13600k / MSI Z690-A / 32gb DDR5 5600mhz / 1TB Samsung 890 Pro NVME / Windows 11 / Cubase 12.
3XS TZ790 system - Windows 11 Home - 13900K, ASUS Prime A Z790, 64GB DDR5 RAM, RME HDSP PCIe card using 4.46 driver
195 Instances at 32 buffers .... 195!!
I was blown away :)
Hi, I have similar pc configuration, except worst cpu (i5 13500), Asus Z790 plus wifi, 64GB DDR5 RAM, NVMe ssd disks, RME HDSPe AIO on 4.46 driver, Windows 11 Pro. I have spent last 3 weeks with system optimization settings (bios, Windows, etc) but my numbers are below my expectation - 39 trasks @32 and 61 @512. Can't believe that only cpu makes that huge difference. I have some wrong setting. . . . Can you advise any tips of your PC setting? I tried almost everything I have found on forums RME, Steinberg, gearspace etc. Also tried to change other PCIe slot, did not help. Using Cubase 12 Pro. When I use 50 tracks with different Kontakt instances, it starts glitching....thanks
@@vladislavlicko1387 Hi, unfortunately I can't help - I bought a pre made system from 3XS and I don't know what they do to make the set up so efficient with Cubase. Hope you get it sorted.
@@vladislavlicko1387Try Cubase 11 and tell me if your scores double? Mine did... C13 has a dodgy engine
MacBook Max M3 48GB Headshones through a USB dock (internal sound as away from home) = 186 tracks at 128 Buffer. Not sure how much difference internal card and dock would make but will test again in house.
So tried with scarlet 2i4 and an additional 10 track. 196 @ 128 buffer. Scrolling, opening menus within cubase all fine.
Nice to see results across platforms. Loving the cheesy funk, too Dom!
CUBASE 11 :
98 tracks @ 512 buffer
CUBASE 13 :
71 tracks @ 512 buffer
Confirms my testing that Cubase 13's audio engine is terribly optimised and badly needs a fix.
Also, good to see 8 year old CPUs still smashing it - you can now buy a matched pair for £84!
HP Z840 workstation, audio optimised, 2 x Xeon 2667v4, 64GB Ram, RME MADI FX
Win 10 PC (2014)
i7-4770K @ 3.50 GHz
32 GB ram
Steinberg UR22 audio interface
16 tracks @ 32 samples
22 tracks @ 512 samples
My PC is showing it's age. Thanks heaps for this benchmarking tool.
I ran the test on my rig. It started to drop out with 35 channels but what I found curious is that increasing the buffer size to 512 made only 2 channels difference. I didn't try to choke the CPU
Specs:
Windows 10 Pro
Intel i7-8700K@3,7HGz
32GB RAM (Kingston HyperX Predator @3333MHz)
Focusrite Scarlett 8i6
35 tracks @ 32 samples
37 tracks @ 512 samples
Excellent, been hoping for something like this.
Here's the results from my seven year old Scan built system (I know but it's lasted better than I expected but I am a mainly audio tracks person).
i7-4770K CPU @ 3.50GHz 32 Gig RAM Win 10
RME Digiface interface
18 tracks at 32 samples
24 tracks at 512 samples
Looking forward to seeing the test results on all of your current systems as I get ready to upgrade (finally).
MacBook Pro M1 8GB RAM
Cubase 11 Pro (running under Rosetta 2)
Mackie Big Knob Studio
32 Samples: 39 Tracks
512 Samples: 45 Tracks
Probably gets better in the future when Steinberg releases some natively compatible version for M1 Macs
Hi.
My recommendation for your tests is to also first check if you have any programs run in the background and close as many as possible. For my tests, I first checked the background tools in the system tray and the task manager and closed all things possible.
I also wonder if activation of the Control Room in Cubase might have any effect on performance?
I have an "older" Ryzen system. My specs are:
AMD Ryzen 7 1700 (8 cores @3.2 GHz), MSI X370 SLI Plus Mainboard, 16 GB RAM, Nvidia GeForce GTX 1050, SSDs: CT1000MX500SSD! and CT525MX300SSD1
Cubase 11 Pro, Steinberg UR44 interface, CC121 connected
ASIO guard activated in setting "normal"
My results:
for 32 samples buffer size 33 tracks without dropouts, total choke with 34 tracks
for 512 samples buffer size 37 tracks without dropouts, total choke with 39 tracks
Intel i7 6700K, 16 GB Ram, nVidia GeForce GTX 1070
Windows 10 Pro
Cubase Pro 11.0.41
-------------------------------------------------------------------------------------
Soundcard: RME Fireface 800 (Firewire 400)
Buffer size (samples): 48 Latency, reported in Cubase Studio Setup: input: 2.109 ms, output: 3.265 ms Result: 22 Tracks
Buffer size (samples): 512 Latency, reported in Cubase Studio Setup: input: 12.630 ms, output: 13.787 ms Result: 24 Tracks
Soundcard: Steinberg UR28M (USB 2)
Buffer size (samples): 32 Latency, reported in Cubase Studio Setup: input: 2.562 ms, output: 2.540 ms Result: 21 Tracks
Buffer size (samples): 512 Latency, reported in Cubase Studio Setup: input: 15.442 ms, output: 16.417 ms Result: 24 Tracks
Hi All
Apple Mac M3 Max - 16 core 16" with 48 GB Ram, 1 TB SSD
Ran test with these results:
Buffer 32 = 196 track
Buffer 512 = 215 tracks
I assume this means my new Mac is pretty decent?
BTW did the same test with a Mac M1 Pro 10 core with 32gb RAM @ 32 buffer and got 80 tracks and 512 buffer 92 tracks - also not bad.
I did this test on a minisforum nbp7 with 32 gb ram and a 1 tb drive, with windows 11 and cubase 13 pro.
Got 65 tracks with 512 buffer and 52 with 32 buffer.
Today have done the test. My System is an I711700KF, ASUS Prime B560M-A MB, 128RAM Kingston DDR4 3200, Kingston M2 SATA 1GB, PNYT600 Graphic, Noctua Cooler, BeQuiet Big Tower, Win 11, Cubase 12 newest Update. System is optimized by my self, curently runnin less than 100 Processes in background. Everything not needed is disabled in BIOS.
My Computer done 60 instances without dropout!!! BS 32 Samples, 44.1 24Bit!!!
Have a nice day!
Sending love from Baltic Seacoast Vivien
News just in
Intel 12900K with 5200MHz DDR5 Ram,
RME Babyface Pro FS
90@32 samples
108@512 samples
It’s Incredibly stable right up until the last instance, no noticeable peaking, from perfect to zero sound with nothing in between.
It's pretty amazing to me just how little performance difference there is here on 12900k between 32 - 512 buffer sizes. I'm wondering if that is more to do with the Babyface? as it's known for the best low latency drivers out there. I use an Apollo and in my projects, i get a big performance increase going from 32 - 512 buffer. Have you published your 12900k findings on the scan site? can't find them if you have. I notice you have an Apollo behind you in this video, would love to know how it compares to the Babyface latency performance wise
@@mrchrisbeaver
Performance will vary more dependent on the amount of VSTi usage compared to dsp.
12th gen round up here .
www.scan.co.uk/info/proaudio/presszone/intel-12th-gen-roundup
@@mrchrisbeaver How much ram?
got 24 tracks at 32 buffer, 31 tracks at 512. also got 9 tracks at 4 buffer and 30/31 seemed consistent for any buffer from 128 and above right up to 2048. i5 8500, 16gb of ddr4 2666, nvmw ssd, windows 11 and Cubase 13 pro. interface was a Tascam us20x20. this test just loaded the asio guard until it failed.
I9 13900K , RME Digiface at 32 samples 139 tracks at 44.1, more it I don't limit the cpu to 253 watts
Since the version 5 cubase has been very easy on any computer . What I want to see is you using any DaW but real monster plugs like NI Reactors ,Rounds at least 8 of those. If not more. Also massive 16 time with heavy patches. Scheps Butchvig voc. 10. H-Verb , Serum 20 instances. Reveal Sound Spire 20 also. Then we will see what your PC is capable of
MBP M2 MAX, Cubase 13, Core Audio
@32: 112 tracks
@128: 131 tracks
Macbook 16 - M1 Pro - 94 tracks
Wow 94 tracks - laptop, silent no heat - these new M2's CPU's must be crazy good!
i7 4790k / msi z97m-g43 motherboard / 16gb ripsaw ddr3 ram xmp 1600mhz / on-chip video / 256gb corsair force gs ssd / 12tb hds / cubase pro 11 / steinberg ur824 + 2 ada8000 / 2 video monitors / I get 24 instances before dropouts and with 25 instances the system is already choked. 32ms buffer. 44khz.
Win 10 pro
Nuendo 11
AMD Ryzen Threadripper PRO 3955WX 16-Cores
128g ram
1x RME HDSP AES 16 (Master)
1x RME HDSPe RayDAT (Slave)
Synced via world clock
102 tracks at 32
128 tracks at 512
the test basically is the same as Dom's all your doing is reducing the individual load of each track. as you add more tracks your increasing the load again at the same rate, so it will just be a ratio percentage difference between the two. staggering the start of notes isnt really real world either as many tracks especially midi get duplicated in large compositions across multiple instruments.
You basically could just add individual midi notes starting at the same time until it drops. that would be more of a performance test of the cpu.
Just for fun, old MBP 15, late 2013 - Windows 10 vs macOS Mojave/Catalina same hardware (i7 - 4th gen, PreSonus AudioBox22VSL, 8GB) 32Buff/2ms
Win10 - 9 Instances at 32 Buffer
macOS - 3 Instances at 32 Buffer
Mac Pro 2019 - 3,2 GHz 16-Core Intel Xeon W, 763 GBs RAM, Cubase 11 Pro, RME UFX+
67 at 32 Buffer // 109 at 512 Buffer
I9-10900K
RME AES audio interface
32Gb, Cubase 11 Pro
32 samples: 28 tracks without red performance light or audible drop outs
512 samples: 60 tracks without red performance light or audible drop outs
Got up to 51 tracks at 32 samples by doing some more performance tuning.
@@LanewoodStudios What did you tune? I have the same machine as you but with UAD apollo. I got 29 tracks same as you. Let me know what you tweaked please!!
@@chrisyking I have to say that the results were not always consistent though, and I also don't know which tuning actually gave what improvement :).
Ok, I now see that the reply I posted with a lot of links is not visible. It may be waiting for review by a moderator. You can also find my insta or facebook via my youtube channel and ask me there via a message.
Hi, just tried this exact test on my Intel 10980XE using a RME Firface 400 @ 48 Samples (lowest) and was able to run 94 channels without absolutely any dropouts. 95+ channels become a struggle though.
impressive numbers there! did you overclock the cpu or base 3ghz?
@@Groovetrapped I'm not a big overclocker, just know the basics really. But I've tried on multiple cpu's to overclock, and never could get one to work without flaws in any DAW I tried(Pro Tools, CuBase, Reason, Reaper, and Ableton Live). And I haven't met anybody yet that has. From New York to Nashville and Chicago to LA! The reason being, something fucked up always happens when it's overclocked! Just like with using Hyperthreading and all that jazz.
@@assshakerstudios549 thanks for the reply! I overclocked my i7 9700k to 4ghz and been running it for over a year now but as you say, nothing has gone wrong but i dont feel i have gotten much extra out of it anyway
W10, Cubase 12pro, Intel i7-9700k-Asus Z370-A-II - 32GBRAM DDR4 - NVIDIA RTX4060 - SAMSUNG SSD 970PRO/860 EVO - Native Instruments Komplete Audio 6 (my TC StudioKonnekt 48 just died :,( )
32tracks at 32 buffer
I need to point out that, in normal condition with an everyday project, there's NO WAY I can use this interface at a buffer of 32. Best I can do is a stretched out 256 but usually nothing less than 512.
I plan to update the PC after I close my latest project (isn't that what we always say?), if and when I will actually be able to do it I'll run the test on the new setup.
Ryzen 3600:
33 instances at 32 Buffer (1.338 ms In / 4.240 ms Out)
44 instances at 512 Buffer (12.222 ms In / 16.939 ms Out)
Ryzen 5900:
76 instances at 32 buffer
89 instances at 512 buffer
32GB DDR4 3200, Windows 10, Cubase 11, MOTU Ultralite Mk 5, ASIO Guard on Normal
Acer Aspire M7720 Win 10, i7 920 2.67GHz 8 core, 12Gb RAM (a 12 year old PC). Cubase Pro 11.0.41, Focusrite Clarett 4 Pre. I run with a buffer of 1024 usually and freeze everything when not being edited.
Buffer 32 Glitches above 7 instances, dies above 8.
Buffer 512 Glitches above 8 instances, does above 16.
I'm looking forward to benchmarking my new Scan XPS Intel 12900K, with 64Gb RAM and Win 11 to see if the improvement is genuinely better that the AMD 5950X from Dom's video as mentioned on this channel.
Apple M1 Mac Mini : 43 duplicates @ 32 buffer. RME UCX (same result @ 128 buffer) RME UCX.
Win 10 Intel 9900k: 64 duplicates @ 32 buffer. (same result @ 128 buffer) Z390 mobo :Antelope Zen Tour SC via Thunderbolt 3
MacBook PRO 2018 6CORE I7 with UAD apollo Twin. RESULTS: 32 buffer: 13 Tracks.512 buffer: 36
Mac Mini M1 16GB/1TB with RME HDSPe AIO Pro. Medium ASIO Guard:
36 instances @ 32 buffer
43 instances @ 512 buffer
Year later, and I'm posting my current numbers today (12/27/22) because I'm building a new DAW tower, this week and I'll report back with those results. once I test that (13900KF, DDR5, etc)
Current system: i9-7940X, 14-core. 28 thread | 128GB DDR4-2667 RAM (8x16GB) | C: drive is Samsung 970 Pro NVMe M.2, PCIe 3.0 | EVGA GTX 1080 8GB FTW2 | MSI X299 M7 Gaming ACK Mobo
Results: same exact results for buffer rates of 48 and 1024 (where I mix): 86 tracks
Interface: RME Babyface Pro FS
OS: Win 10 Pro, v 22H2
We'll see what happens next week!
New tower done. i9-13900KF, PCIe 4.0 Z790 Mobo, 128 DDR5 RAM (5600), all Samsung 4.0 NVMe drives, RTX 3080, Win 11 Pro. Results:
48 Samples 135 Tracks (RME lowest buffer is 48, couldn't do 32)
1024 Samples 149 Tracks (I mix at 1024)
Bi-Xeon E-2580 v3 at 2.5Ghz with 64 GB RAM and a GTX1070 Card : 32 : 62 instances 256 : 76 instances and 512 : 78 instances (not much difference) running Cubase 12 in Windows 10. It's a rather old workstation ( a DELL T7910) great for 3d renderings though with 24 cores and 48 threads
Update : With Cubase 13 the number of tracks before audio cracks fell down to 48 in 32. Going back to Cubase 12. No regrets.
Interesting nerdy test! thanks for that! :-)
Here are my results:
Specs: i9-10980XE @ 3.00GHz (standard settings, no OC) / 64GB DDR4-3200 Crucial 4x16GB / M.2 SSD 500GB Samsung 970EVO plus (System drive) / Audiointerface - RME Multiface 2
@32 samples: it can handle 76 tracks without dropouts and total collapse at 80 tracks
@512 samples: it can handle 84 tracks without dropouts and total collapse at 88 tracks
Not to bad i think, and it still shows, that the intel CPU are a little bit more stabel at freaking high loads, cause when you check some benchmark results, the AMD shuld be way MORE performant in compereson.
When we got Dom to re-run with the new test ( 8:07 ) it was 82 tracks at 32 buffer on the 16 core AMD, that’s exactly what we would have expected in comparison.
IMO, the real problem with the Ryzen 9 is the lack of PCIE lanes available for expansion, especially for fast storage. The only way I see to get the additional storage is to go with a thread ripper board, preferably with thunderbolt, and a 24 core processor. Trouble is, these are as rare as hen's teeth at the moment, and AMD seem to have stalled development on the HEDT platforms. They have announced a new Zen 3 workstation processor, but that's probably not going to be available until next year. (Hopefully, it will use the same chipsets etc., so the current crop of workstation motherboards won't have to be changed.)
Mac Pro 4/5,1 [cheesegrater]
30 instances @ 32 buffer
58 instances @512 buffer
Not bad for a machine built in 2009 [upgraded to 2012 spec]
I didn't reduce background tasks, running apps etc, just sat it right on top of my regular workflow with Photoshop, browsers running.
Cubase CPU monitor pegged but not glitching. Mac Activity Monitor registering 1,950%CPU [which is about 20 cores, give or take, of 24 inc HT.] Overall user activity shows at 85% which agrees with the above.
C10.5 Pro, Win10 Pro, i7-9700k (8C/8T), Base - 3.6ghz, Overclocked to 4.5ghz, 32gb DDR4 RAM OC to 4200mhz. Focusrite Clarett 8Pre USB. (lowest buffer is 16).
32 buffer/44.1khz
32 Tracks - First signs of a drop out
36 tracks - Total Choke
512 buffer/44.1khz
35 Tracks - First sign of a drop out
48 Tracks - Total choke
I'm quite surprised you are only managing 2 tracks more than me at 512ms buffer with double the ram size 32 vs 16 and a 9th gen i7 compared to my 6th gen @4.4ghz. Guess I won't be upgrading any time soon.
@@LemonesqueStudio indeed no need unless you see the benefits. What audio interface do you have? I have a feeling the focusrite drivers are also subpar.
@@Groovetrapped I have the Clarett 4 pre thunderbolt version
Any chance you could do a video on what and how you got your shit overclocked and stable? It is stable right with NO funky shit happening?
@@assshakerstudios549 yep no issues with it. Been running this for over a couple of years. But as I said in another post I doubt if it is bringing as big of a difference or not
Would love to see the 5950X vs the 16/18 core x299 chips overclocked. I get the feeling the latency hit from the chiplet architecture will prevent Zen 3's IPC gains from taking over from those Skylake-X chips (especially running at ~5ghz all core).
Hoping for new Intel HEDT based on Alder Lake with Quad Channel DDR5 next year (haven't been any leaks/rumours indicating that yet though unfortunately).
the 3000 series zen chips had already all but closed the gap (well, when tested with reaper anyway. there's a reason dawbench swtiched lol).
The 5000 series zen chips are even tighter. Zen 3 crushes those intel chips in DAW applications as long as the daw (i.e. cubase) isn't screwing up thread management.
@@misterringer The 3950X was close to a non-overclocked 10980XE at some buffer settings but lost substantially at others. A 5ghz all core 10980XE or 10960X would demolish the 3950X. With Zen 3 bringing ~15-20% IPC gains and lower latency with 8 cores per chiplet I'd love to see the results.
@@AJDOLDCHANNELARCHIVE www.scanproaudio.info/2020/02/27/2020-q1-cpus-in-the-studio-overview/
scan pro's own testing. yes, you can get more from the intel chips with heavy OC, but you need monstrous cooling and power to run those chips at near 5ghz+ speeds.
and again, the 5000 series brought huge IPC and latency improvements. I'd also love to see some hard numbers but I can't imagine Zen 3 got worse :D
@@AJDOLDCHANNELARCHIVE
The 3950X (and all Gen 2 Ryzen) performance at low buffer sizes improved substantially over a year ago once Pete found the sweet spot for memory and timings from the initial testing.
m.ua-cam.com/video/zUDTu6m2ra8/v-deo.html
The IPC gains follow pretty closely against those results, so they definitely jump again. Problem has been that the high end chips from both sides were too powerful for the dawbench test, so it has had to be re-worked (and now only runs in reaper, since cubase started disabling cores with Asio guard off, so couldn’t compare between reaper and cubase in the same tests any more)
@@ScanProAudioTV Thanks for the info! Still if the software is utilising more than 4 cores on Zen 2, or more than 8 cores on Zen 3, there is going to be a latency hit compared to the monolithic-architecture Intel chips (why I'd personally never consider buying AMD CPUs apart from the 5800X, since so many of my programs/workflows rely on low-latency real-time performance).
My bets would be on a 5ghz 10960X (or even delided 7960X from 4 years ago) beating out the equivalent 16 core 5950X at tight buffers in VST heavy DAW projects. My nearly 10 year old 6 core/12 thread X58 Xeon OC'd to 4.2ghz all core, and 12 core/24 thread X99 Xeon (3.3ghz all core boost with BIOS mod) just about keep up with my more strenuous projects of about 100 VST/Sample Library instruments without having to bounce things to audio, but I'm looking forward to the next Intel HEDT to take the strain off... in the meanwhile I'm going to buy a used 7960X and have it delided to tide me over for a couple of years.
Love your guys' testing into these sort of things, I don't think there is anyone else out there doing it (at least to this level) for whatever reason... Cheers!
Mac Pro 2013 - 12 Core Xeon, 64 GB Ram, 1 TB SSD, Steinberg UR824, Cubase Pro 11
29 Instances at 32 Buffer
60 Instances at 512 Buffer
Cubase Pro 11
Windows 10
Intel i7-9700F
16GB RAM
Steinberg AXR4U
32 samples
43 "Duplicate me" tracks, drop outs start at 44th.
I'd like to add a question: I have electrical noise thru my selfpowered studio monitors when CPU works high, I must set the CPU usage at 50% on the Windows energy plan to get them silent. This doesn't happen with my old laptop. I'm looking for another motherboard/CPU/power sypply to solve this. Do you recommend some models ensuring there's no electrical noise as in my Gigabyte 390Z?
Sounds like What you have is a ground loop. Try putting a ART DTI box in line between your interface and your speakers and that should solve it by giving a clean ground signal to the speakers.
@@ScanProAudioTV Thank you. The noise appears when turning PC on, I've just tried some things this morning and the noise disappears plugging the monitors with a no ground (2 pins) adapter. I guess this may be dangerous, so I'll try what you suggest.
99 instances @ 16 buffer size (5.1ms round-trip delay, 222 samples)
101 instances @ 32 buffer size (5.8ms round-trip delay, 254 samples)
111 instances @ 512 buffer size (50.6ms round-trip delay, 2228 samples)
If I overclock my CPU from 8-core performance 52x multiplier to 54x I get 116 instances @ 512 buffer
Specs:
Focusrite Scarlett 2i2
Windows 11 + Cubase 12.0.40
Intel i9-12900KS (stock speeds, .060V undervolted for temperature control, 360 AIO cooling) on an Asus Z690 Hero
32GB DDR5-5200 40CAS
GeForce 3070 OC 8GB
Corsair RM850x PSU
Looking forward to see the upcoming tests from i5 13600k and i7 13700k, so I take the decision between those two!! Any answer will be much appreciated!!!!
Just put together a 13600k rig. I got 88 tracks at 48 Samples (the lowest my Fireface 802 will go). Plenty of juice in the baby. 😁
I5-13600K
32GB RAM 5600
Scarlett SOLO 3rd gen
Win 11 pro
Cubase 12 pro (Asio guard High, 64 bit float, Steinberg power scheme, audio priority normal)
All Cores
93 Instances at 32 Buffer
96 Instances at 512 Buffer
Without E-cores
77 Instances at 32 Buffer
79 Instances at 512 Buffer
Intel Core i7 3770K 3.50 Ghz, 24 GB RAM, Focusrite scarlett 2i4 1st gen.
16 tracks on 512 buffer
Scan 3xS with i9 12900k, 32GB RAM, Win 11 Cubase Pro 10, Focusrite Clarett 8pre TB interface
-32 samples - 79 instances
-512 samples - 98 instances
New build with fresh windows 11 instance - hoping to improve on this as latencyMon is showing DPC issues that I need to work out how to resolve
AMD Ryzen Threadripper Pro 3975WX, 128GB RAM, MODX8, Windows 11, CUBASE 11 - performance was terrible until I disabled ASIO guard, but then I got 110 instances with a buffer of 512 samples or higher, and 72 instances with a buffer of 32 samples
Yes, there’s far too many cores on threadripper for cubase to handle. Asio-guard disabled turns half of them off, so it can start allocating properly.
Ryzen 3900x, 64gb 3200 LPX Ram, Samsung evo 970 500gb m.2, SSL 2, Cubase 10.5 Pro
32 buffer 70 tracks
512 buffer 80 tracks
I'm thinking of getting the same CPU as you.
What motherboard do you use it with, and which graphics card?
By the way,, I have Cubase 10. Were you an ex-user of cubase 10? is there any performance differencein 10.5 ? or is it really just added extras ?
Thanks :)
Hi! Imac 27 2020 i5 6 core CPU,64 ram
Apogee Symphony Desktop,Nuendo 13 pro
46 Instances at 256 Buffer
Mini PC: (712€)
Geekom IT 13, I9 13900H 32GB DDR4 RAM
Windows 11
Wifi ON
Taskmanger Open
Focusrite Scarlett 8i6 Gen 3
Cubase Pro 13
Asio Guard: High
Optimized for Audio in Cubase
Buffer Size: 32
57 Copys
Buffer Size: 128
61 Copys
Buffer Size: 512
71 Copys
My i9 9900KF can do 62 tracks with a buffer size of 64, but I am running an highly customized version of windows 10
I9 13900K RME Digiface USB, 64 GB Ram, Win 11 Pro, Cubase 12 Pro, @32Samples 137 Tracks @ 512 Samples 140 Track Open Retrologue via Audio Gridder..... 279 Instances !!!!.....then the computer had to much background services and struggled...... but Asiometer was 25%
I7 - 12700f 12 core , 16 gb ram , Cubase pro 12 , RME Babyface (old)
Mac os Monterey
136 instances at 512 buffer size.
hello! if you also have windows on that machine could please share the result on cubase/windows just to compare cubase perfomance on mac os vs windows?
@@ardierr.2525yea, would be nice to see on Windows
Excellent video.
I am running the Doctor Mix edition PC and an RME HDSPe AIO. I got 6 channel copies at buffer size 32 and 46 channels at buffer size 256. Running on Cubase Pro 11 on latest Win 10 build 21H1. Should I be hitting higher than that with this system? Spec i9 11900K, Prime Z590-A, 64GB DDR4
Macbook Pro 15 touchbar late 2016 - i76820HQ - Cubase 12.0.30 - _RME UCX - 512 - 17 tracks
2x Intel Xeon Gold 6134 - 544GO Ram - SSD Sata
222 instances at 48 buffer sample (min buffer of my babyface)
i7-9700K , 64 ram = 37 Retrologue+37 Dom (Folders) 512 Samples, UAD Apollo Twin. Win 11.
i7 7820X / ASUS Prime X299-A / 64gb DDR4 / M.2 / RME Fireface 802 (USB) / on NUENDO 11 - chokes at 41 instances, 43 still running but basicaly unsusable.
48ms buffer (can't go lower for a reason) 44.1khz
Which studio desk do you have in the background there?
To be fair you're comparing an 8 core i9 9900k CPU which is now around £300 next to a 16 core 5950x which is around £750 it's still a good improvement jump though and shows what the extra coin could bring to the table. I think a lot of us would be interested in seeing tests comparing Intel 8 core vs AMD 8 core or other like for like (same price range) CPUs. It would be great to see more videos like this with a more refined real world project. It's interesting that Intel had a RTL advantage upto now but it seems AMD 5000 series unified pool has closed that gap. I'm looking at upgrading at the end of the year but waiting to see what Intel has to offer by then especially if the rumors that both 11 and 12th gen will be released this year are true
Now we have decided what a fair test is, the next step is to start to measure all of the 3XS systems (with the same interface, at the same buffer sizes) and publish the results.
We’ll also be doing tests in reaper, studio one and ableton in the future as there are definitely some processors that favour different DAW’s and so users can compare to their existing PC’s.
Very true, it's nearly double the core's, but I still think it's somewhat good comparison since people for quite a long time have been saying that brute single core speed was king with music applications, but now you are being able to run circles around the former 9900k king with a 5600, and a standard R7 37/3800 can match the 5ghz Intel with just 4.3Ghz and that is with the intercore latency of 2 chiplets (wich the 5950 also have, just much better optimized than the Zen2 cores). Will be interesting to see tests with the upcomming next gen TR processors wich of cause is much more expensive, but still cheap compared to Mac Pro's based on intel Xeons.
Its still my opinion that intel is kinda standing still compared to AMD. Neither the 10 or 11 series has managed to catch very much up to AMD, and unless intel puts a rabbit out of the hat with the 12 series, from what we hear about Zen 5, AMD is gonna obliterate intel, unless intel starts moving forward.
On another thing, even they shift the notes in the projects, you still just copy several instances in this test. How many runs projects where you have 4-6 tracks that is copied playing the same notes 10 times ? Would be interesting to see if you made an offset on each new track in a test template up to 128 tracks and see how many tracks more you could squeeze out of having all tracks NOT playing unison with another track
when you build a whole computer, it comes out 1200 vs 1600, only fool intel fanboys (or hackintoshers) would save 25% of the budget to get half the performance. The CPU does not work alone. Not to mention the power usage (thus, thermal output in the studio room and noise) are the same. AMD has double the efficiency.
Interesting to read the scores when looking at pc’s will you publish for mac systems too????
Cubase 11
Mac Pro (Mid 2010)
2 x 3.06 GHz 6-Core Intel Xeon 64GB
Mojave
43 tracks @32 Buffer
Steinberg URC 44C
Asio Guard normal.
i9 9900k at stock speed (No OC), ASUS ROG Strix Z390-E Mobo, 32gb Corsair Vengeance xmp @ 3200mhz, rtx 2070 super, 500gb, Samsung 970 evo 500gb M.2, 1tb Evo 860 and 2tb Evo 870 Samsung, Cubase Pro 10, Steinberg UR44C. 60 instances before dropouts and glitches done on 32 samples buffer size ( 1.2 ms input latency and 2.2ms output latency and 44.1hz sample rate.
hello, i ll do the test with my desktop and laptop, but could you please think about a test with more serial processing (sidechain,plugins on bus, on send, on master track) ?
regards
Apple Mac Mini (2018) T2 Chip, 16GB RAM, 3,2 GHz, 6-Core i7, Catalina 10.15.7, UAD Apollo Twin - Cubase 10.5.2
32 Buffer - 10 channels - % CPU 260,0 - Cubase Average Load 25% - Choked at 44 channels.
512 Buffer - 47 channels - % CPU 1050,0 - Cubase Average Load 100% - Choked at 52 channels.
I think there is something fundamentally wrong with my system. It would be VERY interesting to see if someone with the same machine gets different results.
Appreciate any comments!
Does duplicating the instrument track also duplicate the VSTi instance in the VST rack? If not then you're only testing the increment of compressors and reverbs, no?
Side note: your voice sounds wooly, pop-y/spittyand over-compressed.
can't a cheaper graphics card having a lower VRAM cause glitches in the project with respect to audio tracks
Thanks for testing! Could you also make a video (or maybe share some data/advice) about multi-threading on 5950x in Cubase? This is a bit tricky to test by myself, because when you enable/disable multi-threading, the system sees it as a new/different CPU, and all your system-bound licenses (meaning non-USB-key, so things like Native Instruments stuff etc.) need to be re-activated, which is a big pain. And the reason I'm asking is because some other applications much prefer less but stronger cores (including utilizing mostly just one core!), so I'd love to know if splitting 16 cores into 32 threads gives you a big boost on 5950 in Cubase - big enough to sacrifice sticking with 16 cores=16 threads, which is still quite many cores btw! Thank you and best wishes! :-)
If you are running 16 cores or below there is absolutely no instance that will give you more performance with multi threading off.
Multi threading on, Asio-guard on is the best way to get as much performance as you can in cubase with a 5950x.
Some plugins (such as Kontakt & Acustica Audio ones) do their own load balancing across cores and can limit the potential of the system, especially if you have more than 10 cores.
With plugins that let cubase do the load balancing, you should be able to get 90% + cpu utilisation at any buffer size.
@@ScanProAudioTV Thank you for your reply, that totally makes sense and I do understand that multi-threading 'on' will always be better in those circumstances in Cubase. However what I'm wondering about, is how big of a boost (or difference) is there between multi-threading on and off - because if it would be something like 10-15%, then that might be something worth sacrificing to get 50%+ in single-threaded applications (like gaming, to give an example).
@@HenrykIwan assuming all cores are fully loaded and it’s not limited by plugins, the performance hit will be 40+ % hit without hyper threading . It’s not like you can overclock much more (if any, dependent on cooling) with it turned off.
@@ScanProAudioTV Thank you kindly for this info, much appreciated! :-)
Mac Pro 5.1, 2x3.2 GHz Intel Xeon, 48GB Memory Mojave, Cubase 11 Pro - Run 48 duplicated Instrumental tracks no issues at Buffer 128 Samples...
Asus Prime X570-P MB Ryzen 3950x:
88 tracks at 32 buffer Digiface HDSP latency input .771 output 1.497
Macbook air m1 16gb ram 512 ssd
30 tracks 32 buffer onboard audio (I don't have a compatible audio interface yet) latency input 3.311 output 4.807
With both computers I'm running Cubase with 32bit float
Hello , is that a PCI card HDSP digiface or PCIe card ? In case it is a PCI card , how did you get that running ? Thanks in advance
@@saschakuhn2660 I'm using a startech PCIE adapter with a HDSP PCI card, on my new i9 13900k build as well.
Win 10
i7-2700K @ 4.10 GHz
16 GB ram
RME Babyface Pro FS audio interface
Cubase pro 10.5.30
10 tracks @ 48 samples
24 tracks @ 512 samples
For a PC system... what is a test to run using lots of typical 3rd party plugins in a maxed out mix session say? What is an i7 capable of - does it struggle? or what can be expected?
@dayzaudio I don't think this precisely answers your question, but I found this video generally helpful in getting a grasp on what effect core numbers have on overall performance: ua-cam.com/video/PmCEyLg5bpU/v-deo.htmlsi=W3NzVsR4V77WHS27
Hi Scan, would it be possible to upload a test for Ableton Live 11. I have a 5950x with 128gb vengeance 3600 MHz ram, mobo ASUS proart creator x570, dark rock pro 4 cooler and would like to determine if the system is performing at optimal level. Can you also recommend optimal bios settings for this system? Thanks guys
Do you have any benchmarks on the i7 12700 or something similar? Looking for low power quiet and performance. Would be nice to see Reason benchmarks
i7 12700f 16 gb ram here.
136 tracks at 512 .
Is it really a fair test if you barely use midi instrument but rather insert fx/dsp intensive?
Hackintosh MacOS 14.3.... i9 7940x, 64GB RAM, Steinberg UR22 C, Cubase 10 Pro
120 Instances at 32 Samples
135 Instances at 512 Samples
Win 10 i9 7980xe 2,6Ghz (native)
32GB DDR4 1800Mhz
M-audio Firewire 410 (This vintage not supporting 32 buffer. Yes I know, i know. I will buy something new soon to my home studio :D )
Cubase 10 Pro
null instances at 32 Samples (ai not supporting)
91 instances at 512 Samples (fluency)
Summary, what if I overclock the processor to stable frequency 4,2Ghz and run this test on hackintosh?
Does anyone have M1 and can show the test result?
Well. I think hackintosh is the best choice.
Ryzen 7 3700X, 16G C16 LPX Corsiar RAM, Gigabyte B450 DS3H Motherboard, 500 GB Crucial P2 NVMe - Ultra-cheap £50 motherboard let's the rest down (total system cost £617 incl Corsair case and Corsair 550W Gold Modular PSU purchased Amazon UK June 2021). Old GT 710 Graphics Card
Audient iD14 USB C Interface
512 Samples: 62 Total Tracks
64 Samples: 35 Tracks Total
32 Samples: 22 Tracks Total
AMD 5900 , 64 GB RAM .My RME FF UC cannot do 32 samples , at 48 I can run 99 , ASIO meter pinned , but does play glitch free repeatedly
My old 15-4670 ran Cubase really well. My 12700k is absolutely useless due to clicks pops and coil whine. You move the mouse and you can hear it through the speakers. Everyone tells me it's easy to fix. But nobody can tell me how to fix it.
It’s not any of that, it’s a ground loop.
Plug your computer and pc into the same plug board. If that doesn’t work,
Have a look for a ART DTI box and put it in line with your speakers.
It might be more complicated than that but that solves 90% of issues by breaking the audio ground and making a new, clean one.
@@ScanProAudioTV Appreciate the help - will try the ART DTI Box if I can track one down.
Had the same problem WIN 10 - LGA 1500 (i7_6700K) - My solution was to install a (budget) PNY Quadro P600 graphic card (similar to GTX 1050?), but anyone will probably do - and then disable graphics from CPU in BIOS. In fact right now I am using an very old NVIDIA GeForce GT610 running this bench-test. using a I also chose the most stable NVIDIA driver (enterprise and not newest gaming driver to my card)
I5-13600K
32GB RAM 5600
Scarlett SOLO 3rd gen
Win 11 pro
Cubase 12 pro (Asio guard High, 64 bit float, Steinberg power scheme, audio priority normal)
All Cores
93 Instances at 32 Buffer
96 Instances at 512 Buffer
Without E-cores
77 Instances at 32 Buffer
79 Instances at 512 Buffer
is it good?
CPU Intel Core i5-12400F
RAM 32Gb DDR4-3200
SOUND Audient ID4 MKI
Cubase 12.0.60
Windows 10 Pro 22H2:
60 tracks at 32 samples
60 tracks at 512 samples
macOS Monterey 12.6.4 (Hackintosh):
74 tracks at 32 samples
77 tracks at 512 samples
Has anyone looked into the issue with UAD cards not "communicating" with AMD Ryzen 9? I had to switch back to the i9 for that reason.
seriously?
There was some early issues especially with the low count cards (duo, solo) We've certainly had Quads and Octos running on our 3XS Ryzen machines without a problem on both PCIE and thunderbolt.
@@ScanProAudioTV What about Apollo twins (mk1), any issues with that still you think? See my entry for Apple Mac Mini above. I think there might be something really wrong with the machine, am hoping to get a warranty on it and maybe switch back to PC!
Also you recommend to use 32 samples not All instruments can be run at 32 samples Kontakt some bank just gets disable and few more.
i7 10700k @3.8ghz. 64gig ddr4 2666 corsair ram. asus prime z490-a mobo. RME pcie hdspe aio sound card set at 64 samples 44.1khz. input and output latencies at 2.4 ms
windows 11 latest, cubase pro 11.
64 bit float, asio guard high, audio priority boost. multi process enabled.
50/54 tracks before drop out.
although start using the gui, change window sizes work around the interface and track count drops quite a bit. real world, you wont be looking at loads of tracks whilst they play. you will be in the mixer or arranger, messing with plugins etc. try your tests again whilst in a real world scenario and see the difference. also try running the test with all your track folders expanded with track contents in display, watch the track count drop again. all these things add to the cpu stress. just saying like.
Does it work well with Ableton live 11??
What about temperature?
I got to 22 instances on my old i7 4770k 3.4Ghz. ASUS Z87-k. 32GB Ram. RME AIO Pcie. Crucial CT480 M500 SSD as main programme drive. Is this really weak, even for such an old machine? I usually overcome this by using render in place. I'm looking to upgrade soon, but thinking of waiting new intel chip. This machine has served me well, but looks like it's ready to get retired.
next intel chips will be big.LITTLE - considering that DAWs like CUBASE in 2021 are still not optimized to recognize hyperthreading / SMT, it will take years for them to adapt to the new intel chips. Just get a 5950x or even better a 5600x / 5600g for the best price/perf now and then you will be able to upgrade to 5950x in the future.
HELP!!!...I have a really strange thing happening.I have done this test and when I get to just 2 tracks I get clicks and pops but it keeps playing up to 60 tracks before it wont play anymore.All settings are exactly the same as doms(i.e. 32 samples@44.1khz advanced options all the same ect)Im usin a ryzen 5800x/32gb ram@3666mhz X570 PRO MB M.2 1TB windows 10 pro.Can anyone help im at a loss.
Also task manager is showing CPU 6% usage and RAM 16% usage when I get the overload.
New PC: MB chip B660, CPU i9-12900, 32GB DDR5 RAM - Roland Quad Capture
buffersize 32 samples: 70 tracks /instances
buffersiz 512 samples: 104 tracks /instances
CPU i7-6700K 16GB RAM - RME Hammerfall
buffersize 32 samples: 24 tracks /instances
buffersiz 512 samples: 29 tracks /instances
Specs:
Windows 11 Home
Cubase 12 Pro
AMD Ryzen 5600x (SMT ON)
16GB RAM (Corsair Vengeance RGB Pro) 3600 MHZ
MSI X570-A PRO (Mainboard)
Samsung M.2 NVME SSD 3500 Mb/s
Asio4All Driver
Ultimate Performance Power Plan(Windows Energy Power Plan)
Bios Energy Saving Plan Enabled, not disabled.(Cool 'n' Quiet,C-States etc)
AMD Ryzen Stutter Fix Bios is not Installed
GPU Onboard Sound:(GTX970)
49 tracks @ 2048 samples
49 tracks @ 1024 samples
48 tracks @ 512 samples
47 tracks @ 256 samples
44 tracks @ 128 samples
38 tracks @ 64 samples
MOBO Onboard Sound (Realtek MSI AudioBoost 4)
50 tracks @ 2048 samples
50 tracks @ 1024 samples
48 tracks @ 512 samples
47 tracks @ 256 samples
45 tracks @ 128 samples
39 tracks @ 64 samples
Is there an easy way to automatically boost performance? I just bought a 32gb 3600mhz / 5950x.. Its obviously enough. But any more I can squeeze out of it would be great!
An auto overclock for AMD or something?
The extra processing I'm not too bothered about, it's the latency and IC I'm interested in.
***
If anyone wants a benchmark doing on this system, let me know.
Most of the testing we did actually showed that overclocking reduced the performance, but the biggest difference we did do was to use CAS18, ryzen optimised ram :
ua-cam.com/video/zUDTu6m2ra8/v-deo.html
@@ScanProAudioTV Thank you so much for your reply. I am using cas 18 Vengeance LPX so that's fantastic.
I appreciate your help!
@@ScanProAudioTV Thats also what I have noticed with my R7 3800x. Just run the AMD's with standard AMD optimization. Experimented a lot with Ryzen Master, and my overall test results was worse when trying to force different memory timings and higher Boost settings. There is so little overhead on the newer Ryzens that OC'ing is not worth it, unless you have a highend MB, the best possible ram and a rock solid cooling system of both CPU, Ram and VRM's. For a system that needs to be stable, just get a MB with a VRM that you know wont run too hot and has substantial headroom for the CPU you are using, and run it at the AMD optimized settings. in 99% of the cases its not worth risking stability for 2-3% extra performance wich is the typical number you will be able to OC your system, dependent on how lucky you got in the silicon lotto of your CPU.
13700k
64gb RAM / 5200 Mhz
512 Samples
MSI Z790 Tomahawk
Windows 10 Pro
115 Instances!
Mac Pro 28core XEON 96GB Ram
132 tracks 32 Samples
136 tracks 512 Samples 🤪
i7 12700k (Extrem Power Mode) - 128GB ddr5 - Cubase 12 - RME Fireface 802 - AsioGuard High - 64bit float Mulit-Processing
80 Instances @ Buffer 48 (Lowest Possible) (1.678ms in / 2.063ms out)
83 Instances @ Buffer 512
Get a high freqency in track when asio-guard is on...
My old PCI EMU 0404 is a lot less cpu intensive than Yamaha MG10XU. Same system specs i7 3770. I found very unlikely to be using more than 20 VSTi, so this is more than plenty for music production. Then for a youtuber stresming hi res video and reviewing heavy artificial music workloads, or a hollywood established composer yes, Ryzen 9 cpu horsepower, and 64 gb ram are common use.
(DELL XPS 8500) Intel(R) Core(TM) i7-3770 CPU @ 3.40GHz 3.40 GHz ; 24,0 GB RAM ; Windows 10 Home ; NI Komplete Audio 6 Inferface (512 Samples 44.1 kHz at 16ms) ; 16 Instances ok
Got Asus NV750JV i7-4700HQ 16 GB RAM, Win10, Babyface Pro got 16 instatnces on 48 samples (44.1 / 24 bit) 64bit internal Cubase 11, ASIO guard High. I think asus make great motherboards usable for DAW (its laptop from 2013 😊 and I using cooler pad otherwise its head up fast) But thinking About Zenbook Pro Duo 2021 🤤
hi if someone could compare cubase on the same latest cpu between windows and hackintosh, the cubase results on m1 pro are stunning vs results on windows x86, i would like to understand if it s m1 pro that are wonderful or if it's the cubase/windows combo that is not so good.
x86? No one is running 32 bit any longer, and of course the M1 would beat that even if it could run (razzing you, I knew what you meant, lol). So I'm building my new DAW tower this week, 13900F, DDR5, PCIe 4.0, 128GB ram at 5600, all samsung 980 PRO 4.0's, RTX 3080 (ok, that's for video editing and flight sim 2K, lol), etc etc. I'll post back with tis test results. Hopefully someone can post the M1 Hackintosh score and we'll see whats-a-happenin.
@@PaulieDC sorry i meant ×86_64 or x64 if you want :-)
One year old PC from Scan,
Win 10,
Cubase pro 11
MSI Creator X299
i9-10900X
64gb ram,
2 x ur824 and a ADA8000.
The older UR824 don't do well at 32 buffer setting?
17 instances at 32
50 instances at 128
54 instances at 512
intel 13700k z790 DDR5 - 108 channels @ 32 Buffer Size with no glitches
what speed of ddr5 and how much?
@@betoblanco603 5600 32g
@@yanuka7 Thanks! Would you say that number could grow with 64gb?
@@betoblanco603 More ram is always nice, but in this particular use case, I don't think you will benefit.i use less then half of the 32gb overall with this project playing. but if your projects has a lot of sample based plugins like kontakt more memory could help.
Another thing you should consider is a good CPU cooler that the CPU wont throttle and drop clock speed when getting hot on heavy loads, i use Dark Pro 4 air cooler which is good but i think I could squeeze more with better cooling even without overclocking.
@@yanuka7 Great, nice to hear that. I think I'm going for a 360 AIO for powerful cooling, maybe Arctic Liquid II or NZXT Kraken, hope it helps in performance!
I got 108 channels at 256 buffer size, in a Apollo Twin X duo, Win 10 pro, 4.0 Ghz on a Intel 7980XE 18 core. Beat that if you can...he he he
Macbook Pro 14" - M2 Max - 64GB Ram - 2TB SSD: 113 instances at 32 buffer size.😁
23 for me
Is there someone with 14900k or 14700 results ?
14900k 177 tracks at 128, still tweaking windows 11 pro and the bios settings
@@JoeKataldo how many on lowest buffer?
@@wizmos74I never tried lower than 128 samples. 128 is good for me when recording midi
@@JoeKataldo ok thanks for info,did you use very hungry CPU vsti like Diva,Repro or even Serum plus vst effects and how's effect performance ,i looking for 14900k also second quwstion,did is overclocked ?
@@wizmos74no those I can probably do 10, 20, I used the test they provided in the video
I'm sorry but duplicate in cubase don't specially create a new instrument. You must create those from the add track menu, and disable them until you're ready to use them and plus you want to talk about emulating a real session,? Well in real session we use different instruments patches but you're copies only use the same sound who does that? You can keep the midi but change the patches.
It definitely does create a new instance of the instrument if it is an instrument channel that is being duplicated.