1 Topic by domfortu (edited by domfortu 2023-12-31 14:15:23)

  • Registered: 2011-04-26
  • Posts: 7

Topic: Cracks with Babyface USB-A (via ASIO Fireface interface, old PC)

Hi, I am having cracks (not much glitches, but audio cracks every few seconds) when using ASIO Fireface playing via MIDI with my Babyface with my old PC. It happens only with specific VST (mainly KeyScape) but not with all VSTs and when it happens, as expected, CPU indicator in Ableton 11 shows >100% numbers.
Even if I increase "buffer size" from 128 to 256 even to 1.024 (with latency going then up to 40-50 ms, not ideal for real-time piano playing) some minor cracks still occur on some VSTs (Keyscape, Omnishpere, etc) making it almost unusable. I use to keep Frequency=44100 Hz.
More HW details on my PC and Babyface below, but the main question to me is:
Will I solve cracks by just buying a modern mid-range PC or a Macbook Air M1?
Or should I consider other alterative paths insted (specially if less expensive) ? Thinking to some specific driver updates that I am currently missing, or thinking to upgrade the Babyface soundcard itself with a new RME card, upgrading to a new USB-A cards, or eventually other hardware-related or software-related settings / ideas that I am not considering right now to avoid audio cracks?
In few words, whatever reassurance you can give on succeeeding in avoiding cracks is more than welcome !
Thanks

Actual configuration:
PC Model: ACER M5810 (year: 2009)
CPU: Intel     Core i7 860 @ 2.80GHz
OS: Windows 10 Pro - latest patches.
RAM: 6,00GB Double Channel DDR3 @ 664 MHz (9-9-9-24)
HD: SSD 500 Gb
DAW: Ableton 11 Intro
VST: latest version of Keyscape, many other VSTs.

Babyface Driver: 1.246 (Windows WDM and ASIO 2.3 support)
Babyface Driver Date: 11.07.23
Babyface HW Revision: 226

2 Reply by ramses (edited by ramses 2023-12-31 15:41:14)

  • ramses
  • DAW Mastermind
  • Offline
  • From: UTC+1
  • Registered: 2012-10-14
  • Posts: 13,161

Re: Cracks with Babyface USB-A (via ASIO Fireface interface, old PC)

This is a 14y old CPU offering only low performance. Look at the passmark mixed benchmark and single thread results.
https://www.cpubenchmark.net/cpu.php?cp … z&id=6

As you noticed already, a higher ASIO buffersize cannot compensate for the lack of CPU power if you are using e.g. a CPU hungry VSTi. Get a new PC with at least 6–8 cores and a base clock of 3.6-4.0 GHz at minimum.

VSTi can run multithreaded (utilize multiple CPU cores), but most VST run on a single thread (on one CPU core).

Therefore, do not only look for benchmarks numbers which utilize all CPU cores, take also a look at the single thread performance. The passmark results show both, giving you a good overview.

Regarding Apple: very expensive, especially when ordering with more DRAM and storage. Then you will see quite many postings of people also having issues with pops and clicks, where the root cause is still not 100% clear (although the systems have a performant CPU). Seems to be on either Apple or user side (upgrade installations). Therefore, you might have a happier time with getting a Desktop PC.

I would buy from a company who offers turnkey systems with tested components and drivers to be on the safe side and to have fewer/no hassles. Ask them whether AMD or Intel would be an advantage for your setup/applications (working with VSTi).

Regarding OS: If I were you, I would stick to Win10 as long as possible.
Win 11 only has 1 month (only!) more life in terms of support, but gives you nothing but downsides. First, in the Pro version you now get online accounts (which you don't want). And the new core virtualization, which is poorly designed because it hogs so much performance, severely hurts performance.
Other problems, there are so many things that were first deleted and then reprogrammed. There is no clear concept. Win11 looks more like an itinerant construction site and a concept optimized for SaaS (online accounts), I fear that if this continues, you will be able to take out a subscription for the operating system in the future. Nobody really wants that.
But at the end of the day, it's still better than having to pay Apple's inflated hardware prices. Since you can't upgrade, you shouldn't buy too small, and then it will be even pricier than it already is.

Don't forget a good backup concept / application: Macrium Reflect Home is currently the best.

Look here how I could upgrade my PC system for a very long time:
https://www.tonstudio-forum.de/blog/Ent … mponenten/

and what performance it still has, although it is also quite a few years old:
https://www.tonstudio-forum.de/blog/Ent … cks-de-en/

It's all there … 8-core CPU now … 64 GB DRAM … Only SSDs inside … meanwhile, 2×10 GBit network adapter.
Internal 20 TB storage based on three Samsung 870 EVO QVO as RAID-0 (quick backup cache, NAS alike).
Redundancy by two external USB3 connected 10 TB disks … therefore, no issue to have this implemented as RAID-0.
I needed to take three of the QVO's as they become slow after the cache is full (approx after 30s sustained read/write operation). Three of them in a RAID-0 make them much faster. And the QVO are the only "payable" to get 20 TB storage.
Still a silent PC … my room has 28 dB.

BR Ramses - UFX III, 12Mic, XTC, ADI-2 Pro FS R BE, RayDAT, X10SRi-F, E5-1680v4, Win10Pro22H2, Cub14
  • Registered: 2011-04-26
  • Posts: 7

Re: Cracks with Babyface USB-A (via ASIO Fireface interface, old PC)

Thanks Ramses. All useful information and yes, in line with what I was expecting (buying a 2024 mid-to-high end PC that will realistically solve all of this audio cracks to me). But before doing that, I just noticed an unexpected new berhaviour in "Windows Task Manager". In the same moment when Ableton goes constantly at 200% / 400% (playing hard on my MIDI piano sending to VSTi and yes, lots of audio cracks) my task manager (CPU performance real-time graph in Windows 10) never shows more more than 20%-30% ! Also by monitoring each of the four CPUs, none of them goes above 30%.
Technically, if this real-time info on CPU from Windows 10 task manager is correct, It looks like there is a sort of "cap" / "limit" of CPU throughput assigned to Ableton (30%) and that in the end it is not using the whole CPU power (wich is still 14 years old, as you correctly noted!). I need reassurance that I don't have a "limit" / a block somewhere (even if I understand I should ask this question more into an Ableton forum rather than here, probably).
Thank you again !
D.

4 Reply by ramses (edited by ramses 2024-01-01 14:09:53)

  • ramses
  • DAW Mastermind
  • Offline
  • From: UTC+1
  • Registered: 2012-10-14
  • Posts: 13,161

Re: Cracks with Babyface USB-A (via ASIO Fireface interface, old PC)

Check with LatencyMon.
Maybe one or more CPU cores are blocked too long by a (possibly bad) driver who claims a CPU core for too long.
Low-level routines (drivers) have the highest priority in a system.
They may not be interrupted by the process scheduler to ensure data integrity.

As Windows and macOS are not real-time Operating Systems, the drivers are coded based on "programming conventions" how long they may work on a CPU core, before they detach from the core themselves. EDIT: some drivers might misbehave here if not properly coded, possibly to run a little longer for achieving better benchmark results than the competition, who knows ...

As you only have 4 cores and 4 hyper-threads the likeliness is high, that audio related threads are scheduled to a CPU core which is also processing low-level routines. Then it needs to wait until the low-level routine detaches from this CPU core. This wait time is the DPC latency.

LatencyMon you best execute on a completely IDLE system as it induces a typical DAW load already during measuring.
Although LatencyMon allows latencies up to 1000 µs, it is better if this latency is not too high.
For my system under Windows 7, the values had a minimum of 1,8µs with an average of 40 and rare peaks of up to 100µs.

This means the CPU cores run agile and efficient and work quickly on any upcoming audio load and as a tendency is able to work on a DAW load with lower ASIO buffersizes.

With Windows 10, this value went up a bit, but also the measuring possibility inside Win10 changed, so the numbers are not really comparable any more for my understanding.

Let me say, if LatencyMon reports values below 100 or 200µs, then everything should be really fine. Up to 1000µs is allowed, but then you may expect that the CPU cannot handle higher audio loads "very well" especially not with small ASIO buffer sizes. And to be able to play a VSTi without lag, you need to use ASIO buffer sizes ≤ 128 samples, sometimes even 256 samples work, but you might get the feeling during playing, that the instrument doesn't react immediately to your playing.

BR Ramses - UFX III, 12Mic, XTC, ADI-2 Pro FS R BE, RayDAT, X10SRi-F, E5-1680v4, Win10Pro22H2, Cub14
  • Registered: 2011-04-26
  • Posts: 7

Re: Cracks with Babyface USB-A (via ASIO Fireface interface, old PC)

Thanks again Ramses. Being used to 128 samples in buffer size for reactiveness, as you pointed out,  LatencyMon gives values that are all above 200µs. In specific:
- Higest IPS: 572 µs
- Higest ISR: 220 µs (DirectX kernel)
- Higest DPC: 691 µs (ntoskrnl.exe, but previously was NVidia Windows Kernel Mode driver 457.51 with value around 550 µs).
Also, I checked also this, and I don't have any CPU Throttling enabled in Win10 (pecentages are all 100%) or in BIOS.

I believe the conclusion remains the same: with a proper new system, maybe with a Core i5-13500 (single thread rating = 3.899) and proper SSD and DDR4/5 and yes, with a more solid Win10 instead of 11, I should solve it all. Thank you again for your help today !

6 Reply by ramses (edited by ramses 2024-01-01 14:25:40)

  • ramses
  • DAW Mastermind
  • Offline
  • From: UTC+1
  • Registered: 2012-10-14
  • Posts: 13,161

Re: Cracks with Babyface USB-A (via ASIO Fireface interface, old PC)

I wouldn't use this CPU because for my taste the base clock is quite low with only 2.5 on power and 1.8 GHz on efficiency cores.
https://www.cpubenchmark.net/cpu.php?cp … mp;id=4993
https://geizhals.de/intel-core-i5-13500 … 71148.html €255

What about this one (without GPU, I expect you buy a graphic card anyway, or?)
https://www.cpubenchmark.net/cpu.php?cp … mp;id=5061
https://geizhals.de/intel-core-i7-13700 … 10175.html €359

Otherwise, this with GPU:
https://geizhals.de/intel-core-i7-13700 … 20437.html €410

A good overview about Core I CPUs and their options here:
https://de.wikipedia.org/wiki/Liste_der … rozessoren

BR Ramses - UFX III, 12Mic, XTC, ADI-2 Pro FS R BE, RayDAT, X10SRi-F, E5-1680v4, Win10Pro22H2, Cub14
  • Kubrak
  • DAW Mastermind
  • Offline
  • Registered: 2023-08-03
  • Posts: 541

Re: Cracks with Babyface USB-A (via ASIO Fireface interface, old PC)

I would not go for Intel at all. Or to 11th gen at most..... There is also AMD and it has superior CPUs to Intel's ones....

I would get mad from Intel's big.little and what it brings to realtime applications like audio is. 

Concerning mismatch between DAW CPU usage and Win Task Manager.... DAW CPU usage does not show CPU usage. It shows the percentage of audiobuffer used. That figure is more related to singlecore CPU load than to overall CPU load. And not only CPU load plays the role, but also delay by interupts and so on.... DAW CPU usage figure is only very loosely related to real CPU usage.

  • Registered: 2011-04-26
  • Posts: 7

Re: Cracks with Babyface USB-A (via ASIO Fireface interface, old PC)

Thanks again. Yes, DAWs should not even call this as a “CPU load” to avoid confusion, but should call it “System load” instead, or a generic load indicator.
PS I will consider AMD vs Intel for sure.
Best !

9 Reply by ramses (edited by ramses 2024-01-01 18:32:08)

  • ramses
  • DAW Mastermind
  • Offline
  • From: UTC+1
  • Registered: 2012-10-14
  • Posts: 13,161

Re: Cracks with Babyface USB-A (via ASIO Fireface interface, old PC)

Kubrak, have you actually had any personal experience with Intel Big/Little? The Big/Little issue, if there is one for the specific use case, can be resolved using Process Lasso Pro. There, you can allocate applications to cores if the default distribution isn't optimal.

I have certain reservations about AMD because in some areas, they also seem to engage in certain tricks (for example, the 7950X3D where only half of the chiplets have access to the larger cache, or issues with USB 3 implementation in the chipset). This might be outdated information, but I do believe that Intel has more experience in this particular area. Intel might not currently perform as well or efficiently in benchmarks, but for audio applications, what matters more to me is that they can serve applications with near-real-time requirements well and ensure USB operates smoothly. While you can still upgrade with a Sonnet card, it's becoming increasingly difficult with today's consumer systems due to fewer PCIe expansion slots, unless you're considering workstation or server motherboards.

Therefore, I would suggest reaching out to a company that offers both CPU types and can run LatencyMon on Intel and AMD systems of similar performance levels. I find the market somewhat confusing at the moment, and I wouldn't solely rely on forum statements. Currently, I would find it quite challenging to make a selection, as I'm unsure whom to trust or which statements are reliable in today's landscape. One of the reasons why I upgraded my performance wise still good working 9y old system a lot now.

All I want to say is simply careful, its difficult to do a good purchase decision if you have nothing to compare.

BR Ramses - UFX III, 12Mic, XTC, ADI-2 Pro FS R BE, RayDAT, X10SRi-F, E5-1680v4, Win10Pro22H2, Cub14
  • Kubrak
  • DAW Mastermind
  • Offline
  • Registered: 2023-08-03
  • Posts: 541

Re: Cracks with Babyface USB-A (via ASIO Fireface interface, old PC)

No, as I have said, I am not big.little fan. I have AMD and Intels that are pre-big.little.

Yes, one may use Lasso or deactivate e-cores in BIOS. But why.

I know, that people keep reporting problems with big.little on Native Instruments forum. The both on Intel and Apple Silicon. And it seems to me that RME driver developers "fight" with Apple CPU since very beginning... My guess it is also because of big-little. But only RME developers could say for sure.

I do not have any problems with USB on my AMD computer. I use that computer just for audio.

I have not checked, but I think that in case of 7950X3D the both chiplets have access to 3D cache. But the cores that are on the chiplet that carries also 3D cache run on lower clock because of heat dissipation partly blocked by 3D cache. That is design decision. They could also place 3D cache on second chiplet, but they would have lower singlecore results....

One may avoid that if does not want it. One may buy similar CPU without 3D cache, or if wanting the same core performance and 3D cache one may use model with single CPU chiplet.