Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

Rhobes

macrumors 6502
Original poster
Oct 28, 2004
357
7
Bigfork, MT
Hello All-
I just opened a new 2019 iMac (custom optioned loaded) including 2TB SSD storage and 76 GB 2666MHz DDR4 memory
and have it on line now. I want to add an external back up drive. The iMac has 2 Thunderbolt(TB) 3 USB-C, 40 Gb/s inputs which can also be used with USB-C 3.1 Gen 2 (10 Gb/s). I'm looking at WD & Samsung drives which have Read/Write speeds around 7450 MB/s and an enclosure: Acasis 40Gbps M.2 NVMe SSD. The enclosure states, an Intel Cert. Chip JHL7440 & compatible with: TB 4/3; USB 4/3.2/3.1/3.0/2.

From what I've gathered from the internet (FWIW) 7450Mbps equates to 7.45 Gbps. So, if this be true, then no sense in buying TB connections just go with 3.1 Gen2 USB-C connectors?

Thanks for any comments-
Rhobes
 
@Rhobes "I'm looking at WD & Samsung drives which have Read/Write speeds around 7450 MB/s..."

That speed is only possible if you put the SSD into the M.2 slot on the motherboard of a fast PC computer.

In use with a Macs you can get up to 6000 MB/s using a Thunderbolt 5 external enclosure, if the Mac is a 2024 TB5 Pro/Max/Ultra model.
For TB4/USB4 Macs the maximum will be up to 3400 MB/s.

"So, if this be true, then no sense in buying TB connections just go with 3.1 Gen2 USB-C connectors?"

The maximum a 2019 iMac can achieve is about 2800 MB/s using one of it's USB-C TB 3 ports.
Both TB 3 ports share this bandwidth, so that is the maximum.

Using current versions of MacOS you can connect a USB4 enclosure, but in practice the speeds won't be quite as high as you get with TB 3.
On a 2019 iMac a 'USB4' enclosure is unlikely to perform as fast as a proper old style TB 3 enclosure.

The Acasis enclosure is quite possibly marketed as 'USB4', but isn't.
It was designed before proper USB4 controller chips were available.

It's a TB3 enclosure which will work in a TB4 port, and on PCs where a lower speed version of USB4 is being used, where is will revert to USB 3.1 Gen2...

So to sum up: Yes, the 2019 iMac will support one TB3 enclosure at up to 2800 MB/s, so the latest NVMe SSDs are no faster than the older, slower, cheaper SSDs.

To be cost-effective, it's usual to use a TB3 external SSD for 'scratch disk' work in progress while using the Mac, but use slower USB 3.* SSDs for backup.

In a 2019 iMac, which only has one TB 3 controller, with two ports, its often best to attach slower SSDs to the USB 3.0 USB-A ports.
 
Last edited:
  • Like
Reactions: Rhobes
@Rhobes "I'm looking at WD & Samsung drives which have Read/Write speeds around 7450 MB/s..."

That speed is only possible if you put the SSD into the M.2 slot on the motherboard of a fast PC computer.

In use with a Macs you can get up to 6000 MB/s using a Thunderbolt 5 external enclosure, if the Mac is a 2024 TB5 Pro/Max/Ultra model.
For TB4/USB4 Macs the maximum will be up to 3400 MB/s.

"So, if this be true, then no sense in buying TB connections just go with 3.1 Gen2 USB-C connectors?"

The maximum a 2019 iMac can achieve is about 2800 MB/s using one of it's USB-C TB 3 ports.
Both TB 3 ports share this bandwidth, so that is the maximum.

Using current versions of MacOS you can connect a USB4 enclosure, but in practice the speeds won't be quite as high as you get with TB 3.
On a 2019 iMac a 'USB4' enclosure is unlikely to perform as fast as a proper old style TB 3 enclosure.

The Acasis enclosure is quite possibly marketed as 'USB4', but isn't.
It was designed before proper USB4 controller chips were available.

It's a TB3 enclosure which will work in a TB4 port, and on PCs where a lower speed version of USB4 is being used, where is will revert to USB 3.1 Gen2...

So to sum up: Yes, the 2019 iMac will support one TB3 enclosure at up to 2800 MB/s, so the latest NVMe SSDs are no faster than the older, slower, cheaper SSDs.

To be cost-effective, it's usual to use a TB3 external SSD for 'scratch disk' work in progress while using the Mac, but use slower USB 3.* SSDs for backup.

In a 2019 iMac, which only has one TB 3 controller, with two ports, its often best to attach slower SSDs to the USB 3.0 USB-A ports.

Thanks so much for your great reply, a huge help for me, it clears up so much of my confusion-
Rhobes
 
"So, if this be true, then no sense in buying TB connections just go with 3.1 Gen2 USB-C connectors?"
...
So to sum up: Yes, the 2019 iMac will support one TB3 enclosure at up to 2800 MB/s, so the latest NVMe SSDs are no faster than the older, slower, cheaper SSDs.

To be cost-effective, it's usual to use a TB3 external SSD for 'scratch disk' work in progress while using the Mac, but use slower USB 3.* SSDs for backup.

I agree with your summary though I think my clearer for the original question if broken into two issues:
1. That iMac 2019 can handle an NVMe SSD connected by TB3 (which TB4 and presumably TB5 controllers should fallback to) and gain benefit over an SSD connected via USB 3.x Gen2 with probably a >2x performance advantage all else being equal; plus potential for better SMART compatibility and TRIM support.

2. For backups, an SSD connected via USB 3.x Gen2 is probably sufficient and the performance gain of TB3+ is unlikely to be valuable

In a 2019 iMac, which only has one TB 3 controller, with two ports, its often best to attach slower SSDs to the USB 3.0 USB-A ports.

Just to highlight for the original poster the USB-A ports on that model are USB 3.0 and so limited to 5 Gbit. Any recent model NVMe SSD will be >4x faster on the TB3 port than through the USB-A. Again for backups this may not matter. However, for a full restore or transfer of those backups to a future machine, it might be nice to have the option for TB3+. Also if there is no plan for an external scratch/archive/etc disk, there's no upside to not using the TB3 ports.

The Acasis enclosure mentioned (presumably the TBU401x or TBU405x) could be connected to the TB3 port of that iMac with the bundled cable as well as to the USB-A ports with a USB-A<->USB-C converter cable. Then any recent model of a good quality NVMe SSD used in that enclosure should be able to reach the ~ 2800MB/sec potential of that configuration. If of good quality, the NVMe SSD and/or enclosure could also be reused in the future for purposes besides backup or this particular iMac.
 
  • Like
Reactions: Rhobes
Hello All-
I just opened a new 2019 iMac (custom optioned loaded) including 2TB SSD storage and 76 GB 2666MHz DDR4 memory
and have it on line now. I want to add an external back up drive. The iMac has 2 Thunderbolt(TB) 3 USB-C, 40 Gb/s inputs which can also be used with USB-C 3.1 Gen 2 (10 Gb/s). I'm looking at WD & Samsung drives which have Read/Write speeds around 7450 MB/s and an enclosure: Acasis 40Gbps M.2 NVMe SSD. The enclosure states, an Intel Cert. Chip JHL7440 & compatible with: TB 4/3; USB 4/3.2/3.1/3.0/2.

From what I've gathered from the internet (FWIW) 7450Mbps equates to 7.45 Gbps. So, if this be true, then no sense in buying TB connections just go with 3.1 Gen2 USB-C connectors?

Just to clarify this issue, if you are looking at recent models of Samsung and WD NVMe SSD, they are probably showing peak performance of 7450 MB/s where MB=megabytes/second so 7.4 gigabytes/second. USB 3.1 Gen2 USB-C is 10Gbit/sec ~ 1 gigabytes/second. A USB 3.1 Gen2 connection would be a bottleneck for these drives.

Connecting via TB3 would still be a bottleneck to peak performance with these drives but less so. As PaulD-UK noted, you could expect these drives to peak at 2.8 gigabytes/second in an external enclosure connected via TB3. Which to his point is probably overkill if all you plan to use this drive for is backups.

If you do continue with the plan to assemble an NVMe SSD in that Acasis enclosure, I would get the WD SN850X. Recent versions of macOS have had issues with the Samsung drives after the 970 EVO. Unclear if resolved with the Samsung 9100 Pro, etc.
 
  • Like
Reactions: Rhobes and PaulD-UK
As an Amazon Associate, MacRumors earns a commission from qualifying purchases made through links in this post.
  • Like
Reactions: Rhobes
I must thank all who replied here, I think the replies were top notch & very descriptive, informative. Definitely feel I have made well informed choices here. As a follow up I went with the Acasis TBU 405 Pro M (Fan) which has a newer silent fan over the TBU 401 Pro otherwise identical. I went with the WD Black 2TB SN850X NVMe SSD. Also, to keep the background Apple cable hauntings out of my head, believe I found a good deal on a 31" TB-3 cable connector for 19.95 incl S/H.

I will be using this for back ups, tons of photos and video editing. In a week or so should get the products and start the process of a double boot to the "new" 2019 iMac. I have previous back ups on CCC and Time Machine of 2010 iMac. I have photoshop CS5 & couple other photo/video apps currently opening/running on the Mojave (CS-5 32/64 bit, other apps 32/bit). Plan to copy the Mojave to ext disc and later (after checking that Mojave is indeed not crashing when actually working with them) upgrade that to likely Sequoia. So, thinking to have Mojave stay on iMac internal drive and have bootable external with Sequoia. Also have FCPX updated to run on Mojave, now. But, if don't work well there can further upgrade it to v. 11 & run on Sequoia. Lots of large file photo/video clips going back & forth when saving and working on projects. Thanks again & will be back, I'm sure-
 
  • Like
Reactions: bzgnyc2
@Rhobes "I'm looking at WD & Samsung drives which have Read/Write speeds around 7450 MB/s..."

That speed is only possible if you put the SSD into the M.2 slot on the motherboard of a fast PC computer.

In use with a Macs you can get up to 6000 MB/s using a Thunderbolt 5 external enclosure, if the Mac is a 2024 TB5 Pro/Max/Ultra model.
For TB4/USB4 Macs the maximum will be up to 3400 MB/s.

"So, if this be true, then no sense in buying TB connections just go with 3.1 Gen2 USB-C connectors?"

The maximum a 2019 iMac can achieve is about 2800 MB/s using one of it's USB-C TB 3 ports.
Both TB 3 ports share this bandwidth, so that is the maximum.

Using current versions of MacOS you can connect a USB4 enclosure, but in practice the speeds won't be quite as high as you get with TB 3.
On a 2019 iMac a 'USB4' enclosure is unlikely to perform as fast as a proper old style TB 3 enclosure.

The Acasis enclosure is quite possibly marketed as 'USB4', but isn't.
It was designed before proper USB4 controller chips were available.

It's a TB3 enclosure which will work in a TB4 port, and on PCs where a lower speed version of USB4 is being used, where is will revert to USB 3.1 Gen2...

So to sum up: Yes, the 2019 iMac will support one TB3 enclosure at up to 2800 MB/s, so the latest NVMe SSDs are no faster than the older, slower, cheaper SSDs.

To be cost-effective, it's usual to use a TB3 external SSD for 'scratch disk' work in progress while using the Mac, but use slower USB 3.* SSDs for backup.

In a 2019 iMac, which only has one TB 3 controller, with two ports, its often best to attach slower SSDs to the USB 3.0 USB-A ports.
That is a good outline Paul and help’s me put TB into perspective generally. I use a 2019 Intel iMac with TB3.0.
I have recently purchased a Z666 with ‘true’ TB4 and intend using a 4TB Lexar NM790 PCIe 4.0 inside, but find using the TB3.0 port my resulting speeds far below expected 3000Mb/ps. In fact they are Read 2599 Mb/ps Write, a disappointing 1457Mb/ps.


Research explains that TB implemented on the 2019 iMac is shared between ports and encryption of TB transmissions consumes about 8Gb/ps. of capacity. A further 22Gb is used for the monitor thus leaving 10Gb for other applications. A bottle-neck limiting external SSD speed.

I am hoping an OWC TB 4.0 mini port circumvents the iMac TB limitation and I will achieve significantly better speed performance from the an external Z666. The OWC mini port which has 60 Watts of power to drive 32Gb of TB4.0 to the Z666 therebye achieving around 32Gb/ps throughput. That should put me closer to the 4000+ Mb/ps speed area.


Am I on the right track or am I missing some critical piece of understanding?
 
@Terraaustralis "I am hoping an OWC TB 4.0 mini port circumvents the iMac TB limitation and I will achieve significantly better speed performance from the an external Z666."

No. If you read the reviews of that OWC TB 4.0 hub, the second one is clear:
"Compatibility problems with PC. Not backwards compatible to TB3, even with a Gigabyte branded motherboard Thunderbolt adapter card."

That shows two things:
1. The OWC TB4 hub doesn't work with a TB 3 host chipset. and
2. the Gigabyte adapter card doesn't seem to have a USB 3.x host controller chip, so the OWC hub doesn't work at USB 3.x speeds.

On a pre-TB4 Intel computer the OWC TB4 hub should revert to USB 3.x mode (at 10Gbps), and that is what will happen if you connect it to an Intel iMac.

I tested a Qwiizlab USB4 enclosure with 4TB Samsung 990Pro on an iMac Pro, and got identical speeds to your Zike, so I think that the two-lane PCIe 3x2 ~1500MB/s write speed is a built-in limitation to the USB4 compatibility of MacOS on Intel Macs.

iMacProTB4ssd.png
 
Wow!
Thank you Paul, that is most valuable information. Your analysis seems spot on. There are more tricks and traps with TB connectivity than contained in a barrel full of monkeys!

When is TB3 not TB3? - When its an iMac.

You have saved me from making a very disappointing and expensive error.

My most sincere thanks,

Tony
 
Last edited:
  • Like
Reactions: PaulD-UK
When is TB3 not TB3? - When its an iMac.

I don't believe it would be accurate to say the iMac 2019's Thunderbolt 3 ports are not true TB3. They do use Intel's JHL7540 controller and presumably certified as meeting Thunderbolt 3's specs (at least at one point -- firmware/software updates can probably break things after original certification).

Also note that the iMac Pro uses an older Intel controller (JHL6540) for Thunderbolt and may have more limitations over the one used in the iMac 2019. Issues seen there may not carry over to the iMac 2019.

In any case, note that the performance reported above, though not ideal for TB3, would still not achievable with USB 3.x Gen2 (Gen2x1 -- your results are possibly within rounding error of USB 3.2 Gen2x2 performance but Apple has never supported USB 3.2 Gen2x2). I don't believe you are seeing a fallback to USB 3.x.

Similarly, I don't believe the numbers reported above are due to dropping back to two-lane PCIe 3x2 as that would also limit read speed well below 2GB/second (e.g. what is being seen on write). I don't believe it would be normal for the system to negotiate PCIe 3x4 in one direction and PCIe 3x2 in the other direction. By chance do you have a Thunderbolt hub or monitor connected to your iMac's other Thunderbolt 3 port?

Otherwise, what is going on in this case is hard to say. I do note that the Zike uses the ASMedia 2464PD USB4 bridge controller (which has been certified as TB4) which generally works well when connected to the latest USB4/TB4+ host controllers but may not be as well tested with older Thunderbolt 3 host controllers (perhaps fixable with a firmware update to one or the other though that also may not be easy to obtain).

And/or could there also be a compatibility issue with the Lexar NM790 and the ASMedia and/or iMac and/or macOS? I do note that is a DRAMless/Host Memory Bus (HMB) design SSD . Those aren't officially supported with macOS nor in Thunderbolt enclosures but I've heard of other people using HMB SSD of other brands without issue so I wouldn't definitively say that is the issue either.

Which is all to say your iMac is capable of more but I can't explain why you aren't getting it without testing various combinations of hardware and software that I don't have. On the upside, this enclosure/SSD is at least forward-looking to if/when you move to an M2+ Mac with native USB4/TB4+.
 
  • Like
Reactions: PaulD-UK
Hi bzgnyc2,
Thank you for a most thoughtful analysis. I was being glib with my remark about TB3. As your response rightly points out, I am achieving significantly better figures than USB 3.1 Gen 2 which suggests TB3 is not just defaulting to 10Gbs. However, the disparity between Read/write (Read 2599 Mb/ps /Write 1457Mb/ps.) seems utterly uncharacteristic of the equipment used and PaulD has a persuasive argument. Further, as you point out, there are many opportunities for incompatibility between components. Incidentally, all figures quoted are based upon one TB3 port for the external Zike and the other TB3 port vacant.

To give a bigger picture; I had the Zike666 tested, with the Lexar loaded, on an M2 MBP which gave Read 3370.15Mb/ps /Write 2847.96Mb/ps. This would seem to rule out any incompatibility between the Zike and the Lexar. As for Dramless NVMe being incompatible with O/S or iMac, I have upgraded the iMac internal PCIe with a SN770 PCIe 4.0 Dramless SSD and get good PCIe 3.0 Read 2957Mb/ps/Write 2922 with AJA System Test. So the iMac and OSX Ventura seem happy. Zike have their own test results which are set out below.

When I embarked upon the task of obtaining better external SSD performance I did so with an eye to the future. With the iMac limited to PCIe 3.0, it made no sense to spend on fast externals when the host was limited. In my ignorance, I did expect better external speeds than I am receiving on the iMac.

For the time being, I am resolved to live with the compromise and be grateful for all I have learned. In the distant future, I see the M4 mini as my next upgrade. The equipment I have purchased will at that stage, return the kind of performance I was seeking.

Thank you very much for your thoughtful contribution.
 

Attachments

  • Screenshot 2025-03-22 at 4.34.49 am.png
    Screenshot 2025-03-22 at 4.34.49 am.png
    550.2 KB · Views: 18
Last edited:
Hi bzgnyc2,
Thank you for a most thoughtful analysis. I was being glib with my remark about TB3.

Of course I understand. I just didn't want to leave the thread there as I found from my own postings that other people come back later and find our old threads when trying to solve a problem. More than once I've thrown out broad statements in the context of a thread only to learn that it threw someone else completely off...

As your response rightly points out, I am achieving significantly better figures than USB 3.1 Gen 2 which suggests TB3 is not just defaulting to 10Gbs. However, the disparity between Read/write (Read 2599 Mb/ps /Write 1457Mb/ps.) seems utterly uncharacteristic of the equipment used and PaulD has a persuasive argument. Further, as you point out, there are many opportunities for incompatibility between components. Incidentally, all figures quoted are based upon one TB3 port for the external Zike and the other TB3 port vacant.

To give a bigger picture; I had the Zike666 tested, with the Lexar loaded, on an M2 MBP which gave Read 3370.15Mb/ps /Write 2847.96Mb/ps. This would seem to rule out any incompatibility between the Zike and the Lexar. As for Dramless NVMe being incompatible with O/S or iMac, I have upgraded the iMac internal PCIe with a SN770 PCIe 4.0 Dramless SSD and get good PCIe 3.0 Read 2957Mb/ps/Write 2922 with AJA System Test. So the iMac and OSX Ventura seem happy. Zike have their own test results which are set out below.

When I embarked upon the task of obtaining better external SSD performance I did so with an eye to the future. With the iMac limited to PCIe 3.0, it made no sense to spend on fast externals when the host was limited. In my ignorance, I did expect better external speeds than I am receiving on the iMac.

Thanks for sharing the additional data points. I agree you should be seeing better -- especially on the write. I don't rule out any explanation out yet as I don't really understand why you are getting the performance you are getting. Though it may be one of those things that we see with consumer technology that it won't get solved before it's time to move on anyway...

Yes the SN770 is one of the is the DRAMless/HBM SSD in particular that I've heard people get good performance from. My understanding is that HBM doesn't work over Thunderbolt/etc and macOS doesn't support even when internal (e.g. in a Mac Pro) either. Then I understand these drives go into a fallback mode when HBM is not possible. I don't know if some drives (e.g. your SN770) do better than others in this fallback mode or the issue with onboard DRAM for SSD is overrated.
 
@bzgnyc2 "I don't really understand why you are getting the performance you are getting. Though it may be one of those things that we see with consumer technology..."

When USB4 became a thing that people could buy, Apple seems to have done a minor rewrite of MacOS (Ventura at the time I think) to allow MacOS on Intel TB3 Macs to mount TB4/USB4 storage.
I guess the result is what we see now, and there is no incentive to resist the problem further?

Quote: "My understanding is that HBM doesn't work over Thunderbolt/etc and macOS doesn't support even when internal (e.g. in a Mac Pro) either. Then I understand these drives go into a fallback mode when HBM is not possible."

I've read that WD seem to have solved this on their DRAM-less blades by having SRAM in the controller chip (which performs the same fast database function as DRAM), and so allows their NVMe blades to work well with MacOS (in external enclosures).
 
Last edited:
  • Like
Reactions: bzgnyc2
Hi Guys,
As an average consumer trying to understand technologies for which I have no training, your comment "one of those things that we see with consumer technology that it won't get solved before it's time to move on anyway..." seems most appropriate. I do like to get to the bottom of a problem and understand where possible, but with IT, it seems there is always another layer of possibilities which demand tests for which there is neither available time or equipment. A bit like Russian dolls!

I very much appreciate the analysis and thoughtful considerations which you guys have offered which has been an education for me.

Regarding the internal 2TB SN770 performance on the iMac 19.1; for my graphic work it is fast enough at around 2900MB/ps read/write near the practical PCIe 3.0 speed limits. The weakest point of the SN770 is sustained data transfer as it drops to 500Mb/ps or less after 110 seconds when the cache is full. Fortunately sustained read/write file transfer is a less frequent eventuality for me. On the other hand, with mixed work, cache recovery is fast to give a high average performance. So for one reason or another it has behaved well on this iMac model.

Thanks a million.
 
Last edited:
Hi Guys (PaulD and bzgnyc2)

Given the inexplicable speed results (Write 1457Mbs Read 2599Mbs) from a port which should be operating at 10 Gbs I asked for assistance on Apple Forum. Since you have provided so much help I thought these final comments below fitted with your observations and offered a convincing explanation.

g_wolfman:
A 2019 iMac's T3 ports support USB 3.2 Gen 2 (originally named USB 3.1 Gen 2, or Superspeed+ USB). It is rated by the USB consortium or whatever they call themselves for a 10Gbps transfer speed and a theoretical throughput of 1250 MBps (1250MBps x 8 b/B = 10Gbps).

1457 MBps is already exceeding that speed the even higher read speed is probably due to the drive ability to transfer larger chunks into a direct memory access buffer. It's free chicken, basically, and an unexpected bonus.

The limiting factor is the port on the 2019 iMac being only Thunderbolt 3/USB3.2, not USB4.


I immediately asked myself what the performance would be with a WD Black SN850x in the Z666. Apart from the different NVMe specs, would the 850x with dedicated memory perform better or worse than the Lexar with HMB?

Probably not, as with his iMac Pro, Paul D used a Samsung 990 Pro which has dedicated memory, and received similar speed results to the Lexar.

So the Z666 seems to enable greater iMac port effectiveness despite USB limitations?
 
Last edited:
Hi Guys (PaulD and bzgnyc2)

Given the inexplicable speed results (Write 1457Mbs Read 2599Mbs) from a port which should be operating at 10 Gbs I asked for assistance on Apple Forum. Since you have provided so much help I thought these final comments below fitted with your observations and offered a convincing explanation.

g_wolfman:
A 2019 iMac's T3 ports support USB 3.2 Gen 2 (originally named USB 3.1 Gen 2, or Superspeed+ USB). It is rated by the USB consortium or whatever they call themselves for a 10Gbps transfer speed and a theoretical throughput of 1250 MBps (1250MBps x 8 b/B = 10Gbps).

1457 MBps is already exceeding that speed the even higher read speed is probably due to the drive ability to transfer larger chunks into a direct memory access buffer. It's free chicken, basically, and an unexpected bonus.

The limiting factor is the port on the 2019 iMac being only Thunderbolt 3/USB3.2, not USB4.


I immediately asked myself what the performance would be with a WD Black SN850x in the Z666. Apart from the different NVMe specs, would the 850x with dedicated memory perform better or worse than the Lexar with HMB?

Probably not, as with his iMac Pro, Paul D used a Samsung 990 Pro which has dedicated memory, and received similar speed results to the Lexar.

So the Z666 seems to enable greater iMac port effectiveness despite USB limitations?

Based on the data points we've seen so far, I can't predict the performance a specific Mac/controller/OS -> specific enclosure/firmware -> SSD/firmware...

I will note that Arcasis had similar performance issues as you had when they tested the WD SN850 in their TB3 (Intel JHL7440 controller) enclosure several years ago but resolved the issue with a firmware update for the WD SN850:

Arcasis also found similar performance issues with the Samsung 970 EVO Plus but not the Samsung 970 EVO despite looking pretty similar hardware-wise. No indication that there was a firmware update for the Plus that resolved the issue.

Then I have seen results of 2500-2700 MB/sec R/W for both the Samsung 980 Pro and SN850X in an TB3 enclosure (i.e. one using the JHL7440 chipset such as the Arcasis TBU401/405) connected to an Intel Mac.

As such given the unknowns and interactions that would probably only make sense if we got Apple hardware and software engineers, Intel engineers, ASmedia engineers, and Lexar (and/or maybe even their controller manufacturer) all in the same room, I think we're stuck with educated guesses aka trial and error. I could see that an older enclosure with the JHL7440 chipset (e.g. the Arcasis TBU401/405), might bridge that Lexar SSD to your iMac with read/write speeds in the 2500-2700 MB/sec range. If so that would suggest an odd interaction between the Lexar SSD, the ASMedia bridge chip, and the Intel host controller. Or maybe it is just that Lexar SSD that doesn't work well with Intel host controllers for some reason. And/or maybe the relevant drivers are different enough between macOS for Apple Silicon and for Intel and the Intel one doesn't like the Lexar as much. In either of those cases keeping the Z666 but swapping in the SN850X would get you to the expected 2500-2700 MB/sec speeds.

Unfortunately I can only offer speculation and maybes at this point...

P.S.You might contact Zike and see if they have a firmware update. I have seen reports of substantial (e.g. 50%) performance improvements from updating the firmware of enclosures that use the same ASM2464PD chipset as your Zike. Supposedly 241129_85_00_00 is the latest. Also the ASM2464PD has a "U4 mode" setting and one could try tweaking that setting when installing new firmware. Unfortunately it appears all this needs to be done from Windows.
 
Based on the data points we've seen so far, I can't predict the performance a specific Mac/controller/OS -> specific enclosure/firmware -> SSD/firmware...

I will note that Arcasis had similar performance issues as you had when they tested the WD SN850 in their TB3 (Intel JHL7440 controller) enclosure several years ago but resolved the issue with a firmware update for the WD SN850:

Arcasis also found similar performance issues with the Samsung 970 EVO Plus but not the Samsung 970 EVO despite looking pretty similar hardware-wise. No indication that there was a firmware update for the Plus that resolved the issue.

Then I have seen results of 2500-2700 MB/sec R/W for both the Samsung 980 Pro and SN850X in an TB3 enclosure (i.e. one using the JHL7440 chipset such as the Arcasis TBU401/405) connected to an Intel Mac.

As such given the unknowns and interactions that would probably only make sense if we got Apple hardware and software engineers, Intel engineers, ASmedia engineers, and Lexar (and/or maybe even their controller manufacturer) all in the same room, I think we're stuck with educated guesses aka trial and error. I could see that an older enclosure with the JHL7440 chipset (e.g. the Arcasis TBU401/405), might bridge that Lexar SSD to your iMac with read/write speeds in the 2500-2700 MB/sec range. If so that would suggest an odd interaction between the Lexar SSD, the ASMedia bridge chip, and the Intel host controller. Or maybe it is just that Lexar SSD that doesn't work well with Intel host controllers for some reason. And/or maybe the relevant drivers are different enough between macOS for Apple Silicon and for Intel and the Intel one doesn't like the Lexar as much. In either of those cases keeping the Z666 but swapping in the SN850X would get you to the expected 2500-2700 MB/sec speeds.

Unfortunately I can only offer speculation and maybes at this point...

P.S.You might contact Zike and see if they have a firmware update. I have seen reports of substantial (e.g. 50%) performance improvements from updating the firmware of enclosures that use the same ASM2464PD chipset as your Zike. Supposedly 241129_85_00_00 is the latest. Also the ASM2464PD has a "U4 mode" setting and one could try tweaking that setting when installing new firmware. Unfortunately it appears all this needs to be done from Windows.
Very interesting analysis bzgnyc2. It is wonderful that such well informed contributors are prepared to spend time supporting lay users such as myself. I thank you all. The education is invaluable. The complexity of IT and each individuals particular circumstances are such that community support is essential if we are to improve our self management.

I am confirming Z666 Firmware status with Zike and will respond when I receive a reply.

Curiosity aside, using an SN 850x to recover a few hundred Write Mbps for an under-performing external boot/clone/storage/drive is to me financially unjustifiable. Now I have a better understanding of the issues I shall be happy make do with the Lexar Write deficiency until future host upgrade (M4 Mac mini?), allows improved external performance. At least I am receiving better external speed than the Samsung T7 which I sold.

Much appreciate your support. Long but very interesting and worthwhile journey for me. Moving forward to USB 4 and putting this confusing TB/USB transitional period behind us will be welcome.
 
Last edited:
  • Like
Reactions: bzgnyc2
Based on the data points we've seen so far, I can't predict the performance a specific Mac/controller/OS -> specific enclosure/firmware -> SSD/firmware...

I will note that Arcasis had similar performance issues as you had when they tested the WD SN850 in their TB3 (Intel JHL7440 controller) enclosure several years ago but resolved the issue with a firmware update for the WD SN850:

Arcasis also found similar performance issues with the Samsung 970 EVO Plus but not the Samsung 970 EVO despite looking pretty similar hardware-wise. No indication that there was a firmware update for the Plus that resolved the issue.

Then I have seen results of 2500-2700 MB/sec R/W for both the Samsung 980 Pro and SN850X in an TB3 enclosure (i.e. one using the JHL7440 chipset such as the Arcasis TBU401/405) connected to an Intel Mac.

As such given the unknowns and interactions that would probably only make sense if we got Apple hardware and software engineers, Intel engineers, ASmedia engineers, and Lexar (and/or maybe even their controller manufacturer) all in the same room, I think we're stuck with educated guesses aka trial and error. I could see that an older enclosure with the JHL7440 chipset (e.g. the Arcasis TBU401/405), might bridge that Lexar SSD to your iMac with read/write speeds in the 2500-2700 MB/sec range. If so that would suggest an odd interaction between the Lexar SSD, the ASMedia bridge chip, and the Intel host controller. Or maybe it is just that Lexar SSD that doesn't work well with Intel host controllers for some reason. And/or maybe the relevant drivers are different enough between macOS for Apple Silicon and for Intel and the Intel one doesn't like the Lexar as much. In either of those cases keeping the Z666 but swapping in the SN850X would get you to the expected 2500-2700 MB/sec speeds.

Unfortunately I can only offer speculation and maybes at this point...

P.S.You might contact Zike and see if they have a firmware update. I have seen reports of substantial (e.g. 50%) performance improvements from updating the firmware of enclosures that use the same ASM2464PD chipset as your Zike. Supposedly 241129_85_00_00 is the latest. Also the ASM2464PD has a "U4 mode" setting and one could try tweaking that setting when installing new firmware. Unfortunately it appears all this needs to be done from Windows.
To complete the thread,
Zike have confirmed the Z666 was returned to me after checking and testing the firmware.
Thanks again bzgny2. Will live with what I have for the time being.
 
Last edited:
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.