Sabrent Launches Massive 16 TB Variant of XTRM-Q Thunderbolt+USB3 External SSD

[btarunr]

Sabrent on Monday launched a massive 16 TB variant of its XTRM-Q external SSD that debuted last year in capacities of up to 8 TB. The 16 TB XTRM-Q encloses two 8 TB SSD subunits. Out of the box, these are striped to create the 16 TB volume, but you can re-configure these to run in RAID 1, for data redundancy, or even JBOD/sequential.

What makes the Sabrent XTRM-Q stand out is its Thunderbolt 3 + USB 3.2 dual-mode operation. Unlike the single-drive based XTRM-Q 8 TB, the 16 TB version features a larger body, and requires external power through an included power brick, in addition to the Thunderbolt/USB connection. In striped mode, the driver offers sequential transfer rates of up to 2500 MB/s in Thunderbolt mode. The enclosure is made of aluminium, and doubles up as a heatspreader for the drives underneath. The enclosure is covered by a removable silicone outer sheath. Without this, the drive measures 114.3 mm x 65 mm x 17.2 mm (LxWxH). Available now, it is priced at USD $2,899.



View at TechPowerUp Main Site

 

[kayjay010101]

2,5GB/s when striped is really bad. Too fast to be SATA (which would be around 1GB/s with two drives striped) but way too slow to be NVMe.

What's even the point of it being thunderbolt if it's limited to 20gbps? Or is it TB2?

 

[Valantar]

That power brick is a joke, right? Right? It's 2x the size of the SSD, ffs ... Couldn't they just run this off two 15W USB-C/TB ports? Or at least give it USB-PD support so people could use their laptop chargers?

 

[InVasMani]

The more you buy the more you save...

 

[bonehead123]

Well, the guys that built the pyramids told me that they made something like this, back in the day, like, you know, 821k years ago, give or take a few centuries or so...

And FYI, they also said they were slow as crap then, and apparently still are, hehehe

 

[kpderkp]

Separate power adapter for PSSD??

 

[BluesFanUK]

$3k? Do one. Utterly ridiculous.

 

[yotano211]

2,5GB/s when striped is really bad. Too fast to be SATA (which would be around 1GB/s with two drives striped) but way too slow to be NVMe.

What's even the point of it being thunderbolt if it's limited to 20gbps? Or is it TB2?

What are you complaining about. Its fast. Dont like it, design your own equipment

$3k? Do one. Utterly ridiculous.

A 8tb nvme is about 1200-1400, just double that price.

 

[kayjay010101]

What are you complaining about. Its fast. Dont like it, design your own equipment


A 8tb nvme is about 1200-1400, just double that price.

What a dumb remark. The "don't like it, do it yourself" thing is so dumb I don't even know how to respond to that.

Bottom line is it's not fast enough to warrant being TB3 and that cost certainly gets passed onto the consumer, and Sabrent probably uses the excuse of it being TB3 to add some extra margin. It's lying and directly anti-consumer.

It doesn't warrant being TB3 with that speed, all that does is make people believe it's high end (cuz thunderbolt gud) to increase prices. It's slower than single entry-level QLC drives, never mind the fact that it's supposedly striped so there's two drives in there. I'd expect it to saturate the TB3 40Gbps spec easily but it caps out at 20Gbps which makes no sense if it's NVMe. It must be artifically limited somewhere, the drives have to be faster than that. As I said, even the slowest single NVMe drive is still faster than these TWO STRIPED drives. If they were SATA based it's too fast (two striped SATA drives would cap out at around 1GB/s, not 2,5GB/s), so it must be NVMe, but the rated speed just makes no sense.
And if it IS TB3 and the limit is really in the drives or controller; why advertise it being TB3 or implement the required hardware for it to be TB3 if it can't reach it? It just ups the cost for zero reason...
This is like advertising HDMI 2.1 on a 1080p 60Hz monitor...

 

[Valantar]

What a dumb remark. The "don't like it, do it yourself" thing is so dumb I don't even know how to respond to that.

Bottom line is it's not fast enough to warrant being TB3 and that cost certainly gets passed onto the consumer, and Sabrent probably uses the excuse of it being TB3 to add some extra margin. It's lying and directly anti-consumer.

It doesn't warrant being TB3 with that speed, all that does is make people believe it's high end (cuz thunderbolt gud) to increase prices. It's slower than single entry-level QLC drives, never mind the fact that it's supposedly striped so there's two drives in there. I'd expect it to saturate the TB3 40Gbps spec easily but it caps out at 20Gbps which makes no sense if it's NVMe. It must be artifically limited somewhere, the drives have to be faster than that. As I said, even the slowest single NVMe drive is still faster than these TWO STRIPED drives. If they were SATA based it's too fast (two striped SATA drives would cap out at around 1GB/s, not 2,5GB/s), so it must be NVMe, but the rated speed just makes no sense.
And if it IS TB3 and the limit is really in the drives or controller; why advertise it being TB3 or implement the required hardware for it to be TB3 if it can't reach it? It just ups the cost for zero reason...
This is like advertising HDMI 2.1 on a 1080p 60Hz monitor...

"Doesn't warrant being TB3" is rather meaningless though. There are likely no USB 3.2G2x2 controllers that can connect two m.2 drives. As for performance, they're likely splitting the lanes from the TB3 controller out with two for each drive (there's no way there's a PLX swich in there), so that likely explains that. Sure, that means it's slower than a single-drive solution, but (not that it matters seeing how I don't have £3000 to spend on this, nor would want to) IMO I'd rather have a split lane solution than waste the extra power on a PLX.

 

[yotano211]

What a dumb remark. The "don't like it, do it yourself" thing is so dumb I don't even know how to respond to that.

Bottom line is it's not fast enough to warrant being TB3 and that cost certainly gets passed onto the consumer, and Sabrent probably uses the excuse of it being TB3 to add some extra margin. It's lying and directly anti-consumer.

It doesn't warrant being TB3 with that speed, all that does is make people believe it's high end (cuz thunderbolt gud) to increase prices. It's slower than single entry-level QLC drives, never mind the fact that it's supposedly striped so there's two drives in there. I'd expect it to saturate the TB3 40Gbps spec easily but it caps out at 20Gbps which makes no sense if it's NVMe. It must be artifically limited somewhere, the drives have to be faster than that. As I said, even the slowest single NVMe drive is still faster than these TWO STRIPED drives. If they were SATA based it's too fast (two striped SATA drives would cap out at around 1GB/s, not 2,5GB/s), so it must be NVMe, but the rated speed just makes no sense.
And if it IS TB3 and the limit is really in the drives or controller; why advertise it being TB3 or implement the required hardware for it to be TB3 if it can't reach it? It just ups the cost for zero reason...
This is like advertising HDMI 2.1 on a 1080p 60Hz monitor...

Dont like it, design your own equipment

These are not NVMe drives, they are external drives that can be taken anywhere with a power supply to a outlet.

 

[kayjay010101]

"Doesn't warrant being TB3" is rather meaningless though. There are likely no USB 3.2G2x2 controllers that can connect two m.2 drives. As for performance, they're likely splitting the lanes from the TB3 controller out with two for each drive (there's no way there's a PLX swich in there), so that likely explains that. Sure, that means it's slower than a single-drive solution, but (not that it matters seeing how I don't have £3000 to spend on this, nor would want to) IMO I'd rather have a split lane solution than waste the extra power on a PLX.

2 lanes per drive is still 2GB/s per drive at 3.0 speeds so it should still be close to 4GB/s striped if it was TB3.
I don't see why splitting the lanes would make it slower than a single drive, you'll have to explain a bit further on that.
Plenty lower-end motherboards also restrict certain M.2 slots to x2 and striping those together would still end up being faster than a single drive.

Dont like it, design your own equipment

These are not NVMe drives, they are external drives that can be taken anywhere with a power supply to a outlet.

Still sticking to it, huh?
Just because I don't have the engineering expertise to actually fundamentally design a solution to a product like this - that doesn't necessarily mean I lack the knowledge on when the end product they do provide is bad. What I do know is that the interface could be 3.2 Gen 1 and they would lose no bandwidth and reduce costs quite a lot, and open up to the much larger market that only has USB instead of Thunderbolt.

The drives have to be adressed by some protocol, so they're either SATA or NVMe. All drives have to be adressed by some storage protocol. Just because they're external, that doesn't mean they don't get adressed in their protocols. Practically every external drive that exists is just a USB to SATA controller adapter, the rest are USB/TB to NVMe.
SATA is ruled out simply by speed - a striped pair of SATA drives cap out at 1GB/s; too slow to ever possibly be this solution. That leaves NVMe as the only possible protocol, and it being TB3 with its 4 lanes of bandwidth backs that up.
Either these are the worst NVMe drives ever created, or whatever controller they're using is ultra-garbage and leaves 50% of the performance on the table due to overhead. That really is the only thing I can think of that would make these perform that bad.

 

[Valantar]

2 lanes per drive is still 2GB/s per drive at 3.0 speeds so it should still be close to 4GB/s striped if it was TB3.
I don't see why splitting the lanes would make it slower than a single drive, you'll have to explain a bit further on that.
Plenty lower-end motherboards also restrict certain M.2 slots to x2 and striping those together would still end up being faster than a single drive.

-These are QLC drives. They aren't speed demons in the first place.
-It's a given that a striped layout over TB will be significantly slower than a striped layout over direct PCIe (whether CPU or chipset) due to latency
-Striping across inherently slow QLC drives might not be as efficient
-There's no guarantee that these drives have the same SLC caching behaviour as desktop drives
-It's likely the firmwares of these drives are tuned differently than desktop drives, which are often tuned towards (utterly meaningless) drag-race sequential transfer metrics. These SSDs are most likely tuned to avoid excessive thermal loads and to be better suited for external drive use cases.

Etc., etc.

 

[yotano211]

2 lanes per drive is still 2GB/s per drive at 3.0 speeds so it should still be close to 4GB/s striped if it was TB3.
I don't see why splitting the lanes would make it slower than a single drive, you'll have to explain a bit further on that.
Plenty lower-end motherboards also restrict certain M.2 slots to x2 and striping those together would still end up being faster than a single drive.


Still sticking to it, huh?
Just because I don't have the engineering expertise to actually fundamentally design a solution to a product like this - that doesn't necessarily mean I lack the knowledge on when the end product they do provide is bad. What I do know is that the interface could be 3.2 Gen 1 and they would lose no bandwidth and reduce costs quite a lot, and open up to the much larger market that only has USB instead of Thunderbolt.

The drives have to be adressed by some protocol, so they're either SATA or NVMe. All drives have to be adressed by some storage protocol. Just because they're external, that doesn't mean they don't get adressed in their protocols. Practically every external drive that exists is just a USB to SATA controller adapter, the rest are USB/TB to NVMe.
SATA is ruled out simply by speed - a striped pair of SATA drives cap out at 1GB/s; too slow to ever possibly be this solution. That leaves NVMe as the only possible protocol, and it being TB3 with its 4 lanes of bandwidth backs that up.
Either these are the worst NVMe drives ever created, or whatever controller they're using is ultra-garbage and leaves 50% of the performance on the table due to overhead. That really is the only thing I can think of that would make these perform that bad.

You can say all the fancy technical words you can, at the end of the day, this item is still an external SSD device that requires a external power supply.
Not all storage devices have to be address with SATA or nvme connection. Some times, a company will make their own propriety connector. I cant think of the company's name, but years ago, maybe over 6-7 years, I shucked a external HD. I checked the connector and it wasnt SATA. I returned it to the store, bought another brand.

 

[Valantar]

You can say all the fancy technical words you can, at the end of the day, this item is still an external SSD device that requires a external power supply.
Not all storage devices have to be address with SATA or nvme connection. Some times, a company will make their own propriety connector. I cant think of the company's name, but years ago, maybe over 6-7 years, I shucked a external HD. I checked the connector and it wasnt SATA. I returned it to the store, bought another brand.

Hm? This is an external TB3 enclosure with two m.2 NVMe drives inside. It's literally on the pictured product box, and is clear from the press release. I still find it rather silly that it needs a power supply, but ... it is what it is.

 

[yotano211]

Hm? This is an external TB3 enclosure with two m.2 NVMe drives inside. It's literally on the pictured product box, and is clear from the press release. I still find it rather silly that it needs a power supply, but ... it is what it is.

I know this product is a stripped nvme device, I read the 1st post.

It might be at the device has so many NAND on the PCB board it requires extra juice.

 

[kayjay010101]

You can say all the fancy technical words you can, at the end of the day, this item is still an external SSD device that requires a external power supply.
Not all storage devices have to be address with SATA or nvme connection. Some times, a company will make their own propriety connector. I cant think of the company's name, but years ago, maybe over 6-7 years, I shucked a external HD. I checked the connector and it wasnt SATA. I returned it to the store, bought another brand.

I'm not talking about the port, I'm talking about the protocol. The SATA protocol doesn't need to use SATA ports (you've probably seen plenty of SATA m.2 devices, for example), this applies to all the other protocols as well. All external devices, yes, even those with proprietary connectors, use one of the standard storage protocols (SAS, IDE, NVMe, SATA, etc.), because if they didn't, your OS wouldn't be able to recognize them and use them as storage devices.

I know this product is a stripped nvme device, I read the 1st post.

It might be at the device has so many NAND on the PCB board it requires extra juice.

TB3 can deliver upto 100W and two NVMe drives are like 25W max, with controller overhead it's probably around 30-35W, so there's another baffling aspect about Sabrent opting for TB3 but not using it.
Also, it can't be that dense as there's a thermal limit as well. NVMe drives can't really draw more than 10-12W before they physically can't be cooled down properly anymore on the M.2 form factor, which given the size of this external enclosure, they are 100%, or at least a similar size. So thermally they're really limited to about 10-12W per drive, so about a 25W budget for both.

 

[Valantar]

I'm not talking about the port, I'm talking about the protocol. The SATA protocol doesn't need to use SATA ports (you've probably seen plenty of SATA m.2 devices, for example), this applies to all the other protocols as well. All external devices, yes, even those with proprietary connectors, use one of the standard storage protocols (SAS, IDE, NVMe, SATA, etc.), because if they didn't, your OS wouldn't be able to recognize them and use them as storage devices.

TB3 can deliver upto 100W and two NVMe drives are like 25W max, with controller overhead it's probably around 30-35W, so there's another baffling aspect about Sabrent opting for TB3 but not using it.
Also, it can't be that dense as there's a thermal limit as well. NVMe drives can't really draw more than 10-12W before they physically can't be cooled down properly anymore on the M.2 form factor, which given the size of this external enclosure, they are 100%, or at least a similar size. So thermally they're really limited to about 10-12W per drive, so about a 25W budget for both.

That's for powering devices like laptops and tablets. The minimum power output requirement for on-device TB3 ports is 15W, and I've never heard of a device (PC, laptop, tablet) exceeding that. Meaning that if this needs more than 15W at any point, it needs external power.

I know this product is a stripped nvme device, I read the 1st post.

It might be at the device has so many NAND on the PCB board it requires extra juice.

Nah, it's just that it has two m.2 drives plus a TB3 controller, which in sum is likely to exceed the 15W power output of most TB3 ports.