Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Assuming the contents of your data and system fit comfortably on a single SSD, keeping them on a single drive seems easier and more foolproof. You can clone your drive, and in the event your SSD fails you can simply boot from that clone and have everything in the same state it was before the failure. No need to manage backups of two separate drives, or know whether an app is storing its data/configuration on your system vs data drive, which means you're less likely to inadvertently lose data since it's all backed up in one place. If you need more space, you can restore your clone to the new, larger drive. For a Mac Pro, you could upgrade to a larger drive via PCIe or external SSD (not sure if the Apple internal SSDs are upgradable though).
Except that you are increasing the likelihood of failure due to write wearout, and now you can’t replace the drive containing your data without going to Apple to do it because the T2 prevents it.
 
  • Like
Reactions: thekev
whats the difference between enterprise SSD and consumer SSD

It varies a bit depending on which particular SSDs you’re looking at, but generally two major differences are the behavior under sustained load, and reliability in case of power failure.

Consumer SSDs are optimized for bursty I/O or sequential transfers, because that’s what desktop applications generate. They may rely on an SLC cache, for instance, with poor performance once the cache fills. They may have poor random read performance because internal structures don’t fit in DRAM on the drive and have to be read from flash to locate blocks. They may thermally throttle if under continuous load. An enterprise SSD is designed to run continuously, 24/7, with small random I/O requests, and with the drive nearly or entirely full all of the time. Its life span is generally guaranteed assuming all data is written as random 4K writes. (And a longer guarantee may be available for sequential writes.)

A consumer SSD often has a chance of losing data on power failure; some increase reliability by limiting the amount of data in DRAM, or not using it at all. An enterprise SSD generally has an internal power reserve (supercapacitor) sufficient to flush its internal tables to flash if the power fails, with enough DRAM to index the entire drive so that random I/O requires only a single NAND access. (Typically that’s about 1GB of DRAM per 1TB of NAND flash.)

Some features of enterprise SSDs are less relevant here, like dual-porting so that the SSD can be attached to two controllers simultaneously, supporting encryption, or supporting extra per-block information used for end-to-end error detection. They may support SAS interfaces as well as SATA and NVMe, and often support much more advanced features of NVMe like virtualization and large numbers of command queues.

They also come in larger sizes, like 16 and 32 TB, though that’s more because consumers wouldn’t be willing to pay for that than because of anything unique to the design.
 
Except that you are increasing the likelihood of failure due to write wearout, and now you can’t replace the drive containing your data without going to Apple to do it because the T2 prevents it.

At least the drives are replaceable on the Mac Pro, unlike in the notebooks and Mac Mini where the NAND is soldered to the motherboard. Yes, you need an Apple service center to key the new drives to your T2 chip I am assuming, but it's not like you have to throw out the rest of the computer.

Also, I would assume the prospect is there of expanding the amount of storage connected through the T2 chip at a later date in the Mac Pro, since there are 2 replaceable NAND modules. (though I suspect most people will add something like a 970 EVO Plus using a PCIe card adapter instead)

One of the multiple reasons I haven't bought a 2018 Mini (after owning a 2009, 2010, 2011, and a couple 2012s) is the need to buy the lifetime maximum internal storage capacity you think you will need, as there is no changing your mind. (/looks over at his 2012 Mini with 3TB)
 
As many of you no doubt know, the earlier "cheese grater" Mac Pro (2008-2012) had "slots" for 4 internal hard drives. I use all 4 and over the years have upgraded many of them to SSD's. I don't understand whether the 2019 Mac Pro has any provision for me to plug these in somewhere? I know there's the up-to-8TB of Apple's "internal" SSD, but is there any way I could simply pop my existing SSD's into one of these 2019 Mac Pro's?
 
In a drive carrier/sled internally? Only option is via PCIe adapters. There’s a quad adapter from Sedna for SATA SSDs but pay attention to the specs:

USB-C 10Gbps multibay devices are actually faster than the internal SATA sleds of MP5,1 for most stuff. Been testing this one and getting almost the full theoretical max speeds when using with an MBP16,1 via USB-C with proper cable. Need 2.5”>3.5” adapters with this one.
 
As an Amazon Associate, MacRumors earns a commission from qualifying purchases made through links in this post.
Strange that a $500 consumer motherboard is more future proof with PCIe 4.0 than 2019 Mac Pro.

The Mac Pro is not a consumer computer. it's an enterprise computer. And you'll notice most, if not all, of the current generation of PC OEM enterprise computers - many that cost multiples of the $52K Mac Pro, much less the $6K model - all use PCIe 3.0.
 
  • Like
Reactions: J.J. Sefton
As many of you no doubt know, the earlier "cheese grater" Mac Pro (2008-2012) had "slots" for 4 internal hard drives. ... I know there's the up-to-8TB of Apple's "internal" SSD, but is there any way I could simply pop my existing SSD's into one of these 2019 Mac Pro's?

There's the two 3.5" drive sled from Promise that Apple is selling. Promise also makes a PCIe card with four spinners.
 
I can pick up a Samsung 860 Pro 256GB for $85 and a 512GB for $149.
Why is Apple still charging $200 to go from 128 to 256, and $400 to go 128 to 512GB? It’s absolutely insane!

256GB should be starting size and 512Gb upgrade $100 or $149 tops. I don’t know if Apple still uses Samsung drives, I know Dell does at times but at least be reasonable!
 
  • Like
Reactions: mrfezzywig
Except that you are increasing the likelihood of failure due to write wearout, and now you can’t replace the drive containing your data without going to Apple to do it because the T2 prevents it.

Whether write wearout is a concern, drives failing for any number of reasons is nothing new, so your data is only as safe as your backup/recovery system anyway.

As far as T2 Macs go, for me its security benefits do not outweigh the potential downtime if/when the Apple drive fails, so first thing I do is enable booting from external media. Even though I prefer keeping my system and data together on one drive for the reasons I mentioned, I'd enable external booting regardless. When the Apple drive fails you can then boot from a clone and continue work in minutes.

One of the multiple reasons I haven't bought a 2018 Mini (after owning a 2009, 2010, 2011, and a couple 2012s) is the need to buy the lifetime maximum internal storage capacity you think you will need, as there is no changing your mind. (/looks over at his 2012 Mini with 3TB)

Totally agree. I'd much rather go the opposite direction and get the minimum Apple storage in the Mini (128GB) and instead boot from an external Thunderbolt 3 SSD (like the Samsung X5) with the size I need now. You can then upgrade it later if needed, they're a similar speed, and they're cheaper per GB than spec'ing up the Apple SSD.
 
you'll notice most, if not all, of the current generation of PC OEM enterprise computers - many that cost multiples of the $52K Mac Pro, much less the $6K model - all use PCIe 3.0.

Same reason: Intel doesn’t support PCIe 4 yet.
 
Can you provide a link to your assertion the T2 is basically an A10? If memory serves, that was an iFixit claim from a year or two ago, that was later recanted by iFixit.

I suspect it's actually an Apple-designed full-custom ASIC. And thus does not limit SSD performance.


Given that it has an exact package size as A10, it is highly likely T2 is just those A10 that wasn't good enough and binned for T2.

DMI 3.0 is PCIe 3.0 x4. That’s the limiting factor, not the T2.

And if I were Intel I’d be targeting PCIe 5.0 with 7nm. 4.0 can be skipped. Apple could dump Intel at the high end but I wouldn’t hold my breath.

Yes, DMI 3.0 is PCIe 3.0 x4, but again DMI is a specific Intel proprietary Interconnect. No one apart from Intel uses it. The T2 is connected via normal PCIe Lane.

Icelake will get PCI-E 4.0, Intel's last schedule was for Icelake-SP and Cooperlake-SP to be launched in late 2019, looks like this isn't going to happen. The 7nm Saphire Rapid will come with PCI-e 5.0, that is now likely to be 2022 along with DDR5 support, something AMD will be doing in 2021.
 
I can pick up a Samsung 860 Pro...512GB for $149. Why is Apple still charging $200 to go from 128 to 256, and $400 to go 128 to 512GB? It’s absolutely insane!

Because it's a workstation and workstation prices are...insane...I guess. And it's not just "insane" at Apple - HP and Dell both also charge $500-plus for 512GB of M.2 style SSD storage on their workstation families.




I believe a major point of the disconnect on Mac Pro pricing is many/most people are approaching it from a consumer component standpoint and not an enterprise one. For those of us who use enterprise-level products (of which the Mac Pro is very much), we're used to these kinds of prices and we acknowledge (if grudgingly) that this is what enterprise-level components and systems cost.

And since the enterprise is probably one of the few areas the component and system OEMs are making money (considering how cut-throat the consumer section is which is why the prices are so much lower in comparison) it is not surprising that is where they are charging the premiums so prices for components like CPUs are far higher now then they were a decade ago when the 5,1 was around even accounting for inflation which contributes to the massive spike in base configuration pricing for the 7,1 compared to the 6,1 and, especially, 5,1.
 
Last edited:
Given that it has an exact package size as A10, it is highly likely T2 is just those A10 that wasn't good enough and binned for T2.

Yes, DMI 3.0 is PCIe 3.0 x4, but again DMI is a specific Intel proprietary Interconnect. No one apart from Intel uses it. The T2 is connected via normal PCIe Lane.
Sure, DMI is proprietary to Intel. I haven’t said otherwise so it hardly bears repeating (and it’s not particularly relevant). And yes, the T2 is connected via standard PCIe 3.0 lanes.

But the T2 doesn’t connect directly to the CPU. It connects to the PCH, as do a bunch of USB ports, WiFi, and other ports depending on the particular model of Mac, including GbE, SATA and even an SDXC slot in the case of iMac Pro.

There’s a ton of bandwidth available in the T2 on-chip; for instance recall two flash modules are connected to it. But all that I/O has to funnel down to the four PCIe lane-equivalent choke point that is DMI 3.0. Again, the T2 itself is not the limitation; the issue is the DMI connection.

If Apple wants to increase the bandwidth available to SSD storage, PCIe 4.0 is not necessary. Rather, dedicated PCIe lanes, assuming they are available, can be connected directly from the CPU itself to the T2, thus avoiding the bottleneck caused by congestion/saturation of an oversubscribed DMI bus.

re: A10=T2, I agree it’s extremely likely that Apple simply(!) added additional functionality to the A10 to create a dual-use SoC that can be utilized either in its original role in iOS devices or as the T2 for Macs. After all, transistors are cheap, additional pin-outs relatively so, and it’s an excellent way to leverage the design and manufacture of silicon that otherwise already exists.

But there’s no particular reason I can see to think T2s are plucked from parts not good enough for A10. For all we know, T2 are the highest bin; certainty relatively few are needed for that role in any case.
 
Sure, DMI is proprietary to Intel. I haven’t said otherwise so it hardly bears repeating (and it’s not particularly relevant). And yes, the T2 is connected via standard PCIe 3.0 lanes.

But the T2 doesn’t connect directly to the CPU. It connects to the PCH, as do a bunch of USB ports, WiFi, and other ports depending on the particular model of Mac, including GbE, SATA and even an SDXC slot in the case of iMac Pro.

There’s a ton of bandwidth available in the T2 on-chip; for instance recall two flash modules are connected to it. But all that I/O has to funnel down to the four PCIe lane-equivalent choke point that is DMI 3.0. Again, the T2 itself is not the limitation; the issue is the DMI connection.

If Apple wants to increase the bandwidth available to SSD storage, PCIe 4.0 is not necessary. Rather, dedicated PCIe lanes, assuming they are available, can be connected directly from the CPU itself to the T2, thus avoiding the bottleneck caused by congestion/saturation of an oversubscribed DMI bus.

re: A10=T2, I agree it’s extremely likely that Apple simply(!) added additional functionality to the A10 to create a dual-use SoC that can be utilized either in its original role in iOS devices or as the T2 for Macs. After all, transistors are cheap, additional pin-outs relatively so, and it’s an excellent way to leverage the design and manufacture of silicon that otherwise already exists.

But there’s no particular reason I can see to think T2s are plucked from parts not good enough for A10. For all we know, T2 are the highest bin; certainty relatively few are needed for that role in any case.
If it was an a10 Apple would say so. The fact that the package has the same dimension is meaningless.
 
If it was an a10 Apple would say so. The fact that the package has the same dimension is meaningless.
It wouldn’t mean much with respect to the theory of a T2/A10 dual use part, but it seems it might indicate the need for a similar number of pins, or similar power dissipation? An x-ray would be nice. But I disagree that Apple would necessarily be forthcoming.

When you look at the functionality of the T2 compared to an A-series SoC, there is a lot of overlap: CPU; SMC; RTC; PMIC (off-chip?); i/f for camera (and basic ISP and HEIF/HEVC encoding); audio I/O controller and associated DSP (EQ, echo cancellation, active crossovers for Mac); Boot ROM (iPad/iPhone) and secure boot (Mac); TouchID/Secure Enclave; storage controller and NAND chip (module for Mac) i/f; AES-256 crypto engine; Mx co-processor for Hey Siri functionality; and others.

Despite my snark about T2 being higher-binned A10, it could be that A10 with defective CPU and/or GPU cores could be essentially “free” T2.
When/if FaceID comes to MBP and iMac, at least an A11-generation part would be needed for neural network/ML, which would give MacOS another co-processor with some interesting capabilities.

Duo security has done some work looking at the T2:


Blackhat also has a fairly interesting presentation available on the T2, the secure boot process and possible avenues to exploit it:


Looking at a block diagram of the iMac Pro south of the PCH it seemed to me apparent that the T2 could easily be an A10 SoC, with some Intel-specific glue added to the the actual A10. Or perhaps it’s just based on the A10, with many chunks re-used, but designed specifically for T2.

787D7EDE-2D3A-44F0-992D-F2FDD36B2D19.jpeg


Anyway, as I have only a rudimentary understanding of the subject, I’d be interested in hearing from an actual semiconductor design engineer/CPU architect, in his copious free time :)
 
Last edited:
If it was an a10 Apple would say so.

Apple doesn't even say how much RAM iOS devices have. They don't want us to care about the relationship between the Tx and the Ax.

It's clear that those are related CPUs, at the very least, and given the high similarities, it's very possible that the T2 is just a rebadged A10.
 
But what would that 8TB SSD be priced at if Macsales had billions of dollars a month in operating expenses, including $1.5 billion per month in R&D?

People should keep in mind that the cost of components is only one of the factors that go into the selling price of Apple products. They’re a large company with large expenses. Compared to the total of other costs of sales plus operating expenses, BOM cost is the smaller of the two.
 
So a 1TB NVME to upgrade my 2015 MacBook Pro cost me $87

Using an OWC new PCIe adapter for $399. I can put 4TB in aMac Pro for $748

8TB for $1080.

read is 6000mb/s and write is 7000mb/s

You’re telling me the largest company in the world can’t match that price??
[automerge]1577189456[/automerge]
But what would that 8TB SSD be priced at if Macsales had billions of dollars a month in operating expenses, including $1.5 billion per month in R&D?

People should keep in mind that the cost of components is only one of the factors that go into the selling price of Apple products. They’re a large company with large expenses. Compared to the total of other costs of sales plus operating expenses, BOM cost is the smaller of the two.

No. Apple is the worlds largest company. They were making iPods cheaper than any of their competitors in 2005,6,7 because they used their buying power to get unprecedented deals on solid state memory.

Youre telling me that now, 14 years later they suddenly can’t compete anywhere near on price with companies whose only business is retail SSDs.
Give me a break.
 
So a 1TB NVME to upgrade my 2015 MacBook Pro cost me $87

Using an OWC new PCIe adapter for $399. I can put 4TB in aMac Pro for $748

8TB for $1080.

read is 6000mb/s and write is 7000mb/s

You’re telling me the largest company in the world can’t match that price??
[automerge]1577189456[/automerge]


No. Apple is the worlds largest company. They were making iPods cheaper than any of their competitors in 2005,6,7 because they used their buying power to get unprecedented deals on solid state memory.

Youre telling me that now, 14 years later they suddenly can’t compete anywhere near on price with companies whose only business is retail SSDs.
Give me a break.

The answer is simple. Switch brands and reward computer companies with your currency that give you the value you are seeking. Will you do that?
 
So a 1TB NVME to upgrade my 2015 MacBook Pro cost me $87

Using an OWC new PCIe adapter for $399. I can put 4TB in aMac Pro for $748

8TB for $1080.

read is 6000mb/s and write is 7000mb/s

You’re telling me the largest company in the world can’t match that price??

Your 2015 MBP doesn’t do 6000 MB/s speed.

But yeah, Apple could match that price. But why? They’re not a brand for price-conscious customers. And BTO options on a high-end workstation are never particularly good deals. You can just put your own SSD in. As PCIe card, in an internal bay, externally through Thunderbolt, or thorough USB… tons of options.

No. Apple is the worlds largest company. They were making iPods cheaper than any of their competitors in 2005,6,7 because they used their buying power to get unprecedented deals on solid state memory.

No they weren’t. You could get players with flash memory and a screen for the price of a shuffle.
 
You’re telling me the largest company in the world can’t match that price??

Sure they could.

But they don't have to (people will either pay it or go third-party) so why would they?

I mean HP, Dell, Lenovo and others could charge a hell of a lot less for their workstation RAM and storage, too, but they don't for the same reasons.

Shareholders would not be happy with lower revenues dropping the value of their holdings and it's all about "Shareholder Value" nowadays.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.