OWC's Thunderbolt 5 External SSD for Latest Macs is Impressively Fast

[jphall]

I like to hear the result of the 1M2 drive. It maybe good enough? But twice the speed for the TB5 drive may still be worth it if I even spend a good chunk of money for the 1M2 with fast SSD.

I did a few tests, using the drives connected to the MBP M4 Max. I used a short Thunderbolt 5 cable for the 1M2, and the attached cable on the Envoy Ultra.

That Ultra is the fastest external drive i've used, but the real test will be in video editing which i'll be doing over the next few days.

 

[OWC_TAL]

Roughly 1 GB/s write, while read is around 2.8 GB/s. Should be 3 GB/s in both directions. This is the 1M2 with 8.0TB OWC Aura Ultra IV connected to a 16" MBP M1 Max with an 40 Gbps OWC cable, and System Information shows it connected via TB/USB4 @ 40 Gbps.

Yeah, I don't think it is throttling. I'm not sure the exact amount of SLC for the cache in this drive, but once it fills up, you will see the write speed drop until it has a chance to offload. I think it is about 10% of the capacity. This is how all SSD work, regardless of manufacturer. Some have a larger cache than others.

Read does not drop since the flash can read the TLC at the same rate as the SLC cache. So the first 10% (.1 * 8TB = ~800GB) will be about 3000 MB/s. Then the the drive has written all cache and will write directly to TLC NAND, at around 1000 MB/s. Once the drive has time to offload the cache, the speed will return back to full. QLC drives will be even slower. All 8TB M.2's on the market have this same behavior.

You can actually prove this by running a test in AJA system test. Set it to write once, read many. The read speed will not drop even if run for hours at a time.

 

[OWC_TAL]

I did a few tests

I recommend trying AJA system test. Settings: 4k-Full, 16bit RGBA, 64GB file size, dual DMA in settings. You can download the app from the appstore for free. It stress tests the drives more than Blackmagic speed test does, and is more indicative of real world.

File size makes a large difference and also better matches real world usage.

 

[altaic]

Yeah, I don't think it is throttling. I'm not sure the exact amount of SLC for the cache in this drive, but once it fills up, you will see the write speed drop until it has a chance to offload. I think it is about 10% of the capacity. This is how all SSD work, regardless of manufacturer. Some have a larger cache than others.

Read does not drop since the flash can read the TLC at the same rate as the SLC cache. So the first 10% (.1 * 8TB = ~800GB) will be about 3000 MB/s. Then the the drive has written all cache and will write directly to TLC NAND, at around 1000 MB/s. Once the drive has time to offload the cache, the speed will return back to full. QLC drives will be even slower. All 8TB M.2's on the market have this same behavior.

You can actually prove this by running a test in AJA system test. Set it to write once, read many. The read speed will not drop even if run for hours at a time.

The 8.0TB OWC Aura Ultra IV lists "Up to 6500 MB/s" write speed and "Up to 7800 MB/s" read speed. 1 GB/s = 8 Gbps, nowhere near either that SSD's advertised write speed of 52 Gbps or TB4's max of 32 Gbps.

If it has a single layer NAND for benchmarks and much slower triple/quad layer for actual bulk data, that seems quite deceptive. Do I just leave my drive mounted for a day and wait for it to redistribute its data? I've never heard of having to do that before.

Edit: Just looked up sustained write speeds for the WD_BLACK SN850X and indeed it (along with other popular consumer SSDs) also falls off dramatically after about 45 seconds due to the dynamic write cache filling up. Huh. Maybe I should be looking at enterprise SSDs or a RAID setup.

 

[OWC_TAL]

The 8.0TB OWC Aura Ultra IV lists "Up to 6500 MB/s" write speed and "Up to 7800 MB/s" read speed. 1 GB/s = 8 Gbps, nowhere near either that SSD's advertised write speed of 52 Gbps or TB4's max of 32 Gbps.

If it has a single layer NAND for benchmarks and much slower triple/quad layer for actual bulk data, that seems quite deceptive. Do I just leave my drive mounted for a day and wait for it to redistribute its data? I've never heard of having to do that before.

Edit: Just looked up sustained write speeds for the WD_BLACK SN850X and indeed it (along with other popular consumer SSDs) also falls off dramatically after about 45 seconds due to the dynamic write cache filling up. Huh. Maybe I should be looking at enterprise SSDs or a RAID setup.

Sequential speeds of 6500 MB/s & 7800MB/s are definitely feasible, but that exceeds the bus of TB3/TB4/USB4. Those speeds in 6,000-7000MB/s range are what a direct attach to motherboard will achieve, or installed in a PCIe slot in something like an Accelsior card.

TB3/TB4/USB4 are a 40Gbps connection, but only a portion of that is for non-display data. I can't recall the exact amount off the top of my head, but it is something like 22-24Gbps. The rest is reserved for displayport traffic.

Do I just leave my drive mounted for a day and wait for it to redistribute its data?

Given the cache is about 800GB, and that TLC can write at about 1000MB/s-ish, it should take about 15 minutes for the cache to fully offload. But even a short period of time allows that cache to offload and opens up the full speed again. Also need to account for Garbage Collection as well depending on what is written to the cells (essentially what TRIM is).

also falls off dramatically after about 45 seconds due to the dynamic write cache filling up

800GB at 3000 MB/s is 5 minutes still any speed changes.

Every single 8TB M.2 SSD out there, no matter what vendor, will act the same. Drives with QLC flash will do even worse. That's just how SSDs work with regards to cache. I don't think there is a single MLC SSD that is 8TB, yet alone at a reasonable price. Enterprise U.2 drives that go up to 32/64TB+ use TLC or QLC flash, also sustaining these drop in speeds.

For most situations people don't run into issues, as you must write 800GB before the cache fills up (continuously). And then continue to write without pause. People that need to sustain an entire disk transfer tend to look at our solutions like the Thunderblade or Express 4m2. The Thunderblade with 4 or 8 SSDs will sustain the speed that fully saturates the bus throughout the entire transfer. This is because it is a RAID, splitting the bandwidth across multiple SSDs.

One last thing to note (an apologies this is already a long reply): Enclosures that cannot properly manage thermals for these drives will actually drop in speeds lower than this. A friend of mine has a inexpensive enclosure (something from Aliexpress) and what happens with 8TB drives is that speed decreases to 600-700 MB/s or less. That decrease is from thermal throttling.

 

[Adora]

I did a few tests, using the drives connected to the MBP M4 Max. I used a short Thunderbolt 5 cable for the 1M2, and the attached cable on the Envoy Ultra.

That Ultra is the fastest external drive i've used, but the real test will be in video editing which i'll be doing over the next few days.

View attachment 2451441

View attachment 2451442


View attachment 2451443

I can't really test my external drives speed at the moment because a Thunderbolt 4 Dock is taking both Thunderbolt 4 ports and on that Dock is another TB4 Dock connected and the DisplayPort is in use and a Creative SoundBlaster is also connected, then the four TB4 drives and 4 USB drives are all also connected at the same time. That seems to throttle all connected drives.

My internal M3 iMac Drive has only this speed (set to 5GB stress):

I remember when I got a 2015 5K iMac and this Black Magic test couldn't even display the speed correctly. Those round indicators where both at the end of the red part the whole time. I think they changed it in the mean time for faster drives, but it shows that Apple isn't state of the art anymore.

My M3 has a 1TB drive that mostly is even faster than the base storage.

I won't say it's not fast enough for me. But the upgrade prices are insane if you just get this with all the cheap much faster drives out there.

 

[jphall]

I recommend trying AJA system test. Settings: 4k-Full, 16bit RGBA, 64GB file size, dual DMA in settings. You can download the app from the appstore for free. It stress tests the drives more than Blackmagic speed test does, and is more indicative of real world.

File size makes a large difference and also better matches real world usage.

I have both, and have used both. I've found that they both report quite a range, I had some high numbers were 3x greater than the lowest numbers in a test.

Real world use, while quite limited so far, seems like the new Thunderbolt 5 drive is faster in Final Cut. As always, I still wish for even faster transfers! lol

 

[OWC_TAL]

As always, I still wish for even faster transfers! lol

You could always try 2 of them in a RAID0. That should about double your speed. Though no redundancy so good to keep diligent backups (well in general it is good to keep backups, even more in this case).

 

[Adora]

You could always try 2 of them in a RAID0. That should about double your speed. Though no redundancy so good to keep diligent backups (well in general it is good to keep backups, even more in this case).

I bought two small USB 3.1 Gen2 10Gbit/s RAID enclosures in 2017 and put two M.2 SATA 3 6Gbit/s in each. That really made a big difference in RAID0.

It didn't even need external power.

It just worked with my 2017 MBP without losing connection like it's getting more and more with USB powered devices now for me. I still have those and tried them with only one drive in it. They always ejected by themselves before SuperDuper was finished with the backup of my main drive. No matter what cable I use or if directly connected or to a dock.

I can only rely on Thunderbolt drives even without external power or my big USB HDDs with external power now. Something has changed there and got really bad.

 

[OWC_TAL]

I bought two small USB 3.1 Gen2 10Gbit/s RAID enclosures in 2017 and put two M.2 SATA 3 6Gbit/s in each. That really made a big difference in RAID0.

It didn't even need external power.

It just worked with my 2017 MBP without losing connection like it's getting more and more with USB powered devices now for me. I still have those and tried them with only one drive in it. They always ejected by themselves before SuperDuper was finished with the backup of my main drive. No matter what cable I use or if directly connected or to a dock.

I can only rely on Thunderbolt drives even without external power or my big USB HDDs with external power now. Something has changed there and got really bad.

Oh I mean you could take two Thunderbolt 5 SSDs and RAID them in a RAID 0 if you really wanted. Would be insane performance. Though one TB5 ssd is already pretty darn fast.

Unfortunately cheaper enclosures tend to have that issue where they just fall of the bus. Likely a thermal thing or poor build quality/firmware.

Take a look at the Thunderblade. It is a super fast NVMe Thunderbolt RAID.

 

[jimthing]

Why do Thunderblades work to passively keep cool only OWC M2's inside?
Couldn't you release an empty Thunderblade enclosure without any M2's? So people could update the thing with larger drives as needed. Adding pads and paste, obviously.

There was a thread on here with some people taking them apart and adding thermal paste to larger M2's and putting them together again and getting good results. But apparently one user received a request from OWC not to do a video on it:

https://forums.macrumors.com/threads/owc-thunderblade-upgrade-ssds-possible.2387108/post-33475889

 

[VanVanWang]

Just newly arrived in my city this Monday.
I got mine 2TB version Enovy Ultra and compared with 1M2 and TB3 Envoy Express.
Just like my guess, the tb5 cable can be replaced and it's softer and thinner than the tb3 cable with Express.
Quite expansive after the imported tax added (about 589$ from local retailer), but still cheaper than upgrade MBP's SSD from Apple Store.
Will replace the 2TB ssd with a nexstorage 8tb ps5 version ssd.It's now in the 1M2 ;-)

 

[drrich2]

I got mine 2TB version Enovy Ultra and compared with 1M2 and TB3 Envoy Express.

If I understood that correctly and you now have both Thunderbolt 5 and 3 drives in your possession, please comment if and when practical as to whether opening applications, files, libraries and such 'feels' different or snappier with the 5. For example, Photos with a really large photo library.

I have seen speed test results, but by way of analogy, I would unable to tell the difference between something moving at light and 1/10th light speed. And most people don't do extended data transfers with huge files or large numbers of them.

But a lot of people may wonder what an external drive for files, libraries and some ap.s is like when using USB-C, Thunderbolt 3 or Thunderbolt 5. And that's hard to get from Black Magic speed tests.

 

[OWC_TAL]

Why do Thunderblades work to passively keep cool only OWC M2's inside?
Couldn't you release an empty Thunderblade enclosure without any M2's? So people could update the thing with larger drives as needed. Adding pads and paste, obviously.

There was a thread on here with some people taking them apart and adding thermal paste to larger M2's and putting them together again and getting good results. But apparently one user received a request from OWC not to do a video on it:

https://forums.macrumors.com/threads/owc-thunderblade-upgrade-ssds-possible.2387108/post-33475889

Thunderblade is sold as a solution. That is to say, it is sold as a completed, tested, highly functioning plug and play unit. The SSDs in Thunderblades, depending on configuration, are also tuned to produce less heat. Even with the great heatsink design, heat is still a concern. I don't think the Thunderblade will ever be sold as a 0GB DIY enclosure from what I gather.

I actually did comment on this in a different post: https://forums.macrumors.com/threads/owc-thunderblade-upgrade-ssds-possible.2387108/post-33475399 TLDR the device is yours- you own it. But your results may vary or damage may occur (including data loss) depending on what you do.

We sell the express 4m2 as an alternative option for those wanting to build their own device. We also sell the U.2 shuttle that can be placed in a variety of our enclosures. Think of those as a Thunderblade in the slot of your Thunderbay (flex). We are definitely committed to keeping the 0GB DIY option, including future versions yet to be announced.

 

[OWC_TAL]

tb5 cable can be replaced

Yes. We will be offering replacement service as well as I believe selling the cable for those wanting to replace it themselves. Though on note, even on our previous Envoy EX with a captive cable, very very few if any RMAs or repairs came in for damaged cables. In the real world, it is super low.

Will replace the 2TB ssd

Uh just a word of caution (and yes the device is yours- you bought it so you can do what you wish with it)- but 8TB blades often exceed the power draw that the Envoy Ultra has. These bus powered drives have a limited power budget. Also, not all PCIe devices/controllers are supported by this revision of the Barlow ridge chipset. These are two of the reasons we currently do not offer such a solution yet. Don't be surprised to see random disconnects, shortened lifespan or corrupted data if/when such events occur. Just wanted to be up front about that.

 

[VanVanWang]

Yes. We will be offering replacement service as well as I believe selling the cable for those wanting to replace it themselves. Though on note, even on our previous Envoy EX with a captive cable, very very few if any RMAs or repairs came in for damaged cables. In the real world, it is super low.


Uh just a word of caution (and yes the device is yours- you bought it so you can do what you wish with it)- but 8TB blades often exceed the power draw that the Envoy Ultra has. These bus powered drives have a limited power budget. Also, not all PCIe devices/controllers are supported by this revision of the Barlow ridge chipset. These are two of the reasons we currently do not offer such a solution yet. Don't be surprised to see random disconnects, shortened lifespan or corrupted data if/when such events occur. Just wanted to be up front about that.

Many many thanks for the info.
Today I replaced the Ultra with a 8TB Nextorage PS5 ssd (encrypted APFS volume) which previously installed in the 1M2. And install a new PNY CS3140 8TB into the 1M4.
I currently only have an M1 Max MBP with TB4 ports, so the Ultra read/write speed just as the same as the 1M4 about 2800MB/s~3200MB/s.
Make a full disk backup from Ultra to 1M2 with 3.5TB files (>20w files, from xxKB documents to xxGB ProRes videos) takes about 45mins, the average speed is about 1300MB/s.
And the continued to made an other backup to a LaCie 1big SSD Pro with a WD SN640 7.68T U2 SSD which is a TB3 device and average speed is about 1100MB/s.

The Nextorage SSD is less power efficient:

NEM-PA Series｜M.2 2280 PCIe®4.0 NVMe SSD + Heatsink – Nextorage

NEM-PA Series｜M.2 2280 PCIe®4.0 NVMe SSD + Built-in Heatsink 8TB / 4TB / 2TB / 1TB See Full Spec. Where to buy...

www.nextorage.net

Max power consumption (READ/ WRITE)​

NEM-PA8TB: 11.5 W / 11.0 W
NEM-PA4TB: 11.3 W / 10.9 W
NEM-PA2TB: 10.5 W / 10.0 W
NEM-PA1TB: 9.5 W / 7.7 W

The PNY CS3140 looks more power efficient than Nextorage one but still more then OWC Aura Pro IV ssd...


😂😂😂

Just have a quick test, the Ultra's Read/Write power consumption with a Nextorage 8TB is 8.1W/9.1W
The 1M2 with PNY CS3140 8TB is 8.6W/9.1W

Looks fine?!

 

[VanVanWang]

If I understood that correctly and you now have both Thunderbolt 5 and 3 drives in your possession, please comment if and when practical as to whether opening applications, files, libraries and such 'feels' different or snappier with the 5. For example, Photos with a really large photo library.

I have seen speed test results, but by way of analogy, I would unable to tell the difference between something moving at light and 1/10th light speed. And most people don't do extended data transfers with huge files or large numbers of them.

But a lot of people may wonder what an external drive for files, libraries and some ap.s is like when using USB-C, Thunderbolt 3 or Thunderbolt 5. And that's hard to get from Black Magic speed tests.

Still waiting my M4 Max MBP for further testing...😅

 

[OWC_TAL]

Many many thanks for the info.
Today I replaced the Ultra with a 8TB Nextorage PS5 ssd (encrypted APFS volume) which previously installed in the 1M2. And install a new PNY CS3140 8TB into the 1M4.
I currently only have an M1 Max MBP with TB4 ports, so the Ultra read/write speed just as the same as the 1M4 about 2800MB/s~3200MB/s.
Make a full disk backup from Ultra to 1M2 with 3.5TB files (>20w files, from xxKB documents to xxGB ProRes videos) takes about 45mins, the average speed is about 1300MB/s.
And the continued to made an other backup to a LaCie 1big SSD Pro with a WD SN640 7.68T U2 SSD which is a TB3 device and average speed is about 1100MB/s.

The Nextorage SSD is less power efficient:

NEM-PA Series｜M.2 2280 PCIe®4.0 NVMe SSD + Heatsink – Nextorage

NEM-PA Series｜M.2 2280 PCIe®4.0 NVMe SSD + Built-in Heatsink 8TB / 4TB / 2TB / 1TB See Full Spec. Where to buy...

www.nextorage.net

Max power consumption (READ/ WRITE)​

NEM-PA8TB: 11.5 W / 11.0 W
NEM-PA4TB: 11.3 W / 10.9 W
NEM-PA2TB: 10.5 W / 10.0 W
NEM-PA1TB: 9.5 W / 7.7 W

The PNY CS3140 looks more power efficient than Nextorage one but still more then OWC Aura Pro IV ssd...


😂😂😂

Just have a quick test, the Ultra's Read/Write power consumption with a Nextorage 8TB is 8.1W/9.1W
The 1M2 with PNY CS3140 8TB is 8.6W/9.1W

Looks fine?!

There are very strict tests for TB certification. I'm not sure how you measured your power draw, but for certification, engineers use an oscilloscope to check waveforms, voltage and current rather than strictly using a power meter to gather power consumption over a long period of time. They measure peak power on the device as a whole. They are extremely thorough.

IF we could offer a 8TB model, we gladly would. Our measurements from testing exceeded what is allowed to be certified.

Secondly, it looks like you are testing on an TB4 machine, not TB5. I'm not sure the power draw difference, but I would presume that the faster TB5 speeds draw more power given those communicate at PCIe 4 rather than PCIe 3.

At the end of the day, it is your device and your data. It could work, it could not work in the long term, but that is the risk you take. We can't and don't take that risk with our customers data.

 

[VanVanWang]

There are very strict tests for TB certification. I'm not sure how you measured your power draw, but for certification, engineers use an oscilloscope to check waveforms, voltage and current rather than strictly using a power meter to gather power consumption over a long period of time. They measure peak power on the device as a whole. They are extremely thorough.

IF we could offer a 8TB model, we gladly would. Our measurements from testing exceeded what is allowed to be certified.

Secondly, it looks like you are testing on an TB4 machine, not TB5. I'm not sure the power draw difference, but I would presume that the faster TB5 speeds draw more power given those communicate at PCIe 4 rather than PCIe 3.

At the end of the day, it is your device and your data. It could work, it could not work in the long term, but that is the risk you take. We can't and don't take that risk with our customers data.

Just got my M4 MBP.

Here is a try...

owc envoy ultra with Nextorage 8TB on right vs mbp internal ssd on left

the Nextorage 8TB max power state according this website(https://theoverclockingpage.com/202...pa-8tb-the-best-solution-for-the-ps5/?lang=en) is PS 0 which is 8.8w, and I think the tb4/tb5 port can provide 15w for the attached external storage.

 

[OWC_TAL]

Just got my M4 MBP.

Here is a try...

owc envoy ultra with Nextorage 8TB on right vs mbp internal ssd on left

the Nextorage 8TB max power state according this website(https://theoverclockingpage.com/202...pa-8tb-the-best-solution-for-the-ps5/?lang=en) is PS 0 which is 8.8w, and I think the tb4/tb5 port can provide 15w for the attached external storage.

I'm not quite sure what I can do to help or explain more than I did in my previous comment. Our 8TB is also an E18 based SSD and real world testing it exceeded the allowable power budget. The power draw is measured in a very specific way for TB certification. We will keep working to bring an 8TB model to the market, hopefully sometime in early 2025.

Because SSDs can have such a wide range of power draw, it is not possible/allowed to certify a 0GB/DIY TB5 enclosure that is bus powered.

 

[jimthing]

I'm not quite sure what I can do to help or explain more than I did in my previous comment. Our 8TB is also an E18 based SSD and real world testing it exceeded the allowable power budget. The power draw is measured in a very specific way for TB certification. We will keep working to bring an 8TB model to the market, hopefully sometime in early 2025.

Because SSDs can have such a wide range of power draw, it is not possible/allowed to certify a 0GB/DIY TB5 enclosure that is bus powered.

Interesting, thanks.

Will it be possible to make a passively cooled (or even actively cooled, but with all important near-silent fans!) empty externally powered non-bus NVMe multi-drive enclosure with an Tbolt 4 or 5? The 4M2 unfortunately has many reports of being very noisy fan-wise.

It also doesn't have to be the fastest thing ever, either, but many users want to use a multi-NVMe boxes for non-scratch mass storage uses instead of HDDs.

 

[OWC_TAL]

Interesting, thanks.

Will it be possible to make a passively cooled (or even actively cooled, but with all important near-silent fans!) empty externally powered non-bus NVMe multi-drive enclosure with an Tbolt 4 or 5? The 4M2 unfortunately has many reports of being very noisy fan-wise.

It also doesn't have to be the fastest thing ever, either, but many users want to use a multi-NVMe boxes for non-scratch mass storage uses instead of HDDs.

We've got some great stuff coming in 2025 is all I can say. I think you'll like what we release.