OWC has released their OWC Envoy Ultra external Thunderbolt 5 SSD drive. They’re claiming Real World performance of:
TB 5: > 6,000 MB/s
USB4: 3,800 MB/s
TB 4: 2,800 to 3,800 MB/s
TB 3: up to 2,8000 MB/s
Thunderbolt 3 is only supported for Macs.
I’ve been waiting to see the specifications. I’m still waiting, because they don’t make any sense:
4 TB drive only has 1,9290 GB usable capacity
Has 3 TB 5 and 1 USB A downstream ports
Supports up to 3 displays
If this means that they cut and pasted the specs for some other product, then do we trust the system requirements?
Minimum requirement of macOS 14.x Sonoma or later
Connecting to Intel Macs requires Sequoia
Thunderbolt 3 connections require Sequoia
If true, that means I couldn’t connect it to my Intel Mac with TB 3 ports that tops out at macOS 13 Ventura.
(I want to buy a second external SSD now because I’m currently running booted off of my external SSD backup drive. But after I replace the iMac next year, then I’d be able to take advantage of TB 5. This drive won’t work for that plan if I can’t connect it.)
This sounds weird to me. They say so in several places on the product page, but I don’t understand why a drive can or should know or care about what it is connected to. I wonder if this is a case of “unsupported” or “incompatible”.
They also say that it can only be a boot device on Macs. Again, I don’t understand. Bootability on a PC should depend on the PC’s ROM BIOS/EFI firmware. I would think that some will work and some will not. But maybe this is another case of “unsupported” so they don’t have to deal with BIOS configuration that will vary greatly depending on the motherboard used.
Agreed. Those specs make no sense. Definitely content that belongs with some other device. It looks like these specs are for some kind of docking station, not for an external storage device.
This may also explain the TB3 support restriction. If those specs are meant for a dock with three downstream ports, then it would make sense that TB3 will either not work or have limitations.
I contacted OWC about the Envoy Ultra spec page. They fixed the capacity and removed the downstream ports and DisplayPort info, but the system requirements remain the same.
Yeah, and they still make no sense. Like saying that only APFS is supported on Macs or that only macOS 14 and later is supported (and only 15 on Intel Macs).
If these are real technical requirements and not just a statement about what they’re supporting, then there’s something really weird going on in this device.
More, Thunderbolt 5 offers excellent future-proofing in the event that one’s internal SSD malfunctions. Prices for Thunderbolt 5 drives will go down in time, so being able to eventually purchase a more affordable external to use as a boot drive clone backup means that one can get back to work fast if the mini’s internal storage suddenly dies.
Note that an Apple Silicon Mac will not boot without a working internal SSD. When booting from an external, it first boots into the internal, then boots from external.
True. But yhe 6-7 years since macOS required handshaking from the internal T-series security hardware/firmware hasn’t shown to be a problem for users – unless the internal SSD fails. And if that happens you’d be in trouble regardless of which drive is your boot drive.
Interesting how TB4/5 peripherals aren’t compatible with TB1/2 hosts, even with an adapter. I wonder what the reason might be. I could understand power delivery issues, since Apple’s adapter doesn’t provide power, but I’m truly surprised that a device with its own power supply (e.g. a storage array) isn’t compatible.
He points out an interesting macOS file caching behavior that can lead astray. Even using his own highly customizable Stibium, he was able to make it appear as if a M4P mini could over TB5 read 6 GB/s when in fact it really was doing much closer to 3.7 GB/s. But properly measuring that required unmounting/mounting a drive and restarting his tool to get macOS to flush the caches. I seriously hope OWC didn’t fall into that same trap. They happen to be marketing their brand new Envoy Ultra as reaching that same mysterious 6 GB/s over TB5.
OTOH, since we know TB4/USB4 can push north of 3 GB/s real world to/from fast NVMe flash, it is not out of the question to expect TB5 to be able to double that, under good circumstances at least, considering it has a twice as wide PCIe pipe at its disposal for that.
And other great TB5 article by Howard Oakley. He explains why you never saw 40 Gbps to a TB3-attached disk, why although you’ll see 6GB/s with TB5, you’ll also never see the advertised 80 Gbps, and why the marketing hype around TB5’s “120 Gbps” will not lead to 12 GB/s to/from an SSD. He also clears up where TB5 will really shine: b/w between dock and Mac - provided it works stable and as intended (he has some question marks there).
I think, as with all new technology, the “version 1.0” products that are being released now will have problems. Some may be fixed by firmware updates, but I think we can expect that it will be a year or two before devices all live up to the promises from Intel and Apple.
My 2017 Intel iMac is booted from a OWC Envoy Express, connected to the iMac’s TB3 port. Yesterday I cloned it to an OWC Express 1M2, connected to a USB 3 hub, and booted from it to turn on FileVault. Since the iMac doesn’t have USB4, this connection is at USB 3.1 (max 10 Gb/s). Update: actually was USB 3.0 at 5 Gb/s.
Does this article explain why the Express 1M2 connected to USB 3.1 (claimed 900 MB/s real-world) seemed to be faster than the Envoy Express connected to TB3 (real-world 1,553 MB/s)?
If not actually faster than the Envoy Express, it was certainly fast enough. I was surprised.
I suppose it is possible it was just faster because at that point, FileVault wasn’t on! This Mac doesn’t have a T2 chip. There’s no hardware acceleration of encryption/decryption.
The overhead from FV2 on external flash is small, typically several %.
However, what we perceive as “seems fast” is tricky. For one, the feel of using an external volume depends on a lot more than just peak b/w (IOPS for example is pretty important for a boot volume), but also, and perhaps more importantly, without actual benchmarking, it’s very easy to deceive yourself. Something BTW Howard also points out in his latest article on performance tuning.