Bitwig ARM Version if Apple leaves Intel?

[jonljacobi]

https://www.cpubenchmark.net/

The main list is multi-core performance, however, the detail view for each CPU will give you the single thread score. As of 3rd gen AMD Ryzen, they and Intel Core processors are pretty much even in per core performance. Previously, Intel had a distinct advantage in single core performance. It will be very interesting to see how some of the multi-chip ARM stuff stacks up.

[machinesworking]

Ah, and don't worry about ARM. If a company knows about multi-platform development, then its Bitwig.

Yeah. Not to mention iPad version of Bitwig will be a given if that happens!

it very well may be that 4 years from now it's all ipads. Looking at the iPad Pro with the magic keyboard and a track pad even I see hardly any reason for them to continue to split the line if the ARM chips become standard and iOS and OS X continue to cross over.

[lunardigs]

RAM is really important too. The main counterpart of the CPU is RAM. The CPU's main use of bandwidth is for RAM.
RAM is also an area where sheer clock speed may not betray a true speed advantage. For instance, 2 channel vs. 4 channel RAM is a big difference--possibly half the r/w bandwidth--despite clock speed of either the RAM or CPU.

Furthermore, Registered RAM (RDIMMs)--which implies a controller between the CPU and RAM--is almost always more performant than unregistered RAM (UDIMMs) of a higher clock speed. Also, counter-intuitively, higher Rank count on RDIMMs makes for greater bandwidth, whereas the opposite is generally true for UDIMMs.

In my experience, selection of motherboard and RAM is really important. Don't skimp in this area, even if it means affording a lesser CPU ... my 2 cents.

[teilo]

RAM is really important too. The main counterpart of the CPU is RAM. The CPU's main use of bandwidth is for RAM.
RAM is also an area where sheer clock speed may not betray a true speed advantage. For instance, 2 channel vs. 4 channel RAM is a big difference--possibly half the r/w bandwidth--despite clock speed of either the RAM or CPU.

Furthermore, Registered RAM (RDIMMs)--which implies a controller between the CPU and RAM--is almost always more performant than unregistered RAM (UDIMMs) of a higher clock speed. Also, counter-intuitively, higher Rank count on RDIMMs makes for greater bandwidth, whereas the opposite is generally true for UDIMMs.

In my experience, selection of motherboard and RAM is really important. Don't skimp in this area, even if it means affording a lesser CPU ... my 2 cents.

How is this relevant to ARM vs. Intel on Apple systems?

Also, as to registered RAM, you have this backwards. It's never faster. It's usually slower. All systems have a memory controller between the CPU and RAM. RDIMMs additionally have memory registers in between the RAM and the memory controller that reduces the power load on the memory controller. This makes it possible to address larger amounts of RAM than would otherwise be possible. This is great in a server, but it also makes RAM run SLOWER because every write operation has to write to the registers first, which then transfer the data to RAM.

[don_looney]

RAM is also an area where sheer clock speed may not betray a true speed advantage. For instance, 2 channel vs. 4 channel RAM is a big difference--possibly half the r/w bandwidth--despite clock speed of either the RAM or CPU.

The difference in 2 vs 4 channel RAM for DAW usecase is not important.
Only RAM Speed could speed up the CPU. As example Ryzen 3000 series runs faster when RAM is running at 3600 MHz, you could push it even harder with 3866 Mhz but this is not easy and takes a lot of adjustments.

[lunardigs]

RAM is really important too. The main counterpart of the CPU is RAM. The CPU's main use of bandwidth is for RAM.
RAM is also an area where sheer clock speed may not betray a true speed advantage. For instance, 2 channel vs. 4 channel RAM is a big difference--possibly half the r/w bandwidth--despite clock speed of either the RAM or CPU.

Furthermore, Registered RAM (RDIMMs)--which implies a controller between the CPU and RAM--is almost always more performant than unregistered RAM (UDIMMs) of a higher clock speed. Also, counter-intuitively, higher Rank count on RDIMMs makes for greater bandwidth, whereas the opposite is generally true for UDIMMs.

In my experience, selection of motherboard and RAM is really important. Don't skimp in this area, even if it means affording a lesser CPU ... my 2 cents.

How is this relevant to ARM vs. Intel on Apple systems?

Also, as to registered RAM, you have this backwards. It's never faster. It's usually slower. All systems have a memory controller between the CPU and RAM. RDIMMs additionally have memory registers in between the RAM and the memory controller that reduces the power load on the memory controller. This makes it possible to address larger amounts of RAM than would otherwise be possible. This is great in a server, but it also makes RAM run SLOWER because every write operation has to write to the registers first, which then transfer the data to RAM.

Yes, true, so due to the register--which is a memory controller, not be confused with the PCH or northbridge--RDIMMs are technically not faster than UDIMMs. That's my point: Nonetheless, RDIMMs are typically able to achieve greater practical throughput (bandwidth) than UDIMMs. This is especially true when using multiple sticks & channels.

I point this out as an example of parallel vs. serial, which came up. Also, to contextualize 'whole system' vs. component(s) performance. Performance is a whole system picture. Even mundane components like the PSU matter a lot for system stability (realized performance).
To that end, I think emphasis on clock speeds can be misleading. Rather, realized performance it's multi-factorial. Which I suppose is no surprise.

[cnt]

If Apple goes the ARM route, they should lower the prices by 50% for the hardware to match what you get for a normal PC. Otherwise I don't see the point? For consumer type of work ARM is probably a great choice, but for creators. Eh.. no.
ARM processors has never beaten or even been close to the performance of AMD or Intel.
If they go the ARM route for their "cheaper" / computers / laptops, it would be a smart move to use the latest AMD Ryzens for the other ones, those aimed at creators.

No Apple hardware can beat the AMD 3950x with 32 GB 3600 Mhz RAM, 1000W power, quality motherboard. And add to that it's only a fraction of the cost. I paid around $1500 for my new machine. MacOSX does support the latest AMD Ryzen (some ppl run Hackintosh on it), so it is not far fetched that they will start selling AMD based computers as well as ARM... Interesting to see what will happen.

[teilo]

Yes, true, so due to the register--which is a memory controller, not be confused with the PCH or northbridge--RDIMMs are technically not faster than UDIMMs. That's my point: Nonetheless, RDIMMs are typically able to achieve greater practical throughput (bandwidth) than UDIMMs. This is especially true when using multiple sticks & channels.

I point this out as an example of parallel vs. serial, which came up. Also, to contextualize 'whole system' vs. component(s) performance. Performance is a whole system picture. Even mundane components like the PSU matter a lot for system stability (realized performance).
To that end, I think emphasis on clock speeds can be misleading. Rather, realized performance it's multi-factorial. Which I suppose is no surprise.

Since your point is now that RDIMMs are simutaneously "more performant" and "slower," I'm not going to respond further to this piece of irrationality. Further, none of this has anything to do with the topic.

[don_looney]

Soooo back to topic. Apple announced the move to ARM starting at the the end of this year and the transition is done in 2 years. Maybe we see Bitwig on ARM and then Bitwig on iPadOS.
I think todays annoucements are great for creative working people. Running apps like Procreate on a MacBook is huge. Iam really hyped, and iam on Win10

[lunardigs]

Yes, true, so due to the register--which is a memory controller, not be confused with the PCH or northbridge--RDIMMs are technically not faster than UDIMMs. That's my point: Nonetheless, RDIMMs are typically able to achieve greater practical throughput (bandwidth) than UDIMMs. This is especially true when using multiple sticks & channels.

I point this out as an example of parallel vs. serial, which came up. Also, to contextualize 'whole system' vs. component(s) performance. Performance is a whole system picture. Even mundane components like the PSU matter a lot for system stability (realized performance).
To that end, I think emphasis on clock speeds can be misleading. Rather, realized performance it's multi-factorial. Which I suppose is no surprise.

Since your point is now that RDIMMs are simutaneously "more performant" and "slower," I'm not going to respond further to this piece of irrationality. Further, none of this has anything to do with the topic.

Geez ... thanks for the effort

Well, if Apple goes ARM and you're not along for the ride, you can always build an x86_64 Linux system. In which case, you might consider a serverboard with registered RAM and Xeon(s).

How is this to do with the topic? Well, perhaps we have some Mac users here who might be considering an exit strategy and could use some help. I've built many systems and I'm willing to help.

By the way, maybe you can answer this, teilo: Why does Apple use RDIMMs and Xeons for their Mac Pro? I suppose they could have chosen virtually any configuration, but they chose this; why? Performance, maybe?

[Tj Shredder]

OS X 10.16 aka Big Sur will be like 10.4 (Tiger). The last to support Intel processors directly. No problem, because intel based software will run seamless also in 10.17 and 10.18... this is likely around 2022... Then users will stick to the last version which can run intel software, like Snow Leopard lasted 4 more years... Thats a lot time to finally switch to Linux...(which supports the Arm architecture for a long time already...)

[pdxindy]

OS X 10.16 aka Big Sur will be like 10.4 (Tiger). The last to support Intel processors directly. No problem, because intel based software will run seamless also in 10.17 and 10.18... this is likely around 2022... Then users will stick to the last version which can run intel software, like Snow Leopard lasted 4 more years... Thats a lot time to finally switch to Linux...(which supports the Arm architecture for a long time already...)

Those are guesses on your part? I didn't see any comment from Apple about it.

Big Sur is also not going to be 10.16 but rather OS 11

[Tj Shredder]

Yes guesses. It worked in the past, and Apple will deliver Rosetta 2. OS 11 or 10.16 - I don't care... I always thought the X stands for UniX and was a word/number game which fitted...
I hope my beloved Max/MSP will move to Linux, but I don't have much hope... Running a VM on an ARM based MacOS doesn't sound convenient. Lets see what happens, I will be able to make my music the way I prefer for quite some time...
Usually the only reason to upgrade MacOS in the past was software I wanted but required a newer OS, or it became so aged that it would be a security risc. Never ever it was about OS features...

[Hanz Meyzer]

No, Apple can't let Intel users stick to an outdated macos version soon, because of the new mac pro.

[antic604]

Interesting rumour here - apparently during the transition to ARM Apple might move production of "their" Intel processors to TSMC, which would mean they're gonna be 5nm tech, instead of current 10nm. They're going with a bang!

https://www.youtube.com/watch?v=vbkFfo7w3II