The No.1 Website for Pro Audio
Apple to announce ARM chips for all Macs
Old 7th August 2020
  #1171
Lives for gear
 
juiseman's Avatar
Quote:
Originally Posted by juiseman View Post
For me, I'm just trying to learn peoples workflows and view points. I think we can all be more transparent at times. But Being a Windows and Mac user I'm not biased. Then again, you can't change the facts that you can almost always get more performance for cheaper on a PC. Plus, the fun experience of building your own rig; or building a custom computer that fits your workflow, with your needs can't be always done with an Apple Computer.

Overall, going back to the OP, after reading a little more about the switch to ARM, I see exactly why they did this. I still will take the stance that it will be a longer transition than some are expecting. Also, we don't know how developers will support the new changes.

So, its good discussion in my view..

lol.. pc vs mac was not my goal here; but I guess I did it anyhow..
I like Mac Os & the customization on the PC..I guess you can't have it both ways.

I did state:
"Being a Windows and Mac user I'm not biased"
&
"almost always get more performance for cheaper on a PC"
Old 7th August 2020
  #1172
Quote:
Originally Posted by bace View Post
It is a mesh-topology CPU. It's not a SMP.
It's just a different approach with its own pros and cons. Mesh happens to work better with unrelated workloads (like VMs or containers) out of the box, but it's entirely possible to code for this, should it become popular in workstations.

It's one way to get very high core counts without taking the kinds of CPU latency hits you get with SMP.

Intel has also been doing this with the Xeon Scalable processors.

Personally, I'd love to see some more of the high-throughput server optimizations brought down to the workstation level. Ampere's previous offering, the eMAG, is offered in workstations, but it's not quite the same type of processor. Also, most server optimizations don't benefit client workloads. Many do, though. There was a time when more than one CPU [core] was considered something only servers would need.

Anyway, mesh is even something Apple may consider in their high-end offerings, assuming the core counts are up in the double digits.

Anyway. No idea if any of the players will adapt this kind of tech for the workstation, but the thought of it is interesting at the least.

Pete
Old 7th August 2020
  #1173
Lives for gear
 
crufty's Avatar
Quote:
Originally Posted by uOpt View Post
This goes into the whole speculative computing area. But speculative computing increases total CPU usage (and power used) dramatically.
Indeed—but that js the advantage.Why not render the entire track in memory even when not playing back. With 128 cores of compute and 1.5TB of memory, why not precompute the entire session in memory? People are limited to 1-2 adjustments at a time...maybe ten with a fader pack. The future could easily be one that doesnt require realtime plugins except for whatever happened to be tracked / played / tweaked.
Old 7th August 2020
  #1174
Lives for gear
 
uOpt's Avatar
Quote:
Originally Posted by crufty View Post
Indeed—but that js the advantage.Why not render the entire track in memory even when not playing back. With 128 cores of compute and 1.5TB of memory, why not precompute the entire session in memory? People are limited to 1-2 adjustments at a time...maybe ten with a fader pack. The future could easily be one that doesnt require realtime plugins except for whatever happened to be tracked / played / tweaked.
Render all possible outcomes of a song and the musician's actions just remove the stuff not picked.

Last edited by uOpt; 7th August 2020 at 11:18 PM..
Old 7th August 2020
  #1175
Quote:
Originally Posted by crufty View Post
Indeed—but that js the advantage.Why not render the entire track in memory even when not playing back. With 128 cores of compute and 1.5TB of memory, why not precompute the entire session in memory? People are limited to 1-2 adjustments at a time...maybe ten with a fader pack. The future could easily be one that doesnt require realtime plugins except for whatever happened to be tracked / played / tweaked.
This is similar to how some of the remote game rendering tech works. The rendering is just done up in the cloud, where lots of compute units can be spun up or down quickly. No reason you couldn't get there on the client side at some point.

Pete
Old 8th August 2020
  #1176
TNM
Lives for gear
Quote:
Originally Posted by detritusdave View Post
This is utterly anecdotal, but I was bored and just screenshotted Audio Performance in Cubase (same file - one in Windows(Bootcamp)/one in MacOS). So, exactly the same hardware/plugins,same buffer size etc... Not a lot of difference between the two....

While this is not in any way scientific it shows what I've experienced in real life.... variance between the two is not worthy of note outside of pointlessly partisan b*llocks....

I will not be conducting any more scientifically rigorous examinations.... cos I just can't be arsed....
That means nothing cause I can keep my windows laptop with nuendo showing absolute 100% asio usage on that meter (which has nothing to do with cpu usage) and still have no pops or clicks with asio guard disabled at 128 buffer.

I can not do that on any mac I have ever owned including my imac pro. It will start having dropouts and gaps.

A video is much more useful than a screenshot of a FLUCTUATING asio performance bar, no offence.

Anyway I have done all this to death (videos included) for multiple DAWs and haven't done it for a year or so but I have proven my point over and over, so I feel I have nothing to prove anymore.

And also, what are the asio guard settings in windows and OSX? If it's 1024 samples it's going to make no difference. Turn Asio guard off and then talk to me or put at lowest possible setting. Or leave asio guard where it is and arm some VI's at 64 buffer or 32 audio tracks at 64 buffer with plugins.

Have you tuned that windows partition on the mac to make sure there are NO dpc latency spikes? (each dpc spike = dropout).
The imac pro has incredibly low dpc latency in windows which does make a huge difference. If that's a macbook or something, then it's almost impossible to get low dpc latency in windows with it. You need to get a windows notebook with similar specs to compare.

I'd be curious what the specs of that system are so I can compare with my windows laptop if they are at all close. If the project is possible to be shared i'd be interested in doing a benchmark for a few minutes but I don't want to go *too* deep down that rabbit hole again as I have literally spent THOUSANDS of real hours doing daw vs daw performance testing and mac vs pc performance testing.

I mean it's just simple Logic.

Right now imac pro's and mac pros can not stay at their factory spec all core turbo under DAW load, even though they are running cool as a cucumber at the time. In windows, those same machines are sitting at their max all core turbo without fluctuations.

How can my imac pro which is constantly hopping between 3.2 and 4ghz compete with windows on the same imac pro which never, ever even goes a megahertz under 4ghz?

That stability on the core frequency makes such a massive difference it can not be overstated.
Old 8th August 2020
  #1177
Lives for gear
Quote:
Originally Posted by crufty View Post
Indeed—but that js the advantage.Why not render the entire track in memory even when not playing back. With 128 cores of compute and 1.5TB of memory, why not precompute the entire session in memory? People are limited to 1-2 adjustments at a time...maybe ten with a fader pack. The future could easily be one that doesnt require realtime plugins except for whatever happened to be tracked / played / tweaked.
I'm sure it's just marketing, but MOTU literally calls whatever buffering/RAM usage they use on unarmed tracks Next Gen PreGen. Mentioning Prerendering the tracks etc.

So how far off is any of that from what you're saying? One thing I thought right away is why are we not able to then just select a key command to instantly convert VI's etc. to audio tracks? It's fast enough as it is for most uses but people with huge templates etc. It would be pretty cool if instead of saving a bit of CPU by "offline" rendering or whatever other tricks various DAWs are using, a DAW would just render tracks to RAM until it was tweaked, and instantly converted 30+ tracks to audio with a command.

Less heat, noise and busy work, hopefully it's what we see from all this.
Old 8th August 2020
  #1178
Lives for gear
Quote:
Originally Posted by uOpt View Post
Render all possible outcomes of a song and the musician's actions just remove the stuff not picked.
That is what a quantum daw computer will do.
Old 9th August 2020
  #1179
Gear Maniac
 

When a $399 iPhone SE2 can out perform flagship Android phones costing thousands of dollars more, I can't wait to see what capabilities Apple Silicon brings to laptops and desktops.

I am still using my late 2012 iMac with TB2 and UA Apollo interfaces which have been rock solid for the past years, I am glad to skip over the intel/T2/BridgeOS hybrid mess and go straight to Apple Silicon next year.

If you start with the marketing and work back the new macs will be twice as fast as the previous intel version, that's why I think the MBP 13" will be the first to be updated, faster, quieter, thinner, last longer and more secure, and hopefully only a little bit more expensive than current intel models.
Old 9th August 2020
  #1180
Lives for gear
Quote:
Originally Posted by scottkrk View Post
When a $399 iPhone SE2 can out perform flagship Android phones costing thousands of dollars more, I can't wait to see what capabilities Apple Silicon brings to laptops and desktops.

I am still using my late 2012 iMac with TB2 and UA Apollo interfaces which have been rock solid for the past years, I am glad to skip over the intel/T2/BridgeOS hybrid mess and go straight to Apple Silicon next year.

If you start with the marketing and work back the new macs will be twice as fast as the previous intel version, that's why I think the MBP 13" will be the first to be updated, faster, quieter, thinner, last longer and more secure, and hopefully only a little bit more expensive than current intel models.
Do you pay thousands of dollars for android phones down under? Is there a open-source tax or something? In Europe android are the cheep ones.
Old 9th August 2020
  #1181
Gear Maniac
 

Quote:
Originally Posted by bace View Post
Do you pay thousands of dollars for android phones down under? Is there a open-source tax or something? In Europe android are the cheep ones.
Ha ha, yes we have 'cheep' android phones down under, so let me put it another way, the cheapest iPhone has a faster processor that the most expensive Android phone.

Intel have been milking the x86 cash cow for years, AMD have become more competitive in x86 space but the future looks decidedly ARM, from AWS in the server space to Chromebooks, MS Surface and now Apple in the laptop/desktop space.
Old 9th August 2020
  #1182
Lives for gear
 
charlieclouser's Avatar
 

Quote:
Originally Posted by crufty View Post
Indeed—but that js the advantage.Why not render the entire track in memory even when not playing back. With 128 cores of compute and 1.5TB of memory, why not precompute the entire session in memory? People are limited to 1-2 adjustments at a time...maybe ten with a fader pack. The future could easily be one that doesnt require realtime plugins except for whatever happened to be tracked / played / tweaked.
This is exactly how early Propellerheads software like ReBirth worked - it had very few realtime controls and no ability to map onscreen knobs to incoming MIDI CC for realtime knobbery, so it was able to perform amazing feats on a G3 Mac in 1998.

It's also basically what Logic's "process buffer" is doing - pre-rendering any elements that are not subject to external realtime control. At one point around 2001 or so, when Logic's soft synth performance was so far beyond what any other DAW could do, I was told that the secret sauce was the pre-rendering of anything and everything that was outside the stream of realtime control. Now it's commonplace, but at the time it was revolutionary. And it still works just fine by me. Why overheat two dozen cores to calculate 900 tracks at 32 buffer, when only a few tracks / sends / busses actually need to be calculated that way? Pre-render the rest and play 'em back from RAM while the realtime tracks, all their sends, and busses they feed go onto the live cores.

With Apple silicon they'll have total control over core scheduling or whatever, so I wouldn't be surprised if we see performance gains in Logic that are absolutely bonkers and just don't "make sense" or translate to the way Intel silicon does things.

I, for one, welcome our new robot overlords.
Old 9th August 2020
  #1183
Lives for gear
Quote:
Originally Posted by charlieclouser View Post
This is exactly how early Propellerheads software like ReBirth worked - it had very few realtime controls and no ability to map onscreen knobs to incoming MIDI CC for realtime knobbery, so it was able to perform amazing feats on a G3 Mac in 1998.

It's also basically what Logic's "process buffer" is doing - pre-rendering any elements that are not subject to external realtime control. At one point around 2001 or so, when Logic's soft synth performance was so far beyond what any other DAW could do, I was told that the secret sauce was the pre-rendering of anything and everything that was outside the stream of realtime control. Now it's commonplace, but at the time it was revolutionary. And it still works just fine by me. Why overheat two dozen cores to calculate 900 tracks at 32 buffer, when only a few tracks / sends / busses actually need to be calculated that way? Pre-render the rest and play 'em back from RAM while the realtime tracks, all their sends, and busses they feed go onto the live cores.

With Apple silicon they'll have total control over core scheduling or whatever, so I wouldn't be surprised if we see performance gains in Logic that are absolutely bonkers and just don't "make sense" or translate to the way Intel silicon does things.

I, for one, welcome our new robot overlords.
I think most DAW do that today, But it is quite complicated. Say that you would like to side chain your drumbus with lead guitar. Then you need to figure out what part of the drumbus that also need to be realtime when the guitar plays. And the guitar also share one of the reverb buses with all other tracks. So now you have many channels that needs to be at least partly in realtime.
Old 9th August 2020
  #1184
Lives for gear
 
charlieclouser's Avatar
 

Quote:
Originally Posted by bace View Post
I think most DAW do that today, But it is quite complicated. Say that you would like to side chain your drumbus with lead guitar. Then you need to figure out what part of the drumbus that also need to be realtime when the guitar plays. And the guitar also share one of the reverb buses with all other tracks. So now you have many channels that needs to be at least partly in realtime.
Yes, that's exactly how it was explained to me by the Logic devs many years ago. They weren't suggesting that it was a total cure, but that it could provide meaningful improvements depending on the complexity of the session and routing - and sure enough, it has.

My normal scoring template uses seven stem sub masters that sum into a final composite mix. Each stem has its own reverbs, delays, and mastering effects on the sub masters. So a lot of the time, six of those seven stems lie entirely "upstream" (or maybe "beside-the-stream"?) of the portion of the path that needs to be "live". Only the stem containing my current live track, as well as the composite mix sub masters, needs to be calculated at the actual buffer setting. So in my case I see ridiculously great system performance even on massive projects with hundreds of tracks, since a great chunk of the stuff is "over to the side" and can be calculated ahead of time.

That's why I won't be surprised if Apple silicon lets them get down to the molecular level with core scheduling and "optimized for real time cores" or whatever, and coax even better system performance out of what might, at first glance, appear to be an unimpressive clock speed or core count on their own chips.
Old 9th August 2020
  #1185
Lives for gear
 
crufty's Avatar
Quote:
Originally Posted by machinesworking View Post
Mentioning Prerendering the tracks etc.
how far off is any of that from what you're saying? One thing I thought right away is why are we not able to then just select a key command to instantly convert VI's etc. to audio tracks?
it's one and the same! Only--why even select a key command? The computer could know when it has spare cycles and do it automatically!

The primary reason we don't see this today, I'm guessing, is that 4 core computers can't keep up.

Quote:
Originally Posted by charlieclouser View Post
Why overheat two dozen cores to calculate 900 tracks at 32 buffer, when only a few tracks / sends / busses actually need to be calculated that way? Pre-render the rest and play 'em back from RAM while the realtime tracks, all their sends, and busses they feed go onto the live cores.
exacto-mundo, as the Fonz would say!

The beauty is, it's not really ARM, Intel or AMD specific. The standard DAW mode of operation caters very nicely to laptops and tablets, which are the starving children of the computer kingdoms. however, when a computer has more RAM then disk storage, and can easily peg 120+ CPU at 100% utilization w/out interrupting the creative process, it could really open the door to optimizing the expensive things in audio processing.
Old 9th August 2020
  #1186
Lives for gear
 
crufty's Avatar
Quote:
Originally Posted by bace View Post
I think most DAW do that today, But it is quite complicated. Say that you would like to side chain your drumbus with lead guitar. Then you need to figure out what part of the drumbus that also need to be realtime when the guitar plays. And the guitar also share one of the reverb buses with all other tracks. So now you have many channels that needs to be at least partly in realtime.

yes, this much is true. the key behind speculative compute is that any cache miss requires real-time handling at that instant. Complexity for sure.

There is a genealogy of audio rendering, and as soon as one sample gets the axe, all bets are off for that sample, it's relatives, for that moment thru the future.. So no doubt some trade offs...maybe oversampling gets put on an audio render backlog.

But--that would also be the advantage. Suppose one is recording a full band, live. For armed tracks but receiving noisefloor audio, compute could be spent casting ahead fx chains, oversampling, or whatever, lessoning the potential future load. Maybe some sticks, maybe some does not. What does it matter?

The key is CPU time is not money--we can't save it for a rainy day. It's more like a waterfall--as water passes us by, if we aren't using it then and there, then it's potential utility is lost to us as it goes hurdling over the edge.
Old 9th August 2020
  #1187
Lives for gear
Quote:
Originally Posted by bace View Post
I think most DAW do that today, But it is quite complicated. Say that you would like to side chain your drumbus with lead guitar. Then you need to figure out what part of the drumbus that also need to be realtime when the guitar plays. And the guitar also share one of the reverb buses with all other tracks. So now you have many channels that needs to be at least partly in realtime.
This is why I wish all DAWs had Digital Performers Effect Performance Meter. You get a live readout of which tracks are Real Time, and which are Pre Generated. It's much more accurate than just a CPU performance meter. Reaper has a similar thing, but it's horribly optimistic, the exact opposite of DP's.
Old 10th August 2020
  #1188
Lives for gear
Quote:
Originally Posted by machinesworking View Post
This is why I wish all DAWs had Digital Performers Effect Performance Meter. You get a live readout of which tracks are Real Time, and which are Pre Generated. It's much more accurate than just a CPU performance meter. Reaper has a similar thing, but it's horribly optimistic, the exact opposite of DP's.
That seems to be good. I would have done some colour scheme for the mixer part that is real-time vs preprocessed. I think there is a lot of things you could do, however Im not sure that is good to do it. It should be transparent and would easy confuse non software engineers.
Old 10th August 2020
  #1189
Lives for gear
 
charlieclouser's Avatar
 

Quote:
Originally Posted by machinesworking View Post
This is why I wish all DAWs had Digital Performers Effect Performance Meter. You get a live readout of which tracks are Real Time, and which are Pre Generated. It's much more accurate than just a CPU performance meter. Reaper has a similar thing, but it's horribly optimistic, the exact opposite of DP's.
That is awesome, I wish Logic had that. Actually, I wish Logic had a LOT of stuff from DP now that I think about it....
Old 10th August 2020
  #1190
Gear Maniac
https://www.macrumors.com/roundup/im...%204%2C%202020.


And they did it again, 16GB RAM for $200 wtf ? But atleast they dropped the Fusion Drive yay


Last edited by daslicht; 10th August 2020 at 11:04 AM..
Old 10th August 2020
  #1191
TNM
Lives for gear
Quote:
Originally Posted by charlieclouser View Post
This is exactly how early Propellerheads software like ReBirth worked - it had very few realtime controls and no ability to map onscreen knobs to incoming MIDI CC for realtime knobbery, so it was able to perform amazing feats on a G3 Mac in 1998.

It's also basically what Logic's "process buffer" is doing - pre-rendering any elements that are not subject to external realtime control. At one point around 2001 or so, when Logic's soft synth performance was so far beyond what any other DAW could do, I was told that the secret sauce was the pre-rendering of anything and everything that was outside the stream of realtime control. Now it's commonplace, but at the time it was revolutionary. And it still works just fine by me. Why overheat two dozen cores to calculate 900 tracks at 32 buffer, when only a few tracks / sends / busses actually need to be calculated that way? Pre-render the rest and play 'em back from RAM while the realtime tracks, all their sends, and busses they feed go onto the live cores.

With Apple silicon they'll have total control over core scheduling or whatever, so I wouldn't be surprised if we see performance gains in Logic that are absolutely bonkers and just don't "make sense" or translate to the way Intel silicon does things.

I, for one, welcome our new robot overlords.
Samplitude/Magix and Emagic shared code and both Samplitude and logic were the first two DAWs using the same hybrid buffer technology.

Logic isn't pre rendering, there's just a bunch of bad legacy code in there.. If it was pre rendering every audio change, then quickly hitting space bar to stop and start wouldn't effect the change immediately as it does now, it would still have to complete the render process if that's what it was doing. The massive delays when changing audio clips realtime seems like the audio file has to re-render each volume change, tempo change etc, so I can understand why there is some confusion there.
AFAIK Logic is the only DAW in existence still with this massive delay on every realtime audio adjustment if one persists with keeping the project playing back (I just got used to stopping and starting playback on every major edit).. However VI tracks are not subject to this. The process buffer simply means the internal latency they are playing at. (512/1024/2048).
PT has a fixed 1024 samples or non record armed tracks.
Logic has 3 choices as stated.
Cubendo and S1 have multiple choices.
Bitwig and Ableton play at one true buffer as did Reason up to a couple years back.
Reason I haven't used since its dual buffer implementation so don't know what choices if any there are or if it's just one standard size like pro tools.
Reaper is not using a dual buffer but some sort of pre render however only the default setting of 200 or 250ms max keeps the DAW feeling snappy enough and "realtime". Putting something like 500ms has massive performance gains for non record armed tracks but makes every action sluggish, like even moving the play head to another song position. AFAIK Reaper is the only DAW in existence to actually have a real cpu meter that correlates directly to actual CPU usage and is the only DAW that can actually use over 90 or even 95% of all threads of your CPU. The best I can get out of logic and PT and Cubendo with a perfectly balanced load is about 80% (if lucky).

On another general note (not aimed at you Charlie), for anyone following my mac vs pc performance notes, if you ever do comparisons of your own, make sure you have high performance mode enabled in windows.. The whole point is not to compare apples to apples (npi) but compare the best possible config of each OS vs the other.
Old 10th August 2020
  #1192
Lives for gear
 
crufty's Avatar
Quote:
Originally Posted by TNM View Post
FAIK Reaper is the only DAW in existence to actually have a real cpu meter that correlates directly to actual CPU usage and is the only DAW that can actually use over 90 or even 95% of all threads of your CPU.
Reaper is the best worst daw for sure.

On the pc side reaper is trash at handling small dimension hidpi. I am sure it will get there but I cant spend all day trying to figure out how to get things readable. It hurts my heart a bit but until i can setup.exe to something readable, I have to wait.

Below is a comparison of arm v amd v intel. The one to look at is ‘bare metal’—which is most similar to a DAW. Graviton2 is Amazons ARM server. Important to note the tested workloads are more dsp Adjacent and ARM compilers may not just be up to snuff.

https://www.phoronix.com/scan.php?pa...chmarks&num=12

Results like the above are why I am doubtful of claims of ARM superiority for dsp. Evidence can change that; It will take dedication from Apple to get perf parity. They certainly have the hw/sw chops to do so.

Tl/dr - a $3k 128 core mbp will get my attn
Old 10th August 2020
  #1193
Gear Maniac
 
BCProject's Avatar
 

Wow. This entire thread is 50% wild rampant (mostly uninformed) speculation, 25% armchair DAW reverse-engineering and 25% Apple-bashing.
First - No one is forcing anyone to buy anything.
Second - Seems smart to spend money on things that can do what you want *today*. If the manufacturer has a good track record for support, great.
Third - Apple has ZERO obligation to make "the thing" you specifically want them to make.
Fourth - why don't we wait until there is an actual shipping Apple ARM product to evaluate?
Finally - what is the point of dropping into a pure Apple thread to bash everything Apple and post lengthy missives about the wonders and virtues of Windows? What's your thinking here? Are you expecting readers to slap their heads in sudden epiphany: "My God! TNM's treatise on the unassailable superiority of Wintel is breathtaking and utterly convincing!! What have I been thinking?!?!?" ? Not gonna happen.
Old 10th August 2020
  #1194
Lives for gear
Quote:
Originally Posted by charlieclouser View Post
This is exactly how early Propellerheads software like ReBirth worked - it had very few realtime controls and no ability to map onscreen knobs to incoming MIDI CC for realtime knobbery, so it was able to perform amazing feats on a G3 Mac in 1998.

It's also basically what Logic's "process buffer" is doing - pre-rendering any elements that are not subject to external realtime control. At one point around 2001 or so, when Logic's soft synth performance was so far beyond what any other DAW could do, I was told that the secret sauce was the pre-rendering of anything and everything that was outside the stream of realtime control. Now it's commonplace, but at the time it was revolutionary. And it still works just fine by me. Why overheat two dozen cores to calculate 900 tracks at 32 buffer, when only a few tracks / sends / busses actually need to be calculated that way? Pre-render the rest and play 'em back from RAM while the realtime tracks, all their sends, and busses they feed go onto the live cores.

With Apple silicon they'll have total control over core scheduling or whatever, so I wouldn't be surprised if we see performance gains in Logic that are absolutely bonkers and just don't "make sense" or translate to the way Intel silicon does things.

I, for one, welcome our new robot overlords.
It is my experience, that ARM has better performance per clock already just with the basic ARM silicon. You can get pretty good performance out of a Raspberry Pi, both tried and that i've seen lots of crazy things done with it (like granular synthesis which is fairly compute heavy) and that's pretty low on the totem pole when it comes to the architecture.
Old 10th August 2020
  #1195
Lives for gear
Quote:
Originally Posted by crufty View Post
Reaper is the best worst daw for sure.

On the pc side reaper is trash at handling small dimension hidpi. I am sure it will get there but I cant spend all day trying to figure out how to get things readable. It hurts my heart a bit but until i can setup.exe to something readable, I have to wait.

Below is a comparison of arm v amd v intel. The one to look at is ‘bare metal’—which is most similar to a DAW. Graviton2 is Amazons ARM server. Important to note the tested workloads are more dsp Adjacent and ARM compilers may not just be up to snuff.

https://www.phoronix.com/scan.php?pa...chmarks&num=12

Results like the above are why I am doubtful of claims of ARM superiority for dsp. Evidence can change that; It will take dedication from Apple to get perf parity. They certainly have the hw/sw chops to do so.

Tl/dr - a $3k 128 core mbp will get my attn
These stats are almost pointless, DSP stuff is usually a special case of like 3-5 instructions (ie multiply-add). Sophie Wilson (the designer of the ARM CPU) designed it with multimedia in mind as it was supposed to go to the Acorn Archimedes which had GUI (and computers were well into this trend by then). She's also made quite a bit of contributions to DSPs later on. It's how they could get bitmap graphics in a PDA in 1993. The thermal qualities which went far and beyond Intel's was always going to bite them in the ass. Intel knew it as well, they tried fixing the "heater problem" with Itanium but X86 had so much dominance their replacement never gained any traction.
Old 10th August 2020
  #1196
Lives for gear
Quote:
Originally Posted by crufty View Post
Results like the above are why I am doubtful of claims of ARM superiority for dsp. Evidence can change that; It will take dedication from Apple to get perf parity. They certainly have the hw/sw chops to do so.

Tl/dr - a $3k 128 core mbp will get my attn
I think a 64 core mbc should get peoples attention. A dual 64 core Mac Pro etc.

To me anyway there's no doubt that Apples chips will take years if ever to beat dedicated server chips, and probably never will. I don't think anyone truly believes that you're not going to be able to build an AMD Threadripper or equivalent that outperforms at the same price whatever silicon Mac Pro they put out. I suppose it's possible, but IMO unlikely, as a layman, I see years of development on AMD and Intels side VS a (in abstract terms) relative newcomer taking a huge gamble.

What I do see is Apple putting out a MacBook pro with 32 or maybe 16 cores, before AMD and Intel can. This is where Arm should make a difference. Apple aren't going to put out a thicker laptop because Intel can't keep heat down. I'm not worried about them beating out desktop top end computers, I'm more interested in how they fare with workstation laptops since this is where you see Intel and them having issues. It's also where you see Apple compete a little better price wise, high end Windows laptops are brutally expensive as well, so if they can gain ground there, they probably won't care if their desktops are always beaten by the competition.

Personally I'm not that power hungry, there's just a few heavy on one CPU plug ins that bring down my 12 core Mac Pro. if Apple puts out a solid 16 or 32 core laptop that doesn't have the thermal, battery, and noise issues of the latest 2.4ghz 8 core in the Zenbook and Macbook Pro then I'm in.
Old 10th August 2020
  #1197
Gear Addict
 

I don’t think Apple goal is to outperform any chip on pure performance, but to produce a better efficient CPU for the end-user purposes. I don’t care if my DAW can run 500+ tracks, as I don’t care if my car can drive at rocket speed to take my kids at school. What I want is a better experience. For example I can’t stand my brand new MBP 16 to max out the CPU with... 3 VIs and 2 AA plugins. That’s a shame ! iOS is really more stable, though really limited audio-wise. So, if Apple Silicon can bring me that experience, I’ll resale my Intel MBP for an ARM as soon as possible.
Old 11th August 2020
  #1198
TNM
Lives for gear
Quote:
Originally Posted by BCProject View Post
Wow. This entire thread is 50% wild rampant (mostly uninformed) speculation, 25% armchair DAW reverse-engineering and 25% Apple-bashing.
First - No one is forcing anyone to buy anything.
Second - Seems smart to spend money on things that can do what you want *today*. If the manufacturer has a good track record for support, great.
Third - Apple has ZERO obligation to make "the thing" you specifically want them to make.
Fourth - why don't we wait until there is an actual shipping Apple ARM product to evaluate?
Finally - what is the point of dropping into a pure Apple thread to bash everything Apple and post lengthy missives about the wonders and virtues of Windows? What's your thinking here? Are you expecting readers to slap their heads in sudden epiphany: "My God! TNM's treatise on the unassailable superiority of Wintel is breathtaking and utterly convincing!! What have I been thinking?!?!?" ? Not gonna happen.
and who the hell do you think YOU are?
Forum police or something?

ANY member can come in this topic and discuss "apple's move to arm" and how they perceive it in any way they like.
Your entire post shows you are beyond rose coloured glasses.. you are literally an apple zombie. It's sheep like you why they keep pulling the garbage they do and get away with it. Yeah, there'll be a shipping arm product soon enough and my 2yo 8th gen win laptop will be whooping it for DAW performance every way imaginable. Can't wait to be proven right.
Old 11th August 2020
  #1199
Lives for gear
Quote:
Originally Posted by charlieclouser View Post
That is awesome, I wish Logic had that. Actually, I wish Logic had a LOT of stuff from DP now that I think about it....
DP is vastly underrated IMO. Probably mostly because it took them a while to adopt other DAWs standards, they still stick by completely separate MIDI tracks for instance, which turns a lot of people off to it. OTOH I get turned off by Logics insistence on only separating MIDI input to instruments by channel, so assigning a dedicated controller to a track involves assigning a discrete channel to that controller. I’m sure you could go into the environment and do something about it, but last time I tried this it messed with the Mackie Control I had. Pretty much everything else about Logic is top notch though. I wish DP had the ipad controller Logic does etc.
Old 11th August 2020
  #1200
Lives for gear
 
ponzi's Avatar
Quote:
Originally Posted by BCProject View Post
Wow. This entire thread is 50% wild rampant (mostly uninformed) speculation, 25% armchair DAW reverse-engineering and 25% Apple-bashing.
First - No one is forcing anyone to buy anything.
Second - Seems smart to spend money on things that can do what you want *today*. If the manufacturer has a good track record for support, great.
Third - Apple has ZERO obligation to make "the thing" you specifically want them to make.
Fourth - why don't we wait until there is an actual shipping Apple ARM product to evaluate?
Finally - what is the point of dropping into a pure Apple thread to bash everything Apple and post lengthy missives about the wonders and virtues of Windows? What's your thinking here? Are you expecting readers to slap their heads in sudden epiphany: "My God! TNM's treatise on the unassailable superiority of Wintel is breathtaking and utterly convincing!! What have I been thinking?!?!?" ? Not gonna happen.
Oh my goodness. Such terrible things happening right here in River City... Good thing you brought the commandments down off the mountain so nobody says anything against apple ever again.
📝 Reply

Similar Threads

Thread / Thread Starter Replies / Views Last Post
replies: 4589 views: 576426
Avatar for stixstudios
stixstudios 20th May 2020
replies: 4524 views: 272109
Avatar for S_A_P
S_A_P 3 minutes ago
replies: 93 views: 6050
Avatar for dapz
dapz 1 week ago
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…
🖨️ Show Printable Version
✉️ Email this Page
🔍 Search thread
🎙️ View mentioned gear
Forum Jump
Forum Jump