The Studio Ultra has six TB4 ports.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
MP 14,8 Mac Pro 2023 is a scam
- Thread starter Xenobius
- Start date
- Sort by reaction score
Benchmarks are only good for showing how well a system can run said benchmark.
That only reinforces my point. Max Tech don’t have the system to test or review, so they’re in no position to call it a ‘scam’.
Not being able to beat a 13900k, a _consumer CPU_, is kind of disastrous.
Not being able to beat a 4080 or a 7900 is also kind of disastrous.
The only reason anyone would have to talk about a 6950 XT at this point is to make the new Mac Pro look good.
Because their points are only valid if they actively obtain a Mac Pro, benchmark and compare it, and demonstrate real world usage.
That doesn't answer my question. As far as I can tell my statement has nothing to do with the validity (or not) of Maxs statements.
Your original point was “Benchmarks are only good for showing how well a system can run said benchmark.”
So if they don’t have the system, how their opinions of the value valid?
Wow I know a lot of folks view some of what they put out as clickbaity, and I don’t agree with his theories here, but he did a good job of presenting a lot of the rumors etc that have been totally ignored by the rest of the lackey apple press. Bottom line, this is not a Mac Pro. It’s a Mac Studio pro, or a Mac ‘doh!
People need to start putting pressure on apple pointing out this machine is a failure and demanding, 3rd party gpu support, upgradable Ecc ram, and upgradable cpus.
They can because they're his opinions. That said my statement about benchmarks was wrt your statement:
...the GPU has a metal score that beats the Radeon 6950 XT.
Completely different discussion.
6950XT is old however - does it even work in MacOS? 6800 and 6900 yes but I thought the 6950 wasn’t supported?
If it could be compared with current generation GPUs like the W7900 I’m guessing the results would be different.
But we can’t because Apple refuses stubbornly to do drivers in macOS for those cards.
I used the 6950 XT as a baseline for comparison. And since that itself is an excellent card, the fact that Ultra scores higher suggests that the performance should be good.
But the fact remains that until real world feedback is provided, no one’s opinion including my own is gospel.
Is the 6950 properly supported???
See post above? Doesn’t seem like it.
I don’t think comparisons of unsupported GPUs (spoofed) are valid.
The M2 Ultra is damn fast and you don't need benchmarks for that. We know the results of the M2 Max and the M2 Ultra is more or less a bit slower than the 2x M2 Max. That's not the point. The point is that Apple used an engine from a motorcycle for a rally car. Simply put, Apple doesn't have a Workstation-class processor - the equivalent of a Xeon, Epyc, or Threadripper. Epyc had 128 PCIe4 lines as early as 2019, and as of last year 128 PCIe5 lines for a 1-processor configuration, yet there are multi-processor configurations available.
It is likely that Apple's accountants did not allow the creation of a Workstation-class processor, and unfortunately this pathetic pimple defiling the name Mac Pro was created.
It is likely that Apple's accountants did not allow the creation of a Workstation-class processor, and unfortunately this pathetic pimple defiling the name Mac Pro was created.
Further to that point, the old 6900xt’s scores (around 235000) beat the m2 ultras 220000 score. And the 69000 is decently ancient at this point.
No one's opinion should be considered gospel because it's an opinion.
IMO workstation class processors are in opposition to SoC class processors. It's my opinion the M-series speed is the direct result of the tight integration of everything in one package. Offloading functionality will slow it down, implementing certain features, such as ECC, will slow it down.
The M-series SoC is the result of years of experience with mobile devices and it fits well with portable and even mid-level systems. However for high end systems it appears to fall short.
It’s a shame that every new Mac update brings the wave of videos from the clown class on YouTube, most of whom would fail at a simple task like holding a sign on the side of a street advertising a restaurant. And yet people take the clowns seriously…
Yes and no. I can easily imagine additional 12-channel DDR5 memory in the expansion slots to complement the Mx processor-integrated memory and act as a kind of cache. But this would require a redesign of the memory controller. Adding support for multiple PCIe lines would also require major changes to the processor. Working out the optimal use of multiple graphics cards running on the slow PCIe5 (compared to transfers inside the Mx chip) could also be challenging and expensive. I think Apple unfortunately has come to the conclusion that it won't all be worth it to them.
Which is inline with what I wrote:
"Offloading functionality will slow it down, implementing certain features, such as ECC, will slow it down."
An SoC, with external functionality bolted on, can function as a workstation class CPU but it will not operate as the SoC by itself.
So to keep up with the new tech a PC user can replace the 6950 with a 7950xtx or with a future Radeon card but with the Mac Pro you're stuck with what's inside. To upgrade the CPU/GPU to an M3 or M4 gotta sell the whole thing and buy a new one.
May be I haven't make it clear enough.
AFAIK, ECC isn't mainly about "how much difference it make in daily use / specific task", but more on "the data integrity is guarantee".
No matter use ECC or non ECC RAM, any task "should be" finished flawlessly. However, ECC RAM simply allow the user to make sure the result is valid.
e.g. for a 100hours scientific computation. No matter a computer with ECC or non ECC RAM, both of them should give out the same answer (so, ECC or not should makes no difference on this specific task). But on the non ECC computer, the user has no way to make sure the answer is really valid. Even though ECC RAM isn't 100% bullet proof, but we know, in real world, we can assume there is zero undetected memory error on the ECC enabled computer.
Also, when the computer crash etc. ECC RAM allow the user to know if the problem is coming from the RAM straight away. On the other hand, the non ECC RAM computer user need to guess if there is any hidden memory error.
So, the difference is mainly data integrity / reliability, but not ECC can make any difference on any real world task.
Also, on an ECC machine you can check your logs to see how often ECC hit occurs, I believe.
I’m delighted Apple have brought out a Silicon Mac Pro… now I can at last make an informed buying decision… The very reason I didnt buy an M1 Ultra MS was because of the speculation around the M1 extreme Mac Pro now we know that isn’t happening at least not with 5nm… maybe 3nm will herald a M3 Extreme… I dont know… nor do I care that’s 18 months away… we’re at the crest of the M2 Ultra wave time to get on board… and buy a Mac Studio M2 Ultra….
Will Apple sell many M2 Ultra Mac Pros? probably not… but now we are going to see Mac Pro updates a lot more regularly then we saw in the past in 2-3 years time the Mac Pro will have morphed into something truely worthy of the name… or it’ll have been moth balled completely after Apple decides to double down on the Mac Studio and abandon the tower…
Who knows… but its not a scam, its a pretty decent base line tower… if that’s what you want.
Will Apple sell many M2 Ultra Mac Pros? probably not… but now we are going to see Mac Pro updates a lot more regularly then we saw in the past in 2-3 years time the Mac Pro will have morphed into something truely worthy of the name… or it’ll have been moth balled completely after Apple decides to double down on the Mac Studio and abandon the tower…
Who knows… but its not a scam, its a pretty decent base line tower… if that’s what you want.