Upgrading an iMac Pro to 256GB of RAM Only Possible at Time of Purchase [Updated]

[tizeye]

"Pro" stands for "proficient"

So, non-pro models are NOT "proficient"? Proficient vs what...Microsoft??...surely not Apple.

 

[Unami]

I wonder if I can put this on my new Titanium Apple Credit Cardtm

View attachment 829037

One could buy quite a faux Mac or Windows machine for this handsome sum.

Well, you could get an apple watch for the 3% cashback then.

 

[konqerror]

If you need > 256GB of RAM, you don't need more RAM. You need a new workflow/process/application architecture. At some point, if you need ~60GB/s throughput for that much data, you probably need to consider a distributed and scalable version of whatever it is that you're doing.

Nope. Scientific computing where most of the problems are solving a large matrix. You have to solve the whole thing at once and it has to sit in memory at the same time.

Scalable Xeon gives you 128 GB/socket of memory bandwidth. Put that in a 8 socket system and you got 1 TB/s. The moment you to a different system, the best you can get with Inifiniband is 200 gigabit, 64% of a single socket. Plus latency is up by at least a factor of 10.

 

[Naaaaak]

> Be Apple
> Sell a "Pro" Mac
> Require total unit disassembly to install RAM
> Make sure only the BTO option gets the most RAM

They've become the dumbest computer company company that sells computers.

 

[Stephen13]

Hello, I am a time traveler from the year 1997. I just upgraded my Power Mac 8500 to a whopping 512MB of RAM.

Pardon me, but did you say...256 gigabytes of RAM? I have to believe that is a typo.

 

[nazaar]

My solution was to look outside of the Apple brand.
Apple has dropped the ball too many times for me to think they'll be reliable now.

 

[BigBoy2018]

Oh no! Limited to a plebian 128GB of RAM on the most RAM-efficient OS available! What ever will we do? /s

It does make me wonder what app or process could possibly use 128gb of ram, let alone 256. I'm not saying there isn't something that needs it, but I am curious what would call for such a hellcious amount.

 

[Precursor]

Will hold until iMac Pro comes with 512 GB

 

[bjet767]

I wondered why or what might require such a huge amount of RAM (not SSD). Someone wrote Adobe Premiere needed it, so I check what Adobe officially says.

After reading this I still have no clue, unless one needs to run multiple programs at the same time, why so much?

https://helpx.adobe.com/premiere-pro/system-requirements.html

Processor

minimum - Intel® Intel 6thGen or newer CPU

recommended - Intel® Intel 6thGen or newer CPU

Operating system
minimum - macOS v10.12 or later (v10.13 or later required for hardware-acceleration)

recommended - macOS v10.12 or later (v10.13 or later required for hardware-acceleration)

RAM
Minimum - 8 GB of RAM

Recommended:

• 16 GB of RAM for HD media
• 32 GB for 4K media or higher

GPU
minimum - 2 GB of GPU VRAM

recommended - 4 GB of GPU VRAM

Hard disk space

• 8 GB of available hard-disk space for installation; additional free space required during installation (will not install on a volume that uses a case sensitive file system or on removable flash storage devices).
• Additional high-speed drive for media

• Fast internal SSD for app installation and cache
• Additional high-speed drive(s) for media

 

[xgman]

This whole article probably applies to about two people, if that.

 

[adamw]

I like my iMac Pro, but unless I get a windfall, I won't be able to buy this 256GB version.

 

[AnakinMac]

Technical guess: 64 GB registered DIMMs are usually quad rank and I believe the Xeon W can't support them. I know they are not supported in similar PC platforms.

Part of the issue is the slim design of the iMac limits Apple to 4 DIMMs, comparable Xeon W workstations from other manufacturers have 8 slots with a max of 256 GB via 32 GB DIMMs.

I think Skylake W does support LRDIMMs at 64 GB but LRDIMMs require more power and cooling (and have more latency) so the system has to be designed for them. The 256 GB system probably has changes for that. The larger PC workstations are designed for airflow over the DIMMs.

Another guess is < 256GB model uses unbuffered ECC RAM, which can make the motherboard incompatible with each other.

 

[nguyenhm16]

The new 256GB RAM option consists of four 64GB memory modules, and in an internal document distributed to Apple Stores and Apple Authorized Service Providers, Apple states that the new configured-to-order iMac Pro is the only model designed to work with those 64GB memory DIMMs.

This information suggests that Apple Authorized Service Providers will not be able to upgrade an iMac Pro to 256GB of RAM, nor will third-parties like OWC. In other words, purchasing an iMac Pro with 256GB of RAM will only be possible via the CTO option on Apple's online store during the checkout process.

I'd like to see the document because the first sentence quoted above does not imply the second.

What the first sentence says is that those Apple supplied 64gb RDIMMs will only work on an iMac Pro. Ok, sure, that makes sense, I don't expect to be able to put those into, say, an 27" iMac which takes non-ECC SO-DIMMs.

As worded, it says nothing that would lead me to conclude that only the 64gb RDIMMs that Apple puts in will work, or even that only those RDIMMs are the only ones validated to work. But you can't expect Apple to say that third party RAM will work (even if it will).

Technical guess: 64 GB registered DIMMs are usually quad rank and I believe the Xeon W can't support them. I know they are not supported in similar PC platforms.

Part of the issue is the slim design of the iMac limits Apple to 4 DIMMs, comparable Xeon W workstations from other manufacturers have 8 slots with a max of 256 GB via 32 GB DIMMs.

I think Skylake W does support LRDIMMs at 64 GB but LRDIMMs require more power and cooling (and have more latency) so the system has to be designed for them. The 256 GB system probably has changes for that. The larger PC workstations are designed for airflow over the DIMMs.

The Intel Xeon-W marketing material says that you can use up to 128gb RDIMMs of some sort (for a total of 512gb of RAM), and like you said 64gb+ RDIMMs are usually quad-ranked:
https://www.intel.com/content/dam/w...-briefs/xeon-scalable-w-workstation-brief.pdf

However, it is fair to conclude that you may not necessarily be able to do that in an iMac Pro for firmware, cooling, power or other such reasons. So similarly, there may be technical issues with any random 64gb PC4-2666 RDIMM working, but that's a risk you take with any third party memory.

My guess is MacRumors misconstrued the document.

 

[edgonzalez32]

Ok, that's greedy. But I guess there is a hidden reason that makes sense.

Like what? It's a pro machine. Chances are professionals are buying it. I should be able to run out and swap out a faulty stick of ram OR upgrade the ram when I feel like it's time to.

 

[peter2]

Wow, 256 GB of RAM what a time to be alive!

Is there a significant performance difference between 128 & 265? Has anyone seen benchmarks? I'm very curious.

Try diagonalising a 2^N by 2^N matrix with a sufficient precision (e.g. to find the energy spectrum of a [sufficiently badly behaving] quantum physical system of N spin-1/2 particles). With 128GB of RAM you can go up to, say, 16 particles before you run out of memory while doing the computation. With 256GB you can go up to 17 particles. Is crashing your software and not getting results vs. not crashing and finishing a computation with an actual result difference enough? ;-)

By the way, even the 256GB maximum on RAM in a 2019 "Pro" workstation is a really bad joke. To get a perspective, our old Mac Pro 5,1, used for numerical computations supports (and has installed) 128GB of RAM. 10 years and $10k spent later, you can actually get something that would beat that ancient hardware... Sigh...

 

[kcslc]

I am building a PC for everything related to graphics and video, and my Mac Book Pro for music. Macs are not using Nvidia anymore. If you need so much ram and processing power is to do 3D, and Macs do not use Navidia and nobody is going to get an expensive Mac to then get an external eGPU! All those apps run way faster on PC anyway.

So... for daily stuff... I can use the Mac I have and for rendering Adobe Suit, and 3D... a PC. Always upgradable and way cheaper.

View attachment 829039

Neat! But can it play fortnite?

 

[avanpelt]

I wonder if I can put this on my new Titanium Apple Credit Cardtm

View attachment 829037

Yep, and you can get $475 cash back.

 

[1369281]

I like my iMac Pro, but unless I get a windfall, I won't be able to buy this 256GB version.

Humble brag.

 

[elmateo487]

I’m sure we’ll hear from many of them shortly. Just like all those who complain they need 32GB on the MBP.

I use all 16GB of ram on my MBP 90% of the time I use my MBP

 

[ToneWilliamsUSA]

It's almost impossible for me to support Apple any longer. I buy only its cheapest products any longer, not that I cannot afford more and want more, but rather they simply do not meet my needs. I hang on to Apple, barely, only with the hope that the next iteration of Apple will return to its roots of delivering the very best user experience at a modest premium. I demand excellence, and Apple simply no long delivers. It's so sad; it almost breaks my heart to see what Apple has become.

 

[ksnell]

In desktops, RAM is mostly needed for apps not OS.

I think is what he is trying to say is native Apple apps are very efficient. The second you run anything cross platform this tends to change drastically.

 

[bladerunner2000]

Ok, that's greedy. But I guess there is a hidden reason that makes sense.

To force consumers into buying a whole new Mac even when only one component is the bottleneck. Apple is greed obsessed.

 

[Jaekae]

For the price Apple wants, it must be gold plated .

Its server grade memory so not the same as the consumer memory you get if you for example build your own windows 10 desktop (or maby possible find that kind memory also but probably close in price too)

 

[zubikov]

Nope. Scientific computing where most of the problems are solving a large matrix. You have to solve the whole thing at once and it has to sit in memory at the same time.

Scalable Xeon gives you 128 GB/socket of memory bandwidth. Put that in a 8 socket system and you got 1 TB/s. The moment you to a different system, the best you can get with Inifiniband is 200 gigabit, 64% of a single socket. Plus latency is up by at least a factor of 10.

I get it, there are always ways to max out RAM usage: you can run 10 VMs instead of 5 simultaneously, or you can have 15 variables in a model instead of 10, or run an optimization algorithm on 10 billion records instead of 5. But at some point, wouldn't 512GB, 1TB, 2TB RAM simply become overkill for processor capabilities, not to mention any limitations imposed by OS? At what amount of ram does the processing power become a bottleneck? You will have to focus all efforts on writing incredibly well threaded apps for marginal returns. For true scalability, not incremental gains, I still think a better architecture money well spent. And yes, of course I'm guilty of generalizing every possible use of RAM in a desktop machine, and also I'm by no means an expert in the field.

 

[mikethemartian]

Hello, I am a time traveler from the year 1997. I just upgraded my Power Mac 8500 to a whopping 512MB of RAM.

Pardon me, but did you say...256 gigabytes of RAM? I have to believe that is a typo.

You have a freaking time machine and yet you are more amazed by the size of computer memory in the future?