Jump to content

Jim Roseberry

Members
  • Content Count

    693
  • Joined

  • Last visited

  • Days Won

    8

Posts posted by Jim Roseberry


  1. 13 minutes ago, Teegarden said:

    Thanks for the clear answer!

    I guess that's not the case when your PC power profile is set to maximum performance all the time (and good cooling to avoid throtling)?

    And what is the influence of the soundcard regarding latency?
    I've got a ten year old RME PCIe AIO card that gives effective latency of 2.7 ms with 256 buffer, 24bit, 96kHz, 64 bit double precision buffer (ASIO reported latencies:  input 3.1, output 3.7, total roundtrip 6.8ms). That seem relatively good figures to me (working without glitches when all unnecessary background processes are eliminated)? 

    And the fastest PCIe 4.0 SSD's? Don't they also improve latency? The 10900k has only PCIe 3.0...

    I agree regarding the cost/performance of the 10900k, but I'm quite interested in the new AMD, will just wait till its available at reasonable price🙂

    Power profile is certainly part of the equation... but not all of it.   

    CPU cooler has to be able to keep up (or course).

     

    Your audio interface determines the latency your DAW can achieve.

    The CPU has to be able to sustain the load (glitch-free)... but has no direct affect on the actual latency figure.

     

    6.8ms total round-trip latency isn't bad.

    If you want lowest possible round-trip latency, Presonus Quantum can achieve sub 1ms (96k using a 16-sample ASIO buffer size).

    Right now, only the 10900k can sustain any type of load at those settings... and that's do to the super high clock-speed (5.3GHz).

     

    PCIe 4.0 SSDs have zero impact on audio latency.

    PCIe 3.0 SSDs sustain 3500MB/Sec.

    PCIe 4.0 SSD sustain about 4000-5000MB/Sec.

     

    You may be surprised to hear that the 5950x is currently performing slightly better with the B550 vs. X570 chipset.

    X570 is also active-cooled... which means small (annoying) high-RPM fan.  Noisy!

     

    5950x is an $800 CPU.  

    If you're waiting for significant discount, it's going to be an extended wait.  😉

     

     

     

    • Thanks 1

  2. 15 hours ago, Teegarden said:

    My impression (after reading tons of topics) in our case is that it has to do with the impaired internal latency of the former generations AMD CPUs, unfortunately. I guess @Jim Roseberry can say more about this.
    I've managed to improve performance by tweaking my Threadripper system to the max, but at the cost of way too many unproductive hours and now my hardware is starting to wear out. So now, when I can finally record something decent, I might need to replace things again🤔

    CPU speed is certainly an important factor... but there are numerous facets that affect performance.

    Just because you see "CPU headroom", that doesn't mean the machine is going to perform flawlessly (for DAW purposes).

     

    Threadripper's multi-threaded performance is off-the-chain good, but it's ultra low-latency performance is poor (even the 3970x).

    The new 5950x (Vermeer) is the first series of AMD CPUs where that's finally been resolved.

    The 5950x can run loads at a 32-sample ASIO buffer size... that the 3970x just can't sustain (glitch-free).

    That said, the new 10th Gen i9-10900k can run Helix Native (with a substantial patch using two 2048-sample Cab IRs) completely glitch-free at 96k using a 16-sample ASIO buffer size.  The first CPU to be able to effectively do this (no glitches).

     

    Even with a 10900k or 5950x, if the machine's DPC Latency is high... you'll experience audio glitches/drop-outs.

    Lets say you want to run Helix Native at 44.1k using a 64-sample ASIO buffer size.

    That means the machine has 1.5ms to process the next audio buffer and get it in cue for playback.  

    Anything that interrupts this process will cause an audio glitch/dropout.  (ie: Poorly written drivers can monopolize the CPU.)

     

    Processes constantly running in the background (backup/sync, A/V, etc) can negatively affect performance.

     

    There's performance and power throttling:

    Say you have a typical song structure... where the song starts with maybe 8-16 tracks of drums, guitar, bass, keys, and lead vocal/melody.

    When the song reaches the bridge, let's say it breaks down to just the kick and a single bass part.

    At this point, CPU use (demand) is extremely low... so the system decides to throttle CPU clock-speed down to 1/4 speed... as well as parking several cores.

    When the stripped-down bridge ends, here comes the massive chorus-out-vamp.  Huge stacks of backing vocals, synths, etc.

    That massive CPU load now falls on the CPU running at 1/4 clock-speed... with several cores that have been parked.

    Glitches at best... or a complete transport dropout.

     

    The harder you're pushing the machine (heavier loads, lower latency), the more important all the details.

    A general-purpose user wouldn't notice a few millisecond hiccup in data-flow.

    For someone wanting to run at a 64-sample ASIO buffer size, those few extra milliseconds can result in glitches/dropouts.

     

     

    • Like 1
    • Thanks 1

  3. With proliferation of (live) streaming, many companies don't feel the need to be bound to NAMM's schedule (especially when it's virtual).

    D.I. from Line-6 (owned by Yamaha) mentioned this recently on The Gear Page.

     

    • Thanks 1

  4. On 1/8/2021 at 6:45 AM, Bruno de Souza Lino said:

    Of course the Cab Clone IR is great. It was licensed and built by Two Notes (the guys that make the Torpedo stuff). The first Cab Clone was almost unanimously disliked for sounding like crap.

    The IR loading section was licensed from Two Notes.  The reactive-load is not made by Two Notes.

    I have a pair of Captor-X boxes.  They're great (especially for the cost).

    The reactive-load in the Badlander (to me) sounds slightly better than the one in Captor-X

    The IR loading in Captor-X as far more advanced:

    • Can load/mix a pair of Cab IRs in realtime
    • Can determine the length of the imported IR (which determines latency)
    • Onboard Reverb, EQ, Enhancer, Tuner (which is erratic), Gate

  5. Just picked up a Badlander 50w head.

    The onboard reactive-load and CabClone IR sound excellent.

    Lots of features and versatility... and not over-the-top cost wise

     

    It's always a little sad to see a heritage company merge or get bought.

    • Like 5

  6. Last time I was testing the 3950x and 3970x, I remember seeing a performance boost using the "Aggressive" Thread Scheduling Model.

    I also remember some issues when using the Aggressive Mode. 

     

    I've been testing the new 5950x.

    AMD has ***finally*** solved their performance issue at ultra low latency settings.

    I'll revisit the Thread Scheduling Models... and give some details about performance.

    • Like 5

  7. There is a Eucon control surface plugin for Cakewalk.

    You need that, the Eucon installer from Avid, and Eucontrol for iPAD.

     

    I currently don't have Eucon installed on my machine.

    I had it installed/working just fine with CbB.

    It's a little more convoluted to configure (than native support for Eucon)... as you have to manually configure the various options in the Eucon control surface plugin.

     

    You can use Eucontrol to control any Eucon enabled application (ProTools, Cubase, Samplitude ProX, etc).

    Capabilities go far beyond a simple control surface.

    ie: You can directly access menu items via a single touch.

    Even setting up some basic things (undo, redo, quantize, Views, Process menu options, etc) can greatly enhance workflow.

     


  8. On 10/21/2020 at 2:36 PM, Tony Carpenter said:

    I want to know when mickeysoft are going to emulate Core Audio and allow aggregate devices!!.. It's the 21st century, we are on 64bit OS and with silly amounts of RAM and CPU power... drivers from major audio hardware makers are stable.... excuses now still??... It's the one part of Windoze that still piiiii sss es me off :).

    Unless those multiple audio interfaces are running via a single clock source, the aggregate device is all but useless.

    Audio tracks recorded/played via two separate audio clocks will drift apart over time (due to the small differences between the clocks).

    A more robust setup is two audio interfaces (same model/series) that were designed to operate (simultaneously) under the same driver (essentially adding more I/O).

    Those with multiple audio interfaces all running from a single clock-source (those who'd actually/practically benefit from aggregate devices) are a minuscule part of the overall DAW using community.  The DAW using community is a tiny fraction of the overall number of computer users.  This is why MS doesn't spend development time... to allow aggregate devices. 

     

     

    • Like 2

  9. 18 hours ago, craigb said:

    So, per Jim above, the entire World must be currently in "Long Battery Life" mode... 😏

    In nearly 30 years, I don't think I've ever been contacted by a client who wanted to work at high ASIO buffer sizes (higher latency)... or at lower CPU clock-speed (to conserve battery-life or power).  😁

    Folks have different levels of needs/expectations.

     


  10. Live Nation recently publicized that they're going to ask performers to take a substantial pay cut.

    This (naturally) upsets performers... 

    If the venue can only be 1/4 capacity, that means 1/4 the intake.

    Everyone loses... and ticket prices skyrocket even higher

     

     


  11. 9 minutes ago, craigb said:

    no one will know,

    until the "World leaders" are under control.

    Fixed x2   😁

     

    In Ohio:

    Gigs (only large clubs can afford full bands) start at 6-7pm... and end at 10:30-11pm

    Last-call is at 9:45pm.

    Alcohol sales end at 10pm.

    Bars have to be closed at 11pm.

    I don't mind the early gigs... but does the "Rona" only come out at 11:01pm?

    People are "loading up" at last call... leaving hammered... and/or going to private after-parties to continue.

     


  12. 18 minutes ago, Scott Kendrick said:

    I want to make sure I can map the captured midi to the individual tracks for mixer track in Ezdrummer - e.g. a kick, a snare, overheads - etc... that's the way I record now and I do this so I can independently apply effects, adjust volumes etc. The only way I've approached "routing" the captured midi is to insert the EZD instrument track after the fact and copy paste the midi from the base midi track into that instrument track. Hope that makes sense.

    Also I've been learning towards the FocusRite Scarlett 2i2 for my audio interface - seems like a pretty popular one, but would it work for these purposes we are discussing? (if you're familiar)

    You'll need a MIDI Drum Map that changes the MIDI notes (recorded from the TD25)... into the MIDI notes that trigger the desired sounds from EZ Drummer.

    If one doesn't already exist, I'm sure you could create one in 30-60 minutes.

     

    Create a test project with the following:

    • MIDI Drum performance (recorded from the TD25) - on a MIDI Track
    • MIDI Drum Map (to map the TD25 pad notes to articulations in EZ Drummer)
    • Instrument Track with EZ Drummer

    You could (also) change the MIDI notes the TD25 pads send... to match the desired sounds in EZ Drummer (thus eliminating the need to use a MIDI Drum Map).

    There are also MIDI Drum Maps in EZ Drummer (there is a "Roland" Map but not sure it matches the TD25).

    There's no right/wrong way to get to the proper MIDI notes (to trigger the desired EZ Drummer sounds).  Pick what makes most sense to you.

     

    With the above test project:

    On the MIDI track with your MIDI drum performance, select the MIDI Drum Map (you created for the TD25) as the MIDI output

    If you use a MIDI Drum Map in Cakewalk, that Drum Map allows you to select the EZ Drummer instance for the MIDI output of each mapped MIDI note

    • If you use EZ Drummer's MIDI Maps, skip this step.
    • If you changed the MIDI notes sent by the Nitro pads (on the TD25 itself), skip this step.

    If not using MIDI Drum Maps in Cakewalk, set the MIDI Output (of the recorded MIDI Drum track) to EZ Drummer.

    The recorded MIDI drum performance should now be triggering EZ Drummer (with the desired articulations).

     

    When you insert EZ Drummer into your Cakewalk project, you have two options for the Instrument track.

    1. Simple Instrument Track - creates a single stereo output for EZ Drummer
    2. Instrument Track Per Output - creates a track for each audio output from EZ Drummer

    With the first option, you'd have to mix the drums within EZ Drummer.

    With the second option, you've got individual mic channels that can be mixed within Cakewalk (Kick, Snare, HiHat, Toms, Overheads, etc).

    There's no right/wrong... just how you prefer to work 

     

    The Focusrite Scarlett series is OK when it comes to round-trip latency (~6ms at the smallest buffer size). 

    It would be better for triggering EZ Drummer samples in realtime (than the TD25's onboard audio interface)... as it offers lower round-trip latency.

    Some drummers would find even 5-6ms round-trip latency too much lag to comfortably trigger EZ Drummer samples in realtime.

    Get that lag (round-trip latency) sub 3ms... and it feels a lot more immediate.

     

     

     


  13. 30 minutes ago, Bapu said:

    @Jim Roseberry

    Please define the range of large buffer sizes. 1024? 2048? 4096? larger?

    Is smallest 512 and lower?

    The lower the latency (smaller the buffer size), the greater the advantage of the 10900k's significantly higher clock-speed.

    The higher the buffer size (higher the latency), the greater the advantage of the 3970x's many additional cores.

     

    I don't consider 512-sample buffer size to be small.

    I wouldn't work at anything higher than 256-samples.

    Most times, I'm working at 32/64/128-sample ASIO buffer size.

     

    Insert a single instance of Helix Native... running with Presonus Quantum set to 96k using a 32-sample ASIO buffer size.

    That's running at 1ms total round-trip latency.  An extreme example...

    • AMD 3950x and 3970x - audio will glitch
    • Intel 10900k - no audio glitches

    If I'm paying $2000 for a CPU, I want it to excel at ultra low latency performance.

    10900k is about 1/3 the cost of the 3970x... and offers a great balance of features:

    • 5.3GHz all-core clock-speed
    • Runs quiet with large/quality air-cooking
    • 10 cores/20 processing threads
    • TDP = 125w

    There's no such thing as a quiet 3970x machine (TDP = 280w)... unless you allow it to thermal-throttle (which negates the purpose)

    • Like 1

  14. Your audio interface will determine the lowest latency you can achieve.

    Of course, the machine has to be able to keep up with the load (or you'll hear glitches or experience drop-outs).

     

    Are you using the TD25 as your audio interface... or are you using the Laptop's onboard audio?

    The TD25 (if it's like the large TD50) functions as a USB audio interface.

    Between the Laptop's onboard audio and TD25, the TD25 has a proper ASIO driver (which will out-perform ASIO-4-ALL).

    If you're after super low round-trip latency, you won't be happy using either the TD25 or Laptop's onboard audio as your audio interface.

    If getting a dedicated audio interface (that offers low round-trip latency) isn't an option, use the TD25.

     

    If you're dealing with an audio interface that doesn't offer 5ms round-trip latency or lower, forget triggering samples from EZ Drummer in realtime (there will be too much lag).  

    In this scenario, I'd record the MIDI performance from the TD25 (monitoring its internal sounds while tracking against the click-track).

    Once the TD25 MIDI performance is recorded, you can then route the MIDI track to an instance of EZ Drummer.

    You'll miss being able to play the EZ Drummer sounds in realtime... but it's the most effective work-around (to avoid latency).

     

    If you had a Thunderbolt audio interface like the Presonus Quantum... paired with a fast machine, you could trigger EZ Drummer in realtime... with no perceptible lag/latency.  

    Quantum can run at 96k using a 32-sample ASIO buffer size (resulting in total round-trip latency of 1ms).  

    At these settings, the machine has 1/3 of a millisecond to process the next buffer and get it in cue for playback.

    If anything interrupts this process, you'll hear glitches.

    This can be done, but (needless to say)... it's extremely demanding.

     


  15. 16 minutes ago, Starise said:

    There has been talk of a 10 core + eventual ARM processor. That doesn't sound like small device territory any more. 

    How does adding more cores affect latency and potential clock-speed?

    Generally speaking:

    • To make effective use of significantly more cores, you're talking higher buffer sizes.
    • The more cores on the CPU, the harder it is to achieve high clock-speed
    • Super tight enclosures mean low clock-speed (thermal limitations)
    • Long Battery Life means low clock-speed 

     

    AMD has 32 cores on 3970x Threadripper, but the all-core clock-speed is about 4GHz.

    Intel 10900k has 10 cores, with all-core clock-speed of 5.3GHz.

    • At large buffer sizes, the 3970x will smoke the 10900k
    • At smallest buffer sizes, the 10900k will smoke the 3970x

     

    In my experience, Mac laptops currently provide longer battery-life than most PC laptops.

    When using a laptop, I'm almost always plugged-in.  🤷‍♂️

     


  16. Just a couple of points:

     

    "Long Battery Life" and "High Performance" are diametrically opposed.  You can have one or the other... not both.

     

    General-purpose use (Office Apps, Surfing the Internet, etc) doesn't require much in the way of CPU.

    ARM CPUs are already fine doing these tasks (iPads).

     

    We work with a lot of Mac users who were essentially abandoned (as power-users) by Apple.

    Most often they're professional composers... who need specific components in a high-performance machine; a machine that can be expanded/upgraded as need demands.

    The new Mac Pro???  7k for the base model.  Specs our composer clients need would be ~$10,000... and still not offer the same speed/performance and custom/expansion/upgrade options.

     

    • Like 1
×
×
  • Create New...