Hi,
I am dealing with a decision to make and hoping to get some advice: I want to buy a new audio interface that will be used only for my studio daw - a desktop.
The question is: Do I go the pcie-way or the safe and comfortable usb/firewire-way. I am thinking about a RME hdspe AIO (pcie) or an RME UC (usb). If I buy the additional hw-i/o's for the AIO, the price will be the same. An UCX (usb/firewire) will be even more attractive, but it is an additional 30% investment. I use an external pre-amp - which means, the line-in-only of the AIO is not a dealbreaker.
 
The point is: I read, that the pcie solution is way more performant than I can expect from an usb-solution. But the AIO is at least five years old - so probably the newer hw kicks it ….
 
Does anybody has some experience with an AIO and/or UC? Or is the reputation better than the solution and I should check a different piece of hw?
 
Thanks in advance!

More important than PCIe vs. USB is the specific choice you make.
 
The best USB audio interfaces (RME) yield 4.3ms total round-trip latency at a 48-sample ASIO buffer size 44.1k.
The best PCIe units (RME, MOTU, Lynx) allow you to go down to a 32-sample or even 16-sample ASIO buffer size.
This yields 2-4ms total round-trip latency.
Of course, the lower the latency... the bigger the CPU hit.
Your machine has to be able to sustain the load.
 
If your primary focus is working with virtual instruments... the RME AI/O is an excellent choice (very popular with composers).
If your primary focus is recording audio tracks, I'd go with something that offers a little more flexibility... and gets the A/D out of the PC (which results in a lower noise-floor).
 
There's no reason to be put off by the best USB audio interfaces.
ie: The RME UFX sounds great, it's rock-solid, and offers round-trip latency on par with the best units available.
 
From a performance standpoint, you'd do well with anything RME, MOTU, or Lynx.
 
 
 

Jim,
 
Can you clarify a bit more since didn't understand some of the intention. Always trust your advice. Never been on this part of the forum and was about to open a thread. Looks like vdd is dealing with the same issue.
 
Qualitative assessment is that I have an old quad core XP 32 bit rig with a PCI M-Audio 1010 LT sync'd to a 2 channel M-Audio 2496. This performs well, 10 channels, but obsolescence has me up against a wall and will have to upgrade the system at some point.
 
Have been beta testing a synth that is a real CPU hog but performs well on the XP rig.
 
Have a 64 bit win 7 i7 laptop 6G RAM connected via USB 2 to a Presonus 22VSL box. The thing sounds like a popcorn popper unless I increase the buffer to over 2 sec. All drivers current.
 
The fire power of the laptop is overshadowed by the bottleneck of the USB 2 connection. Feel like hard wired PCIex slot is the way to go on a new computer.
 
Any recommendations on a PCIex with 8 channels.
 
Many thanks,
John

I doubt your problem lies in the bandwidth of USB2. I can quite happily run 12 channels @24/44.1 monitoring them all via Sonar not the interface and get 4.8 seconds round trip latency with no clicks or pops. I can run 20 channels if I up the latency a little, still under 10ms round trip.

I get similar performance running the same interface into a Mac using Firewire400.

Unless you're running a huge number of tracks in and out you won't be overloading the USB2 bus, assuming you don't have other stuff hanging off the same port and handling a lot of data at the same time. If you have something causing serious PCI bus latency issues that will affect any audio interface and cause drop-outs no matter what the interface is or how it connects.

If you've not done so, I suggest you download a free PCI bus latency monitor and see what it reports.

Thanks tlw,
 
I realize this is a complex issue and maybe something interfering with the pathway aside from that which is on the surface. Will analyze with the latency monitor.
 
I can fire up a lean and clean Reaper session, one track only and overload the system with a single instance of Nitroflex. Granted, a CPU hog in the extreme, but the laptop has all the power necessary otherwise to carry the load if there were not a bottleneck somewhere between the motherboard and the USB box.
 
Somewhere far to the south, I have my XP setup and will test out the USB box if I can ever get there long enough to compare. Dynamite system. So sad that it is being squeezed out of commission by forward progress.
 
John

JohnKenn
...
I can fire up a lean and clean Reaper session, one track only and overload the system with a single instance of Nitroflex. Granted, a CPU hog in the extreme, but the laptop has all the power necessary otherwise to carry the load if there were not a bottleneck somewhere between the motherboard and the USB box.
...

Three possibilities to consider:
 
1) the 22VSL driver (even the "current" v1.3, which claims Win10 compatibility). I have a 44VSL and ran it on the hardware in my sig below. I always had latency issues and ran into the occasional bug. The driver totally failed under Win10. PreSonus have dropped support for the VSL interfaces and have just hung owners out to dry.
 
2) the 22VSL is buss-powered. There can be issues with it (and many other 2-input buss-powered i/f's) if you're using condenser mics and trying to supply phantom power to them. In that case, a powered USB hub is required.
 
If your driver works for you, great, but I think it highly likely that it is at least partly responsible for your snaps, crackles, pops and poor latency.
 
3) the driver for the USB port(s) in your computer.

Thanks Maartian,
 
Not dealing (yet) with win 10 and don't have an alternative reference point for the Presonus box to compare with. Just that it is a USB connection and delivering inferior quality response in comparison with a way older computer with a hard wired PCI card.
 
Knee jerk reaction is to assume that Presonus is cool and the block is in the USB versus PCI connection, but could be in the Presonus drivers. USB 2 could be cool still.
 
John

JohnKenn
Jim,
 
Can you clarify a bit more since didn't understand some of the intention. Always trust your advice. Never been on this part of the forum and was about to open a thread. Looks like vdd is dealing with the same issue.
 
Qualitative assessment is that I have an old quad core XP 32 bit rig with a PCI M-Audio 1010 LT sync'd to a 2 channel M-Audio 2496. This performs well, 10 channels, but obsolescence has me up against a wall and will have to upgrade the system at some point.
 
Have been beta testing a synth that is a real CPU hog but performs well on the XP rig.
 
Have a 64 bit win 7 i7 laptop 6G RAM connected via USB 2 to a Presonus 22VSL box. The thing sounds like a popcorn popper unless I increase the buffer to over 2 sec. All drivers current.
 
The fire power of the laptop is overshadowed by the bottleneck of the USB 2 connection. Feel like hard wired PCIex slot is the way to go on a new computer.
 
Any recommendations on a PCIex with 8 channels.
 
Many thanks,
John

Hi John,
 
FWIW, I'm going to repeat what others have said.   
The problem isn't USB2's bandwidth.
Eight channels of I/O is nowhere near saturating the USB2 bus.
 
FWIW, I run a RME Fireface UFX (which is a USB2 unit with 30 channels of I/O).
I can run substantial loads glitch-free at a 48-sample ASIO buffer size.
That's 4.3ms total round-trip latency at that 48-sample ASIO buffer size 44.1k.
 
Lots of throttling going on to keep CPU temps in check with most laptops.
Have you measured the laptop's DPC Latency?
My guess is it's pretty high.  If that's the case, no audio interface will work well at lower ASIO buffer sizes.
 
Lynx, MOTU, and RME all make excellent PCIe audio interfaces.
Lynx doesn't make a PCIe unit with 8 channels of analog I/O.
MOTU's PCIe 424 controller can be used with the 2408mkIII.  Converters are a little long-in-the-tooth... but performance is good.
RME's HDSPe with Multiface-II performs well.  You may have trouble finding this new.
 
MOTU have announced true "PCIe via Thunderbolt" drivers for their new AVB audio interfaces.
You'll need a motherboard that supports Thunderbolt 3 via USB-C port... and you'll need to run Win10.
This combination will yield performance equal to PCIe units.  That's the whole point of Thunderbolt (external access to the PCIe bus).
 
 

As mentioned in another thread, USB is fine unless you're like me and are asking far too much of it.  Having dedicated vendor-supplied drivers also helps as they will have most likely addressed the issue of the latency which is introduced by the USB Clock Front & Back buffers.

"Asking too much of it" amounts to something like the following set-up, using M$ provided horribly-not-that-good drivers for many of the devices, all off a rather ageing-now MOBO with the odd USB hub to provide the additional ports:

USB3: external HDD
USB3: razer mouse
USB2: microphone
USB2: keyboard (PC)
USB2: keyboard (MIDI/controller)
USB2: Faderport
USB2: 2-channel audio I/O to the Behringer mixer
USB2: Wifi dongle
USB2: Digitech GSP1101
USB2: Lexicon MX400
USB2: Behringer FX2000

Pretty much any hard drive activity destroys even the 2-channel audio output to the mixer (although admittedly that is using the somewhat crappy default M$ drivers).  In the above set-up (i.e. mine), a USB-based audio interface would not be a good idea.  In particular, the Wifi dongle is a Very Bad Idea (TM).  I'm thinking of getting some new IP-over-mains adapters and ditching the dongle tbh, in fact as I write this I've just ordered some from Amazon.

Jim, Ken,
 
Appreciate the feedback. Will go thru latency checks and figure out what is causing the jam. Guess it stands to reason that if the USB route was this clogged in general because of USB 2 limitations, they probably wouldn't have a product on the market. Laptop works well for guitar, but a couple Kontaks libraries can overload.
 
Next computer will be a desktop with Thunderbolt. Not sure how well the PCIe to PCI slot converters work, but would be an option to keep the 1010 LT. Has always been a great PCI card. Sad to see it go.
 
John