MP 1,1-5,1 Bummer... cMP won't run Massive X

[Mactrunk]

I've been waiting for Native Instruments to release the highly anticipated update to their virtual instrument, "Massive".
From the specs the new version, "Massive X" looks very cool.

It just came out. I went to install it and got the error message...
"Massive X currently requires an AVX compatible processor. Your computer does not support the AVX instruction set."

I'm running a cMP and am very disappointed.

Does anyone know of a workaround?

This is the first piece of software I've encountered that won't even allow installation on a cMP.

Thanks for any insight.

 

[startergo]

None of the CPU's for cMP has Advanced Vector Extensions (AVX). The only way to install software requiring AVX on cMP AFAIK is to install an OSX KVM machine over a Debian installation which emulates different SMBIOS and passing through the PCI devices you need.:
https://passthroughpo.st/new-and-improved-mac-os-tutorial-part-1-the-basics/

 

[thekev]

No. Your machine probably supports SSE4 or some earlier SSE variant depending on the model. It lacks the ability to decode VEX prefixed instructions. Even if they didn't actually use 256 bit data types anywhere, the encoding differs from SSE. Things which were 2 operand instructions became 3 operand instructions, and the instruction length is slightly longer. AVX itself actually comes in different versions. Intel supports AVX2 and FMA3 on Haswell and Skylake based architectures. AVX is also supported on Sandy and Ivy Bridge types, although these don't benefit quite as much.

They either have some low level code sections that they don't want to rewrite, or they don't want to compile and ship for multiple instruction sets. Apple hasn't sold anything in more than five years that didn't support AVX, so I see the logic with it. It's still unfortunate. Running a VM means that you're effectively hosting an imaginary cpu, which knows how to decode VEX prefixed instructions. It might be considerably slower. If your concern is performance rather than features, I would suggest sticking with the older version.

 

[davidec]

Yeah Sandy Bridge is a minimum unfortunately. The cMP Xeons won’t do it. I’ve had a play on my MBP and of all the instruments it’s the one I can live without. It is amazing but I’ll happily wait for the nMP

 

[deconstruct60]

None of the CPU's for cMP has Advanced Vector Extensions (AVX). The only way to install software requiring AVX on cMP AFAIK is to install an OSX KVM machine over a Debian installation which emulates different SMBIOS and passing through the PCI devices you need.:

Blowing pure smoke at the VM pretending to be something that the systems is not isn't going to solve the problem here. For low latency / real time applications you can't just "emulate" AVX . AVX represents a expansion of the instruction set , The BIOS/Firmware doesn't really cover it. And not just in instruction set but in how the load/store system is provisioned. Emulating it is possible but not at anywhere near full speed.

One of the primary points of adding AVX to the instruction set was to do things that the normal add/subtract/etc instructions really couldn't do as efficiently. ( and that SSE didn't particularly do well enough on. )


Massive X is forward looking application. It isn't meant to be seamlessly compatible with anything before this.
" ...built for a new decade and designed to evolve ..."
https://www.native-instruments.com/en/products/komplete/synths/massive-x/

The cMP was designed over a decade ago. That is a decade in the opposite direction.
[doublepost=1561727954][/doublepost]

....Running a VM means that you're effectively hosting an imaginary cpu, which knows how to decode VEX prefixed instructions. It might be considerably slower. If your concern is performance rather than features, I would suggest sticking with the older version.

Virtualization and emulation are technically two different things. Modern virtualization hypervisors and vm foundation systems don't make the CPU "opaque". All they do is provision the context where a guest operating system thinks it actually have control of the whole system. To make it look like it is in charge when often it isn't when in kernel mode. There is a trap when try to do some subset of instructions, but there is no wholesale emulating of instructions.

There is no "might be" in the considerably slower even if manage to extend the trap and emulate aspect.

 

[thekev]

There is no "might be" in the considerably slower even if manage to extend the trap and emulate aspect.

Fair enough on both points. You are correct that supporting AVX at the hypervisor level would require software decoding of vex prefixed instructions, and I hadn't thought about that. In hindsight, that would be unbearably slow.

 

[mikehalloran]

I know that CPU upgrades for the 6.1 are easily done. I’ve not followed which ones are compatible, however. Is there a “bolt on” that will allow the cMP to run Massive X?

I take it this applies to the 5.1 as well. Again, people have been upgrading the CPUs on those, too, so the same question applies.

I’m running an iMac Pro so am not directly affected but have a number of friends who are not going to be happy when they try to download after the announcement we were emailed this morning.

I’m not a big synth guy but this upgrade was promised when we bought Komplete 12 last year.

 

[startergo]

None of the CPU's for cMP has Advanced Vector Extensions (AVX).

I take it this applies to the 5.1 as well. Again, people have been upgrading the CPUs on those, too, so the same question applies.

 

[deconstruct60]

I know that CPU upgrades for the 6.1 are easily done. I’ve not followed which ones are compatible, however. Is there a “bolt on” that will allow the cMP to run Massive X?

You can't just "bolt on" a future CPU to an older system that was not designed for it. The CPU has to fit into a socket of a specific design. It also has to couple to a I/O chipset ( "Platform Control Hub" (PCH) in Intel parlance ). You can't just stuff a new CPU into any older logic board.

Technically, the 6.1's CPUs have some of the AVX instructions. I think it is the initial set ( which is a subset at this point as AVX has evolved 2+ generations since then . )

I take it this applies to the 5.1 as well. Again, people have been upgrading the CPUs on those, too, so the same question applies.

The CPUs used to upgrade the 5.1 are just as old as the 5.1. They are just cheap because in end-of-life and (at this point very likely ) used status. They aren''t 'new' in any way on the dimesnion of supporting the modern extensions to the x86_64 instruction set at all.

While Intel was on the tick-tock upgrade progression path the CPU socket changed for Mac Pro class processors every 18-24 months (along with the PCH). Even the 6.1 there is nothing new that will fit/operate. The 6.1 was at the end of a Xeon E5 tick-tock cycle so there is nothing "newer" that fits there becasuse the processors that existed when the 6.1 was launched in late 2013- early 2014.

I’m running an iMac Pro so am not directly affected but have a number of friends who are not going to be happy when they try to download after the announcement we were emailed this morning.

Download before reading the specs ??

" System Requirements
... MASSIVE X requires an AVX-compatible CPU to run. .."
https://www.native-instruments.com/en/products/komplete/synths/massive-x/specifications/


macOS has had support for AVX since 2011. For folks who have bought a modern system since 2012 this is basically a non issue. Folks living on machines on the Vintage and Obsolete list are mainly those impacted.

I’m not a big synth guy but this upgrade was promised when we bought Komplete 12 last year.

It is delivered. But they aren't delivering it for systems in the vinatge/obsolete status. I"m not sure why folks would expect them to. That isn't normal high end software business practice at all.

 

[handheldgames]

Has anyone tried the opcode-emulator-opemu-plug-in-project from the Hackintosh world?

 

[startergo]

Has anyone tried the opcode-emulator-opemu-plug-in-project from the Hackintosh world?

So this should be in L/E/, right? Nice project.
I did not see any success in that thread though. Only KP.

 

[Mactrunk]

Deconstruct60 said...

"macOS has had support for AVX since 2011. For folks who have bought a modern system since 2012 this is basically a non issue. Folks living on machines on the Vintage and Obsolete list are mainly those impacted.
It is delivered. But they aren't delivering it for systems in the vinatge/obsolete status. I"m not sure why folks would expect them to. That isn't normal high end software business practice at all."

Hey DC...
I got my new cMP in 2012.
In fact the 'About This Mac' says, mid 2012.
O.K. I agree this is a 7 year old mac.

That said, this is the first time I've encountered a VI that won't install.

Maybe it is so great that it needs to obsolete 2012 macs.

For me it is very disappointing.

 

[deconstruct60]

....

Hey DC...
I got my new cMP in 2012.
In fact the 'About This Mac' says, mid 2012.
O.K. I agree this is a 7 year old mac.

The 2012 cMP is mostly a mirage. In terms, of instruction set support that is really a 2009-2010 CPU. Pragmatically making that a 9 year old Mac at this point. ( That's why the 2012 is on the Vintage/Obsolete list this Spring). The 2009 and 2010 models had respectively gone onto the Vintage/Obsolete list in each of the preceding two years. There are haven been some pretty clear indicators for the last 2 years that these systems with a shared foundation were being phased out. That wouldn't have been lost on any software vendor paying attention.

There was nothing 'new' about the 2012 model except configuration changes (of 2010 era options) and a firmware bump. Apple needed a stop gap while they got to work on a solution and the 2012 model was far more so "we aren't quitting the Mac Pro" placeholder. It is also there so that Apple could retire the 2010 at a more normal (for Apple) service life window. (they could start the "ceased manufature" clock on Vintage countdown. )

The MP 2012 value was primarily just being closer to being a "box with slots". It was essentially empty at providing a more modern foundation. The foundation was 2 years old at the time. In terms of modernization it was a "form over function" value proposition.

Around June 2012, Intel launched the Xeon E5 1600 v1 series. Apple entirely skipped that (while competing systems all adopted Xeon E5). That puts a large new instruction set coverage gap in the older Mac Pro systems. If Apple had done one more iteration on the Mac Pro in 2012 to Xeon E5 it would probably be still alive for another year or so. They didn't.

 

[DPUser]

Fortunately, everything in Komplete 12 except Massive X will run fine on cMPs. The AVX requirement only applies ot Massive X. The pre-X version of Massive will still run on older Macs.

As noted above and as far as I know, this is the first VI that requires AVX. Power users running Vienna Ensemble Pro with more modern slave PCs should be able to host Massive X on a slave and use it successfully.

 

[fastlanephil]

Fortunately, everything in Komplete 12 except Massive X will run fine on cMPs. The AVX requirement only applies ot Massive X. The pre-X version of Massive will still run on older Macs.

As noted above and as far as I know, this is the first VI that requires AVX. Power users running Vienna Ensemble Pro with more modern slave PCs should be able to host Massive X on a slave and use it successfully.

This sounds like a good solution for most users. I see VSL has dropped the price of VEP 7 back down to what VEP 5 was, which I have. I was last using it to run a 64-bit host in my old 32-bit DAW.

You could always build a basic PC using hackintosh specs for more options.

 

[howiest]

The 2012 cMP is mostly a mirage. In terms, of instruction set support that is really a 2009-2010 CPU. Pragmatically making that a 9 year old Mac at this point. ( That's why the 2012 is on the Vintage/Obsolete list this Spring). The 2009 and 2010 models had respectively gone onto the Vintage/Obsolete list in each of the preceding two years. There are haven been some pretty clear indicators for the last 2 years that these systems with a shared foundation were being phased out. That wouldn't have been lost on any software vendor paying attention.

There was nothing 'new' about the 2012 model except configuration changes (of 2010 era options) and a firmware bump. Apple needed a stop gap while they got to work on a solution and the 2012 model was far more so "we aren't quitting the Mac Pro" placeholder. It is also there so that Apple could retire the 2010 at a more normal (for Apple) service life window. (they could start the "ceased manufature" clock on Vintage countdown. )

The MP 2012 value was primarily just being closer to being a "box with slots". It was essentially empty at providing a more modern foundation. The foundation was 2 years old at the time. In terms of modernization it was a "form over function" value proposition.

Around June 2012, Intel launched the Xeon E5 1600 v1 series. Apple entirely skipped that (while competing systems all adopted Xeon E5). That puts a large new instruction set coverage gap in the older Mac Pro systems. If Apple had done one more iteration on the Mac Pro in 2012 to Xeon E5 it would probably be still alive for another year or so. They didn't.

Very well said, and to the point. I appreciate that.

 

[Ludacrisvp]

So my thoughts on the opcode emulator.
Sounds great!
But it looks like the people using it or trying to have all been AMD users. Perhaps this is the reason for what looks like issues.
And a main thing about forums and features like this provides. Most of the time you’re just going to see the negative replies not a bunch of “it works”. Also with this being targeted more toward the AMD crowd I’m not sure that many intel users have seen it nor knew they would benefit from it.
I’m not in a position to test it right now but I feel like some people on here should if they have the ability to do so.

 

[deconstruct60]

So my thoughts on the opcode emulator.
Sounds great!
But it looks like the people using it or trying to have all been AMD users. Perhaps this is the reason for what looks like issues.

AMD relatively recent processors have all had AVX. AvX is not an Intel only set of opcodes.

https://en.wikipedia.org/wiki/Advanced_Vector_Extensions#CPUs_with_AVX

AMD staring picked up AVX in their desktop line up in 2011-12 timeframe. Again before much of the span of the Mac Pro 2012 (and after the even older Mac pro 2009-2010 models).

Emulating AVX is more about being 'old' than it is about being Intel or AMD implemented.

And a main thing about forums and features like this provides. Most of the time you’re just going to see the negative replies not a bunch of “it works”.

In the context of apps with real time , low latency constraints , that is probably closer to "happens to work" than "it works".

 

[Ludacrisvp]

Emulating AVX is more about being 'old' than it is about being Intel or AMD implemented.

I don't see the goal of this emulator being just for the AVX instructions as it has plenty of others that it emulates.
The only people I see replying were having issues with this and their AMD processor is the point I'm making.
We haven't heard from those using Intel to know if it is viable or not.

 

[mikehalloran]

"macOS has had support for AVX since 2011. For folks who have bought a modern system since 2012 this is basically a non issue.

Uh... yea, right. If you say so. So Massive X should install, right?

Except that it doesn't.

 

[AidenShaw]

Emulating AVX is...

...mostly pointless.

AVX (and AVX2) uses 256-bit hardware registers to operate on eight 32-bit data values in parallel. It's "loop unrolling" in hardware - doing math at eight times the performance of scalar.

Since at best an AVX emulator will be about 8 times slower than AVX - it's very possible that emulation will be too slow to make the program viable.

So, a car with a one cylinder engine might get from Point A to Point B - but the ride with an eight cylinder engine will be more fun.

 

[Mactrunk]

Gentlemen,

Consider me bonked.

Massive X is the first piece of software that has refused to install on my vintage 5,1.
But surely not the last.

Yes, of course, the writing is on the wall.
Time to give up the ghost.
5,1 is sliding in to destiny.

That said, I love the machine and everything else is working so well.
Flawless daily driver.

I have a year or two to transition to something else.
Maybe 7,1 will be sorted out by then.

Thanks for all of your great, informative posts.

 

[DPUser]

I have a year or two to transition to something else.
Maybe 7,1 will be sorted out by then.

This is a realistic approach. Might even be more than a year or two...

I have so many amazing tools in my audio toolbox, not being able to use Massive X or some other iterative software is the least of my worries. If a truly indispensable and unique piece of code that requires AVX comes along sooner than I expect, that will be the day my 5,1 gets retired.

 

[norda72]

I bought my Mac Pro 5.1 2012 in 2013. Some said already then that the computer was old, but I didn’t listen. Now I see how old it really was if other Macs have been supporting AVX since 2011. Last year I upgraded it to 12-core and 128 GB RAM, I wasn’t aware of this. I regret it now.

 

[thekev]

Now I see how old it really was if other Macs have been supporting AVX since 2011.

If you're concerned about support for the most recent instruction sets, you would have to purchase something that supports AVX512 today, which doesn't leave you with that many options (maybe the most recent ones). Not much software used AVX in 2011, because it was quite different from SSE. It maintained the assumption of 128 bit lanes in spite of 256 bit data paths, so things like palignr, which could be used to shift streams of data an operand at a time lacked a "good" replacement. It's possible to emulate them, but it takes too many register names, which is bad because most people who are interested in AVX would want to also batch enough operations at a time to hide instruction latency.