NEW Studio One v3 updates incoming.

[Jim Roseberry]

The reason why Thunderbolt hasn't taken off under Windows:

Thunderbolt provides access to the PCIe bus (nothing more/less).

Microsoft literally just offered full driver support for "PCIe via Thunderbolt" with the latest Thunderbolt 3 chipset. Currently, only a small number of motherboards offer Thunderbolt 3 (via a USB-C port).
Microsoft claims that the Thunderbolt 3 support should be backward compatible with TB1 and TB2... but they don't guarantee it.

Now that we have PCIe via Thunderbolt drivers available... it's up to manufacturers to take advantage.

With USB-3.1 emerging, it'll be interesting.
The latest Intel USB-3.1 controller uses four PCIe lanes and provides a total bandwidth of 32Mbps.
10Mbps for each USB-3.1 port.

The best Thunderbolt audio interfaces (like PCIe units) will allow you to take the ASIO buffer size down to 32 or even 16-samples. Otherwise, there's no performance advantage compared to a quality USB-2 device (like the RME Fireface UFX).

[fedexnman]

Users are really happy with there Thunderbolt interfaces so I'm a tad jealous , when USB-C comes out , is Thunderbolt3 , and on PC hopefully we will see some good cross platform audio devices ? I see a lot of solid comments about how nice these TB audio interfaces are with practically zero latency and no hiccups , pops or clicks , so I'm a little jealous makes me feel my interface and OS is junk . The Presonus Studio 192 looked promising but the research I've seen is that they are basically no different than a USB2 device even though they are USB3 ...That's disappointing . Hopefully the 192 mobile gets it right ? Who knows ? Seems like Mac is beating Windows/PC in the audio interface department lately with TB devices .... Anyways back on topic , hopefully the S1V3 update will be exciting .

[Uncle E]

Focusrite Clarett may end up being the 1st Thunderbolt audio interface with Windows we will see...

Lynx Thunderbolt interfaces (Hilo-TB, Aurora 8-TB, and Aurora 16-TB) already work on Windows. I think Clarett works on Windows, too.

[Uncle E]

The best Thunderbolt audio interfaces (like PCIe units) will allow you to take the ASIO buffer size down to 32 or even 16-samples. Otherwise, there's no performance advantage compared to a quality USB-2 device (like the RME Fireface UFX).

I'm sure you know already, Jim, but for those that don't, 16 samples translates to 1ms latency at 96K. Focusrite, Lynx, and Apogee Thunderbolt interfaces are all capable of it.

[Jim Roseberry]

Yes, that's extremely low round-trip latency (just like a good PCIe audio interface).

Forgot to mention...
To achieve "PCIe via Thunderbolt" under Windows, you need:
- Z170 motherboard that supports TB 3 (via USB-C - connector)
- Windows 10

[ObsoleteAcc99]

[DELETED]

[bungle]

i wonder if they are going to fix the gui and performance.. that could be a big change, as well as making a bunch of midi improvements.
At least very curious to hear the announcement.

The GUI does not need fixing ? If you mean you do not like the GUI then say so, do not call things fixes if you just dont like them (Personally i think it is the first full time usable GUI for S1)
The performance does not really need fixing, OSX version needs fixing, performance on Windows, while not as good as something like Reaper, is however within usable ratios, OSX version is just broken full stop (But so is Maschine OSX, Reaper OSX and i suspect a few others on OSX)

[LawrenceF]

The best Thunderbolt audio interfaces (like PCIe units) will allow you to take the ASIO buffer size down to 32 or even 16-samples. Otherwise, there's no performance advantage compared to a quality USB-2 device (like the RME Fireface UFX).

I'm sure you know already, Jim, but for those that don't, 16 samples translates to 1ms latency at 96K. Focusrite, Lynx, and Apogee Thunderbolt interfaces are all capable of it.

My PCI card wlll run largish audio projects reliably at 32 samples with no dropouts, not so much at 16 samples, it depends, and not so much with really heavy VI's. So that makes a little sense I guess, if Thunderbolt is really just much the same as PCIe.

You guys know way more about the internals of all that than I do but my understanding of my 2408 card is that it has DSP chips on it that actually manage or help to manage the audio I/O, allowing much lower latency in general than what I get with USB devices, I suppose because the motherboard CPU doesn't have to also do that job?

No clue.

[aMUSEd]

Even USB 2 RME interfaces can go just as low, my Fireface UC can manage 32 samples easily for non cpu demanding plugins like Sunrizer, and 64 for most of the rest (I usually need more for granular plugins but low latency doesn't matter as much for them anyway). If I'm playing my Virus Ti, which is a zero cpu hit plugin, I can go down to 16 samples. The quality of the drivers makes a huge difference.

[Uncle E]

Even USB 2 RME interfaces can go just as low, my Fireface UC can manage 32 samples easily for non cpu demanding plugins like Sunrizer, and 64 for most of the rest (I usually need more for granular plugins but low latency doesn't matter as much for them anyway). If I'm playing my Virus Ti, which is a zero cpu hit plugin, I can go down to 16 samples. The quality of the drivers makes a huge difference.

Maybe these Thunderbolt interfaces can get down to 16 samples even with CPU demanding plugins?

[Jim Roseberry]

Even under the absolute best of conditions, it's going to be hard to run *dense* projects at a 16-sample ASIO buffer size.

With current generation hardware (and well optimized software), 32-samples is now practical for many folks.

As a point of reference (if my math is correct):
At a 16-sample ASIO buffer size (44.1k), the ASIO buffer = .363ms

Round-trip latency is the sum of the following:
- ASIO input buffer
- ASIO output buffer
- Driver's safety-buffer (usually hidden)
- A/D D/A

Round-trip latency from just the ASIO input/output buffers is 0.726ms.

Safety-buffer is the wild-card.
It can vary considerably between different units (especially USB models).
PCIe and Thunderbolt units typically use a small safety-buffer.

At a 16-sample ASIO buffer size (44.1k), I'd expect round-trip latency to be ~2ms.
That's extremely low... even for the most finicky of individuals/situations.

As Uncle E mentioned, at a 16-sample ASIO buffer size (96k), you're at ~1ms total round-trip latency.
Bear in mind that CPU use will be exceptionally high at these settings.
Even so... it's good to see these options becoming available.

[LawrenceF]

Round-trip latency is the sum of the following:
- ASIO input buffer
- ASIO output buffer
- Driver's safety-buffer (usually hidden)
- A/D D/A

Given that, what's the likelyhood of most systems running 1ms or less "true" round trip? I know many or most converters these days are pretty low latency wise but is 1ms round trip (literally speaking) something that is a reasonable expectation? What's the typical latency of converters these days?

(just asking the tech-wise guys here like Jim, not suggesting anything either way)

[Jim Roseberry]

Even USB 2 RME interfaces can go just as low, my Fireface UC can manage 32 samples easily for non cpu demanding plugins like Sunrizer, and 64 for most of the rest (I usually need more for granular plugins but low latency doesn't matter as much for them anyway). If I'm playing my Virus Ti, which is a zero cpu hit plugin, I can go down to 16 samples. The quality of the drivers makes a huge difference.

Agreed that the driver quality makes a HUGE difference.
RME units are exceptional.
My only question is how are you getting RME USB units to go down to 32-sample and 16-sample ASIO buffer size?

The Fireface UFX offers a 48-sample buffer size as minimum.
At that 48-sample ASIO buffer size (44.1k), total round-trip latency is 4.3ms
You can run substantial loads completely glitch-free at these settings (as long as your machine can sustain the load). This is excellent performance.

[Jim Roseberry]

Even USB 2 RME interfaces can go just as low, my Fireface UC can manage 32 samples easily for non cpu demanding plugins like Sunrizer, and 64 for most of the rest (I usually need more for granular plugins but low latency doesn't matter as much for them anyway). If I'm playing my Virus Ti, which is a zero cpu hit plugin, I can go down to 16 samples. The quality of the drivers makes a huge difference.

Maybe these Thunderbolt interfaces can get down to 16 samples even with CPU demanding plugins?

Under ideal circumstances, I'd expect Thunderbolt to offer performance equal to PCIe.

[fedexnman]

I think I'm gonna go with a RME 9632 with the added XLR in an out breakout cable accessory . I guess Thunderbolt will be the equivalent to PCI and PCIe sound cards like you said , awesome stuff really , I'll wait it out for that . That particular RME card has been around forever and they still make drivers for it and also works with Linux . Anyhow 5 more days till StudioOne3 update news .