Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Instead, we got the 6,1 in a different form factor. In the 6,1 everything connected via thunderbolt - in the 7,1, everything connects via MPX. It is proprietary, and limits what I can do.

MPX is just an extension to PCIe to use the machine's power supply and cooling system. You can just insert PCIe instead. MPX doesn't "limit what you can do".
 
But apparently Workstation-class Xeons, 50% more memory slots supporting 12x as much memory, and twice as many PCIe slots isn't quite enough. I guess it really needs to have an optical drive so you can watch those blockbuster rentals right?

Are you just trolling or do you not understand what how MPX works? It's a regular PCIe slot, with an additional connector. A standard PCIe card will still work, so what exactly is it limiting?

Ah yes, the 'sealed box' where you have more internal expansion than any previous model, and they introduce user-replaceable parts for the T2-secured SSDs that are not replaceable on other models.

Let me break it down some more. If I go too fast, pop a flare, I'll find you.

It is a mediocre system until you factor in price/performance - then it is a TCO fail.

Xeons can use ECC memory - Every AMD processor from a Ryzen 3 to an Eypc will also use ECC memory - just have to get a board that uses it. Mine is an ASUS WS-570-ACE. Added bonus - no RGB.

$6,000 gets you an 8 core/16 thread system with a 3 year old video card, 32gb of ram, and 256Gb of system memory.

$6,000 gets you 24 cores/48 thread system with a modern Navi video card, 128Gb of ram, and 1tb of system memory.

Which system do you think will provide more performance?

My system ran me about $3,500 - 16 cores/32 threads, 128Gb ram, 1Tb NVMe SSD, RX570 8gb. I didn't have to buy an MPX module ($2,600 to add 4 hard drives - which are smaller than the ones I currently own). If I want to upgrade the video card, I don't have to spend who knows how much for a video MPX module.

To add large HDDs internally, you have to go with a 3rd party solution that costs thousands of dollars and isn't actually covered by Apple care.

To quote the Apple website:

Note: Products sold through this website that do not bear the Apple brand name are serviced and supported exclusively by their manufacturers in accordance with terms and conditions packaged with the products. Apple’s Limited Warranty does not apply to products that are not Apple branded, even if packaged or sold with Apple products. Please contact the manufacturer directly for technical support and customer service.


To add SSDs, need to get an Apple compatible PCIe card. There are some good ones out there, but I don't need to, since the board supports 2 out of the box.

Once you have added your MPX modules, you only have half of those PCIe slots - or the same amount I have currently.

And those PCIe slots are half the speed of the ones that I have.

In 2016, the 7,1 would have been an expensive, but great computer.

Problem is that Zen was unleashed in 2017. The computing world changed - Apple didn't.

TCO Fail.
 
Let me break it down some more. If I go too fast, pop a flare, I'll find you.

Translation: the only valid "complaint" you have (valid in the sense that it could even makes sense, not that I agree with it) is cost.

If I want to upgrade the video card, I don't have to spend who knows how much for a video MPX module.
You don't have to use an MPX video card in a 7,1 either. You keep making claims alluding to not being able to use standard PCIe cards so I have to assume you're just trolling on that point now.

To add large HDDs internally, you have to go with a 3rd party solution that costs thousands of dollars and isn't actually covered by Apple care.

Ok, bit of a tangent here, but yes there are expensive options available. You could also add some slow hard drives with a $200 or $400 bracket that make use of the SATA ports, and use no PCIe slots.

I don't personally think the R4i is particularly great value either - but thats because I'm more realistic about the failure rate of spinning rust, and would rather relegate any mass-storage like that to external devices which can be powered off (if not hot-swappable) to replace a failed drive. But that's just me. I only work with computers for a living, it's not like it's my hobby or anything.

To add SSDs, need to get an Apple compatible PCIe card. There are some good ones out there, but I don't need to, since the board supports 2 out of the box.

So to "add" SSD's beyond the two modules it ships with, you need to add a PCIe card to a 7,1. Beyond the two M2 slots your board ships with, how do you "add" more M2's? Oh right, I forgot you have <checks notes> a whole four PCIe slots available to you.

Once you have added your MPX modules, you only have half of those PCIe slots - or the same amount I have currently.
Ok let's break this down. I'm going to ignore that you didn't specify which MPX modules.

Your board, has 3, x16 slots, and 1 x1 slot. But from the way the specs read to me, two of the x16 slots share 16 lanes? So it's either 1 x16 (and the other unused) or 2 x8 (in x16 mechanical slots)? The third x16 (from the chipset rather than the cpu directly) seems to be fixed at x8 in a x16 mechanical slot. Am I reading that correctly?

So if you want to use all slots, you have 3, x8 and 1, x1. If you want max performance for a single card, you have 1 x16, 1x8 and 1x1. All are PCIe4.0 if you have a modern enough Ryzen.

So to compare (in the 'use all slots' setup) with the PCIe 3.0 slots in the 7,1 it's like having 3, 16x and 1, 2x, right?

Ok so what does the 7,1 have?

3, x16 double-wide slots.
1, x16 single-wide slot
1, x8 double-wide slot
2, x8 single-wide slots
1, x4, half-length slot (pre-allocated to the rear TB3/USB ports so let's ignore that one shall we?)

Now.. If you fill both MPX bays with full-size MPX cards, you "lose" all the double-wide slots.

So you're left with 1 x16 and 2 x8.

I guess technically you do have more with your board unpopulated, than the Mac Pro has with 2 MPX bays full, because you have that little 1x slot left, but comparing a half-loaded board with one that has nothing and saying "look I have just a little more expansion before I plug anything in" is disingenuous at best.

And those PCIe slots are half the speed of the ones that I have.
Yes, and you have a lot fewer of them overall.
 
I am like a LOT of folks in the Mac Pro forum - I wanted an updated 5,1, not an xMac. I didn't want consumer grade CPUs, I need a lot of ram, and I need a number of PCIe slots (Video card, eSata card, and possibly a sound card).

The 5,1 was price competitive with Windows workstations. If they weren't moving to AMD, give me a Xeon, the ability to add whatever PCIe cards I need (ability to add a Blu-ray player would be a "nice to have"), and be priced competitively - like the 1st 5 generations of Mac Pros. The 7,1 is a $1,400 computer in a $4,600 case.

But, that stuff costs money! You keep saying $1400, but isn't the processor alone like $800? An RX580 isn't high end (And, 3-year-old... who cares if it does what it needs to do? The alternative are fairly pricy cards that add some % of performance, typically not relative to the % price increase.), but it adds a couple hundred more, right? We're getting close to your $1400 already and we haven't even hit RAM, storage, power supply, design, etc.

And, then consider that all these components aren't just 'get by' as I often see in comparisons, but are top end stuff.

I get that it isn't what you wanted, but I've also heard from true high-end people like Alex Lindsay, for whom it is almost exactly what they wanted.

Also, a sound card? eSATA card? Why? But, it does have slots for that.

Yes, the Intel-transition early Mac Pros (2000s) were priced pretty competitively (and were probably closest to mainstream PC in Apple history). Just remember, though, that was kind of an anomaly relative to Apple history. For most of the history, the hardware was differentiated and on a bit of a different pricing tier. (We can debate whether that is good or bad... just pointing it out.)

... from what I have seen, moving to subscriptions cuts off the hobbyists.

Yes, that is more likely the case, at least for the higher-end software. It is priced in relation to income-earning potential by the target user-base.

But, what was the alternative? Hobbyists spending $thousands (would they do that?) and then just trying to keep running it for many years w/o upgrades? I suppose some did that, but my experience from my earlier days was that a lot of hobbyists just ran cracked/copied versions until they were making money. I suppose this new mode kind of shuts that path down, yes.

Most of the subscriptions of software I've dealt with so far isn't more expensive, it is just distributing the payment differently. Instead of spending $2k every other year, you spend $85/month or $1000/yr or such. I'm sure that varies, and I don't buy a lot of high end stuff, but that has been my experience so far.

As an aside, when I started in 3D, we purchased Electric Image Animation System when it had dropped from around $8000/license to *only* like $2900. There weren't any things like Blender around. The hobbyist has it way better these days, whether they pick open-source or paid software.

I am with Apple as long as they make a better product - somebody builds a better mouse trap - I am out. The true believers don't understand this mindset - they would rather settle for less performance, because Windows Suks! It is an emotional thing for them; And that is fine, I just don't take them seriously.

AMD has delivered a better mouse trap on the hardware side, and Win10 is on par with OSX - I have run it for 9+ months now and have yet to see the issues I saw with Windows NT 20 years ago. Everything just works.

It isn't so much about working or not, as the experience while doing it. I spent most of my career in IT and certainly put a lot of time in on Windows. It sucks less now, but it still sucks in comparison. That isn't just emotional, or I'd be happily running a Windows box quite some time ago.

It seems to me that Cloud hosting of applications ultimately diminishes the issue of platform dependencies.

Yes, or at least that seems to be what some are talking about these days. We'll have to wait and see how some of the upcoming packages run in actual use, though. As I mentioned earlier, if Stadia (Google's cloud gaming) can work, then so should cloud-3D or CAD. But, I'm not sure Stadia ultimately works all that well (the reviews seem kind of mixed).

There are other workflow issues, and even security ones. Some companies refuse to put their projects into the cloud. And, for big projects, having them local on a GB or 10 GB network is a LOT different than having them out there in some highly bandwidth limited cloud.

Of course, there are some upsides, too. We'll have to see. I just don't want that to be my only option.

To add large HDDs internally, you have to go with a 3rd party solution that costs thousands of dollars and isn't actually covered by Apple care.

To quote the Apple website:

Note: Products sold through this website that do not bear the Apple brand name are serviced and supported exclusively by their manufacturers in accordance with terms and conditions packaged with the products. Apple’s Limited Warranty does not apply to products that are not Apple branded, even if packaged or sold with Apple products. Please contact the manufacturer directly for technical support and customer service.
...
TCO Fail.

Fail for a hobbyist, I suppose. Not for the actual high-end pro target market.

But, which care-system covers your PC at all, much less the 3rd party components you add in?
 
Last edited:
  • Like
Reactions: Stephen.R
But, that stuff costs money! You keep saying $1400, but isn't the processor alone like $800? An RX580 isn't high end (And, 3-year-old... who cares if it does what it needs to do? The alternative are fairly pricy cards that add some % of performance, typically not relative to the % price increase.), but it adds a couple hundred more, right? We're getting close to your $1400 already and we haven't even hit RAM, storage, power supply, design, etc.

And, then consider that all these components aren't just 'get by' as I often see in comparisons, but are top end stuff.

I get that it isn't what you wanted, but I've also heard from true high-end people like Alex Lindsay, for whom it is almost exactly what they wanted.

Also, a sound card? eSATA card? Why? But, it does have slots for that.

Yes, the Intel-transition early Mac Pros (2000s) were priced pretty competitively (and were probably closest to mainstream PC in Apple history). Just remember, though, that was kind of an anomaly relative to Apple history. For most of the history, the hardware was differentiated and on a bit of a different pricing tier. (We can debate whether that is good or bad... just pointing it out.)

Yes, that is more likely the case, at least for the higher-end software. It is priced in relation to income-earning potential by the target user-base.

But, what was the alternative? Hobbyists spending $thousands (would they do that?) and then just trying to keep running it for many years w/o upgrades? I suppose some did that, but my experience from my earlier days was that a lot of hobbyists just ran cracked/copied versions until they were making money. I suppose this new mode kind of shuts that path down, yes.

Most of the subscriptions of software I've dealt with so far isn't more expensive, it is just distributing the payment differently. Instead of spending $2k every other year, you spend $85/month or $1000/yr or such. I'm sure that varies, and I don't buy a lot of high end stuff, but that has been my experience so far.

As an aside, when I started in 3D, we purchased Electric Image Animation System when it had dropped from around $8000/license to *only* like $2900. There weren't any things like Blender around. The hobbyist has it way better these days, whether they pick open-source or paid software.

It isn't so much about working or not, as the experience while doing it. I spent most of my career in IT and certainly put a lot of time in on Windows. It sucks less now, but it still sucks in comparison. That isn't just emotional, or I'd be happily running a Windows box quite some time ago.

Yes, or at least that seems to be what some are talking about these days. We'll have to wait and see how some of the upcoming packages run in actual use, though. As I mentioned earlier, if Stadia (Google's cloud gaming) can work, then so should cloud-3D or CAD. But, I'm not sure Stadia ultimately works all that well (the reviews seem kind of mixed).

There are other workflow issues, and even security ones. Some companies refuse to put their projects into the cloud. And, for big projects, having them local on a GB or 10 GB network is a LOT different than having them out there in some highly bandwidth limited cloud.

Of course, there are some upsides, too. We'll have to see. I just don't want that to be my only option.

Fail for a hobbyist, I suppose. Not for the actual high-end pro target market.

But, which care-system covers your PC at all, much less the 3rd party components you add in?

As a hobbyist, I have bought multiple Apple workstations The G4 Power Macs were reasonably priced, my G5 Power Mac was reasonably priced, my 1,1 was reasonably priced, and my 4,1 was reasonably priced. The 6,1 was most definitely not reasonably priced. I would have to spend an additional $2,000 just to replace the functionality that Sir Idiot Boy removed - and I certainly wouldn't have seen a major uplift in performance - and that was before the video card issues started.

The hobbyist level is a different world now - I am sure there are still folks running cracked software, but it isn't necessary anymore. It is amazing what one can do with even bottom of the stack software.

In the 3d world, hobbyist can and do spend thousands of dollars - I spend about $1,500 a year just on digital assets (15+ years). Many of us have our very own render farms. Not too hard to do. I turned multiple Z210 into render nodes ($250 gets you a Xeon 4c/8t CPU, 32gb of ram, & 120gb SSD). Much cheaper than buying a video card during the crypto-mining apocalypse a couple years back.

And for 3d art - the Mac Pro is the ONLY available option - full stop. I HAVE spent thousands of dollars keeping my Mac Pros up to date (It's not like we had new computers to purchase, now is it?). My 1,1 went through 3 video cards, a CPU transplant, and 2 different memory configurations. My 4,1 received 2 CPU transplants, 2 video card transplants and got to 96 gb of ram (see sig).

I'd would simply point out that this IS what the Mac Pro community has been doing for the past decade - for the last 10 years, we only had the trashcan or a WinTel system as an upgrade path. Go look at the forum and look at the lengths people will go to keep their 4,1s & 5,1s going. Or the people that are still using 1,1s & 2,1s. (Dump a pair of low power Cloverfields in them & they make great HTPCs.)

And then there is the 7,1...... $1,400 dollars of parts in a $4,600 case.

$1,400 gets you an 8 core/16 thread Ryzen system that is comparable to a base 7,1. I'd remind you that you don't actually get much for that $6,000. (8c/16t CPU, 256Gb NVMe ssd, 32gb ram, RX580 video card, in case you forgot.)

My 16 core system is $3,500. That processor (Ryzen 9 3950x) has dropped to under $700, be cheaper by Xmas, since the 4950x will be available. Will I spend $750 for a 4950x? Nope - 15% performance increase isn't worth it.

AFA support - 7,1 breaks - that is a 2 hour drive to the nearest Apple store (assuming it is open in these troubled times) - 7,1 is then shipped off to another location. Who knows how long it will take for it to get back.

Last time I had to use Apple Care, my computer was gone for 3 weeks. The returned product lasted all of 193 days and died again. The netbook I bought to hold me over is still running BIONC. Since that time, Apple QA has gotten steadily worse.

If I have a Dell or HP workstation, a tech will show up within 4 hours. Don't get that with Apple Care - gotta buy the Enterprise version.

For my system, I come from the wild, wild, west era of IT - I used to provide internal tech support for a small part of a company (200+ seats) called International Business Machines. Have you heard of them?

Before that I supported multiple offices that not only had different hardware - every office had a completely different set of software: 3 different operating systems, 4 different word processors, 3 different spreadsheet apps, 2 different presentation apps, 2 different database apps, and then all of our internal applications - Good times.

My current Ryzen system breaks - pull a spare part out, replace and go.

Why an E-sata card? My backup is a Mercury Qx2 - Sata is faster than firewire 800, and I am not dumb enough to replace something that works.

At the end of the day, the Mac Pros have gone from a general purpose workstation to a very specialized workstation. I would be OK with that, if Apple still made a general purpose workstation.

Apple no longer makes computers for creative hobbyists.
 
  • Haha
Reactions: Nightfury326
Let me break it down some more. If I go too fast, pop a flare, I'll find you.

It is a mediocre system until you factor in price/performance - then it is a TCO fail.

Xeons can use ECC memory - Every AMD processor from a Ryzen 3 to an Eypc will also use ECC memory - just have to get a board that uses it. Mine is an ASUS WS-570-ACE. Added bonus - no RGB.

$6,000 gets you an 8 core/16 thread system with a 3 year old video card, 32gb of ram, and 256Gb of system memory.

$6,000 gets you 24 cores/48 thread system with a modern Navi video card, 128Gb of ram, and 1tb of system memory.

Which system do you think will provide more performance?

My system ran me about $3,500 - 16 cores/32 threads, 128Gb ram, 1Tb NVMe SSD, RX570 8gb. I didn't have to buy an MPX module ($2,600 to add 4 hard drives - which are smaller than the ones I currently own). If I want to upgrade the video card, I don't have to spend who knows how much for a video MPX module.

To add large HDDs internally, you have to go with a 3rd party solution that costs thousands of dollars and isn't actually covered by Apple care.

To quote the Apple website:

Note: Products sold through this website that do not bear the Apple brand name are serviced and supported exclusively by their manufacturers in accordance with terms and conditions packaged with the products. Apple’s Limited Warranty does not apply to products that are not Apple branded, even if packaged or sold with Apple products. Please contact the manufacturer directly for technical support and customer service.


To add SSDs, need to get an Apple compatible PCIe card. There are some good ones out there, but I don't need to, since the board supports 2 out of the box.

Once you have added your MPX modules, you only have half of those PCIe slots - or the same amount I have currently.

And those PCIe slots are half the speed of the ones that I have.

In 2016, the 7,1 would have been an expensive, but great computer.

Problem is that Zen was unleashed in 2017. The computing world changed - Apple didn't.

TCO Fail.
Ryzen is not a workstation platform. Unless you use EPYC, you are limited to a paltry 256GB of ECC RAM.

Dell, Lenovo and HP all make Xeon workstations in the $50-$150,000 range, but I don’t need those either. Maybe we’re not the target market? 🤷‍♂️

It’s quite apparent that the Mac Pro is targeted at the business sector: one-person shop, SMB, enterprise, whatever. Customers that need a lot of cores, GPU, RAM capacity and/or memory bandwidth.

Those folks are perfectly willing and able to pay $6k to $60k for a machine that makes them money, whether they need/use the special MPX slots (or video stream coprocessor) or are just using it as a general purpose workstation.

And while Apple doesn’t target the Mac Pro at creative hobbyists, it is a perfect computer for them. I know people who spend $60k on a cool looking pickup truck because that’s their hobby. And many, many people spend more than $200 a month on their hobby 🤷‍♂️
 
Last edited:
As a hobbyist, I have bought multiple Apple workstations The G4 Power Macs were reasonably priced, my G5 Power Mac was reasonably priced, my 1,1 was reasonably priced, and my 4,1 was reasonably priced. The 6,1 was most definitely not reasonably priced. I would have to spend an additional $2,000 just to replace the functionality that Sir Idiot Boy removed - and I certainly wouldn't have seen a major uplift in performance - and that was before the video card issues started.

I somewhat agree here, in that I too owned a Power Mac in the past (and several MacBook Pros which didn't hold up to my attempted hobbyist/pro use). Apple does need to address this segment better, and they've absolutely gone high-end with the recent machines (out of my league anyway).

I think the machines we're *supposed* to be getting (in Apple's thought) are mini/iMac or maybe iMac Pro. I can see somewhat why those aren't appropriate options for you (or me, fully). So, in that regard, I'm with you on Apple needing something more geared for us.

But, even Apple recognizes the issue with the cylinder Mac Pro. It was a great design I'm glad exists, as it inspired stuff like my Blackmagic eGPU, I assume. Maybe the upcoming Xbox? I hope more people learn from Apple's design there, but fix it and take it to new levels.

The hobbyist level is a different world now - I am sure there are still folks running cracked software, but it isn't necessary anymore. It is amazing what one can do with even bottom of the stack software.

Yes, I've just started to realize this recently. I need to start messing with Blender. I've been amazed with some of the ArchVis stuff I've seen come out of there.

In the 3d world, hobbyist can and do spend thousands of dollars - I spend about $1,500 a year just on digital assets (15+ years). Many of us have our very own render farms. Not too hard to do. I turned multiple Z210 into render nodes ($250 gets you a Xeon 4c/8t CPU, 32gb of ram, & 120gb SSD). Much cheaper than buying a video card during the crypto-mining apocalypse a couple years back.

I get that. Some people are willing to put a lot into hobbies or other interests that don't necessarily make money (ie. Pro). I probably put around a $half-million into my education which I'll never recoup, but it was also a step on the path in moving from something that made $ which I didn't enjoy, to hopefully something I do.

And for 3d art - the Mac Pro is the ONLY available option - full stop.

Hmm, I guess that depends. I bet there are people doing 3D art on iMacs, iMac Pros, MacBook Pros, minis, etc.

My problem with all of them, which I'd concede, is that they aren't properly setup to do so (with possible exception of iMac Pro). The thermals just aren't geared to 24x7 100% utilization. That said, I', hoping my 2018 mini is going to be OK. I try to keep it under control though, and don't run 100% 24x7. For that, I think you're right it is either the iMac Pro or Mac Pro.

Note: All my problems in the past were GPU related, or mainboard issues, probably due to switches in solder technology that are more time-specific (the PS3 & Xbox also, famously, had problems, as did a lot of the industry). If that is the case, then with my eGPU, I might be just fine running the mini full-out. Maybe I'll need a fan replacement someday.

And then there is the 7,1...... $1,400 dollars of parts in a $4,600 case.

$1,400 gets you an 8 core/16 thread Ryzen system that is comparable to a base 7,1. I'd remind you that you don't actually get much for that $6,000. (8c/16t CPU, 256Gb NVMe ssd, 32gb ram, RX580 video card, in case you forgot.)

My 16 core system is $3,500. That processor (Ryzen 9 3950x) has dropped to under $700, be cheaper by Xmas, since the 4950x will be available. Will I spend $750 for a 4950x? Nope - 15% performance increase isn't worth it.

Is that really apples to apples though (pardon the pun)? That new Mac Pro is just kind of all-out in every hardware aspect.

If I have a Dell or HP workstation, a tech will show up within 4 hours. Don't get that with Apple Care - gotta buy the Enterprise version.

But, if you have a Dell or HP workstation, you probably aren't buying it for $1400, nor getting a technician to show up in 4 hours. My guess is you'll be spending in that range where you could happily have a Mac Pro as well.

For my system, I come from the wild, wild, west era of IT - I used to provide internal tech support for a small part of a company (200+ seats) called International Business Machines. Have you heard of them?

Before that I supported multiple offices that not only had different hardware - every office had a completely different set of software: 3 different operating systems, 4 different word processors, 3 different spreadsheet apps, 2 different presentation apps, 2 different database apps, and then all of our internal applications - Good times.

Yes, that is the world I came from as well. I spent most of my early career doing network/software installs from the small business to Fortune 100 companies for a top reseller of Novell and Autodesk. Then I spent years in IT consulting between a small business started with a friend, to an independent. Then I worked in a Sr. IT role for a near Fortune 50 for almost a decade.

It isn't that I'm not technically capable of building my own machines, or working within Windows or Linux. I just don't enjoy doing so... which is partly why I left that world to pursue other interests (at great financial cost).

Why an E-sata card? My backup is a Mercury Qx2 - Sata is faster than firewire 800, and I am not dumb enough to replace something that works.

OK, but you'd have to admit, that is a pretty specific use-case.
And... one you could easily solve for not a heck of a lot of money.

need/use the special MPX slots (or video stream coprocessor) or are just using it as a general purpose workstation.

And while Apple doesn’t target the Mac Pro at creative hobbyists, it is a perfect computer for them. I know people who spend $60k on a cool looking pickup truck because that’s their hobby. And many, many people spend more than $200 a month on their hobby 🤷‍♂️

True (as discussed above), though I think it is fair to keep pointing out the MASSIVE hole in Apple's product line for hobbyists and prosumers who don't have quite that much money. ssgbryan is quite correct that Apple used to have such machines at much more affordable prices.
 
Last edited:
  • Like
Reactions: 09872738
A word processor or calculator is less likely to use instructions that are available on the x86 that won't have direct equivalents on the new ARM chips. Apple already called out some of the vector instructions as not working under Rosetta 2 and I'd extrapolate it to mean code leveraging those instructions will require alternate implementations on Apple Silicon (which is interesting for me because the Intel shift lost Altivec support and wasn't quite equivalent at the time). It's more likely that these programs that do 3D image manipulation that is likely to benefit from vector instructions than a word processor or calculator.

Thing is, not every Intel CPU support all the different vector extensions. So the application ask at runtime, which capabilities are present and choses the right code path. In all cases i am aware of, there is an already existing alternate implementation, which just uses C/C++, which can be trivially compiled for ARM.
 
<snip>
True (as discussed above), though I think it is fair to keep pointing out the MASSIVE hole in Apple's product line for hobbyists and prosumers who don't have quite that much money. ssgbryan is quite correct that Apple used to have such machines at much more affordable prices.
There is a hole, but it’s not massive. 80% of Apple’s sales are notebooks, and another 10-15 of sales are iMac. That leaves 5-10% of sales to be split between the mini, iMac Pro and Mac Pro.

According to Apple, a lot of creatives moved to iMac when it became fast enough that they didn’t need to buy a Mac Pro anymore. (Surely that trend was accelerated by the amazing 5K Retina screen on the 27” iMac.)

It’s impossible to give everyone what they want; some are always going to want something different. There are those that want a larger mini with a PCIe slot or two. Those who want nVidia GPUs. Those who won’t buy until they get upgradeable SSDs, USB-A and/or MagSafe on a MBP. Those who want AMD CPUs, etc.

What’s Apple to do, try to satisfy every niche?
 
Thing is, not every Intel CPU support all the different vector extensions. So the application ask at runtime, which capabilities are present and choses the right code path. In all cases i am aware of, there is an already existing alternate implementation, which just uses C/C++, which can be trivially compiled for ARM.

I think every Mac since 2012 (maybe 2011, I'd have to go look) has had a version of the AVX instruction though you're right there are versions that won't have that functionality prior to that. The reason to leverage these instructions is generally to get a performance improvement for what ever operation is happening because obviously every instruction can be implemented more basically via other mechanisms. However those mechanisms might not operate as quickly as simple code and would need to be re-implemented with the equivalent versions that may or may not exist on the final ARM chips to get similar performance. If you're using those tools you're more likely to have an up to date machine with a CPU that has those instructions available and are leveraging that capability. Just see the lamentations of ssgbryan on the inadequacies of the new Mac Pro 7,1.

However the question you posed was what is the difference and the difference is time spent optimising code for a particular architecture to get the performance. Figuring out the equivalent instruction, optimisation or porting to a new library/implementation is the likely difference between the spreadsheet application and something like a 3D modelling application moving to ARM. If it was sufficient to not use the instructions, then likely they wouldn't have bothered to implement it in the first place. However if you are leveraging those instructions then you're also invested in trying to make your software perform better. It looks like the Accelerate framework might provide indirect access to do similar things unlocking features in Apple Silicon that aren't documented or otherwise available but it would require porting to that framework. Depending on where your tooling has it's primary market (e.g. Windows), that may not be something a developert bothers investing in.
 
  • Like
Reactions: SteveW928
There is a hole, but it’s not massive. 80% of Apple’s sales are notebooks, and another 10-15 of sales are iMac. That leaves 5-10% of sales to be split between the mini, iMac Pro and Mac Pro.

According to Apple, a lot of creatives moved to iMac when it became fast enough that they didn’t need to buy a Mac Pro anymore.

This.

I'm sure some people would love a $2k, $1k, $whateverK Mac Pro Lite, xMac, whatever. What they're missing is that they market has moved on. Offices aren't filled with towers any more; they're filled with laptops. For the edge cases where you need high-end GPU power, an eGPU will almost invariably do. The Mac Pro is a high-end niche because it can be.

Now, Apple does choose not to serve every niche. And for those who want macOS but don't quite want any of the Mac hardware Apple is offering, that sucks. But it's not a lot of people.
 
This.

I'm sure some people would love a $2k, $1k, $whateverK Mac Pro Lite, xMac, whatever. What they're missing is that they market has moved on. Offices aren't filled with towers any more; they're filled with laptops. For the edge cases where you need high-end GPU power, an eGPU will almost invariably do. The Mac Pro is a high-end niche because it can be.

Now, Apple does choose not to serve every niche. And for those who want macOS but don't quite want any of the Mac hardware Apple is offering, that sucks. But it's not a lot of people.

Exactly. The only people (IMO) who really need a Mac Pro are audio professionals and video editors who rely on the mac platform. In those cases, it's a business asset, and they most likely don't care about the cost.

In my team of 15, we have 15 laptops. My friend who works in an IB has a thin client and a VM with whatever resources he wants to throw at it. The world has moved on from the "computer under a desk" model.
 
I think every Mac since 2012 (maybe 2011, I'd have to go look) has had a version of the AVX instruction though you're right there are versions that won't have that functionality prior to that. The reason to leverage these instructions is generally to get a performance improvement for what ever operation is happening because obviously every instruction can be implemented more basically via other mechanisms. However those mechanisms might not operate as quickly as simple code and would need to be re-implemented with the equivalent versions that may or may not exist on the final ARM chips to get similar performance.

The question was about the effort of getting a native ARM version, not about the effort to put in every imaginable performance optimization.
Point in case, when i ported 7zip to ARM i left out all the AVX and SSE stuff available in the Intel version and compiled just the available C reference code. Turned out that on the Surface Pro X (7W Machine) i am still close to or beating Intels 15W chips in 7zip benchmark (and most certainly beating the 2020Macbook Air running the AVX version)
Apparently i spent minimal effort - mostly just recompiling and still getting great performance.
 
The question was about the effort of getting a native ARM version, not about the effort to put in every imaginable performance optimization.
Point in case, when i ported 7zip to ARM i left out all the AVX and SSE stuff available in the Intel version and compiled just the available C reference code. Turned out that on the Surface Pro X (7W Machine) i am still close to or beating Intels 15W chips in 7zip benchmark (and most certainly beating the 2020Macbook Air running the AVX version)
Apparently i spent minimal effort - mostly just recompiling and still getting great performance.
Linus Torvalds has complained about Intels crappy FP performance. Having to rely on AVX to get anything close to decent performance annoys him.
 
Exactly. The only people (IMO) who really need a Mac Pro are audio professionals and video editors who rely on the mac platform. In those cases, it's a business asset, and they most likely don't care about the cost.

In my team of 15, we have 15 laptops. My friend who works in an IB has a thin client and a VM with whatever resources he wants to throw at it. The world has moved on from the "computer under a desk" model.

Perhaps your world has moved on - the one I operate in hasn't. The rest of the Apple lineup is thermally throttled.

The people that use a Mac Pro are people that push the CPU/GPU.

If you don't push them, an iMac (pro or non-pro) is fine.
 
Perhaps your world has moved on - the one I operate in hasn't. The rest of the Apple lineup is thermally throttled.

The people that use a Mac Pro are people that push the CPU/GPU.

If you don't push them, an iMac (pro or non-pro) is fine.

I too live in that world.

As far as Apple chasing market segments, I think they've got the "small, throttled, non-expandable, thermally-challenged" market covered.
 
Perhaps your world has moved on - the one I operate in hasn't. The rest of the Apple lineup is thermally throttled.

The people that use a Mac Pro are people that push the CPU/GPU.

If you don't push them, an iMac (pro or non-pro) is fine.
What’s your definition of “thermally throttled”? All Macs run well beyond Intel’s rated speeds.
 
Last edited:
The question was about the effort of getting a native ARM version, not about the effort to put in every imaginable performance optimization.
Point in case, when i ported 7zip to ARM i left out all the AVX and SSE stuff available in the Intel version and compiled just the available C reference code. Turned out that on the Surface Pro X (7W Machine) i am still close to or beating Intels 15W chips in 7zip benchmark (and most certainly beating the 2020Macbook Air running the AVX version)
Apparently i spent minimal effort - mostly just recompiling and still getting great performance.

Did you just equate your porting of 7zip to ARM with porting something like Maya to ARM with the implication being that it would likely be equivalent for them?
 
Base clock or all core boost?

The whole point of boost is that it's temporary.
[automerge]1594758862[/automerge]
Did you just equate your porting of 7zip to ARM with porting something like Maya to ARM with the implication being that it would likely be equivalent for them?

Presumably, Maya makes more use of vector extensions, and thus has more per-arch optimization potential, but that doesn't mean getting it to run at all, relatively "unoptimized", is harder than doing so for 7-zip. This isn't the 1990s where people put tons of inline assembler into their C code.
 
The whole point of boost is that it's temporary.
[automerge]1594758862[/automerge]


Presumably, Maya makes more use of vector extensions, and thus has more per-arch optimization potential, but that doesn't mean getting it to run at all, relatively "unoptimized", is harder than doing so for 7-zip. This isn't the 1990s where people put tons of inline assembler into their C code.
Generally speaking Intel will run at all core boost all the time if thermals allow for it. Or at least that is the observed behavior on Windows.
 
Base clock or all core boost?
Feel free to provide your own definition, per my post.
[automerge]1594764047[/automerge]
Generally speaking Intel will run at all core boost all the time if thermals allow for it. Or at least that is the observed behavior on Windows.
That’s not possible at Intel’s rated specs. MacOS or Windows isn’t relevant, it’s an Intel issue.
 
  • Like
Reactions: MacCheetah3
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.