What's new

PCI audio interfaces improve performance?

Buddy

Member
Trying to figure out a way to better performance--keep my buffer low, no crackles. I have a solid machine, 8700K, 64gb ram, all m.2/ssds. Right now I'm looking at my firewire Apollo as the potential bottleneck. Do not have Thunderbolt on this mobo. Would a PCI-based interface improve my latency/performance?

If so, any recommendations for good converters/low noise/low latency PCI card interfaces? Would need to be able to take my Apollo and other preamps as ins. Thank you!
 
I assume RME RayDat so you can use your Apollo as AD/DA convertor. Personally I run a RME Babyface Pro with the Apollo connected over ADAT to it, clock is also synced over ADAT. Since my routing is pretty much fixed I use it as a 'dumb' 8 channel convertor so I don't use the UA Console. I lost access to UAD processing this way though.
 
I loved my RayDAT. Had a UAD 4-710d (that an Avalon 737 linked in to UAD) and a pair of MR816CSX's connected to it via ADAT. It worked really fine.

I sold all my PreAmps and the RayDAT card and moved to dual Digigrid IOX's. I love this setup and all of the flexabilty way more than my PCIE setup. The preamps in the IOX are really nice. I miss nothing from my analog stuff. Much easier to manage my setup now.

My only reasons for getting rid of my RME RayDAT setup was to get away from having to manually adjust my preamps, and the flimsy ADAT connectors.

If you are going to go the PCIE route, I also highly recommend RME. Always solid drivers and I never had any issues with my setup.
 
I assume RME RayDat so you can use your Apollo as AD/DA convertor. Personally I run a RME Babyface Pro with the Apollo connected over ADAT to it, clock is also synced over ADAT. Since my routing is pretty much fixed I use it as a 'dumb' 8 channel convertor so I don't use the UA Console. I lost access to UAD processing this way though.

This is helpful! I think I'm confused how it all comes together. If RME is handling ASIO, Apollo as A/D, can I still send other preamps to the Apollo's ADAT and SPDIF to make use of it's full 18 channel capacity? (I currently send a Saffire PRO40 to Apollo's ADAT to get 8 more ins.) Or would everybody connect to the RME directly?
 
This is helpful! I think I'm confused how it all comes together. If RME is handling ASIO, Apollo as A/D, can I still send other preamps to the Apollo's ADAT and SPDIF to make use of it's full 18 channel capacity? (I currently send a Saffire PRO40 to Apollo's ADAT to get 8 more ins.) Or would everybody connect to the RME directly?
Hmm, good question, from what I remember I ran into some routing limitations with the UA console, just can't quite remember what it was exactly that was bothering me. Currently setting everything up in a new room, so it's a perfect time for me to revisit the UA console to set up the routing again, so I will report back on my experience either later today or tomorrow.
 
You would connect Saffire to the RAYDAT ADAT inputs. You will have 8 ADAT port in banks of 8 for 32in and 32out, plus SPDIF/AES outputs. All the routing gets done in TotalMix, although it is a limited version of TotalMix for the RME RayDAT card.
 
I do use PCI cards for audio but before you spend a lot of money on it and, even worse, go through the brain damage of resetting your signal flow, you owe it to yourself to find out just how much latency this is really going to save.

It does help, but I don't know that it's "night and day" difference. There easily could be something else that's creating latency in your setup, whether it's your DAW buffer, or you're monitoring through effects, or something else altogether.

Substituting a PCIe/PCI card for a USB interface no doubt will be a plus but, depending on the rest of your setup, such a move might offer negligible improvement against other impediments.
 
Substituting a PCIe/PCI card for a USB interface no doubt will be a plus but, depending on the rest of your setup, such a move might offer negligible improvement against other impediments.
Agreed. The rest of the setup needs to be checked too, software and other hardware.

If the plugins/template allow for a reasonable latency, then adding a PCI card will get the desired result. How about a used RME Hammerfall card? Great bang for the buck and rock-solid drivers. Had one for many years and never had any trouble with it.
 
There is also the Windows 10 issue with trying to kill off firewire support. I recall having to install some old kind of legacy driver to get firewire to work for my Steinberg interfaces. So, I do not think retiring that legacy hardware as the interface is a bad idea. It is kinda what i did with the MR816csx. I was able to keep using the Preamps buy running them into the RayDAT card.

But...I totally agree with JohnG and others said about making sure your system is tuned end-to-end, first and foremost. That will go a long way to figuring out where you are getting latency or drop outs. My last nightmare turned out to me a USB flash drive driver issue. Video cards and Network card drivers are another common source of trouble that i have experienced in the past.

My Didigrid stuff is not faster than my PCIE RayDAT. Basically a toss up. But, the workflow and connectivity improvements were wroth it to me. Actually, it was a push in cost after I sold all the external preamps.

It kinda ends up boiling a bit down to budget. If you have the resources, buy the best you can afford, lock it down, and get back to making music.
 
RME Hammerfall card

Yes it's awesome -- use a lot of them -- but they are PCI, not PCIe, so new motherboards often lack them. There is a Startech adapter available that seems to work well (I tested it) but the hardware fit is not great for my cases. Consequently, I'm still choosing among those few MOBOs that retain a PCI slot to accommodate the Hammerfalls.
 
I am using a x32 as my audio device. Usb performance on that box is horrible. Worse then average and usb in general is worse then pci. I recently spent some bucks for the lynx AES16e-50 pci card which basically connects to the x32 over AES50 and bypasses usb entirely. Best case latency (without dropouts) measured with external loopback dropped from over 11ms round trip latency on usb to 3.5ms going through pci and that’s including AES50 overhead and x32’s overhead.

For me, worth it, but it costs an extra thousand bucks for that luxury.

Usb is very plug and play. Simple. If you can tolerate 10ms latency which is pretty typical for usb, then save the money and complexity. If you want 5ms or even less then you’ll have to spend some bucks and go either pci or thunderbolt.
 
I am using a x32 as my audio device. Usb performance on that box is horrible. Worse then average and usb in general is worse then pci. I recently spent some bucks for the lynx AES16e-50 pci card which basically connects to the x32 over AES50 and bypasses usb entirely. Best case latency (without dropouts) measured with external loopback dropped from over 11ms round trip latency on usb to 3.5ms going through pci and that’s including AES50 overhead and x32’s overhead.

Totally agree on X32 USB being too slow. I have X32R and M32C. I bought the Waves X-WSG cards ($400 each) for them so they integrate in my Soundgrid network. I can run with about 3.5ms latency with Soundgrid, but I tend to run it a bit higher 4.833ms. I pretty much just use them for IEM mixes, since my two IOX's cover most multitrack input situations. X32 works great with the M32Q app on iPhone and wired IEMs.

I did have the X-Adat card when I had my RayDAT. Worked great, with nice low latency, but the ADAT cables are not rugged enough for live rig, or even in my studio, as things get moved around alot.
 
I actually got the x32r for my studio and might use it later for gigging, but in my studio it provides me currently with 24 mic pres, a few more extra inputs and rather complex monitoring scenes to handle things like 5.1 surround mixing or not or feeding stereo to all 4+1 speakers in the room if I feel like it or whatever. Plus hardware monitoring with decent fx and limiting. It’s a great studio device when pci card is used to turn it into low latency audio device.

And in a pinch I can mix a live show with it.

When I was looking deeply and testing RTL I found out that the lynx pci card only adds 5 samples of latency over and above the audio buffer size. This is reported in its driver. The usb driver, on the other hand adds hundreds of samples of safety offset. Thus no way around long latency with it. It has to do that because usb2 couldn’t keep up otherwise. There is no point in running their usb driver below 128 sample buffer size and 256 is better.

Anyway pci is the way to go unless you have a newer Mac with thunderbolt. I hear rumors that the next gen Mac Pro will have a next gen pci bus.
 
I do use PCI cards for audio but before you spend a lot of money on it and, even worse, go through the brain damage of resetting your signal flow, you owe it to yourself to find out just how much latency this is really going to save.

It does help, but I don't know that it's "night and day" difference. There easily could be something else that's creating latency in your setup, whether it's your DAW buffer, or you're monitoring through effects, or something else altogether.

Substituting a PCIe/PCI card for a USB interface no doubt will be a plus but, depending on the rest of your setup, such a move might offer negligible improvement against other impediments.

Very sensible. I've been throwing money at this problem without really figuring out the source.

I've run LatencyMon and I can see buffer times in Cubase but is there a way to understand specifically what performance impact the Apollo is responsible for?
 
I've run LatencyMon and I can see buffer times in Cubase but is there a way to understand specifically what performance impact the Apollo is responsible for?

I don't have an Apollo, but I would think that info is readily available?

There are, alas, innumerable avenues by which latency can sneak in. If you have instantiated a plugin that has automatic delay compensation in your host, you could inadvertently be creating a lot of mush even if you aren't using that track in a particular piece of music -- even if you forgot it's there!

So first thing for me was to disable all plugins in my Pro Tools rig (through which I monitor). Then I could drop the buffer to 64, raising it if I am ready to record and mix.

Sadly, it's an art.
 
Whilst other people provided some excellent info on things that you could look at before spending money. I hooked up the apollo and I realised what my issue was, only having either 4 stereo or 8 mono tracks to route. By default everything goes out to the monitor outs. Here is what I have:

Apollo Inputs->outputs to RME
1/2->Adat 1/2
3/4->Adat 3/4
5/6->Adat 5/6
7/8->Adat 7/8

I also wanted to have all the adat(1-8) inputs go out of the analog(line 1-8) outs of the apollo..I can't do this because the UA console only allows routing of 4 stereo tracks or 8 mono tracks (or any combination). Seems fairly limited to me.
 
I don't have an Apollo, but I would think that info is readily available?

There are, alas, innumerable avenues by which latency can sneak in. If you have instantiated a plugin that has automatic delay compensation in your host, you could inadvertently be creating a lot of mush even if you aren't using that track in a particular piece of music -- even if you forgot it's there!

So first thing for me was to disable all plugins in my Pro Tools rig (through which I monitor). Then I could drop the buffer to 64, raising it if I am ready to record and mix.

Sadly, it's an art.

I understand. I guess what I mean to ask is, the figures reported in Cubase for input latency/output latency with no plugins instantiated, is it correct to assume those are mainly owing to interface only?

If so, I'm getting ~3.5ms in/out latency at 64 samples and 48khz sample rate. What I'm reading is PCIe can get <1, but you have to tack on another 1ms for the A/D. So theoretical latency savings might be on the order of a millisecond. Seems like not a huge distinction in latency, but is it better performance under load? My #1 issue is crackling at low buffer with hungry libs.

My understanding is PCI/PCIe communication is going to be more efficient and so maybe I would see better performance in big projects. Has this been your experience?
 
Yes it's awesome -- use a lot of them -- but they are PCI, not PCIe, so new motherboards often lack them. There is a Startech adapter available that seems to work well (I tested it) but the hardware fit is not great for my cases. Consequently, I'm still choosing among those few MOBOs that retain a PCI slot to accommodate the Hammerfalls.

I would be interesting to know what boards you are using? Are these boards still being made with PCI slots for modern chipsets?`
 
Top Bottom