The No.1 Website for Pro Audio
 Search This Thread  Search This Forum  Search Reviews  Search Gear Database  Search Gear for sale  Search Gearslutz Go Advanced
Tim Cook on Mac Desktop commitment Virtual Instrument Plugins
Old 11th November 2018
  #2101
Gear Guru
 
UnderTow's Avatar
Quote:
Originally Posted by TNM View Post
True, but with the FLS slot issue, there is also no real point using windows with a powerful CPU anyway, as there's no way to actually make use of all that power from a single DAW, which is how many of us like to work. 128 FLS slots get filled in no time with ore than 50% cpu power to spare, even in a single 8 core machine. This is the single reason that stopped me going windows, as well as forced updates that break things, which even staunchest windows advocates will admit. The problem with HP is windows.
I don't think I've ever reached the slot limit on Windows but can easily reach the CPU limit. Not quite as easily on my new 14 Core system but I can still reach it. And I've never had a Windows update break anything. Admittedly I've only recently switched to Win 10.

Alistair
Old 11th November 2018
  #2102
Lives for gear
 
Lady Gaia's Avatar
Quote:
Originally Posted by TNM View Post
if everything works the same, that's fine. If all my audio devs have to port their code, then it may be years before everything is as it is now, and some will just flatly refuse to port yet again.
The "port" effort is likely to consist of recompiling their existing code and testing it, unless they have written hand-optimized assembly. It should be relatively trivial given that we're not talking about a new operating system or any significant structural changes – just a new instruction set for the new CPUs.

The downside is for users who rely on software that isn't being actively developed or maintained. Drivers for older hardware, plug-ins from companies that no longer exist or that have moved on to newer efforts, that kind of thing. Apple has provided emulation in the past to ease transitions but it always comes at some performance cost and it may very well not be practical for drivers.

Of course all of this creates opportunities for active developers, as it's an incentive to make sure there are great options out there. As with everything, there are tradeoffs but I don't see many, if any, active developers deciding they don't want to bother.
Old 11th November 2018
  #2103
Lives for gear
 
ponzi's Avatar
From what I can see, this fls limit business is a tempest in a teapot. A couple of brands of plug ins have been identified that find this a limitation, and Pete says microsoft is looking into it anyway. At the very least, a workaround is to get plugs that have been coded in such a way that this buffer, or whatever it is, does not get all used up. The problem has not been widespread enough for people to systematically identify the exact contours of it.
Old 12th November 2018
  #2104
Lives for gear
 
JSt0rm's Avatar
yeah i just put the business to this computer. Put on a few plug ins known to me to be how should we say hogs. 256 pro q 2, 256 sound toys crystallazer, like 30 one soft synths. Those plug ins I mentioned. A bunch of rx 6 advanced stuff as inserts. Then I printed 1khz tone on 256 tracks and looped that for 2 hours. Nothing popped up although someone mentioned that another instance of the same plug in wont create more issues?

I have a monster conform session I want to open. Im decrypting a drive right now and I want to open that one as it is a sluggish beast.
Old 12th November 2018
  #2105
Lives for gear
Can we like, discuss Windows issues one of the Windows threads, please? Valid PC hardware comparisons and the like are fine I think, but the state of another OS —which is a fluid and elusive entity — than what is commonly running on Macs really warrants its own thread.
Old 12th November 2018
  #2106
Lives for gear
 
JSt0rm's Avatar
Quote:
Originally Posted by Mikael B View Post
Can we like, discuss Windows issues one of the Windows threads, please? Valid PC hardware comparisons and the like are fine I think, but the state of another OS —which is a fluid and elusive entity — than what is commonly running on Macs really warrants its own thread.
respectfully, apple is late.
Old 12th November 2018
  #2107
Tui
Gear Guru
 
Tui's Avatar
Quote:
Originally Posted by JSt0rm View Post
respectfully, apple is late.


Old 12th November 2018
  #2108
TNM
Lives for gear
Quote:
Originally Posted by UnderTow View Post
I don't think I've ever reached the slot limit on Windows but can easily reach the CPU limit. Not quite as easily on my new 14 Core system but I can still reach it. And I've never had a Windows update break anything. Admittedly I've only recently switched to Win 10.

Alistair
You are lucky.. I hit it with like 15 uad plugins and one arturia analog lab in Cubase.

I couldn't add ANY other "different" plugins.

Free CPU was 70%. On a quad core 2012 laptop LOL.

Have you tried the free FLS slot vst plugin counter to see how close you are getting? I'd be very curious.

For those using a lot of waves plugins, they won't have the issue as waves uses a set amount of slots for an unlimited amount of their different plugins.

Some plugins behave and only use 1 slot. Cubase leaves 90 slots free i think after loading.

But most plugins are using at least 2 slots..I think reaper leaves 100 slots.. So.. 50 unique plugins in reaper would max it out, IF they only use 2 slots, and are not waves..

And forget UAD.. 40 unique UAD plugins, with NO vi's, NO other native plugins, completely maxes out all FLS slots in any DAW and makes it impossible to add any other plugins...

So....

it just depends on your usage, the actual manufacturers you use, and even a bit on the daw and how many free slots it has after loading.

I wasn't going to invest in a powerful 12 or 14 core PC and not be able to use the power. The workaround is VE Pro but damn if i don't detest that thing (and it's only really useful for VI tracks anway).

Apparently j bridge can work around it too as J bridge uses one process per plugin (or can be made to).

The other thing could be is that you are very sparse with plugins.. there are videos of people hitting the limit in computers from 5 years ago, and hitting it fast.
Old 12th November 2018
  #2109
TNM
Lives for gear
Quote:
Originally Posted by JSt0rm View Post
Interesting. I've never heard of an fls error before.

Reading this thread...

End of Windows for Pro production - FLS Limit?

It seems its based on certain plug ins hogging that resource?

Can you point me to the forced updates issue? All I am seeing is you are able to disable updates and driver updates.

Im coming from a 12 core macpro with 64gb of ram running an hdx2 system with a aurora16 and a hd i/o. I was never starved for cpu power. My biggest ssue was the lack of modern i/o. Running 500 tracks plus prores video in super sessions feels slower. This is a test But Im welcome to knowing about any issues for I'm neck deep.

I havent installed the hdx cards yet but I will try to create the fls issue with native ultimate.
It's in the windows 10 topic here, about the force updates,. You can only DEFER them, even in win pro. Eventually they will all install and could literally break your entire daw when it happens.

There are no force updates at all on OSX at any time, even security updates can be disabled. Are you still planning to run pro res? If so, you'll need to keep your mac to export to prores, windows can not do it. You can IMport it though I believe.
Old 12th November 2018
  #2110
TNM
Lives for gear
Quote:
Originally Posted by JSt0rm View Post
yeah i just put the business to this computer. Put on a few plug ins known to me to be how should we say hogs. 256 pro q 2, 256 sound toys crystallazer, like 30 one soft synths. Those plug ins I mentioned. A bunch of rx 6 advanced stuff as inserts. Then I printed 1khz tone on 256 tracks and looped that for 2 hours. Nothing popped up although someone mentioned that another instance of the same plug in wont create more issues?

I have a monster conform session I want to open. Im decrypting a drive right now and I want to open that one as it is a sluggish beast.
256 of one plugin is fine, 256 ProQ2 will only use 2 fls slots as fab filter plugins do, for example. More slots are only used by different vst id. (or aax).
Old 12th November 2018
  #2111
Quote:
Originally Posted by Lady Gaia View Post
The "port" effort is likely to consist of recompiling their existing code and testing it, unless they have written hand-optimized assembly. It should be relatively trivial given that we're not talking about a new operating system or any significant structural changes – just a new instruction set for the new CPUs.

The downside is for users who rely on software that isn't being actively developed or maintained. Drivers for older hardware, plug-ins from companies that no longer exist or that have moved on to newer efforts, that kind of thing. Apple has provided emulation in the past to ease transitions but it always comes at some performance cost and it may very well not be practical for drivers.

Of course all of this creates opportunities for active developers, as it's an incentive to make sure there are great options out there. As with everything, there are tradeoffs but I don't see many, if any, active developers deciding they don't want to bother.
Well, DAWs and complex instruments like Kontakt (with 3rd party scripts) or Omnisphere will require a shizzload of QA time to validate everything is working smooth, especially with 3rd party products (plugins, controllers), especially when said 3rd party may not have finished their port yet. It is not as big as a major OS overhaul, yet full QA time needs to be calculated for, including fixes and optimizations, adding QA time again.
Old 12th November 2018
  #2112
People, this is not an FLS thread. Please discuss somewhere else, although it has already been done over and over...
Old 12th November 2018
  #2113
Gear Guru
 
UnderTow's Avatar
Quote:
Originally Posted by TNM View Post
Have you tried the free FLS slot vst plugin counter to see how close you are getting? I'd be very curious.
I just installed it and checked on a pretty busy project. 293 tracks, 241 FX plugins (34 unique plugins), 14 Instrument plugins (4 unique plugins). FLS Checker reports 40 free slots (out of 85 free slots in an empty project in Cubase).

Alistair
Old 12th November 2018
  #2114
Gear Guru
 
UnderTow's Avatar
Quote:
Originally Posted by Lady Gaia View Post
The "port" effort is likely to consist of recompiling their existing code and testing it, unless they have written hand-optimized assembly.
Which is exactly what is used in many plugins. Maybe even most. (Although I have no idea how to verify that). And does/will the Apple chip support the SSE and AVX instruction sets?

I doubt companies like Waves or Native Instruments will be too keen on rewriting and testing their numerous plugins because Apple are yet again breaking compatibility.

Quote:
Of course all of this creates opportunities for active developers, as it's an incentive to make sure there are great options out there.
That is one way of looking at it. The other way is that everyone (but Apple) loses out because developers have to waste time on adapting and testing their plugins because Apple decided to do things differently again (for their own benefit, not for the customer's).

Quote:
As with everything, there are tradeoffs but I don't see many, if any, active developers deciding they don't want to bother.
Really? It all costs time and money. Time and money not spent on new products and real improvements that benefit the users as it is just old products being updated.

Alistair
Old 12th November 2018
  #2115
Lives for gear
 
Lady Gaia's Avatar
Quote:
Originally Posted by UnderTow View Post
Which is exactly what is used in many plugins. Maybe even most. (Although I have no idea how to verify that).
With no way to verify your claim, you’re making a pretty sizable assumption phrased as an assertion. I’d be curious to know what the actual story is if any developers care to speak up.

Quote:
And does/will the Apple chip support the SSE and AVX instruction sets?
SSE and AVX are brand names associated with x86 instruction set extensions. No other instruction set is going to use these names, but ARM v8 does indeed have vector extensions of its own, branded as NEON.

Quote:
That is one way of looking at it. The other way is that everyone (but Apple) loses out because developers have to waste time on adapting and testing their plugins because Apple decided to do things differently again (for their own benefit, not for the customer's).
Apple hasn’t clearly indicated that they will, so the only reason anyone is speculating about it is because there are obvious benefits for users. Thermal constraints have been a limiting factor for desktops and laptops alike, so switching to an architecture that can potentially comparable performance at greater power efficiency is, frankly, pretty exciting. Longer battery life. Less fan noise. How would these things only benefit Apple?

It is, of course, all speculative since there isn’t any actual announcement on the table.
Old 12th November 2018
  #2116
Lives for gear
 
ponzi's Avatar
Quote:
Originally Posted by Lady Gaia View Post
...Thermal constraints have been a limiting factor for desktops and laptops alike, so switching to an architecture that can potentially comparable performance at greater power efficiency is, frankly, pretty exciting...
As syllogisms go, I can't disagree--if more efficient, then less heat. Curious, how does the new architecture reduce heat? Going from cisc to risc--less transistors? Some other means?

I recall the cisc/risc 'debate' has been going on for some decades, and one of the things about risc, is that for the instructions it does not implement, the functionality has to be implemented in software with attendant slowness, so some of the advantages of risc get mitigated a bit.

Not challenging what you are saying, just seeking insight. Thanks.
Old 12th November 2018
  #2117
Lives for gear
 
Lady Gaia's Avatar
Quote:
Originally Posted by ponzi View Post
As syllogisms go, I can't disagree--if more efficient, then less heat. Curious, how does the new architecture reduce heat? Going from cisc to risc--less transistors? Some other means?
Many of the advances in efficiency are a matter of coming up with designs that keep less of the CPU die powered up and active rather than squeezing more efficiency out of the parts of the die that are actually doing work. This is likely much more about implementation decisions than anything inherent in the instruction set.

After all, other ARM v8 implementations aren't really all that close from a performance or efficiency standpoint.

Quote:
I recall the cisc/risc 'debate' has been going on for some decades, and one of the things about risc, is that for the instructions it does not implement, the functionality has to be implemented in software with attendant slowness, so some of the advantages of risc get mitigated a bit.
Complex decoding logic for a complicated instruction set has its own issues that can result in deeper instruction pipelines and therefore greater penalties when the pipeline stalls, as due to a branch misprediction. Most engineering isn't a matter of right or wrong answers but subtle tradeoffs. CPU design improvements these days come as a result of deep analysis of complex workloads, often hand-in-hand with the compiler responsible for choosing what instructions to emit in the first place.
Old 12th November 2018
  #2118
Lives for gear
 
JSt0rm's Avatar
Quote:
Originally Posted by TNM View Post
It's in the windows 10 topic here, about the force updates,. You can only DEFER them, even in win pro. Eventually they will all install and could literally break your entire daw when it happens.

There are no force updates at all on OSX at any time, even security updates can be disabled. Are you still planning to run pro res? If so, you'll need to keep your mac to export to prores, windows can not do it. You can IMport it though I believe.
I wont be exporting any video. I think I have all the codecs I need on the system.

A couple pain points coming from osx so far (to keep it on topic) No sound miner v4.5 must migrate to sound miner hd. I havent loaded that yet as I'm waiting on that decrypt.

spanner is osx only as well but I have a work around for that.

Other then that some gui issues with windows that osx doesnt have. Wasted space mostly. Little things like menus highlighting while doing keyboard shortcuts in osx doesnt happen in windows.

As I said this is a test.
Old 12th November 2018
  #2119
Lives for gear
 
ponzi's Avatar
Quote:
Originally Posted by Lady Gaia View Post
...Complex decoding logic for a complicated instruction set has its own issues that can result in deeper instruction pipelines and therefore greater penalties when the pipeline stalls, as due to a branch misprediction...
Thanks for the insight. I recall some years ago, intel was using a long pipeline to get its clock frequency numbers up, but as you mentioned, the long pipeline and misprediction 'bubbles' caused a performance penalty--so they sort of stopped marketing on clock frequency and shortened the pipeline on the next generation.

Aside from companies having patents on techniques, it looks to me like everyone has the same tools and makes choices on design trade offs.
Old 12th November 2018
  #2120
Lives for gear
 
Lady Gaia's Avatar
Quote:
Originally Posted by ponzi View Post
Aside from companies having patents on techniques, it looks to me like everyone has the same tools and makes choices on design trade offs.
Compatibility with choices made in the past is one of the biggest anchors in technology, because the factors that made something a good decision in the past will have changed and may or may not turn out to be a good idea going forward. Intel got very lucky in one of those transition periods. Their instruction set turned out to be pretty efficient when memory bandwidth constraints became the biggest bottleneck, but now that heat is looking like the primary factor in designs they may not be quite so fortunate.

Patents always play a role. ARM was very focused on efficiency from the outset, and it’s paying dividends now with patents covering much of that space. Apple gets the benefit of being an ARM licensee while arguably having more to gain from investing in taking full advantage than anyone else in the field outside the Intel / AMD. It’s always fascinating to witness shifting fortunes.
Old 12th November 2018
  #2121
Lives for gear
 
JSt0rm's Avatar
Quote:
Originally Posted by Lady Gaia View Post
Compatibility with choices made in the past is one of the biggest anchors in technology, because the factors that made something a good decision in the past will have changed and may or may not turn out to be a good idea going forward. Intel got very lucky in one of those transition periods. Their instruction set turned out to be pretty efficient when memory bandwidth constraints became the biggest bottleneck, but now that heat is looking like the primary factor in designs they may not be quite so fortunate.

Patents always play a role. ARM was very focused on efficiency from the outset, and it’s paying dividends now with patents covering much of that space. Apple gets the benefit of being an ARM licensee while arguably having more to gain from investing in taking full advantage than anyone else in the field outside the Intel / AMD. It’s always fascinating to witness shifting fortunes.
Im dragging along 15k is software so I would like some backwards compatibility please
Old 12th November 2018
  #2122
Lives for gear
 
Lady Gaia's Avatar
Quote:
Originally Posted by JSt0rm View Post
Im dragging along 15k is software so I would like some backwards compatibility please
Completely understandable, and if Apple follows past precedent they’ll provide a mechanism for running existing software (Mac OS Classic apps ran alongside OS X apps for several years, as did PowerPC apps on x86 hardware.) Many vendors produced “universal” binaries that ran on PowerPC and x86 systems without additional cost. Presumably we’d see something similar again at the system level rather than trying to provide support in silicon (witness the latest iOS devices not being able to run 32-bit code at all, following several years where Apple aggressive,y pushed vendors to support 64-bit execution in app updates.)

The question is usually how fast to push to the new technology and what trade-offs are considered acceptable. Apple has been more aggressive than others historically, which may or may not be a good indicator of what the future holds.
Old 13th November 2018
  #2123
Gear Guru
 
UnderTow's Avatar
Quote:
Originally Posted by Lady Gaia View Post
With no way to verify your claim, you’re making a pretty sizable assumption phrased as an assertion. I’d be curious to know what the actual story is if any developers care to speak up.
All plugin code I have seen from experienced developers uses inline assembly or intrinsic functions to optimize things. Even (the more complex) stuff done with approachable tools like SynthEdit uses it. It is kind of standard AFAIAA.

The question is how much of the market uses it. Unless you have any evidence to show something else, I'm going to assume it is the majority of it.

Quote:
SSE and AVX are brand names associated with x86 instruction set extensions. No other instruction set is going to use these names, but ARM v8 does indeed have vector extensions of its own, branded as NEON.
If the instructions sets are not the same, the plugins would need to be recoded for a different instruction set or suffer a huge efficiency penalty. Some of the developer kits and frameworks will of course adopt new instruction sets but that takes time and will probably only happen once it becomes worth doing it due to market share. A typical chicken and egg situation. Avid found out the hard way with AAX-DSP how long it takes developers to adapt to a new format... if they adopt it at all. (For instance Waves said no thank you).

And even if frameworks like JUCE do adopt new instructions sets, this doesn't cover the stuff in the inner loop that really taxes CPUs. Developers will still need to hand code that for efficiency which means recoding old plugins, not just recompiling.

Quote:
Apple hasn’t clearly indicated that they will, so the only reason anyone is speculating about it is because there are obvious benefits for users. Thermal constraints have been a limiting factor for desktops and laptops alike, so switching to an architecture that can potentially comparable performance at greater power efficiency is, frankly, pretty exciting. Longer battery life. Less fan noise. How would these things only benefit Apple?
You have already answered this yourself with these comments:

Quote:
Originally Posted by Lady Gaia View Post
Many of the advances in efficiency are a matter of coming up with designs that keep less of the CPU die powered up and active rather than squeezing more efficiency out of the parts of the die that are actually doing work. This is likely much more about implementation decisions than anything inherent in the instruction set.

After all, other ARM v8 implementations aren't really all that close from a performance or efficiency standpoint.
The way to stress test CPU's is to run things like Intel Burntest which stress the floating-point and AVX portions of the chip. Those are the portions that use most power and produce most heat. Exactly the stuff we need for efficient audio plugins. I don't see why an ARM chip would be more efficient for this kind of stuff. For web browsing, office apps etc, sure, but that is not what interests us for audio.

I would say the main driving factor for Apple is control and increased margins. Not helping their customers or developers although I am sure Apple will come up with some spiel to that effect that many will believe...

Quote:
Complex decoding logic for a complicated instruction set has its own issues that can result in deeper instruction pipelines and therefore greater penalties when the pipeline stalls, as due to a branch misprediction. Most engineering isn't a matter of right or wrong answers but subtle tradeoffs. CPU design improvements these days come as a result of deep analysis of complex workloads, often hand-in-hand with the compiler responsible for choosing what instructions to emit in the first place.
OS-X has always been notoriously bad with short pipelines. IBM wrote a white paper about the subject and we still see this in benchmarks that test applications that require near realtime performance (including audio) to this day. But we will see what Apple come up with if they do indeed decide to go down this route.

Alistair
Old 13th November 2018
  #2124
Lives for gear
 
stratology's Avatar
 

Arstechnica released an article that goes into detail about the new A12X chip, compares its performance to laptops and desktops, and offers lots of interesting technical information.
Old 13th November 2018
  #2125
Lives for gear
 
Lady Gaia's Avatar
Quote:
Originally Posted by UnderTow View Post
All plugin code I have seen from experienced developers uses inline assembly or intrinsic functions to optimize things.
Depending on how extensive the optimizations are, this can be a huge barrier or relatively minor. In any case, Apple has been encouraging developers to use the likes of vecLib or Accelerate for many years rather than CPU-specific intrinsics precisely because they're efficient inline snippets of hand-optimized code that allow the compiler to produce efficient SIMD-optimized code for any supported CPU.

If this isn't common practice then we're going to see a slower transition. Again, I suspect the low end of the product line will be where we'd see the shift happen first. It would put affordable hardware in developer hands and not particularly inconvenience those just looking for typical productivity tools.

Of course it's also possible we'll see something remarkable in binary translation. I haven't kept up with the state of the art on that front, but I doubt it would be anything more than a stopgap measure.
Old 13th November 2018
  #2126
Gear Maniac
I'd just point out that some companies already make iOS versions of their existing plugins (Korg for example). Other companies make iOS-only versions (Moog Model 15 anyone?). Clearly this is all doable. I also don't see every Mac developer just throwing in the towel if Apple does a switch. The market is too large. At any rate I'd be surprised if the desktop Macs transition over anytime in the very near future. The mobile lineup makes the most sense. Plenty of time for an orderly transition if this all happens.
Old 4 weeks ago
  #2127
Gear Maniac
Resurrecting this thread just to point out that Apple's timeline, if the various news organizations are to be believed, has the Mac transition to ARM starting next year...way sooner than I thought would happen for sure.
Old 4 weeks ago
  #2128
Lives for gear
 
didlisquat's Avatar
Quote:
Originally Posted by rezoneight View Post
Resurrecting this thread just to point out that Apple's timeline, if the various news organizations are to be believed, has the Mac transition to ARM starting next year...way sooner than I thought would happen for sure.
All that microdosing must be paying off.
Old 4 weeks ago
  #2129
Lives for gear
Apple has left the building.
Old 4 weeks ago
  #2130
Lives for gear
 
Lady Gaia's Avatar
Quote:
Originally Posted by rezoneight View Post
Resurrecting this thread just to point out that Apple's timeline, if the various news organizations are to be believed, has the Mac transition to ARM starting next year...way sooner than I thought would happen for sure.
It sounds reasonable enough, and I will be very curious how they approach it. Starting with laptops seems like the most obvious move, and I’d expect it to begin with the low end but I could make a case for something more aggressive in the Pro space that can push performance barriers due to thermal advantages. I suspect software migration will be the primary driver to stick with entry-level devices first, though.

This year’s expected announcement of a new Mac Pro could help telegraph the strategy, though. It would be odd to introduce a high-end Intel design in 2019 and then abandon Intel for other Pro gear the following year.

The other related rumors recently suggest a 16+” laptop in the works and a return to Apple branded displays with a new 6K model.
Topic:
Post Reply

Welcome to the Gearslutz Pro Audio Community!

Registration benefits include:
  • The ability to reply to and create new discussions
  • Access to members-only giveaways & competitions
  • Interact with VIP industry experts in our guest Q&As
  • Access to members-only sub forum discussions
  • Access to members-only Chat Room
  • Get INSTANT ACCESS to the world's best private pro audio Classifieds for only USD $20/year
  • Promote your eBay auctions and Reverb.com listings for free
  • Remove this message!
You need an account to post a reply. Create a username and password below and an account will be created and your post entered.


 
 
Slide to join now Processing…
Thread Tools
Search this Thread
Search this Thread:

Advanced Search
Forum Jump
Forum Jump