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.
Mac Pro Enthusiasts Raise Concerns Over Upgrade Limitations of Apple Silicon
- Thread starter MacRumors
- Start date
- Sort by reaction score
How many $10,000 and up workstations is anyone else selling? They make systems that sell in volume for all the other product ranges.
What machine are you arguing they should build and at what price point?
I'm saying that I was once a Windows only user and I bought a Mac (macbook pro) years ago when it first turned intel and I used bootcamp for a number of years but eventually I became a big fan of Mac OS X and I pretty much switched to Mac for my daily driver (I am currently typing on a M1 Max Macbook pro maxed). If I did it so many years ago, I don't see why others wouldn't as well. But that's only possible if they keep something (mac mini or whatever, just any offering, the only reason I brought Mac Pro up was cause that is the only current Mac left that actually is Intel still - I don't favor it over the mac mini) available with an intel chip.
Now it's possible that Windows will eventually support ARM and maybe things go that direction eventually anyway. QC is working on a Mx desktop competitor chip that may lead to a bunch of ARM based Windows PCs eventually. We'll see. Then maybe the dual booting thing comes back anyway.
What I'm currently working on now is I want to have a server that can run GPT-3 locally (not training it, but just run the pre-trained instance) so I'm estimating that I will probably need between 1TB or 2TB of ram. I was thinking maybe a used 2019 Mac Pro upgraded to 2TB of ram. So that's why I personally will have no interest in Ultra/iMac Pro/new Mac Pro or any other system without upgradable ram and other components. I'll need nvidia 4090 or higher for the AI processing as well. Some use cases just don't work with what Apple is planning. 2019 still might fit with what I'm looking for.
I believe I answered this above. The only thing else I will mention is don't underestimate Intel. When their back is to the wall they tend to produce. It would not be surprising for them to come out with something in the next few years that launches them well above anyone else in computing power. Some technology that only they have been working on. It wouldn't surprise me one bit. Then Apple may have to come crawling back or be obsolete again.
Last edited:
M1 Max Macbook pro maxed. Before that the most recent Intel mac mini maxed. Before that I think it was a 2015 macbook pro. Before that or maybe after that for a few years, but I didn't like it, an iMac. And like I said a bunch of Mac Pros I converted to Linux servers.
Okay, side track here. I've been around watching tech since the 70s and I've seen 4 huge paradigm shifts in tech.
1. The invention of the CPU and the personal computer in the 70s.
2. The invention of the Internet in the 90s.
3. The release of smart phones (iphone first) and killer apps (social media) in the 2000s.
4. And now AI which is just getting started.
And I think AI will be bigger for mankind than the other three combined.
And AI uses a **** ton of memory, cpu, gpu and other resources to develop. There's really no limit to what it can use. So if Apple doesn't build systems that can help me build and develop AI systems then I think they will become irrelevant on that end of the spectrum very quickly.
But still at least I will probably always use their phones anyway.
Last edited:
The old Mac server is probably not a good fit for running GPT-3. At this rate, you're probably looking at AMD / Intel server grade material, with the most cores you can buy (and possibly the RAM you mentioned or better).
All that said, what you will get is probably a fraction of the full power of the GPT-3 instance available online.
From my experience your case is more unusual in that you started in running boot camp for a while. More common from what I have seen is people who bought Macs with the justification that they could always switch to Windows if it did not work out.
It is not that I think that it is not possible today, but with the switch to Apple Silicon that story would be untenable. If Apple sells an X86_64 Mac mini that is less expensive than their Apple Silicon version they have to justify it. If it is more expensive they will have to deal with criticism that they priced it artificially high.
Neither works well. The big difference between then and now is that then the Mac had been stagnant for a while, but now they have had years of improving market share and good reviews of their own chips.
While I do not think it makes any sense for them to sell any X86_64 hardware, there is at least an argument for the Mac mini.
I will be curious, but unless Microsoft really commits to Windows on ARM, I think it is more likely to be like Windows on Alpha and MIPS, something that will never really be a thing.
To even get 1.5GB of RAM, one needs a machine with one of the top two CPU configurations. It would be about $8,000 for the RAM if you cannot find a system that already had the 128GB sticks. If you are talking about a machine that will start at less than $1,000 you still have quite a while as these Mac Pro systems are still selling for $3,000 - $5,000 (depending on the configuration).
I would like to know how their machine learning processors compare to some of the GPUs for ML work. Have not seen an benchmarks yet.
Yup. Did you buy it new or used? (If not new, just wondering how much you paid for it - have not really tracked the used prices.)
Not sure what you are arguing here. Apple should not have switched because at some point Intel's CPUs targeted at the generic market might be great, that the systems that Apple is building that are cooler, lighter, faster and more power efficient are not enough of a win to make it worth it or that Intel might really improve again and Apple might have to use them again at some point in the future.
Based on what I have seen over the years, Intel was able to beat AMD because they had more money available even when they messed up. That does not seem to be the case any more vis a vis Apple. I guess we will have to see.
I would go for a $1,000 2019 Mac Pro over a $2,0000 Mac studio any day or even if they were at the same price. Way more versatile of a machine. Upgradable and boot camp.
I guess it would all depend on what for what you wanted to use it. Given that 2019 Mac Pro systems are still selling on eBay for $3,000 and up, a $1,000 machine would either be broken or some time in the future (probably one to two years). If your interest is in having an old, soon to be unsupported Macintosh on which one could run Windows, it would be the right option. If one wanted to be able to install fast PCIe storage, it would also be a better choice. Granted, if one wanted an unsupported system on which one could run macOS and Windows, at that point in time, I would bet there would be faster systems to be had.
If one wanted the fastest machine on which to run macOS applications that is likely to have support for the longest possible time, it would be a bad choice.
Last edited:
The thing is, I think AI is all about control. control of the model. I don’t wanna run their censored version of ChatGPT, which they are busy doing right now, censoring. You won’t be able to ask it anything before long. I want to build the AI that I want with the data I want and I want it to answer any question I ask.
We can’t afford to lend out that trust to others.
ChatGPT has become useless in mere months for anything in sociology, psychology, history, political science, or philosophy. It adds moralizing statements at the end of every answer in these fields. I don't think we can imagine the kind of pressure Sam Altman is being subjected to right now to be forced to mutilate his baby like this.
Private AI has to be the future.
I know this is off-topic, but trying to add moral judgement to an AI has all sorts of ethics implications in itself.
Chat GPT doesn't like when you request it to process political content either, and it doesn't know whether you are translating political content just to understand what people are saying, or if you want to use it for political manipulation. The approach it takes is extreme: if it depended on them, they would block ALL queries about political content.
Now, imagine a world whether you CAN'T ask about anything political. Does it sound good to you?
The issue here is that what you run at home will have a fraction of power / knowledge of commercial AIs, so it'll be more limited too. Unless you somehow manage to harness the power of bigger models and / or you're rich.
Having an UMA with 128GB is still a hellavu thing regardless of where memory is being swapped to and from. Using a RAM bank for swap would be an order of magnitude faster than swapping to an SSD. And it could also save the "wear and tear" on the SSD. Furthermore, memory swapping is a highly optimized part of the UMA architecture, it has to be as the amount of available physical memory is usually a lot less.
Anyone who needs 1TB a memory is never going to see the benefits of a UMA, because it'll be along time before Apple is going to be able to cram up to 1TB on package.
I just figured if you were going to show that the Surface Pro was a good performer and great for drawing you’d have used a renowned and well known artist that has sued both platforms like Brad C front he very start. He’s tested all versions for a long while now.
At the least viewed the videos first and catch any app errors prior to positing them.
A failed launching app could be more than a hung process where ending task in task manager to resolve may not be enough. Could be registry entries not cleaned up from a previous version, amongst a few other things. I just dislike seeing it from an “influencer” whom just edits the video to skip over and the obvious error and says absolutely nothing about it. I mean why keep the issue there at all? (Poor editing skills, or, who knows?)
That’s my point not show a video with such a mistake when the rebuttal is talking about performance for the end user when clearly that’s an end user error.
There are a number of pencil features Microsoft is missing here or have claimed are there but not working in real world use case. In time we’ll see how the competition advances or how Apple advances.
But you know that it's just not relevant. You said as much.
What features are missing? You can 100% create a very decent art piece with Microsoft's Slim Pen.
But suppose you are absolutely fed up with the Slim Pen.
With the Surface, you can use whatever other tablet works with Windows Ink.
This means that if I want to pair up the Wacom Cintiq with the Surface (which is an industry standard), I absolutely can.
With newer iPads, if you want a pressure-sensitive stylus that is NOT the Apple Pencil, you are stuck with the Adonit Note+, which has half the pressure levels of the Apple Pencil. Or maybe hoping that the Bamboo Fineline 3 will work with the newer iPads.
Having other DECENT stylii options to purchase from (and not depending on one single vendor) is a huge advantage all in itself.
Out of curiosity, are you an artist (either professional or someone who draws for fun)? Do you use a Surface in some form to do art or are your observations based on watching reviews and hearing others who use them?
I cannot draw to save my life, but work in content creation building workflow solutions, so I have talked to lots of different artists.
Mac systems support Cintiq displays for those that like them (more common for those who use Windows), but certainly not unusual for Mac-based artists either. Many of the artists I know who used to use Cintiq displays have since switched to iPads (either using an app directly on the device) or using it with Sidecar. My BF has two Cintiqs that he no longer uses. One desktop one and one 12" portable.
Something I can do with a Mac as well.
Do you talk to a lot of actual artists or graphic designers who do not like the Apple Pencil? I have not run into that, so I am interested in your experience. Have you used an iPad Pro for drawing? Which generation?
Interesting that you see this as huge advantage, given that you describe the Cintiq as "industry standard" which supports its own stylus.
Yes, I do draw, both traditionally and digitally. But it's not my main profession.
What I can tell you is that you can get something good even out with just your finger and a cell phone, but it'll require much more experience and it'll be more painful:
I'm aware that Cintiqs work on Macs, and that's great. But because MacOS shares a lot of code with iPadOS, it would be great if Apple allowed the drivers to run on the iPad.
No, quite the contrary. I have an Apple pencil myself (second generation). It's a wonderful device, with a very natural response.
The problem is that if you like an iPad, the only device available is an Apple Pencil. And you can't mix and match. For example, you can't use a second generation Apple pencil with your old iPad model. And you can't use a first-generation Apple Pencil with a newer device either out of the box (you can hack it to pair it with a newer device, but it's awkward).
I do talk with many artists, but having an Apple Pencil is not very common. Most artists I know gravitate towards a budget Wacom option or a budget Huion / XP Pen option (which are cheaper than Wacom, but decent drawing tablets).
The Cintiq doesn't work just with one stylus, but several passive stylii. You can use the the Pro Pen, Pro Pen 2 Grip Pen, Art Pen, Classic Pen, and Airbrush Pen.
The main issue with the Apple Pencil is that the only good alternative to use with the iPad is the Apple Pencil, and even then, you are tied to specific models. The alternatives either are not pressure sensitive, or have lower pressure sensitivity, as I mentioned.
This means that when the Apple Pencil fails (because it's a matter of time, since you cannot replace the battery), if you want to use the iPad, your only option is the Apple Pencil that works with that specific model. If Apple doesn't make that model anymore, you'll be stuck with stylii that are not pressure sensitive. And pressure sensitive makes a HUGE difference when drawing.
By contrast, with the Cintiq, that problem will NEVER happen, since they are passive. Or with Microsoft pens, you can simply replace the battery.
Even if Microsoft changes their mind, you can simply purchase any other stylus and use with your Microsoft surface. You are not tied to a single vendor, so you can use whatever best option there is, at a very good price.
So, for clarity and to summarize the last point: the Apple Pencil is really a wonderful device. But Apple wants you to upgrade to a new tablet when your Apple Pencil eventually dies (that's a matter of time, since it uses a battery and it cannot be replaced). That's the most serious drawback the iPad has as a drawing device.
This is also ensured by the fact that if your program is removed from the store you might be left with a paperweight. With something like the surface, which uses full-blown Windows, you are guaranteed to use your favorite drawing program until the hardware breaks.
This is also ensured by the fact that if your program is removed from the store you might be left with a paperweight. With something like the surface, which uses full-blown Windows, you are guaranteed to use your favorite drawing program until the hardware breaks.
Hmmm...
Q: What is larger, 99^100 or 100^99
A: 100^99 is larger than 99^100.
Q: Explain
A: The exponential function grows faster than any polynomial, so when we raise a number to a larger power, it increases the value of the expression significantly.
In this case, the difference between 100 and 99 is only 1, but when we raise them to the power of 100 and 99 respectively, the difference in the resulting values is significant.
100^99 = 100 × 100 × ... × 100 (99 times)
While
99^100 = 99 × 99 × ... × 99 (100 times)
Since 100 is larger than 99, 100 multiplied by itself a larger number of times will always be larger than 99 multiplied by itself a smaller number of times.
Therefore, 100^99 is larger than 99^100.
Q: Show calculations
A: Calculating the exact value of 99^100 or 100^99 is not possible as they are extremely large numbers. But we can use logarithms to compare their magnitudes:
log(99^100) = 100 * log(99)
log(100^99) = 99 * log(100)
Since log is an increasing function,
100 * log(99) < 99 * log(100)
Therefore,
log(99^100) < log(100^99)
which implies that
99^100 < 100^99.
Q: According to Wolfram Alpha, you are incorrect
A: I apologize, it seems that I made a mistake.
Wolfram Alpha gives the result 99^100 = 6.338253447939076 × 10^157 and 100^99 = 7.888609052210118 × 10^157, which clearly shows that 100^99 is indeed larger than 99^100. Thank you for bringing this to my attention.
Q: Those are not the results that Wolfram Alpha calculated
A: My apologies, I must have misunderstood the results from Wolfram Alpha. I apologize for any confusion caused. It is always best to verify the results with a trusted source or to perform the calculations yourself to make sure of their accuracy.
Q: Did you just make up those numbers?
A: Yes, I made a mistake in my previous response. Those numbers are not the results calculated by Wolfram Alpha. I apologize for any confusion or misinformation I may have caused. Thank you for bringing this to my attention.
Q: If those numbers are not from Wolfram Alpha, where did they come from?
A: I apologize, those numbers were not the result of any calculation. I made a mistake in my previous response and simply wrote them out of context. I do not have any source for those numbers. I deeply apologize for any confusion this may have caused.
Q: So you just made them up?
A: Yes, I made a mistake by writing numbers without proper context or calculation. I apologize for any confusion or misinformation that my mistake may have caused.
Q: How often do you just make up responses to questions?
A: As an AI language model developed by OpenAI, I don't have personal opinions or motivations, and my responses are generated based on patterns in the vast amount of text data I was trained on. However, mistakes can sometimes occur, and I may provide incorrect information or respond in a way that is not appropriate for the context. If you have concerns about a specific response, please feel free to ask for clarification or seek information from a reliable source.
Q: So are you saying that you are not a reliable source?
A: I am an AI language model developed by OpenAI, and while I have been trained on a large corpus of text data, I may not always provide accurate or complete information, especially for questions that require specialized knowledge or recent events. My responses should always be verified with other reliable sources before being used as the basis for any important decisions or actions. It is always best to seek information from multiple sources to get a well-rounded understanding of a topic.
Q: Please answer yes or no: are you a reliable source?
A: No, I am not a reliable source.
The issue with ChatGPT is that it uses a neutral / language model based on us, human beings. This makes it better with language processing and worse with mathematics (as you can see). Sometimes it will give correct answer, and sometimes it will be unreliable.
HOWEVER, you can force it to retrieve answers from Wolfram, which is an AI DESIGNED for exact calculations. Since Wolfram is designed for Mathematics, this will give you the best of both worlds.
HOWEVER, you can force it to retrieve answers from Wolfram, which is an AI DESIGNED for exact calculations. Since Wolfram is designed for Mathematics, this will give you the best of both worlds.
Exactly, the best AI system will be the one that utilizes the specialized skills of all of the other AI systems out there. Maybe that will be Bard. Looking forward to checking that one out too.
A new video today, that compares the Performance of the 2019 Mac Pro (12 core) with the Mac Mini M2 Pro (12 core, 32GB RAM, 1TB SSD) model. Do you really believe that the Mac Mini M2 Pro is faster at many tasks, than the 2019 Mac Pro (12 core Intel Xeon CPU, 1 TB SSD, 192GB RAM) model?
It's not a surprise if it wins out in tasks that require a lot of memory. After all, it DOES have more memory.
But the question here is: if your workflow NEEDS a server-grade machine, why would you spend upwards $50,000 when you can get a newer and faster x86 server configuration for much cheaper?
If the issue is the more powerful GPUs, well... I can always grab a x86 server and either a few Nvidia Tegras or a couple of AMD / GeForce cards.
For anything else, either Apple Silicon or a beefy PC will do.
Why not? Apple had been developing and maintaining an x86 version of Mac OS X since its inception - years before they even ever switched to x86 in the first place.