Unfortunately you’re not getting true expandability
- if I’m counting correctly there are only 16 actual lanes of PCIe, maybe 24 but I don’t think so.All the slots are on switchers so as you saturate the routes you’ll lose performance quickly
Unfortunately you’re not getting true expandability
Tech specs are up but they don’t mention it anywhere:
techboards.net
Unfortunately you’re not getting true expandability
I think we’ll probably have to wait for stress tests of the PCIe lanes to see how bad things get under different loads and devices. I know LTTtech did one for the 10Gb mini and were impressed that Apple managed to effectively get everything its own lane and saturating the device didn’t result in any loss of performance. I can’t imagine that the same will be true here though. We’ll see. You’re right that for many people they may not notice it, but those people are likely to be happy on thunderbolt and the studio - a big draw of PCIe slots is you get the full bandwidth with the internal expansion and here all you’re really paying $3000 for is that the devices are on the inside … I personally can’t say that’s good value. I’m not saying it’s wrong to want that, devices on the inside, but it’s wrong, in my opinion, for Apple to charge that much for it.
SSD raids can saturate it. But let’s say you were right, then who is the Pro for? That’s the problem. It’s basically just for someone who would prefer to stick their devices inside instead of outside with no additional bandwidth benefits and wants to pay an extra $3000 to do that. Like for instance let’s say Apple changed the PCIe controller to allow normal mapping of memory so that under Asahi Linux I could use a dGPU for CUDA. This is an actual use case for me. In fact it’s even the ideal because I can plug in one GPU and can get the full bandwidth as long as I do nothing else. But I’d have to pay >$2500 (taking away some money to pay for thunderbolt enclosure over) over the Studio to get a little extra bandwidth and have the organizational convenience of it being internal. Even for me who might otherwise be considered an ideal use case, needs the internal bandwidth but can get away with just one device, that’s a tough sell - it is vastly overpriced.
It would be for, just as one example, this guy who's a professional music producer and sound designer - going from two older Mac Pros with cards stuffed into external PCIe boxes, to a rack mount Mac Pro with slots. And found happiness.
If I were still in the biz, I'd be using the new Mac Pro as a foundation for custom-designed multiple SDRs on PCIe cards used by customers for collecting and analyzing RF signals of interest. Using an external PCIe box would be, in a word, hokey. And off the table.
No doubt there are many dozens of similar applications for a Mac Pro with slots.
Of course, but really the only thing we’re waiting on is if the CPU/GPU got any kind of speed bump other than that it’s just an Ultra … it’s not going to increase the PCIe lanes available in the Pro relative to the Studio. Make no mistake I hope someone does stress test the PCIe setup under various load outs to see what it can handle and I’m sure someone will eventually will but that core limit of the number of PCIe is intrinsic because it is just an Ultra. In other words ideally such a test would also include an Ultra Studio and the costs of the two setups compared.
You seem to be conflating my argument with one that says the Pro should not exist. That is not where I am coming from. Perhaps my earlier question would have been better phrased as "then who is this Pro for?" The answers you gave are good answers to "are there users for any kind of Mac Pro at all?" In fact when @Joelist I believe phrased that very question these are the answers you gave and I agreed with them (still do). But that is fundamentally different from the question I am posing here. My question's context is not only are there enough use cases to sell this particular device to but also are there enough where those use cases would be willing to fork over an extra $3000 on top of the cost of the studio without additional benefit beyond internal vs external expandability? Yes of course to answer with absolutes we'll need hands on. But we can tell a lot right now and what we can see or at least what I can see is not good. A $7K starting point is unreasonable for what Apple is providing especially since they set the price of the studio at $4K. Again, I'm not criticizing users who want to move PCIe devices inside their case. That's perfectly valid - hell my own ideal use case would benefit from it! I am criticizing Apple charging those users who want that an extra $3 grand to do it with seemingly no other benefits that could possibly justify that level of price increase. Frankly even a speed bump wouldn't fully justify it though that would help a little. In essence, they've got a case and overall setup with this Pro that is wanting, in fact begging for, a far more powerful and connective SOC than what they've put into it, but they're still charging users like that's what they've indeed put in it. I don't want the Mac Pro line to die - I want a better base Mac Pro or for this base Pro to be cheaper with higher tiers. I've actually been impressed with the base prices of the Macs until now, but this one is ... just not justified by what we see.
So I looked it up and of course someone already made the thing I suggested could be made and it is indeed far less money:To some extent. COGS for the Mac Pro for sure would be more - whether floor standing or rack mount. Of course not $3K more. Perhaps Apple's GPM is higher because it's a specialty low-volume product rather than a commodity phone/laptop/MacStudio.
It wouldn't bother *me*in the slightest, though (with the caveat... if I were still in the business).
I wouldnot want tonever go back to designing SDR systems, or any system, using PC-based rack mount computers as a foundation. That $3K would simply be passed onto customers, with an appropriate GPM markup. And life would go on.
www.macrumors.com
