Probably asked before, but I was wondering if native can have as low latency as a PT HD system nowadays? In case yes, with what set up this would be able to achieve? In case no, clear PT HD is still ruling on that...?

Regards,

Roger

What exactly are your needs and budget? Any interface for it's salt has a mixer in it that's pre-ASIO with "virtually no" latency.

running through the DAW, lowest I know about is 32-samples. Most can do 64 samples on a modern PC with to my perception is not noticeable.

Sure can, Apogee Symphony and Logic does the job very well.

I don't think you'll ever get as low with Native as long as one has to go threw asio/host, but close is more like it. The roundtrip is what must be considered.

I can get low latency like 2ms one-way with my Core i7 wich is about 7ms roundtrip, but getting there is'nt the problem, it's staying there that's the problem heh


P.S. Actually i can get 1ms but my system would be extremely limited with performance and i could not use many Native plugins i like. Likewise with 2ms but with 2ms i can operate in a somewhat decent manner tho still very limited. Most high quality plugins i like are too demanding for small buffer-sizes, either by themself or in combination.

It's not about budget. It's about the lowest latency available.

Regards,

Roger

Then it looks that PT HD is still ahead on this.

Regards,

Roger

Btw how much is PT HD's round trip?

Nonsense. EVERYTHING is about budget. Anything done in the computer is a compromise.

I can live with my <5ms latency if it means I don't have to spend an extra $10,000. All interfaces let you monitor through it's internal mixer anyway and are significantly cheaper than HD. Chek out the Soundscape/SSL products, they even have decent plugins.

By your rational analog console is the only way to go since it has the lowest latency possible.

Umm, not really, with a 32-sample buffer you get 1.6ms at 96kHz roundtrip with Symphony. Of course you'll have to use Logic to get those times but it ain't just ProTools on the block anymore.

I'd actually be surprised if some of these new madi interfaces don't rival ProTools as well.

If you're already sold on HD though there's no real point in this thread.

Thanx. This is a good answer.

Oh btw, I'm not sold on HD at all yet. Just want to check out the current state. So far I understand HD and Symphony are really close and maybe some MADI stuff of RME. More input on this maybe?

Also, what about recording and using a reverb or some plugs. HD has it's plugs on the card, don't know how this is with sympony. I do know about UAD etc.

Regards,

Roger

With Symphony all your plugins are coming from Logic, no cards involved. I've been using Logic this year and have been tracking through the software at a buffer of 64 with no noticeable latency. So far I've only been doing 2 tracks at a time, have yet to do a big session. I'm using Apogee but not the symphony setup (am researching that right now). The best thing for me with Logic is the low cpu overhead. I can be at a 64 buffer and have various VIs going (BFD2, etc) and not have it be an issue. When I do have to raise my buffer I can still track VIs easily. There's a lot of horse power there. Not to mention the great plug ins, etc.

But does'nt protools latency decreases at higher samplerates aswell? My problem and many others is not with getting low latency but keeping it with Native.

Sorry, I used 96kHz cause that's what we record at in the studio. At 44.1kHz with a 32-sample buffer I can get 3.1ms roundtrip which is still directly comparable to ProTools. Tracking through plugs really isn't a problem on a native system anymore either so long as you've got one of the latest Mac Pro's. We're using a dual 2.8 quad and have never had a hiccup!!

As has already been mentioned as well, if you're using UAD or Powercore plugins to track through you also won't have any increase in latency either.

I'm seeing 1.3ms with MOTU hardware in Cubase (96khz/128 samples). I'll be darned if I can tell the difference between 1.3ms and 0ms. Seems pretty good to me.

It's the same as sitting 1.3 feet away from the speaker.

I'd like to know that as well.
According to Apogee (who obviously want to sell their own interfaces) they can get lower than a PT rig with their Symphony PCI cards running Logic on a Mac Pro at 32 samples buffersize and 96kHz operation. I think the roundtrip latency of that setup is a little below 2ms (yes, a true audio in-to-out roundtrip).

For anything native not operating at higher samplerates than 44.1 or 48 (which, due to the added CPU drain of higher rates, is still what most people working with native systems are using), the best I've experienced was using an RME Hammerfall (PCI) on a Windows desktop machine back then under Logic 5.5.1 (the same figures should now as well be possible on Macs, back then RMEs Windows drivers were just better). I think rountrip latency boiled down to something around 4ms (maybe 4.5). Motus interfaces are in about the same league. These are figures that are about as good as Apogees.

In any case, if you want the lowest possible latency on a native system, PCI (or PCMCIA, ExpressCard) is the way to go. RME seems to be almost there with their new Fireface UC and USB 3 might bring quite some more interesting developments to the table, but for the time being, PCI based solutions are king. For now, FW and USB solutions still add more (usually undocumented) "safety buffers", which you can't do anything about (they're a fixed number, regardless of your drivers settings).
I'd be highly interested in some Fireface UC measurings.

As far as keeping latency down while CPU usage is raising goes, it has to be said that most hosts suffer from that problem, Cubase (especially on Macs) is known to be notoriously bad. The only real exception seems to be Logic (unfortunately, I should add), but Reaper seems to be on a good way, too.
With Logic, it's really like "set and forget". I keep my buffersize at 64 all throughout whatever projects. And I can push the CPU to the max without noticeable crackles or anything. Raising the buffersize at that point would perhaps give me another compressor or two, but that's about it. In Cubase however, the differences between, say, 64 and 512 samples buffersize are really dramatic. And that's most likely why almost all Cubase users I know are raising their buffersizes once it comes to mixing. Not too great a solution, IMO. I sometimes just like adding the odd track (using audio software monitoring and/or virtual instruments) when I'm already in mixing stage - not a problem at all in Logic, quite a hassle in Cubase (and some others).
And that's exactly one of the topmost reasons why I'm still using Logic. As long as it doesn't crap out for other reasons (yeah, there's plenty...), it ist without any doubt *the* native environment for low latency operation.

- Sascha

Thanx for this very clearing reply!

Regards,

Roger

Thanks for your reply Sascha.

As a Cubase user myself, and since you had to bring up the Logic vs Cubase aspect into this discussion...I thought a small reality check about those "huge" buffer differences might be in order.

So, I just loaded a Cubase 5 project with over 250+ plug-ins, in order to test how the buffer size affects the performance.


In detail I tested the following:

-32 audio tracks (16 tracks of drums + 16 tracks of pad sounds in order to easily hear if it starts to pop)
-272 plug-ins (8 plug-ins for each channel + 8 send FX + 8 master plugs = total of 272)

For the plug-ins I tried to approximate something I’d use in a real-world situation –like I’d ever even need this much of ‘em!

Anyway the plugs were:

32 x Waves Q
224 x Waves RenComp
8 x Waves Rverb
4 x Waves Req
4 x Waves L2
= 272

-First I ran the project with the buffer @ 64 samples, then @ 512 samples.

-With 64 samples the ASIO meter was stable @ 66%.
-With 512 samples the ASIO meter was stable @ 60%.

Mind you, this was on a Windows XP system, so I cannot comment on the mac. But from what I’ve experienced, I’d say your talk about the HUGE buffer differences are not very valid.

On my Cubase 5 setup, I never have to switch the buffer size over 64 samples, even on large projects.

Remember this was on a DUAL-core PC, and I didn’t even push it!


My point is that sometimes those geeky benchmarks with 750 instances of the same convolution reverb on a dual octo-core Mac Pro do not tell everything about the “performance differences in different DAWS”.

Example: ATM I’m doing the full post-production for a 30 min movie in Cubase 5 with 90 + tracks, using lots of different Waves plugs, all running @ 32 samples perfectly.

I doubt Logic would perform any better since that’s the lowest buffer size my interface has to offer.


Peace

StereoPari, you already mentioned the most important thing yourself: You're on a Windows machine. That's making quite a difference.
And well, I would completely agree that Cubase 5 has been improved massively under Windows. Under OSX, low latency performance still seems to be shoddy.

- Sascha

What audio interface do you use?

Regards,

Roger

The main problem i find with Native hosts enviroments in general is with vsti's. Anytime i use instruments in any complex ways my buffer-size raises dramatically. There are however many effects that will do this without needing large quantities of them. Some single effects will do this. I can load many many efficient effects at 1ms or 2ms latency for a complex mix with my Core i7 but they generally are not the effects i truly prefer and that's the problem. Couple that with the vsti problem. I like to work in a no compromise/workaround manner. I mainly use Sonar wich appears to be worse then cubase with latency judging by the benchmarks i seen, but not so dramaticly worse that it appeared as if the plugins that induce alot of latency still would'nt do it to a great extent in Cubase, but i'll need to be fair and demo Cubase at some point.

In the end if your able to go threw all stages of music production with low buffersize, using the plugins you truly prefer and work methods you truly prefer in the Native enviroment then yea you have something really special. You may be one of the few. As long as it works for you then mission accomplished, and its good to see.

I have not tried the new uad-2 yet but i read from different users that the Live-mode still induces a bit of latency, it just does'nt double the latency anymore. It also is very cpu taxing from what i read. I'll need to further investigate and i may even purchase the uad-2 solo before this year is out if i don't settle for uad-1. UA needs to have more info written on the live-mode.

Hi Roger,

I'm using the Steinberg/Yamaha interface, MR 816 CSX.

If you're not familiar with it, it's basically a Yamaha n-series mixer turned into a firewire audio interface (same preamps, AD/DA, etc).

Hi guys,

I ran some tests about 6 months ago using the 1st generation 8 Core Mac Pro and thought you might be interested in the results. The idea was to push the 8 Core Mac Pro using Symphony to see the extent of its performance in offering both maximum plug-in counts and lowest latency at the same time.

Because all plug-ins have different DSP usage I created a new project and attempted to max it out with one particular plug-in.

I have not yet had the opportunity to run this test with a new Nehalem machines but we can assume some measure of improvement.

Here are the results:

Maximum Plug-ins at 44.1k
Round trip analog to analog latency measured at 3.1 ms
with Logic set to 32 sample buffer & Apogee Symphony High Performance Mode.

Space Designer (stereo) 140
Delay Designer (stereo) 207
Reached memory overload before DSP overage
Roughly 30% headroom on DSP usage meter
Guitar Amp Pro (stereo) 169
Waves SSL Channel Strip (stereo) 417
Reached memory overload before DSP overage
Roughly 20% headroom on DSP usage meter
Waves Renaissance Channel (stereo) 588
Waves Renaissance Reverb (stereo) 195
Waves IR1 (stereo) 51
Reached memory overload before DSP overage
Roughly 30% headroom on DSP usage meter
URS Channel Strip (stereo) 420
Sonnox Oxford EQ (stereo) 133
Sonnox Oxford Dynamics (stereo) 120
Sonnox Oxford Reverb (stereo) 168

Maximum Plug-ins at 96k
Round trip analog to analog latency measured at 1.6 ms
with Logic set to 32 sample buffer & Apogee Symphony High Performance Mode)

Space Designer (stereo) 81
Delay Designer (stereo) 130
Guitar Amp Pro (stereo) 40
Waves SSL Channel Strip (stereo) 315
Waves IR1 (stereo) 39
Reached memory overload before DSP overage
URS Channel Strip (stereo) 170
Sonnox Oxford EQ (stereo) 84
Sonnox Oxford Dynamics (stereo) 53

I hope this sheds some light on the performance available using the Symphony system with Logic.

Impressive results.

Roger

Hi Tony,
You were semi active on another thread I started at the logic users group forum. When you did these tests how many audio tracks were you running or was it strictly just loaded with plugins? I'm getting my symphony system going today so I guess I'll find out soon enough!

Hi Dave,

Congrats on getting your Symphony rig! Feel free to drop a line if you have any questions.

Regarding the tests, I ran 20 stereo tracks of audio and stacked the plug-ins.

on all pc's ive tried them all the ESI products give 1ms... even when recording 8 tracks and playing back what's already recorded as a stereo monitor feed

never seen any pci card beat ESI latency. sadly i cant use them with my macs tho

How are you measuring this? An AD and DA conversion alone will typically cause 1-2 ms. On top of that the driver and buffer setting both add their own latency.

Symphony delivers 1.6 ms latency measured from AD -> DAW -> DA with software monitoring ON.

I was just about to ask the same.

Btw, pretty impressive figures you've posted.

- Sascha