Thank you for this detailed report! Notable aspect: afaik TRIM is not working with USB 3.1, apart from 10Gbps issue. It seems Thunderbolt/eSATA still the best alternative for fast external SSD's.
Last edited:
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
USB 3.1 PCIe for cMP info repository
- Thread starter stmp
- Start date
- Sort by reaction score
Thank you for this detailed report! Notable aspect: afaik TRIM is not working with USB 3.1, apart from 10Gbps issue. It seems Thunderbolt/eSATA still the best alternative for fast external SSD's.
TRIM is being added to the ASM chips with firmware updates. Startech has the updates listed on each product's page. Too bad the FW patch is locked per product, so it can't be used on any other enclosure. OS support for TRIM over USB is another matter entirely.
eSATA is the best value, but it's simply not as ubiquitous as USB. It's also not capable of 600-700MB/s for 2x SSD RAID 0.
TB is not an option for cMP.
The Oyen Digital Digital Mini Pro does the same exact thing. Cuts out after a few seconds at 400+MB/s, even with the external power supply. The ASM chip gets up to the exact same 105°C and locks up. Ho heatsink or thermal pad on the ASM chip. It's 1.3mm short of touching the bottom of the enclosure, so that nice, pretty extruded aluminum case really doesn't help much for sinking heat.
There's also a capacitor on the PCB(that is a combo of drive mounting tray and enclosure circuit board) that's inside the footprint of an 840 PRO, MX300, and a WD platter drive. All 3 just barely touch it. That's physically a bad design. They have a huge PCB compared to the other designs, and they had to put components right there?
This one's going back. It costs double what the cheap, plastickey ones do, and doesn't even observe simple board clearance rules.
Yes, it's really that close! I can't see light between the cap and the front of the SSD, even if I push the SSD back against the socket.
Ordered the Caldigit FASTA-6GU3 Plus, just to be sure it's not this card.
Satechi B01FWT2N3K with the Via VL716 arrives Tuesday. Will post something up probably Tues night on that.
There's also a capacitor on the PCB(that is a combo of drive mounting tray and enclosure circuit board) that's inside the footprint of an 840 PRO, MX300, and a WD platter drive. All 3 just barely touch it. That's physically a bad design. They have a huge PCB compared to the other designs, and they had to put components right there?
This one's going back. It costs double what the cheap, plastickey ones do, and doesn't even observe simple board clearance rules.
Yes, it's really that close! I can't see light between the cap and the front of the SSD, even if I push the SSD back against the socket.
Ordered the Caldigit FASTA-6GU3 Plus, just to be sure it's not this card.
Satechi B01FWT2N3K with the Via VL716 arrives Tuesday. Will post something up probably Tues night on that.
Just found: https://www.aliexpress.com/item/LA3...Type-C-Interface-4Pin-10Gbps/32763664731.html
Would this LA31-12C card work in a Mac Pro?
Would this LA31-12C card work in a Mac Pro?
I have yet to read of anyone having working USB-C 3.1 ports in the Mac Pro besides Caldigit.
UPS dropped off the Satechi B01FWT2N3K with the Via VL716 today, surprise!
It's TOTALLY STABLE!!!
5Gb/s USB3.1 with WD platter drive, Samsung 840 PRO, and MX300. 3 hours of non-stop 5GB read at 500MB/s then write at 450MB/s sequential benchmark with the MX300(the higher-wattage SSD), no problems. I forgot I had it running... Was supposed to be 20-30 mins.
While the ASM1351's all hit 105°C then lock up, the VL716 tops out at 125°F/65°C. That's measured with a Fluke 561 and a FLIR. I suspect it's a bigger difference than 100mW, but I only have a USB 2.0 power meter; no 3.0/3.1. Guessing here, but I'd expect to find that the Via chip has better power management, so while it my in fact be 100mW difference at peak, the VIA spends more time at a much lower power state.
The VL716 does look to have about 4-7% lower random 4K IOPS than the ASM1351. Sequential R/W is the exact same.
Still won't show link speed at 10Gb/s. I got the ASM1351 drive dock to show 10Gb/s one time; I think it was immediately after rebooting from Win7 to Sierra.
I used an Anker 3' Male USB-A to USB-C cable to go from the Ableconn PUSB31P2A card's USB-A port to the Satechi B01FWT2N3K's USB-C port.
Now that I have a known-stable enclosure, I think I'll pull the PUSB31P2A and try the PU31-1A1C(with 1x A and 1x C) again and see if the USB-C port works with this drive.
Avoid the ASM1351 enclosures like the plague.
VL716 still appears to not support TRIM in Sierra, but the drive shows as CT1050MX300S
Another piece of the puzzle: I have never seen any of the ASM1351 enclosures negotiate extra operating current. 80% more available power is a huge deal.
Update: The Satechi B01FWT2N3K works on the Ableconn PU31-1A1C thru the USB-C port. Still only 5Gb/s, and with the same extra operating current as with the USB-A port. Sys Profiler shows it as "USB 3.1 Bus" as well.
It's TOTALLY STABLE!!!
5Gb/s USB3.1 with WD platter drive, Samsung 840 PRO, and MX300. 3 hours of non-stop 5GB read at 500MB/s then write at 450MB/s sequential benchmark with the MX300(the higher-wattage SSD), no problems. I forgot I had it running... Was supposed to be 20-30 mins.
While the ASM1351's all hit 105°C then lock up, the VL716 tops out at 125°F/65°C. That's measured with a Fluke 561 and a FLIR. I suspect it's a bigger difference than 100mW, but I only have a USB 2.0 power meter; no 3.0/3.1. Guessing here, but I'd expect to find that the Via chip has better power management, so while it my in fact be 100mW difference at peak, the VIA spends more time at a much lower power state.
The VL716 does look to have about 4-7% lower random 4K IOPS than the ASM1351. Sequential R/W is the exact same.
Still won't show link speed at 10Gb/s. I got the ASM1351 drive dock to show 10Gb/s one time; I think it was immediately after rebooting from Win7 to Sierra.
I used an Anker 3' Male USB-A to USB-C cable to go from the Ableconn PUSB31P2A card's USB-A port to the Satechi B01FWT2N3K's USB-C port.
Now that I have a known-stable enclosure, I think I'll pull the PUSB31P2A and try the PU31-1A1C(with 1x A and 1x C) again and see if the USB-C port works with this drive.
Avoid the ASM1351 enclosures like the plague.
VL716 still appears to not support TRIM in Sierra, but the drive shows as CT1050MX300S
Another piece of the puzzle: I have never seen any of the ASM1351 enclosures negotiate extra operating current. 80% more available power is a huge deal.
Update: The Satechi B01FWT2N3K works on the Ableconn PU31-1A1C thru the USB-C port. Still only 5Gb/s, and with the same extra operating current as with the USB-A port. Sys Profiler shows it as "USB 3.1 Bus" as well.
Last edited:
KT4004 USB 3.0 (PCIe x1 2.0) connected to USB 3.1 enclosure with Samsung 840 Pro:
Ableconn PU31-1A1C USB 3.1 Gen2 (PCIe x2 2.0) connected to USB 3.1 enclosure with Samsung 840 Pro(USB-A and USB-C posted virtually identical numbers): 42% faster than USB 3.0 PCIe x1 2.0. +140MB/s
Ableconn PU31-1A1C USB 3.1 Gen2 (PCIe x2 2.0) through USB 3.0 Hub connected to USB 3.1 enclosure with Samsung 840 Pro(This shows the gain of having 2 PCIe lanes, with the hub artificially capping the enclosure to USB 3.0 speed): 17% faster than USB 3.0 PCIe x1 2.0. +56MB/s
Ableconn PU31-1A1C USB 3.1 Gen2 (PCIe x2 2.0) connected to USB 3.1 enclosure with Samsung 840 Pro(USB-A and USB-C posted virtually identical numbers): 42% faster than USB 3.0 PCIe x1 2.0. +140MB/s
Ableconn PU31-1A1C USB 3.1 Gen2 (PCIe x2 2.0) through USB 3.0 Hub connected to USB 3.1 enclosure with Samsung 840 Pro(This shows the gain of having 2 PCIe lanes, with the hub artificially capping the enclosure to USB 3.0 speed): 17% faster than USB 3.0 PCIe x1 2.0. +56MB/s
This may be irrelevant to this thread but I made a small discovery today, or at least I think I did, and I want to mention it here.
It was my understanding that Apple would not allow an external drive to boot through a third party USB 3.0 port and I always had my portable SSD drive directly connected to a USB port on my MacBook Air to boot from the drive (without a hub in between). Three days ago, I finally got around setting up my CalDigit TS2 (Thunder Station 2) that acts as an external dock which is connected to my MBA through a Thunderbolt cable.
Now I tried connecting my portable drive to TS2 and boot from there and I was thunder struck (no pun intended) when I found out that this TS2 had allowed my external drive to boot through a USB-A port on the device. I have tired booting from some USB 3.0 and 3.1 ports (at 3.0 speed) on PCIe adapters installed in my Mac Pro 5,1 but this was not even close to possible.
I wonder if this is Thunderbolt's built-in spec and, if so, I may seriously consider moving up my main machine to Mac Pro 6,1 or wait for 7,1 to come out.
It was my understanding that Apple would not allow an external drive to boot through a third party USB 3.0 port and I always had my portable SSD drive directly connected to a USB port on my MacBook Air to boot from the drive (without a hub in between). Three days ago, I finally got around setting up my CalDigit TS2 (Thunder Station 2) that acts as an external dock which is connected to my MBA through a Thunderbolt cable.
Now I tried connecting my portable drive to TS2 and boot from there and I was thunder struck (no pun intended) when I found out that this TS2 had allowed my external drive to boot through a USB-A port on the device. I have tired booting from some USB 3.0 and 3.1 ports (at 3.0 speed) on PCIe adapters installed in my Mac Pro 5,1 but this was not even close to possible.
I wonder if this is Thunderbolt's built-in spec and, if so, I may seriously consider moving up my main machine to Mac Pro 6,1 or wait for 7,1 to come out.
It's not about Apple allowing or not allowing anything. The cMP's firmware has no idea what to do with third party USB 3.0 cards. The card starts working after the OS loads the proper drivers.
You should be able to boot from an external drive just fine from any of the native ports on the nMP, whether USB or TB.
You should be able to boot from an external drive just fine from any of the native ports on the nMP, whether USB or TB.
Thank you for your clarification. That's what I have figured out. I realize I may have been buried too deep in my cMP and didn't see the game changing.
Anyway, on my cMP I am booting from my external SSD via an eSATA port on FASTA 6GU3 Plus at 6G speed and while on other iMacs and MBA via their native USB 3.0 ports at 5G. This arrangement has been working just fine. The TB connection to MBA I mentioned above has given me a glimpse of new light.
Incidentally CalDigit has released their FASTA 6GU3 Plus firmware updater which will make the card USB 3.1 Gen2 capable and I will try that on my cMP and see how 10G will work. To get the updater you must contact their support and ask for it as I don't see it readily downloadable on their site. Sierra is require and all cables connected to the back of the card must be removed before the installation takes place. Their information is here:
http://www.caldigit.com/FASTA-6GU3plus/
I would love to hear back from you and add the resulting information to the big USB thread.
http://www.caldigit.com/support/FASTA-6GU3-Plus/CalDigit-FASTA-6GU3-Plus-10G-Firmware.zip
Since it's not on their site yet.
Since it's not on their site yet.
I'm curious with respect to how much RF interference this card generates with respect to BlueTooth (e.g. mouse).
Just ran the 10gbps update on the FASTA-6GU3 Plus. No 10gbps in system profiler.
The USB-C connector never runs at anything but 480mb before update. After update, I only get 5gbps. Tried 2 different enclosures and 3 different cables.
I tried with ASM1351 and VL716 and only still get 5gbps. Benchmarks still unchanged as well.
Mac OS 10.12.2
Aux SATA power cable is connected as well.
Caldigit looks like a fail. Between the lack of 10gbps after firmware update and the USB-C port working at 480mbps, unless you desperately need to have eSATA and USB 3.x in a single slot, the card isn't too attractive, especially at the price.
[doublepost=1483733462][/doublepost]
I've noticed no issues with BT since dropping in any of the USB 3.1 cards with ASM1142 chips.
The USB-C connector never runs at anything but 480mb before update. After update, I only get 5gbps. Tried 2 different enclosures and 3 different cables.
I tried with ASM1351 and VL716 and only still get 5gbps. Benchmarks still unchanged as well.
Mac OS 10.12.2
Aux SATA power cable is connected as well.
Caldigit looks like a fail. Between the lack of 10gbps after firmware update and the USB-C port working at 480mbps, unless you desperately need to have eSATA and USB 3.x in a single slot, the card isn't too attractive, especially at the price.
[doublepost=1483733462][/doublepost]
I've noticed no issues with BT since dropping in any of the USB 3.1 cards with ASM1142 chips.
Last edited:
Okay, here are the BlackMagic benchmark results.
My Caldigit FASTA-6GU3 Plus came standard with USB 3.1 Gen1 (5G) and I connected an external SSD drive in a MiniPro RAID enclosure, with two OWC Mercury Extreme Pro 6G 960GB SSDs at hardware RAID 0 inside, to the USB-C port of the FASTA card.
I got 394.0 Write/427.3 Read. These are the original values before the FW update.
Next I pulled all unnecessary cables out from the back of my Mac Pro and installed CalDigit-FASTA-6GU3-Plus-10G-Firmware 1.0 and updated the card.
The new values I got are 682.4 Write/743.5 Read. The values nearly doubled but are these the USB 3.1 Gen2 (10G) speed?
I had mixed results at first but after choosing a correct cable these speed were attained. You really have to be choosy on the cable you use to connect these devices right.
To clarify, I used a USB 3.1 Gen2 (10G) grade USB-C to USB-C cable. If you use a USB 3.0 cable you only get the USB 3.0 speed. Some USB 3.1 cables may actually be only Gen1 capable. I hear a Thunderbolt 3 cable would work also because it has a pair of USB-C connectors but I don’t know if its true.
The FASTA card also has a USB-A port and I had a chance to test it with a USB 3.1 Gen2 grade USB-A to USB-C cable and I obtained similar values of 693.7/709.2.
Below is the MiniPro's info on USB 3.1 bus.
What does this tell you?
Last edited:
It still shows 5gb/s. I wonder if this is a flaw in system profiler.
USB 3.1 5gb/s with 128/130-bit encoding and protocol overhead should max out around 550MB/s. Max possible theoretical is 610MB/s.
You're definitely getting a link rate over 5gb/s. 743.5MB/s on your benchmark should be around 6.5-7.0gb/s over the link.
10gb/s should max out in the range of 1.0-1.15GB/s. The ASM1142 USB chip runs over 2x PCIe 2.0, which will max out around 700-900MB/s, greatly depending on the chipset of the Mac/PC it's plugged into. You'll max out the PCIe slot before the USB 3.1 10gb/s link saturates.
USB 3.1 5gb/s with 128/130-bit encoding and protocol overhead should max out around 550MB/s. Max possible theoretical is 610MB/s.
You're definitely getting a link rate over 5gb/s. 743.5MB/s on your benchmark should be around 6.5-7.0gb/s over the link.
10gb/s should max out in the range of 1.0-1.15GB/s. The ASM1142 USB chip runs over 2x PCIe 2.0, which will max out around 700-900MB/s, greatly depending on the chipset of the Mac/PC it's plugged into. You'll max out the PCIe slot before the USB 3.1 10gb/s link saturates.
Just for reference, here is the way I have my MacPro5,1 configured and the Caldigit is in slot 2 because I understand Squid would slow down in slot 1 and 2.
Mac Pro 5,1 2010
Intel Xeon X5690 3.46GHz x2 (12 core)
Memory 1333MHz DDR3 ECC 16GB x6 96GB
Nvidia GeForce GTX980 Ti (PCIe Slot 1) 6GB cache
CalDigit FASTA 6GU3 Plus (PCIe Slot 2) USB 3.1 Gen2 10G x2, eSATA 6G x2
Amfeltec Squid Gen2 M.2 x16 (PCIe Slot 3) SM951 x4, SoftRAID 0 (1.9TB)
Highpoint Rocket RAID 4520 (PCIe Slot 4) SanDisk Extreme Pro x8, SoftRAID 0 (7.8TB)
BTW I will show two more benchmarks I did on this setup: Squid and Highpoint. This has little to do with USB 3.1 subject but I guess this is about the max performance a cMP can get.
Squid RAID
Highpoint RAID
Nice. Have you tried the Squid in the second x16 slot? If it's the x16 Squid, you should get more bandwidth from the x16 slot, even if it's at PCIe 1.0 link speed.
May be worth a shot running the Squid in slot 2 and Caldigit in slot 3.
PCIe x16 1.0 max real-world is very roughly 3.0GB/s. PCIe x4 2.0 tops out about 1.3-1.5GB/s. even if the Squid runs at only 1.0 speeds, the wider link should still provide better speed.
May be worth a shot running the Squid in slot 2 and Caldigit in slot 3.
PCIe x16 1.0 max real-world is very roughly 3.0GB/s. PCIe x4 2.0 tops out about 1.3-1.5GB/s. even if the Squid runs at only 1.0 speeds, the wider link should still provide better speed.
Last edited:
Yes, that's what I was thinking and I'm going to try next. I will post the result here again. Thanks for your comment.
Gonna go way outside the box on this thought: the 8-drive RAID 0, if you're simply running Mac OS soft RAID, you may get better speeds putting 2-4 drives on the internal SATA connectors and the remaining 4-6 on the Highpoint.
The internal SATA connectors are good for 265-285MB/s each. The Highpoint over the PCIe slot is giving you about 1200MB/s.
My logic is that with it being striped data, 3 drives on internal SATA will give you ~800MB/s and the remaining 5 drives will still be able to max out the Highpoint's PCIe. At some point, you'll saturate the southbridge's DMI link. Something like 3+5 or 4+4 may get you just over 2GB/s total throughput.
This is assuming the slot the Highpoint is using is on the northbrige. If the slot is using the southbridge's PCIe lanes, you'll max out around 1.5GB/s total.
Going a step further, you could put 1 or 2 of the SSD's on the Caldigit's eSATA ports for another couple-few hundred MB/s. That card's total slot bandwidth is around 1.6GB/s.
You may get close to 2.5GB/s with 3 on Highpoint, 3 on internal, and 2 on eSATA.
Last edited:
I recently bought a thunderbolt to usb A adapter for my iMac 2011 27in , and i found out this adapter from china Alibaba is using asm1142 host controller , does it mean i will get usb 3.1 10g speed? i do not have a USB 3.1 ssd to test, I thought usb 3.1 10g only works with usb C port, but this adapter comes with a USB A port.
I have a ssd enclose with ASM1153e with UASP and trim support (Firmware 140509A182_40), and USB to Msata with VL713 chipset, they both 5g bridge MAX out speed on this adapter at 360MB/s .
right now i am using a USB 3.0(3.1 G1) hub with VL813 chipset. is it a wast, if this adapter can go up to 10Gbp?
I have a ssd enclose with ASM1153e with UASP and trim support (Firmware 140509A182_40), and USB to Msata with VL713 chipset, they both 5g bridge MAX out speed on this adapter at 360MB/s .
right now i am using a USB 3.0(3.1 G1) hub with VL813 chipset. is it a wast, if this adapter can go up to 10Gbp?
Not really a waste.
USB-A ports can do 10gbps; actually getting 10gbps link rate with most cables and with Mac OS is difficult at best.
ASM1142 operates over x2 PCIe 2.0. Most(pretty sure all) USB 3.0 controller chips are x1 PCIe 2.0, so you should get a little better performance there, assuming your TB-PCIe enclosure is x2 PCIe 2.0 or better.
5gbps USB 3.1 is slightly faster than 5gbps USB 3.0. USB 3.0 uses 8b/10b encoding; for every 10bits sent down the wire, 8 bits are data and 2 are 'lost' to error detection. USB 3.1 uses 128b/130b encoding, so you get 128 bits of data to 2 bits of overhead. This means the max data bitrate of USB 3.0 is 4gbps and USB 3.1 Gen 1 is 4.92gbps. You lose a few percent more on each to addressing, commands, packet spacing, etc.
Encoding is 128/132, not 130
ASM1142 configures itself x1 when connected to PCIe 3.0 bus (see specs on ASMedia site) and only in this case it will be able to use effectively 128/132b encoding.
Connected to PCIe 2.0 it will still use 8/10b (2.0 can't do 128/132) and speed gain over USB 3.0 is only due to 2 lane config.
Encoding is 128/132, not 130
ASM1142 configures itself x1 when connected to PCIe 3.0 bus (see specs on ASMedia site) and only in this case it will be able to use effectively 128/132b encoding.
Connected to PCIe 2.0 it will still use 8/10b (2.0 can't do 128/132) and speed gain over USB 3.0 is only due to 2 lane config.
So USB 3.1 encoding is 8b/10b when connected over PCIe 2.0??? Pretty sure one bus's encoding doesn't dictate the other. I.e. USB 3.1 has to use 128/132 by the spec.
I documented a 100 MB/s speed increase further up in this thread by switching to direct USB 3.1 5.0gb versus having a 3.0 hub between. The speeds I was seeing would be impossible at 8/10 5gbps.
I'm aware that PCIe 2.0 is 8/10 and 3.0 is 128/132. I'm referring to the encoding on USB 3.0 vs 3.1.