MP 7,1 LTT: ECC is dumb; 'buy base model nMP if you want a status symbol but don't care about performance'

[OS6-OSX]

What do you think is causing the crashing?

My Avid systems on Mac run near flawlessly, so you may have some OS revision/MC version clash.

Marianna over on the Avid Knowledgebase forum will escalate your complaints pretty quickly if they're repeatable.

Using MC 8.10 OSX 10.13.6 34GB ECC also "near flawlessly"
6-20-2019 MC 2019! See how that new engine runs. If they are smart it better be connected to RAM! Remember they have 16K on their minds! Also, if they want people to stay out of DaVinci via AAF the rendering better be connected to RAM! If you leave via AAF to DaVinci why come back to Symphony?"

 

[ekwipt]

What do you think is causing the crashing?

My Avid systems on Mac run near flawlessly, so you may have some OS revision/MC version clash.

Marianna over on the Avid Knowledgebase forum will escalate your complaints pretty quickly if they're repeatable.

The latest Avid MC 2018.12.3 runs quiet well, not many issues I must admit!

 

[AidenShaw]

https://spectrum.ieee.org/computing...mputer-dirty-power-cosmic-rays-and-bad-solder

… An even more dramatic example of cosmic-radiation interference happened at Virginia Tech’s Advanced Computing facility in Blacksburg. In the summer of 2003, Virginia Tech researchers built a large supercomputer out of 1,100 Apple Power Mac G5 computers. They called it Big Mac. To their dismay, they found that the failure rate was so high it was nearly impossible even to boot the whole system before it would crash.

The problem was that the Power Mac G5 did not have error-correcting code (ECC) memory, and cosmic ray–induced particles were changing so many values in memory that out of the 1,100 Mac G5 computers, one was always crashing. Unusable, Big Mac was broken apart into individual G5s, which were sold one by one online. Virginia Tech replaced it with a supercomputer called System X, which had ECC memory and ran fine....

 

[Andropov]

Well, short of having an extremely well-equipped physics lab with an alpha particle source, it's going to be difficult to generate ECC errors on demand... generally speaking one would have to rely on statistical evidence of Setup A with ECC turned on corrected N errors in X hours, and Setup B with ECC turned off presumably didn't correct ~N errors in X hours.

Actually you don't need an extremely well-equipped physics lab, alpha-particle sources aren't so hard to come by. But there wouldn't be a point, the mechanism and benefits of ECC memory are already well understood. Ultimately you'd get a probability for any bit to be flipped after some given time and you'd know that ECC memory would correct it and non-ECC wouldn't. Wether or not that would be critical for the software is a different matter altogether.

 

[SoyCapitanSoyCapitan]

EXTREMELY interesting commentary from linus regarding Mac Pro on his WAN show

For those who don't know, LTT is one of the most influential tech reporting organizations in the world. They have 8.6 million subscribers on youtube.

Link to timecode 9:17

Some highlights
- starts with a correction from his previous video on the nMP: looks like same mainboard for all new mac pros
- 8 core config should not exist, just buy an iMac or imac pro -it's probably faster even with the iMac throttling
- it probably doesn't matter that the memory bandwidth (6 channel) is crippled by only having 4 sticks on the low end because there isn't enough compute power anyway
- ECC Vs DDR4 -- ECC may not even matter anymore
- "Workstation" is a stupid marketing term that has no meaning anymore
- With the base model Apple is basically saying "buy this if you're the kind of ego, image-is-everything moron that wants to have a Mac Pro as a status symbol but doesn't care at all about what's under the hood"
- buying the base-model and then upgrading later is not economical at all because nobody will buy your 8 core on e-bay because it's crap considering it's only compatible with extremely expensive mobos
- All that said, probably a smart move to have the 6 grand as an option from a marketing standpoint, as only "nerds" like linus will figure out what a rip-off it is.


If you haven't seen it already, this is the original response video

It's important to note Linus is not actually over all negative about the nMP -- especially because of the ability to do 128GB GPU memory. He sees that as a very unique and and potentially cash-saving feature as apparently all competing options are enterprise-level crazy configs. He's basically just poo-pooing most other use-cases for this product. As a side note, he's also extremely excited about the screen and even defends apple a bit on the stand.

Here is the video, advance to 9:17 timecode

[doublepost=1560193732][/doublepost]The thing I found most interesting was the ECC stuff. Apparently he asked Intel to give him a workload to demonstrate the difference between ECC and DDR4 and they literally could not come up with one.

Most influential tech reporter? Maybe for kids who don’t know better. But...

Linus Tech Tips is a YouTuber who is a nobody in the tech world. A complete nobody. He barely had pubes when the real people in today’s tech were already seniors. He has never created anything, invented anything, and barely knows how to use macOS. He has zero effect on innovation and direction.

For illustration, he completely missed the accelerator board option for the Mac Pro, the modular card format’s cable free advances, and the scientific use cases.

 

[Zdigital2015]

Note that many of those comparisons pit a dual socket Z8 against the lower end Xeon-W in the MP7,1.

That's dishonorable.

I did as close to the base configuration 2019 Mac Pro as I could using an HP Z4 G4 and the following:

Z4 G4 Workstation - $4,862.68
* Windows 10 Pro 64 for Workstations
* Intel® Xeon® W-2145 Processor (3.7 GHz, up to 4.5 GHz w/Turbo Boost, 8.25 MB cache, 8 core)
* Z4 G4 1000 W Chassis
* 32 GB (4x8 GB) DDR4-2666 ECC Memory
* Operating System Load to M.2 / 256 GB HP Z Turbo Drive Self-Encrypted (SED) TLC M.2 SSD
* AMD Radeon™ Pro WX 7100 (8 GB GDDR5, 4x DisplayPort) Graphics
* Intel® X550-T2 10GbE Dual Port NIC
* Premium - 2 x USB 3.1 Type C; 2 x USB 3.0 Type A
* HP SD Card Reader
* No included Optical Disc Drive
* USB Premium Wired Keyboard
* USB Premium Wired Mouse
* No Adapters Needed
* HP Z4 Standard CPU Cooling Solution / HP Z4 G4 Fan Front Card Guide / HP Z4 G4 Memory Cooling Solution / HP Z4 G4 Dust Filter and Bezel
* 3/3/3-year warranty / Single Unit Packaging / Z4 G4 1000 W Country Kit

Obviously, its a last Gen Xeon W, but the rest is spot on as close as can be chosen. A closer comparison can be done when HP updates the Z4 with Xeon W-32xx CPUs.

Apple's charging about 30% more than HP here, which sounds about right, if HP is going for an 8% profit margin on this lower end configuration. Once you start adding serious storage and up the CPU or DRAM, HP Tax kicks in.

 

[AidenShaw]

Apple's charging about 30% more than HP here, which sounds about right, if HP is going for an 8% profit margin on this lower end configuration. Once you start adding serious storage and up the CPU or DRAM, HP Tax kicks in.

Thanks, that Z4 looks like a reasonable match - if you need everything that the MP7,1 comes with.

If you don't need an exact match, the price can be even more reasonable:

• Substitute Core i7-9800X (3.8/4.4 GHz, 8 core, 16.5 MiB cache) - save $1050
• Don't have 10 GbE infrastructure? Drop the X550 and save $549

Price is now $3397 - which is much closer to the sweet spot.

And Newegg has the single port X550 for $274 - in case you need 10 GbE but only one port. https://www.neweggbusiness.com/product/product.aspx?item=9siv1578u34813 Newegg has the dual port for about $470 - avoids the $79 HP tax.

 

[a2jack]

Thanks, that Z4 looks like a reasonable match - if you need everything that the MP7,1 comes with.

If you don't need an exact match, the price can be even more reasonable:

• Substitute Core i7-9800X (3.8/4.4 GHz, 8 core, 16.5 MiB cache) - save $1050
• Don't have 10 GbE infrastructure? Drop the X550 and save $549

Price is now $3397 - which is much closer to the sweet spot.

[doublepost=1560364140][/doublepost]And BTW...Love the cat. LOL

 

[AidenShaw]

[doublepost=1560364140][/doublepost]And BTW...Love the cat. LOL

Yes, clearly the "R.I.P. Mac Pro" tombstone was no longer appropriate.

And that isn't my kitten, but I have one that's nearly the same. Assuming that you can call a 20 lb (9 kilo) beast a "kitten".

 

[NT1440]

Most influential tech reporter? Maybe for kids who don’t know better. But...

Linus Tech Tips is a YouTuber who is a nobody in the tech world. A complete nobody. He barely had pubes when the real people in today’s tech were already seniors. He has never created anything, invented anything, and barely knows how to use macOS. He has zero effect on innovation and direction.

For illustration, he completely missed the accelerator board option for the Mac Pro, the modular card format’s cable free advances, and the scientific use cases.

Somebody posted a video earlier in the thread about his custom build that was compared against the iMac Pro. What puzzled me was if this guy is the know-all tech guru some see him as...why did he have this other guy there that actually built it? He kept deferring to that guy for the nitty gritty details, while just covering the specs himself.

This Linus guy strikes me as someone who is extremely skilled at building and using gaming rigs, can throw out tech specs, but doesn’t have the “expert” grasp of technology itself. He reminds me of retail workers who are good at speaking at the level above an average consumer so he *appears* to be an expert for anyone that’s not tech literate, but everything about him makes me think he’s just an above average gamer specific tech person.

 

[thisisnotmyname]

I don't know enough about the hardware to make a judgement -- are you saying there's literally no experimental way to verify ECC is significantly better?

fly your mac pro and some other machine around near the stratosphere and wait for cosmic rays to affect your RAM?

(sarcasm but it's not easy to engineer a bit flip scenario)

 

[Zdigital2015]

Thanks, that Z4 looks like a reasonable match - if you need everything that the MP7,1 comes with.

If you don't need an exact match, the price can be even more reasonable:

• Substitute Core i7-9800X (3.8/4.4 GHz, 8 core, 16.5 MiB cache) - save $1050
• Don't have 10 GbE infrastructure? Drop the X550 and save $549

Price is now $3397 - which is much closer to the sweet spot.

I do not quite understand why Apple shuns the Core X-Series, especially starting with the Core i7-6xxxX Series when pricing and stratification really began in earnest. I understand the benefits of ECC and how Apple might want to have a higher end Mac Pro than the 2006-2012 models, as they are trying to do with the 2019 Mac Pro. Frankly, I think Intel gave those CPUs away to Apple back in 2006-2008, but of course, I have no proof of something that bold, so maybe I shouldn’t say it without proof. I suspect that the Core X-Series is too close to "enthusiast or gamer" who tend to be THE most price sensitive of the "Consumer, Corporate, Connoisseur" trinity, or least amenable to Marcom BS.

I genuinely think Apple doesn’t want enthusiasts bugging them about the lack of Nvidia GPUs and the $6000 entry price does that make the price of entry high enough and narrow enough that the target audience just accepts the Vega II/Vega II Duo and keeps on truckin'.

Here’s the thing...the problem that Apple solved with the MPX module is the same problem that faced them with creating a tower-based successor to the Mac Pro 5,1. The problem was: how does Apple integrate Thunderbolt 1 (it was 2012 after all) onto the MP motherboard AND route it back and forth with the GPU ports to allow the Mac Pro 5,1 PLUS/EVO/MAX whatever you want to call it so that the both motherboard and GPU had Thunderbolt ports present and the user could plug in the Apple Thunderbolt Display into the GPU or to a port on the motherboard and use a DisplayPort monitor plugged into the GPU, as well as mix and match other devices such as storage.

Instead, back in 2012, someone whiffed or determined the tech just wasn't feasible, I am not an EE and not qualified to speak as to what went wrong. That or it was Marcom and Cook wanting to reduce the overall cost by making a computer that went on the desk that you could ship 12 units in the same amount of space as one MP 5,1, use 1/8th th raw aluminum and still charge the PRO the equivalent or more money.

Basically, Apple took a rather ill-advised and expensive detour into nMP land and dual GPUs that I don’t think they ever needed to take. Maybe the technology didn’t exist to do what they with the 2019 Mac Pro, but it’s really just a variation on the AGP+Power/Data that they gave us with the ADC connector. Granted, Thunderbolt is a more complex beast, but muxing video and data signals together seems to be something Apple does just to prove it can be done, so it shouldn't have been a challenge, like "How many waffles can we eat".

Either way, six years later, the Mac Pro is here and 6 years of successive models and gradual price increases that didn't happen were rolled into that $5999 price tag. Also known as "Just throw the frog in the pot! Maybe he jumps out, maybe he don't".

EDIT: My hot take...for the $5999 entry price, Apple might have made things go over easier had they gone with the following for the base model: 12-core Cascade Lake Xeon W-3235, 48GB ECC DRAM (6x8GB), 512GB (2x256GB) Flash and AMD Radeon Pro Vega 56 (8GB). Just my 2¢.

 

[xraydoc]

EXTREMELY interesting commentary from linus regarding Mac Pro on his WAN show

For those who don't know, LTT is one of the most influential tech reporting organizations in the world. They have 8.6 million subscribers on youtube.

Link to timecode 9:17

Some highlights
- starts with a correction from his previous video on the nMP: looks like same mainboard for all new mac pros
- 8 core config should not exist, just buy an iMac or imac pro -it's probably faster even with the iMac throttling
- it probably doesn't matter that the memory bandwidth (6 channel) is crippled by only having 4 sticks on the low end because there isn't enough compute power anyway
- ECC Vs DDR4 -- ECC may not even matter anymore
- "Workstation" is a stupid marketing term that has no meaning anymore
- With the base model Apple is basically saying "buy this if you're the kind of ego, image-is-everything moron that wants to have a Mac Pro as a status symbol but doesn't care at all about what's under the hood"
- buying the base-model and then upgrading later is not economical at all because nobody will buy your 8 core on e-bay because it's crap considering it's only compatible with extremely expensive mobos
- All that said, probably a smart move to have the 6 grand as an option from a marketing standpoint, as only "nerds" like linus will figure out what a rip-off it is.


If you haven't seen it already, this is the original response video

It's important to note Linus is not actually over all negative about the nMP -- especially because of the ability to do 128GB GPU memory. He sees that as a very unique and and potentially cash-saving feature as apparently all competing options are enterprise-level crazy configs. He's basically just poo-pooing most other use-cases for this product. As a side note, he's also extremely excited about the screen and even defends apple a bit on the stand.

Here is the video, advance to 9:17 timecode

[doublepost=1560193732][/doublepost]The thing I found most interesting was the ECC stuff. Apparently he asked Intel to give him a workload to demonstrate the difference between ECC and DDR4 and they literally could not come up with one.

I have to disagree with at least one bullet point above.
The 8-core option has a reason to exist. Even if the processor benchmarks are slightly slower than an i9 6-core or 8-core model, no iMac/iMac Pro is going to have the ability to use multiple internal GPUs the Mac Pro can. People buying the Mac Pro are those that need multiple GPUs for their workflow.

With that said, as much as I'd love a tower Mac with 'desktop' class hardware and not Xeon/server level hardware, it just isn't going to happen.

 

[Ph.D.]

Google (and others) did a study of their servers in which they found about 5 bit errors per 8 GB RAM per hour. They found that these errors were their number one reason for server crashes. Of course this problem is highly dependent on many factors such as altitude and the type of RAM, but things are only getting worse as RAM scales down in dynamic RAM capacitance per bit, etc. Anyway, let's use that number.

The 'Pro comes stock with 32 GB of ram, meaning 20 bit errors per hour. Scale that to 1.5 TB ram, as the Mac Pro is capable of, and you have about 1000 bit errors per hour.

If you are running a game, maybe you wouldn't care if those errors caused occasional glitches or crashes. But would your application experience serious issues with 20 to 1000 bit errors per hour? If so, you would want ECC RAM.

 

[slughead]

Google (and others) did a study of their servers in which they found about 5 bit errors per 8 GB RAM per hour. They found that these errors were their number one reason for server crashes. Of course this problem is highly dependent on many factors such as altitude and the type of RAM, but things are only getting worse as RAM scales down in dynamic RAM capacitance per bit, etc. Anyway, let's use that number.

The 'Pro comes stock with 32 GB of ram, meaning 20 bit errors per hour. Scale that to 1.5 TB ram, as the Mac Pro is capable of, and you have about 1000 bit errors per hour.

If you are running a game, maybe you wouldn't care if those errors caused occasional glitches or crashes. But would your application experience serious issues with 20 to 1000 bit errors per hour? If so, you would want ECC RAM.

I think I saw that study--crazy the amount of errors. They also said some ECC has something called "Chipkill" which reduces it "36x." What's also interesting if that another study still had a significant uncorrected error rate even with ECC (though DDR3 was 15 times higher in the same study)-- I would think it'd be way better because of the rarity of double-bit issues. Also important though is all the studies I've seen including the Google study (the big one I saw was 2009, I suspect that's the one you're referring to) was with DDR3. DDR 4 does have some features specifically for error-reduction and I could not find any data showing how well these work. In fact I could find zero studies looking at the error rate of DDR4.

I also wonder if RAM in general has changed enough over the past 10 years that the numbers are different. I'm sure you're right : ECC still does make a significant difference for certain applications, which ones specifically would be interesting, and if we'd expect anyone to actually purchase a nMP for those use-cases.

Also worth noting is that all AMD CPUs from ryzen to thredripper support ECC at present. The whole "gotta have a WERKSTASHUN for ECC" thing is kinda just an intel thing... but that's none of my business.

 

[Umbongo]

I also wonder if RAM in general has changed enough over the past 10 years that the numbers are different.

Yes it has improved, but you'd need a study at a large scale to get good information.

I think we have firmly established that servers will have it, because who is gonna manufacture Registered DIMMs without ECC. Workstations use server parts so it Is there too. It isn't cost prohibitive to the target market so it isn't a big deal now. Just something for people's to ignorantly complain about.

I'd love another study though. Shut a lot of people up who dismissed it all as a myth.

 

[leman]

I did as close to the base configuration 2019 Mac Pro as I could using an HP Z4 G4 and the following:

Except the MP has more RAM slots, more PCIe slots and a better PSU (not even talking about usability or thunderbolt ports). Z4 is cheaper but also has less features.

 

[johnbono]

fly your mac pro and some other machine around near the stratosphere and wait for cosmic rays to affect your RAM?

(sarcasm but it's not easy to engineer a bit flip scenario)

Just book a flight to the Ukraine, or Fukushima, or a couple of those Pacific Atolls.

 

[ssgbryan]

Except the MP has more RAM slots, more PCIe slots and a better PSU (not even talking about usability or thunderbolt ports). Z4 is cheaper but also has less features.

No need to limit yourself to HP. Velocity Micro will cheerfully sell you an AMD EYPC workstation with triple the cores, four times the memory, four times the M.2 memory, as well as the same video card (or an Nvidia card if your workflow needs it), and PSU for about 70% of the price.

Granted, it only has 6 PCIe slots (3x16, 3x8), but none are wasted on Thunderbolt adapters.

Best of all, no RGB.

 

[AidenShaw]

Some bits (corrected) from the linked article:

"Just how many spurious bit flips are happening inside supercomputers already? To try to find out, researchers performed a study [PDF] in 2009 and 2010 on the then most powerful supercomputer—a Cray XT5 system at Oak Ridge, in Tennessee, called Jaguar.

Jaguar had 360 terabytes of main memory, all protected by ECC. I and others at the lab set it up to log every time a bit was flipped incorrectly in main memory. When I asked my computing colleagues elsewhere to guess how often Jaguar saw such a bit spontaneously change state, the typical estimate was about a hundred times a day. In fact, Jaguar was logging ECC errors at a rate of 350 per minute."​

350 corrected memory errors per minute. Think about that. 350 errors per minute. About one per TebiByte/Minute. I have servers with 2 TiB RAM - that means two corrected errors per minute.

LTT is stupid for dissing ECC. I'm glad that I got a Xeon with ECC for my home PC. LTT's ignorant rants should be ignored.

 

[ekwipt]

Yes it has improved, but you'd need a study at a large scale to get good information.

I think we have firmly established that servers will have it, because who is gonna manufacture Registered DIMMs without ECC. Workstations use server parts so it Is there too. It isn't cost prohibitive to the target market so it isn't a big deal now. Just something for people's to ignorantly complain about.

I'd love another study though. Shut a lot of people up who dismissed it all as a myth.

or not...
[doublepost=1560393197][/doublepost]

Some bits (corrected) from the linked article:

"Just how many spurious bit flips are happening inside supercomputers already? To try to find out, researchers performed a study [PDF] in 2009 and 2010 on the then most powerful supercomputer—a Cray XT5 system at Oak Ridge, in Tennessee, called Jaguar.

Jaguar had 360 terabytes of main memory, all protected by ECC. I and others at the lab set it up to log every time a bit was flipped incorrectly in main memory. When I asked my computing colleagues elsewhere to guess how often Jaguar saw such a bit spontaneously change state, the typical estimate was about a hundred times a day. In fact, Jaguar was logging ECC errors at a rate of 350 per minute."​

350 corrected memory errors per minute. Think about that. 350 errors per minute. About one per TebiByte/Minute. I have servers with 2 TiB RAM - that means two corrected errors per minute.

LTT is stupid for dissing ECC. I'm glad that I got a Xeon with ECC for my home PC. LTT's ignorant rants should be ignored.

that's a study from 10 years ago, surely memory both ECC and non, servers, workstations and gaming builds have been improved upon by now

 

[JoeG4]

AMD just released a 16 core 105 TDP chip with PCIe4 at $799

Apple could've sold the Mac Pro to a much larger base on a lower end Mac Pro

The Mac Pro is. amazing but only once you go over the 16 or is it 18 core chips and get into the 28 core config

The new enclosure and parts are SO modular that they could make a cheap motherboard and stuff it in the chassis and you'd have a lower end machine. They could even replace the outer skin with a cheaper version to save even more money I suppose. Everyone would still happily buy the **** out of it.

 

[ekwipt]

Some bits (corrected) from the linked article:

"Just how many spurious bit flips are happening inside supercomputers already? To try to find out, researchers performed a study [PDF] in 2009 and 2010 on the then most powerful supercomputer—a Cray XT5 system at Oak Ridge, in Tennessee, called Jaguar.

Jaguar had 360 terabytes of main memory, all protected by ECC. I and others at the lab set it up to log every time a bit was flipped incorrectly in main memory. When I asked my computing colleagues elsewhere to guess how often Jaguar saw such a bit spontaneously change state, the typical estimate was about a hundred times a day. In fact, Jaguar was logging ECC errors at a rate of 350 per minute."​

350 corrected memory errors per minute. Think about that. 350 errors per minute. About one per TebiByte/Minute. I have servers with 2 TiB RAM - that means two corrected errors per minute.

LTT is stupid for dissing ECC. I'm glad that I got a Xeon with ECC for my home PC. LTT's ignorant rants should be ignored.

Hey Aiden, would you agree that the Mac Pro is primarily geared towards Video Editing? I can't find any benefits in ECC RAM for video Editing anywhere on the Interwebs. I'd love to see some proof from someone!

 

[filterdecay]

I did as close to the base configuration 2019 Mac Pro as I could using an HP Z4 G4 and the following:

Z4 G4 Workstation - $4,862.68
* Windows 10 Pro 64 for Workstations
* Intel® Xeon® W-2145 Processor (3.7 GHz, up to 4.5 GHz w/Turbo Boost, 8.25 MB cache, 8 core)
* Z4 G4 1000 W Chassis
* 32 GB (4x8 GB) DDR4-2666 ECC Memory
* Operating System Load to M.2 / 256 GB HP Z Turbo Drive Self-Encrypted (SED) TLC M.2 SSD
* AMD Radeon™ Pro WX 7100 (8 GB GDDR5, 4x DisplayPort) Graphics
* Intel® X550-T2 10GbE Dual Port NIC
* Premium - 2 x USB 3.1 Type C; 2 x USB 3.0 Type A
* HP SD Card Reader
* No included Optical Disc Drive
* USB Premium Wired Keyboard
* USB Premium Wired Mouse
* No Adapters Needed
* HP Z4 Standard CPU Cooling Solution / HP Z4 G4 Fan Front Card Guide / HP Z4 G4 Memory Cooling Solution / HP Z4 G4 Dust Filter and Bezel
* 3/3/3-year warranty / Single Unit Packaging / Z4 G4 1000 W Country Kit

Obviously, its a last Gen Xeon W, but the rest is spot on as close as can be chosen. A closer comparison can be done when HP updates the Z4 with Xeon W-32xx CPUs.

Apple's charging about 30% more than HP here, which sounds about right, if HP is going for an 8% profit margin on this lower end configuration. Once you start adding serious storage and up the CPU or DRAM, HP Tax kicks in.

the z4 doesnt have the expansion of the 7,1 macpro.

 

[amedias]

the z4 doesnt have the expansion of the 7,1 macpro.

was just about to say the same thing (again)...

That Z4 spec is only a match to the base config (and some way beyond), but it won't scale to the same place the 7,1 will. If you need to scale that far then to do that comparison it's onto a Z6 or Z8, and then again that's not a direct comparison as they are dual socket capable.

Doing comparisons is hard, not just between the 7,1 but even between other vendors as there's more to it than just matching core count and GB on an arbitrary spec.

If all you'll need is the base spec of a 7,1 then yes there are better/cheaper ways to achieve that spec, but if you'll only need the base spec of a 7,1 you probably don't need a 7,1 at all.

There really should be a Mac Midi/Z4 equivalent from Apple but it sadly is unlikely to happen.