review of acasis TBU42 thunderbolt NVMe enclosure/dock/hub

[whodiini]

Someone was interested in this device, so am starting a new thread on using this on a M1 Studio Max.

First off, I want to reference a prior review, so I dont have to repeat what they say

Acasis Thunderbolt 3 External M.2 NVME Docking Station TBU41...

I recently purchased an Acasis TBU412. This compact Thunderbolt 3 NVME Docking Station retails for approximately the same price as the Jeyi ThunderDoc...

egpu.io

The part number TBU412 and TBU42 I believe reference the same thing.

I tried this with a 2 TB Hynix P31 gold NVMe. I have it running in an Acasis self powered TB3 enclosure, so I can compare.

I dont like that the cover is magnetically attached because you cant provide pressure onto the thermal pad. I prefer screws like my other Acasis. That said, hte thermal pad provided is exactly the right thickness. I am guessing it is 1.5mm thick. But the Hynix is very efficient and doesnt generate a lot of heat, so it should be OK with the Hynix. Others I am not so sure.

I plugged it directly into a back port of the Studio Max so it has all the 40Gbps bandwidth available. I first plugged it in. Nothing. Then I added power to the USB C power. It turned on. So it requires external power, period. I then ran a series of tests with Blackmagic to see how the speed of hte NVMe performed without and with different devices connected to the second thunderbolt port. I rebooted each time I changed hardware so that the new hardware config was recognized. This is where it got interesting and complicated.

Blackmagic NVME only: 2384MB/sec Write, 2776 Read. Decent speeds.

Next I connected a LG 5k monitor to the second TB port.
Blackmagic NVME + LG 5k: 908MB/sec Write, 2754 Read. LG 5k takes 22 Gbps bandwidth, leaving 18 Gbps available. This is lower write than what I expected and higher read than expected. Perhaps the 22Gbps bandwidth is not symmetrically split between read and write? I am just speculating here hte reason. But clearly 1)the write speed of the NVMe is unacceptably degraded 2) the 2nd TB port sets the bandwidth and the NVMe uses whats left of hte bandwidth.

Next I connected a Sabrent USB 3.1R2 NVMe enclosure to the second TB port via a Sabrent TB3 cable.
Blackmagic NVMe + sabrent USB 3.1R2 w/Sabrent TB3 cable: 450 MB/sec write, 2753 read. huh?? The Sabrent TB3 cable degraded the NVMe speed. I dug through my collection of cables and found a Belkin 10Gbps Superspeed labeled cable and replaced the Sabrent TB3 cable with the Belkin SS cable.
Blackmagic NVMe + Sabrent USB 3.1R2 w/Belkin SS 10 Gbps marked cable: 2384MB/sec Write, 2776 Read. Sabrent: 853 write, 924 read
I believe these 2 NVMe max out the bandwidth of the TB bus. Remember that the data bandwidth max is 40 - 8 (reserved for graphics) = 32 Gbps.

There is no reason to connect up a second TB NVMe enclosure since hte bandwidth would be limited. For fun, I connected an envoy express 2 PCIe 3.0 lane TB3 and the performance of both dropped.

Next I will try to reduce the suppled power to see how small a power supply I can use, since I am not using it to provide 60W to a laptop, which is what this is spec'd at.

My bottom line is that for my needs, it really doesnt provide much more than a bus powered enclosure, and given that my acasis cover can be held down by screws, I prefer the $20 cheaper enclosure only.

It also has a displayport, but it is only 1.3 because it uses an older intel chip. The newer intel chip does displayport 1.4. Probably did it to save a few dollars ~ $5?

PS. I will write another thread about enclosures and cables with the M1 Studio Max. The situation with USB 3.1Gen2 is so problematic I would just avoid that whole mess and go with TB3.

 

[joevt]

Blackmagic NVME only: 2384MB/sec Write, 2776 Read. Decent speeds.

Those are not bad. Better than 22 Gbps for read. I would like to see AmorphousDiskMark speeds.

Next I connected a LG 5k monitor to the second TB port.
Blackmagic NVME + LG 5k: 908MB/sec Write, 2754 Read. LG 5k takes 22 Gbps bandwidth, leaving 18 Gbps available. This is lower write than what I expected and higher read than expected. Perhaps the 22Gbps bandwidth is not symmetrically split between read and write? I am just speculating here hte reason. But clearly 1)the write speed of the NVMe is unacceptably degraded 2) the 2nd TB port sets the bandwidth and the NVMe uses whats left of hte bandwidth.

LG UltraFine 5K takes ≈ 29 Gbps when it's doing 10bpc.
2560x2880 59.999898 Hz 177.720 kHz 469.180000 MHz * 2 * 30bpp = 28.15 Gbps.
2560x2880 59.996475 Hz 177.710 kHz 483.370000 MHz * 2 * 30bpp = 29.00 Gbps.

DisplayPort traffic is mostly one direction (transmit) so it affects mostly transmit/write performance of other devices.

With only 11 Gbps remaining, 908 MB/s is pretty good. Maybe AmorphousDiskMark will show a slightly higher number.

Next I connected a Sabrent USB 3.1R2 NVMe enclosure to the second TB port via a Sabrent TB3 cable.
Blackmagic NVMe + sabrent USB 3.1R2 w/Sabrent TB3 cable: 450 MB/sec write, 2753 read. huh?? The Sabrent TB3 cable degraded the NVMe speed. I dug through my collection of cables and found a Belkin 10Gbps Superspeed labeled cable and replaced the Sabrent TB3 cable with the Belkin SS cable.
Blackmagic NVMe + Sabrent USB 3.1R2 w/Belkin SS 10 Gbps marked cable: 2384MB/sec Write, 2776 Read. Sabrent: 853 write, 924 read
I believe these 2 NVMe max out the bandwidth of the TB bus. Remember that the data bandwidth max is 40 - 8 (reserved for graphics) = 32 Gbps.

There is no reason to connect up a second TB NVMe enclosure since hte bandwidth would be limited. For fun, I connected an envoy express 2 PCIe 3.0 lane TB3 and the performance of both dropped.

Strange that the cable would affect write speed like that.

There's No evidence of a reserve for graphics when no displays are connected. There's probably a limit for PCIe traffic on a single Thunderbolt cable even for integrated Thunderbolt host controllers which are not limited by a 31.5 Gbps PCIe gen 3 x4 connection that discrete Thunderbolt controllers have but we don't know the details.

I would use ATTO Disk Benchmark.app to test total bandwidth of multiple storage devices simultaneously. It's the only benchmark that I know that will add them up for you.

 

[whodiini]

Those are not bad. Better than 22 Gbps for read. I would like to see AmorphousDiskMark speeds.

Here is Amorphous with and without the Sabrent connected is the same:

The sabrent:

Runninng ATTO: both together: (3000MB/s combined write = 2000MB/s from the P31 + 1000MB/s from sabrent!)

P31 stats:

sabrent stats:

 

[whodiini]

I will try a test tomorrow chaining 2 TB3 drives and running ATTO to see if I can saturate the bandwidth.

This P31+Sabrent looks like it hasnt yet saturated.

 

[joevt]

I will try a test tomorrow chaining 2 TB3 drives and running ATTO to see if I can saturate the bandwidth.

This P31+Sabrent looks like it hasnt yet saturated.

It appears that ATTO is not properly bypassing the read cache.

 

[whodiini]

It appears that ATTO is not properly bypassing the read cache.

yes, absurdly high read rates. Thats why I only mentioned the writes.

On the windows version, you can check Direct I/O to bypass caches, but there is no option on the mac version

 

[whodiini]

OK, to test the saturation of hte thunderbolt bus. I connected a second TB3 after the ACASIS. It is a Orico TB3 with a XLR CS3030 original components.

First with it connected, I ran diskmagic. I get weird results. The ACASIS NVme gets

Blackmagic NVMe + Orico TB3 w/ TB3 cable: 500 MB/sec write, 2683 read. huh??
I also ran blackmagic on the Orico: 1502 MB/sec write, 2481 read. (Plugged directly into the back of the studio max, the orico tests at 2251 write, 2485 read)
Instead of swapping cables (since I believe in TB3 cables), I ran amorphous on the combo. Again, ignore the reads, cache is not disabled.

Both NVMe, 3000 MB/sec write speed. Since this is not different than the Acasis+Sabrent, I conclude that the TB bus is saturated.

Now when I look at each disk individually, it gets interesting.
The Acasis NVMe is performing at its peak, unlike what Blackmagic reports. It is writing at 2000 MB/sec

However the Orico is throttled down to 1000 MB/sec.

So that means
1) The TB bus is saturated.
2) The speed of the second daisy chained disk is limited to 1000 MB/sec, so putting a TB enclosure or a USB 3.1R2 enclosure have the same speeds.
3) Blackmagic low write test speeds - something strange is happening that isnt reflected in ATTO. I am going back and doing the same test earlier with the cable that got me slow speeds and see if ATTO reflects that.

 

[joevt]

2) The speed of the second daisy chained disk is limited to 1000 MB/sec, so putting a TB enclosure or a USB 3.1R2 enclosure have the same speeds.

I wonder if the order you add the disks to ATTO matters. Also, the second disk shouldn't be limited to 1000 MB/s if the first disk is not being written to. It's not likely that you'll be writing to multiple disks at the same time in normal tasks.

 

[whodiini]

I wonder if the order you add the disks to ATTO matters. Also, the second disk shouldn't be limited to 1000 MB/s if the first disk is not being written to. It's not likely that you'll be writing to multiple disks at the same time in normal tasks.

When I run blackmagic, it runs each disk separately. When I run ATTO, I see both disks are simultaneously doing R/W.

 

[whodiini]

One more interesting test:
Apple charges ~$1200 for 4GB of fast storage. What can one do otherwise?

Apple's storage is fast. Amorphous disk test on my Studio Max:

I took 2 Acasis TB3 enclosures and put 2 identical Hynix P31 Gold 2TB, and used disk utility to raid stripe them. I connected each one to one port on the back of the Studio max so they each had all the bandwidth of the port.
Amorphous disk test of the striped RAID:

The striped RAID is in the same ballpark as the native Studio Max storage!

Cost of this: 2 x ($90 enclosure +$160 P31 gold) = $500. So you can with carefully chosen components, save about half the cost of additional built in storage with a small hit on performance.

 

[whodiini]

Plugged it into a 30W GaN charger the size of an iphone charger and it runs just fine.

I happened to try several on hand. Most worked, one intermittently worked.
If it worked on power up, it would stay working. But if it didnt work on power up, it wouldnt work until I unplugged and replugged it in. Dont know if that one was defective or there is an issue with the design.

With one NVMe installed and another Sabrent USB 3,1R2 enclosure attached with a NVMe, my wattmeter shows 4-6W. When the computer is off, it draws 2W. A 30W charger should be plenty if not providing power to a laptop.

I have tried 4 30W chargers. One was Gan I, another GaN II, the other 2 were mosfet based. Three act similarly. They sometimes work. I havent figured out why, upon plugging them in, they work or they dont. I have to repeatedly plug them in and unplug until it starts powering the acasis. The one that works reliably is an older Anker Atom GaN I from 2019. It has PD 1, no PPS, no QC. All the other 3 chargers are PD3.0. One says PPS, the others dont say PPS, but it doesnt seem to matter. I have tried different cables as well, and no difference. 65W chargers seem to work more reliably. But I do not understand it because it only draws 6W, so a 30W charger should work fine. (I check on a wattmeter). I emailed Acasis and asked the specs on the power supply needed and they replied that the power supply they provide is GaN 20V @ 3A. From that I assume it needs 20V, but that seems high. So for those who are interested, the smallest one that works is an Anker Atom 30W PD 1. All the 65W chargers seem to work fine.

 

[Infavito]

Have you ever checked the SMART data of your SSD? I have noticed that the number of "Unsafe Shutdowns" in SMART DATA increases after every "Safe Disconnect" of storage from my Macbook pro M1. I tested Orico TOM2T3 and Acasis TBU405 and got the same results...

 

[Infavito]

One more interesting test:
Apple charges ~$1200 for 4GB of fast storage. What can one do otherwise?

Apple's storage is fast. Amorphous disk test on my Studio Max:

View attachment 2076371

I took 2 Acasis TB3 enclosures and put 2 identical Hynix P31 Gold 2TB, and used disk utility to raid stripe them. I connected each one to one port on the back of the Studio max so they each had all the bandwidth of the port.
Amorphous disk test of the striped RAID:

View attachment 2076372
The striped RAID is in the same ballpark as the native Studio Max storage!

Cost of this: 2 x ($90 enclosure +$160 P31 gold) = $500. So you can with carefully chosen components, save about half the cost of additional built in storage with a small hit on performance.

Your article inspired me a lot. Thank you! It's a little sad that I've read it just now, and I already have a rather expensive Samsung 980 Pro 2 Tb SSD. Now I see that Hynyx would be a better choice.

 

[whodiini]

Have you ever checked the SMART data of your SSD? I have noticed that the number of "Unsafe Shutdowns" in SMART DATA increases after every "Safe Disconnect" of storage from my Macbook pro M1. I tested Orico TOM2T3 and Acasis TBU405 and got the same results...

I just looked at the smart data on my 2 external thunderbolt drives. Nothing abnormal. One shows about 8 unsafe shutdowns, the other 4. I ejected one of the drives, then remounted it, checked Smart and the unsafe shutdown was still the same number. That drive I ejected was a RAID of a Acasis TBU405 and TBU42

 

[Infavito]

I changed previous SSD for extremely new Samsung 980 pro 2TB. And I've got the same result!
I did the test on another freshly installed MacOS (I have 2 MaсOS on internal SSD) then. And I got +1 “Unsafe Shutdown”! What's the hell?! Video

 

[Infavito]

I changed previous SSD for extremely new Samsung 980 pro 2TB. And I've got the same result!
I did the test on another freshly installed MacOS (I have 2 MaсOS on internal SSD) then. And I got +1 “Unsafe Shutdown”! What's the hell?! Video

I've just received a response from Acasis support. And they admit the problem! Here is what they write: "Sorry for making you wait. We carefully reviewed the videos and pictures you recorded, for this problem, our technical staff also tried very hard to make a lot of attempts, there is still no good solution, other products on the market are also the same, currently only after the shutdown will not be produced this situation , thank you for your support of our products, follow-up questions can contact us."
I hope they will find the way how to prevent these "Unsafe shutdowns".

 

[EncryptedUser]

I received my TBU42 today.
Was planning to use it as a Time Machine and monitor output hub.

But pretty much immediately I ran into some trouble.

It appears that if I route video through the TBU42 - the acasis is unable to wake my Eizo monitor after it’s gone down into standby.

I’m forced to disconnect and reconnect the DisplayPort or USB-C cable to get the TBU to wake the Eizo.

If Use my OWC TB3 hub set up and connected as the TBU - the OWC have no problem at all.

Non working:
Mac mini M1 > TB4 cable > powered TBU42 > DisplayPort > Eizo = unable to wake Eizo.


Non working:
Mac mini M1 > TB4 cable > powered TBU42 > USB-C—DisplayPort > Eizo = unable to wake Eizo.



All works fine with OWC.

Mac mini > TB4 cable > powered OWC TB3 Hub > USB-C > DisplayPort > Eizo = works perfect.


Anyone have a setup like mine or even might be able to test it?

 

[steve123]

I hope they will find the way how to prevent these "Unsafe shutdowns".

Did Acasis ever resolve the unsafe shutdowns issue?

 

[steve123]

I’m forced to disconnect and reconnect the DisplayPort or USB-C cable to get the TBU to wake the Eizo.

Did you resolve your issue with the TBU42? I am looking for an enclosure to use with my MP 6,1 (Thunderbolt 2) with an Apple TB3 to TB adapter. I require an enclosure that can be powered since the Apple adapter does not pass power between the two ports.

 

[EncryptedUser]

Did you resolve your issue with the TBU42? I am looking for an enclosure to use with my MP 6,1 (Thunderbolt 2) with an Apple TB3 to TB adapter. I require an enclosure that can be powered since the Apple adapter does not pass power between the two ports.

No. Unfortunately not.
For a while I simply began turning off my Eizo but soon got annoyed and bought a second OWC mini TB4 dock and connected my Eizo to it.

This is the reply I got from Acasis.
I asked them some follow ups but they never responded again.

According to your description: the product TBU42 can't wake up the extended screen in standby mode. In response, we fed the situation to our R&D staff, and a few days later, we received a reply from the R&D staff, who said that our product TBU42 does not have a hibernation function and can only manually eject the device when it is not in use, and that this situation may be due to the hibernation function set by the computer.

 

[whodiini]

Did you resolve your issue with the TBU42? I am looking for an enclosure to use with my MP 6,1 (Thunderbolt 2) with an Apple TB3 to TB adapter. I require an enclosure that can be powered since the Apple adapter does not pass power between the two ports.

I use this asasis enclosure that REQUIRES external power. I have been using it with a cheap 30W USB C power supply. It does not require a 60W power supply if you arent powering a labtop thru the passthru port.

Acasis 6-in-1 40Gbps M.2 NVME SSD Enclosure DP 4K60Hz Docking Station

📢 This product is sold out, but don't despair. We have launched upgraded and optimized new styles. Welcome to visit our product page and click "Upgrade Version" to learn more. 6 In 1 Multifunctional Expansion ACASIS Thunderbolt 3 40Gbps enclosure, also a great docking station, offers a variety...

www.acasis.com

 

[steve123]

TBU42 does not have a hibernation function and can only manually eject the device when it is not in use, and that this situation may be due to the hibernation function set by the computer.

Which mac are you using? I found this thread that discusses an ongoing problem with waking external monitors ...

https://forums.macrumors.com/threads/macos-ventura-m1-macs-external-displays.2371555/

 

[EncryptedUser]

ahh yes I’ve read those posts but I’ve not have had any of those problems.

I’m on a Mac mini M1 which is set to never sleep. But is set to turn off the monitor instead after 20 minutes of no use. And locks the screen.

Using a ALOGIC Ultra USB-C to DisplayPort cable from M1 Mini directly to Eizo works perfect.
The mini turns off the display signal and Eizo goes into power down/standby.
Any mouse or keyboard activity wakes up the Eizo and asks for password.

If I route the ALOGIC Ultra USB-C to DisplayPort cable via an OWC TB4 mini hub it’s works exactly the same and shows no problem.

Routing the ALOGIC Ultra USB-C to DisplayPort cable or what so ever any DP-DP 4K/8K via the TBU42 and I need to pull and reconnect the cable to get the Eizo to wake up.

So it’s clearly some missing support for protocols in the TBU42.

 

[steve123]

I’m on a Mac mini M1 which is set to never sleep

LOL, that sounds like my MP 6,1 ... cannot let it go to sleep or you will have a kernel panic when it wakes up.