The Best Mac Nobody Needs

IMPORTANT: I'm really dumb and didn't check our export file very well. Half of our benchmark charts are missing and I really apologize for not catching this before I pushed "publish." It won't happen again. Please see this Google Sheets link with all of the data we collected for better insights on how it compares to PC GPUs. snazzy.fm/fx

I can’t believe we got a stopgap Apple Silicon Mac Pro right after the stopgap Intel Mac Pro.

Pegasus made a 32 TB raid array that was compatible with the previous generation Mac Pro. It use the MPX slot, and on Apple own website is advertised as being compatible with the current generation 2023 Mac Pro. Spoiler alert, it isn’t.

Great review and specifics about Mac Pro that actually are helpful. Would love a retrofit for the existing case. Would love to get more life out of it.

Man, that is once again probably the most honest and level headed Mac prospective I've seen.

The thunderbolt future that Phil was so excited about back in 2013 has finally happened now a decade later.

I think they missed a huge opportunity with a „Mac Pro/Mac Studio“ in a 1U rack mount chassis. Considering all the rack mounted thunderbolt hardware and the acceptable success of the Rack mount Mac Pro, a slim 1U M2 Ultra Mac with fat networking and a buttload of thunderbolt would fill that professional gap.

You could do a video with Neil Parfitt on the pro. He's a composer who bought a Mac Pro, and a very amusing guy. His biggest issue was the cross compatibility of software that worked on an Intel Mac Pro but didn't work at all on a Mac studio and also when firmware updates come through, it is a total gamble for them that their software and plugins will actually work at all. Something that could potentially massively affect their ability to work and therefore earn money.

I think they released it just so they can say, "We've finally finished our transition" in the next Apple event or something.

BINGO! My 2019 is safe from replacement for now. You want to see the new Mac Pro completely destroyed, run a comparison with a W6800 duo. And the memory limitations are potentially an issue too for many as there are workflows that need more than 192 GB.

I think for this form of computing (desktop/tower), Windows/Linux PCs may still have the edge. What killed it for me was the PCI-e's underutilized potential. At the moment, the biggest advantage of the awesome Apple silicons are mini-PCs. Small computers that doesn't lack big performance. I see a lot of businesses in the future, moving towards small form factor which in the case of Apple, the Mac Mini. Awesome video Quinn.

Apple needs to bring back the Xserve with Apple Silicon. Imagine a silent, relatively low power, media ingest/render farm server with an M2 Ultra or Steve Jobs forbid, more than one M2 Ultra chips running ala NUMA Nodes. Top of the line media capabilities with top of the line efficiency. That's where I see ARM chips with dedicated media engines like these M-series chips shine.

they really need to bring back GPU Support, internal and external. These integrated GPUs are nice in the Performance per Watt regard but for anything serious they're to compromised.

I have one — maxed out — and I love it. I can't believe the speed and precision of the renders I'm getting. If I'd bought the Mac Studio with the same M2 Ultra I wouldn't have internal SATA ports, upgradable storage, eight thunderbolt ports or the ability to put afterburner cards or RAID storage into the PCI lanes.

This is what I said on MKBHD’s vid and I do agree! I don’t think this is a send off to the Mac Pro. I think this is just the result of getting extremely poor yield on the higher end M2 that they couldn’t feasibly put it in there even if it cost a lot more as they wouldn’t mind it. It would also explain why this machine missed the 2 year mark by a full year of probably trying everything to get the yield up. It honestly reminds me of AirPower and how, technically, they had it working, but they just couldn’t get most of them within the heat tolerance for mass production.

it would be nice to add a "lower is better", "higher is better" caption to your charts, cos not all of us understand these numbers.

Been waiting on this vid. The Mac Pro 2023 could have been somewhat of a solid value add at it's current state, if wasn't $3k just for the internal PCIe slots that are for storage expansion/IO or had a massive performance uptick and higher base specs than the Mac Studio. Since it's neither, it exists in for a very small subset of users. Hopefully, the Mac Pro finds its way to some sort of upgradability in the next iteration.

It is technologically possible to add additional RAM and dGPUs to Apple Silicon. I hope they’ll bring that to a future version of Mac Pro because there definitely are pro workflows that require crazy amounts of memory and graphics.

UAD plugins used to run exclusively on their external DSP processors, or their audio interfaces which included DSP, but a few months ago they added support for their plugins to run natively (causing quite the uproar for those who spent thousands on external DSP processors). Waves also allowed for this, you used to have to use their DSP servers for "live" audio processing (i.e external processing on a digital mixing console), but they also recently released SuperRack performer, again allowing you to do all your live processing on your laptop. Latency is definitely not as good as a dedicated DSP processor, however still in an acceptable range (5ms compared to less than 1ms) Companies are definitely keeping native plugin support alive, and it's probably due to the immense processing capabilities of newer chips.

I think you're right. The super early leaks were discussing the 'Jade-C' (M1 Max) 'Jade-C Chop' (M1 Pro), the 'Jade-2C' (M1 Max) and a Jade-4C. That leak was absolutely bang on the head except for the 4C. It just, never came. However, I'm not sure that the die size is the primary driver, given nvidia's GH100 is at 800mm2, my guess would be the interposer tech being a problem when working with 4 dies rather than 2. Adding to that, they've painted themselves into this corner by steadfastly refusing to support NUMA on macOS-- so they can't just break it out to a slower interconnect and bang another SoC in, without having to make big changes to macOS to prevent the performance tanking