Nuendo record performance vs. PT

Nuendo
1

Hello,

We use lots of workstations. Nuendo, ProTools, Live, Logic, anything the client will pay us for. Here in the US that means we must have ProTools, then anything else that will get us paid or enable faster work. My preference is Nuendo, been using it since it came out for MacOS. We have 6 Nuendo dual- and triple-madi systems, along with HDX2, HDX3, and HDNative systems. We are what many in our small part of the industry call experts (we’ve faked it well!).

In the past I would often gloat over my PT brethren with Nuendo’s record performance. While Avid used to recommend no more than 32 tracks per record drive and most operators did 48, Nuendo was happy with 64 or more. With PT 10.3 Avid did something that leapfrogged well past what Nuendo is capable of, and anecdotal conversations with large scale PT11 users indicates the same.

Systems -
OS10.6, 10.8, 10.9
MacPro various vintages
MacMini and MacBookPro with Sonnet and Magma expansion chassis, many flavors
PT 10.3.8, Nuendo 5.5.6, Nuendo 6.x

All combinations, all supported OSes, all hardware, it’s the same - Nuendo will record 128 tracks per drive on internal Sata but is not happy with many more than 64 tracks per drive using Firewire or eSata. ProTools is happy recording 128 tracks per drive using Sata, eSata, firewire, usb3. What happened?

The cpu utilization on Nuendo recording 128 tracks is around 18% (+/-) total cpu on most of our systems, the same on PT is around 10%; an expected difference due to HDX. You could argue that makes the record difference but that’s not accurate - there’s still a lot of cpu left over. Nuendo is half as efficient at record drive performance, PT is doing something different that enables a much larger margin for record drive performance. I suspect it has to do with drive buffer, interface buffer, and PT’s advertised “ram buffer”.

Factors accounted for -
I/O buffer size
drives and drive interface
OSes
interface type
cpu hardware

For most folks I suppose this doesn’t matter - who the hell records 128 or 192 tracks at the same time? Well, we do. The client wants to walk away with the drive immediately after so they can jump into post overnight, we try to avoid copying if we can. If we must record to two external drives, or three in the case of 192 tracks, that’s more expensive.

Steinberg should take a look at Nuendo’s record performance compared to ProTools, unfortunately my favorite and most profitable platform has fallen behind and in today’s world that makes a difference.

Hugh

2

Do you have the create audio images during record option turned on?
Also, does the audio pre record preference change anything?

Ollie

3

I’ve recently upgraded from N4 to N6 and my record latency on 48samples went from 2.56 ms in and 3ms out to 5.5ms out on the same hardware. That’s a round trip of 8ms! Constrain latency compensation just made it skip audio in the record file. I now can’t use my the control room for monetoring as I did with N4. I now have to put the Mic directly through on the rme total mixer. I have a new machine on the way and hope it will solve my latency issue I never had before.

4

On Mac or PC?

5

Hi Hugh

I’m on Pc, sorry I’ve updated my profile.
I also have a live symphony to mix down to picture . 72 tracks 24bit 96k and it won’t play off a single drive. I need to split the tracks half half to play correctly with no plugs or eq. The recording is 600gig for the 3 nights, so I can’t put it on the solid state drive.
I do have a very old machine but when I was recording one track with one voice no plugs or eq the input latency was only slightly higher but the output latency was massive, over double that of N4 with the rme ff800 set to 64 samples.
Most of my work is either TV or corporate and I have a low track count. Sometimes I track live events and we do a 4 or 5 cam shoot which I also edit in avid mc. I lockup on 9pin machine control so I don’t even have a video load.
We’ll see what happens when the new machine comes.

6

96kHz at 24-bit is ~288k/sec. 72 tracks of that is ~21MB/sec which should be no sweat for an internal SATA or eSATA drive (they are for all intents identical unless your external enclosure has a horrible controller.)

I just imported 53 stereo tracks (so 106 effective tracks on paper, but it’s not quite as bad as 106 independent tracks since the two channels are interleaved…) at 96khz/24bit from an eSATA disk and while it’s a cacophony of noise (they were all songs) they all play back no problem… about 30MB/sec. The disk-o-meter in the VST Performance window (I’m at home on Cubase) doesn’t even light up, I’m not even sure it works :slight_smile: A speed test on the drive gives well over 160MB/sec (sequential), a 3TB Seagate drive stuffed in a generic Sans Digital enclosure I bought.

Now for USB2 and Firewire 400, 20MB/sec is starting to approach the upper limit on sequential reads for most externals I encounter… and writing is always slower. For Firewire 800 or USB3 that many tracks shouldn’t be a problem assuming a decent external drive and controller, not crappy little 2.5" bus-powered drives which are pretty slow no matter what they’re connected to. However there are other issues to consider than just drive speed particularly for USB/Firewire, like if they are on the same physical bus as your audio interface if it happens to be a USB/Fireware-based interface. Lots of variables.

I don’t have 128 independent channels to source, but I just made 128 tracks set to the same mic input and hit record, to the same eSATA drive, and it doesn’t seem to care. Still writing to 128 96khz/24-bit files at once, even though they happen to all contain the same data.

So yeah, I’d analyze your hardware setup a bit – see how your new machine behaves.

(FWIW I did the above tests at home on an HP z600, 6-core 2.66GHz, afore-mentioned eSATA drive connected to the motherboard’s SATA controller, on Windows 7 x64 Cubase 7.5 with a Focusrite Scarlett USB2 interface.)

7

Hi,

If your mac has HT capable CPU (HyperThread) then disable it.
You can do this from the Instruments app that comes with the free Xcode.
Open Instruments, goto preferences, the CPU tab and uncheck “Hardware Multi-Threading”

I have no problem recording 128 mono tracks at 64 samples buffer (Asio guard on) in Nuendo 6.xx here.

8

Hello psvennevig,

I have experimented in the past with HT on and off and didn’t find a worthwhile improvement with it off, I can attempt again.

Are you recording to an internal sata or to an external firewire? I have no issue with internal sata, just external busses, which we frequently need due to reasons listed above. My comparisons are to PT using the same hardware.

Hugh

9

I tested on an external FW800, USB2 and Thunderbolt drive. No problems.

10

Hello,

Unfortunately that does not help with the many hardware and OS combinations we’ve tested (see above for list). It’s still mostly the same result when recording to external high quality drives using the same hardware -
Nuendo 5 + 6 have significantly reduced recording performance compared to PT 10 + 11 at high track counts.

Thanks trying,
Hugh

11

Hello,

I have some results that have solved our issue but still reveal the inefficiency in Nuendo for recording.

After several anecdotal and strange reports from operators I tested the following -
13 - Toshiba 500GB
4 - Toshiba 1TB
2 - Toshiba 2TB
4 - Seagate 7200.14
4 - Seagate 7200.12
2 - Seagate 7200.11
2 - Hitachi 2010 build date
2 - WD Black
2 - WD Blue
2 - OWC SSD 6Gb
3 - pcie flash

Systems tested -
MacPro tower OS10.6.8
MacPro tower OS10.8.5
2 - MacMini 2013 quad OS10.8.5 in Sonnet pcie chassis
MacMini 2013 quad OS10.8.5 in Magma pcie chassis
MacBookPro retina OS10.9.x
MacBookPro 2013 OS10.8.5
MacBookPro 2011 OS10.6.8

Interfaces tested -
Internal Sata where available
eSata where available
USB3 on all systems
Firewire on all systems

Test conditions - any interface available on a system was used for any drive available on that system. Most systems had all interfaces available.
A variety of enclosures were tested, it was verified that a particular brand was good, with a good drive performs well.
Each drive was tested multiple times to verify repeatability.
Nuendo 5.5.5, 5.5.6, 6.something, Logic Pro X.something, and ProTools 10.3.8/HDX2 were all tested.

Weird indeed.
The same model drives varied substantially in performance.
The Toshibas exhibited the worst performance, rarely succeeding above 70 tracks in Nuendo.
The Seagate 7200.x were fine with 128 tracks but oddly the dot12s performed a bit better than the dot11s or dot14s. Strange.
The Hitachis were good for 80 tracks and 112 tracks respectively, still not good.
The WD Blacks were good at 128 tracks, didn’t test them above that.
The WD Blues, suggested by our enclosure supplier, were good at 192 tracks on eSata and usb3 and 128 tracks on firewire.
The SSD and pcie flash were very good of course.

ProTools had no problem recording 128 tracks to any of these drives, even the 2 Toshibas that wouldn’t get to 64 tracks in Nuendo.

So the takeaway here is twofold -
First - drives are varying far too much; even if it benchmarks well it may not perform well in Nuendo, Logic, and other DAWs.
Second - my original point now borne out thoroughly - Nuendo has poor record performance compared to ProTools 10/11, a remarkable change from a few years ago. This should be examined.

Hugh

12

Wow.
Speechless… :confused:

13

This is interesting. I have never needed to record more than 32 channels at a time and Nuendo an PT were always up to the task but I have found that PT can’t keep up with Nuendo on playback with very high track counts (100+). This is with PT10 HD native.

Dean

14

It would be worth repeating the tests with Windows 7, to see if the results were the same.

DG

15

This must be a Mac thing.

I just loaded up 300 tracks and recorded without Nuendo breaking a sweat. (+/- 5% on the performance meter / buffer of 64 samples / Internal Audio Drive)
Next, I grabbed the closest USB-3 portable drive. Recorded 75 tracks seamlessly at 96/24 (buffer of 512 samples).

16

Hello,

I agree, would be interesting to see if the results are similar on Win7. As noted in my results the drive mechanisms vary far more than standard benchmarking would suggest so it’s necessary to use drives that do not perform well with Nuendo on the Mac.

Robin,
I agree, it’s probably a Mac thing and the way Nuendo queues recording information. Since PT 10/11 do it very well now (surprising as hell to me!) it indicates Nuendo is not doing it efficiently.

Hugh

17

Hugh, I agree with you.

I think PT is using the RAM of the machine much more effectively. They are probably buffering and storing the audio recording in RAM and then writing to the drive at a pace that the relevant media can handle. I’m surprised Steinberg aren’t doing the same thing, considering the amount of RAM we have available these days.

I’m not sure how secure the process is though. Perhaps you should look at Nuendo Live. It’s not expensive and from what I can gather is the most secure way for critical live recording. Even if there is a crash all of recorded audio up that point is secure.

Cheers,
Robin

18

Ultimately the hard drive will need to feed the ram or pull from the ram so there will be a time limit to how long ram buffering will work. This could really create a problem with a high track count recording where it starts off working well and then runs out of ram because the drive has reached its limit. The recording that you thought was working suddenly isn’t.

I remember the days of testing hard drives where measurements were always on my radar. It has essentially become a non issue for me now.
Dean

19

Hello,

Ultimately the hard drive will need to feed the ram or pull from the ram so there will be a time limit to how long ram buffering will work

I use the term “buffering” because it’s something everyone understands (and it’s a feature PT touts) - the ram doesn’t actually fill up. Every workstation buffers data on the way to the drive, some do it more efficiently than others. When recording multiple tracks most workstations write a part of Track 1, then write a part of Track 2, etc. up to in our case Track 192, then go back around to writing some more of Track 1. Our drive testing showed that Toshiba drives performed very badly at this and other mechanisms had varying levels of performance despite all these drives benchmarking well. Additionally PT does this much better than Nuendo without filling up the ram. This indicates two things - the drive write access is different between the two programs and is more efficient in PT, and that some drives do not deal well with the particular type of write access Nuendo is doing with large track counts.


Perhaps you should look at Nuendo Live.

Thanks for the recommendation. We’ve looked at Nuendo Live, I was one of the beta testers. There are several reasons why the product doesn’t work for us the most obvious is we need an actual daw. However in those situations where we’re just acquiring it does not work as advertised. If you read the NLive forum posts you’ll see several issues that make it not usable for us - inaccurate and variable timestamps, dropped samples on long recordings, varying file starts with prerecord on, “flexible” timeline making AAF imports not accurate, etc.


I hope Steinberg takes a look at the write performance in Nuendo, it is not efficient and I’m still amazed that PT 10/11 has improved write performance as much as they have; they used to be the worst at it.

Hugh

20

Correct on the buffering.
One thing that hasn’t been mentioned is safety. What happens when the system goes down for some reason? Is all, or some data lost? Or can you count on it that all the data, up to the crash, will be recoverable?
Needless to say that a safety system take up a good deal of performance too …

Fredo