denverdrummer
I can tell you that Apple and Intel came up with a very proprietary solution that is very expensive, and that locked alot of folks out of that market.  So if you're going to go TB, you're going to have to go Mac for the time being it seems.

IMO, That's exactly why Apple forced the Thunderbolt (Dunderbolt???) issue with their new "Cylinder" machines.  

Jim Roseberry
Thunderbolt provides access to the PCIe bus.  That's it folks!
Nothing more... nothing less
 
With a PC tower/rack, we've got PCIe slots... so Thunderbolt offers no performance advantage.
With a new Mac Pro "Cylinder", there are no PCIe slots... so Thunderbolt is an absolute must.
 
We're at an awkward "in-between" moment.
Thunderbolt development for Mac is ahead of PC (for the moment)... out of complete necessity.
IMO, It was a dumb move for Apple to completely eliminate PCIe slots (without more time to transition).
Thunderbolt peripherals offer no performance advantage (compared to PCIe) and they're more expensive.
 
Thunderbolt is the current market buzz-word... with audio-interface ads touting huge bandwidth advantage vs. USB.
That's all fine and well... but most audio interfaces (including higher-end units from RME) are nowhere close to saturating the USB2 bus.
 
I personally don't want to see PCIe slots completely disappear in one-fell-swoop (a la Mac).
If you're going to eliminate a major bus protocol, it needs to be done SLOWLY (a la PCI slots).
 
IMO, The Thunderbolt situation has been handled poorly by all involved.

• Apple pushed the issue by eliminating PCIe slots on their new Mac Pro.
• MicroSoft is in "wait-and-see" mode.
• UA and MOTU are pushing new Thunderbolt audio interfaces... at a time when support for PC is lagging. Instead of offering top-notch USB2/3 performance (like RME), if you're a PC user, you take a performance hit.

PCIe will be around for a while and really Apple has no control over that.  The corporate enterprise market is really dependent of PCIe for other technologies like 32GB FibreChannel, 40GB Ethernet, etc. etc.  Thunderbolt 2 is at 20Gbps, but other technologies are already ahead of it speed wise, they just aren't used for audio.
 
Agreed that Apple tried to force the issue with the Mac Trash Bin Pro, and it's rather stupid to be honest.  HP who is the biggest competitor in big companies/studios against the Mac Pro, does include TB ports, but now it's looking like the AI vendors aren't playing ball with them.
 
In addition to UA and MOTU, FocusRite just introduced their new Clarette series interfaces (Mac Only).
 
Seeing that Gibson/TASCAM are now full in with Sonar, a PC only solution, they have a great opportunity to follow someone like RME with a USB 3.0 solution.
 
The one technology that is already speced for 100GBS performance that could compete with Thunderbolt and is already in the Audio space is Ethernet.  Audio over Ethernet is a well established protocol that is already out and working on 1Gbps Ethernet, and coudl easily be ported to 10Gbps Ethernet.  40G and 100G Ethernet are now too expensive for the consumer market, but 40G is really starting to ramp up in the corporate world so 10G should become cheaper.

Jim Roseberry

Greeny
 
Wrong. Every usb 2 interface (all  about 3 of them) that has good latency has it at expense of cpu usage, a pci or pci-e card with the same latency uses way less cpu.

With a current generation machine, this is absolutely false.
CPU use is almost nil.
I can post some video examples if you want to see side by side comparison.

Start adding  vsti and vst and see which one brings your computer to its knees first, I'll bet you my last £££ its the usb one :)

Greeny
Start adding  vsti and vst and see which one brings your computer to its knees first, I'll bet you my last £££ its the usb one :)

 
FWIW, I have a "wee bit" of experience building/using DAWs... testing them under MANY different circumstances. That's what I've been doing for a living the past ~20 years. 
 
I've played many gigs as a keyboard player... running a RME USB audio interface at a 48-sample buffer size... with *very* substantial loads of virtual instruments (Kontakt, Omnisphere, Ivory-II, etc) hosted by both Forte' and Cantabile.  Absolutely 100% glitch-free.
 
We're not talking USB audio interfaces from the USB1.1 period (which were poor performers).
RME set the standard on what to expect from a USB2 audio interface.
Low round-trip latency... and absolutely rock-solid performance.
With a modern machine, you can run substantial loads (completely glitch-free) at very small ASIO buffer sizes.
 
I can post video if you really need to see it...   
 

Jim Roseberry

Greeny
Start adding  vsti and vst and see which one brings your computer to its knees first, I'll bet you my last £££ its the usb one :)

 
FWIW, I have a "wee bit" of experience building/using DAWs... testing them under MANY different circumstances. That's what I've been doing for a living the past ~20 years. 
 
I've played many gigs as a keyboard player... running a RME USB audio interface at a 48-sample buffer size... with *very* substantial loads of virtual instruments (Kontakt, Omnisphere, Ivory-II, etc) hosted by both Forte' and Cantabile.  Absolutely 100% glitch-free.
 
We're not talking USB audio interfaces from the USB1.1 period (which were poor performers).
RME set the standard on what to expect from a USB2 audio interface.
Low round-trip latency... and absolutely rock-solid performance.
With a modern machine, you can run substantial loads (completely glitch-free) at very small ASIO buffer sizes.
 
I can post video if you really need to see it...   
 

I am not saying you cant and the RME baby face is one of the "about 3" that I previously mentioned. But it cannot keep up with a PCI card so why pretend it can? also what's the rmes lowest latency? 4.4ms or so? at 48 buffer, a decent pci card can do half that at 64 buffers with less strain on the system and then you are able to run more vsti or fx etc on the same machine. I ain't here to argue just saying it as I see it. Don't need to see a video.
 
EDIT: Also worth noting that AFAIK the babyface has the best usb latency out there, and with few exceptions getting below 6ms without crazy sample rates with a usb interface is nigh on impossible which is not a lot of fun if you are a guitarist. So my point still stands that thunderbolt is a solution for a problem that DOES exsist since there are many people who neither want or need a babyface but have no choice but to buy one if they want both usb and low latency.

Greeny
I am not saying you cant and the RME baby face is one of the "about 3" that I previously mentioned. But it cannot keep up with a PCI card so why pretend it can? also what's the rmes lowest latency? 4.4ms or so? at 48 buffer, a decent pci card can do half that at 64 buffers with less strain on the system and then you are able to run more vsti or fx etc on the same machine. I ain't here to argue just saying it as I see it. Don't need to see a video.
 
EDIT: Also worth noting that AFAIK the babyface has the best usb latency out there, and with few exceptions getting below 6ms without crazy sample rates with a usb interface is nigh on impossible which is not a lot of fun if you are a guitarist. So my point still stands that thunderbolt is a solution for a problem that DOES exsist since there are many people who neither want or need a babyface but have no choice but to buy one if they want both usb and low latency.

 
I'm not here to argue.  
I'm here to inform.
 
The RME Babyface (and their other USB audio interfaces) yields the following round-trip latency:
48-sample ASIO buffer size @ 44.1k = round-trip latency of 4.9ms
64-sample ASIO buffer size @ 44.1k = round-trip latency of 5.6ms
 
The best PCIe audio interfaces:
64-sample ASIO buffer size @ 44.1k = 5ms round-trip latency
 
That means the RME USB units are 0.6ms off from the best PCIe units available.
 
 
Most USB2 audio interfaces don't provide ASIO buffer sizes smaller than 48-samples.
 
The best PCIe units allow you to go down to a 32-sample ASIO buffer size (some even down to a 16-sample ASIO buffer size).  Of course, this yields lower round-trip latency... but you won't be able to run heavy loads.
Absolutely *no* PCIe audio interface yields round-trip latency of ~3ms using 64-sample ASIO buffer size @ 44.1k.
That's impossible... here's why:
Round-trip latency is the sum of the following:

• ASIO input buffer
• ASIO output buffer
• A/D & D/A converter
• The driver's hidden safety-buffer

 A 64-sample ASIO buffer size @ 44.1k = 1.5ms
That means just the ASIO input and output buffers alone are 3ms.
Add latency of the A/D D/A and the driver's hidden safety-buffer... and you're at 5ms.
 
There is no "pretending" or "fudging" the figures here... this is real-world performance.
If you think USB consumes significant amount of CPU (on a current generation machine), connect a USB2 or USB3 HD and run a benchmark that measures sustained thru-put and required CPU use to achieve it.
You'll find CPU use is negligible...   
 
Regarding Thunderbolt:
Currently on the PC, both UA and MOTU use Thunderbolt>Firewire (not a true PCIe driver).
Thus, you're really not reaping the "reward" of Thunderbolt.
Thunderbolt provides access to the PCIe bus.  Nothing more... nothing less.
Using Firewire over Thunderbolt instead of PCIe driver, you'll see the same performance as using Firewire-400 or USB2.  Absolutely the same round-trip latency... (witness the MOTU 828x)
On the Mac, UA has proper PCIe drivers for their Apollo series.  Round-trip latency is 4.5ms with a 64-sample ASIO buffer size @ 44.1k.  
Even with true PCIe drivers, Thunderbolt isn't going break major barriers regarding round-trip latency.
 
 
 

https://www.dropbox.com/s...j19v/Capture9.PNG?dl=0
I use that for full projects all the time, heavy load no problems. Good luck doing that on usb.
 
RE Thunderbolt, theres a new breed coming out, focusrite and maudio both had them at NAMM, focusrite are touting 1ms latency on theirs, although that is no doubt at 192khz, which is about right and matches my pci card.
 
And we are still only talking about 1 usb interface here that is well regarded as the best, most can make 10ms seem a struggle.
 
It sucks that no other company but motu and RME can seem to get their act together with the drivers cos I am sure their interfaces could do ~5ms if the drivers were upto scratch.

It's about price for performance.  Most USB interfaces are in the sub $1000 category aimed at the home studio user. The RME interfaces are much more expensive.  The Madiface XT which supports USB 3.0 can probably get even lower than 5ms, but I haven't seen specifics yet, but those start at $2500.  The TB interfaces will all be in the upper $2K or higher range.  That's fine if you're in that price range, but the fact there's not alot of better options for people in the sub $1K territory.
 
The issue that we're bringing up is Apple trying to force it's hand with Thunderbolt.  Right now it looks like all of those TB interfaces are Mac only, and as Jim mentioned there is no real performance boost over a PCIe card, and that the Trash Can Pro was aimed at axing the card market to external A/D converters only.
 
What's frustrating to me is that USB 3.0 has more than enough bandwith to provide 5ms or better, and it's a much cheaper and more ubiquitous solution than TB.  USB 2.0 is probably good enough, but I'd like to see more A/D vendors take a crack at USB 3.0.  Right now RME is the only one.

I just have to say that the term "Wintel" is very strange these days and when someone uses it, it makes me think they are a bit of a Mac zealot who eats up marketing silliness rather than actually evaluates things and is a bit out of date on knowledge.
 
I can understand if you don't like Windows, fine, personal preference is personal preference. But what's with "Wintel" as opposed to what? If you buy a computer, it is x86. Macs are all Intel all the time, PCs have Intel or AMD options both which run Windows, Linux, or any other OS you like.
 
"Wintel" was a thing from back in the PPC days when Apple was attempting to convince people that PPC chips were faster than X86 chips. Of course we all know how that went, Apple switched to Intel, and swept all that under the rug.
 
In terms of being "left behind" no, not at all. PC system are actually more current, since they do updates more often. If your focus is on Thunderbolt then ok, but please understand that's not an Apple technology, it's an Intel technology and as others have mentioned you can get it on a PC if you want. So why don't you see many of them? Because nobody cares. It just isn't very useful for very many applications, and it is a security risk, being a direct PCIe connection in a way USB is not.
 
If you wish to use a Mac, go right ahead, but recognize you can't use Sonar on it without booting to Windows or using an Emulator (which adds a lot of latency). Also don't lie to yourself and tell yourself that it is to get more up to date hardware, because it isn't. If it is just for TB then ok, I guess, but as others have mentioned, that's a little silly since you don't need TB. You want low latency? Get RME and be done.

,