x4 PCIe to NVMe adapter with supercapacitor

[averyfreeman]

Hey,

I just saw these on ebay and thought people on this forum might find them interesting:

https://www.ebay.com/itm/Adapter-Ca...068163?hash=item33b28f4e03:g:wIIAAOSwWTRW1WI5

Looks like a pretty hefty capacitor for power protection... thoughts?

 

[Ericloewe]

That's interesting, but it may not provide as much protection as you'd hope. It relies on the drive detecting that the host died and cleaning itself up. I'm not sure that your average consumer SSD has that kind of logic, because when would it ever be useful?

Looks like a pretty hefty capacitor for power protection... thoughts?

4F, 5V... Yeah, that's probably more than many power loss protected SSDs have. It's probably necessary because the drive would have to wait for the host to timeout, since it doesn't realize that the power went out. That's probably a whole wasted second.

 

[averyfreeman]

Hm, yeah I find that totally interesting - I wonder if there's some way I could try and create a "real-world" situation to test it to see if it works?

Bought two of these then decided later to go with an SM953 for a SLOG which already has PLP. I wonder if I could double-up on the power protection? hehe Pretty good deal on the SM953s over there right now, too...

 

[systemofapwne]

Sorry to revive this rather old thread, but I just considered, if this product might do the job and I concluded "yes. So I ordered it and will report back, once it arrived.

My considerations: Assuming a maximum power draw of 10W (which is reasonable for a NVMe SSD at full load), that capacitor will last roughly 20 seconds - assuming, you can use all of the stored capacity. But the reality is, that once it reaches a certain "low" voltage, the SSD will just give up. Judging on the components and traces on that card, I assume, it will most probably do a 12 V (PCIe-Power) -> 3.3 V voltage regulation (required voltage for the SSD), which feeds the supercap. So the cap will only be charged to 3.3 V and thus will never reach it's full capacity (which for this type is 4 F at 5.5 V). This leaves us with 60% of the maximum capacity at best (12 seconds at 10W). Now lets consider, your SSD will be fine by being powered between 3.0 V to 3.3 V. That means, once it reaches 3.0 V, it stops working. This means (assuming a linear rather than exponential decay of the capacity at constantly drawn current), you only will be able to draw power from the capacity for ~10% of it's full capacity -> 1.2 seconds at 10 W. Will this be sufficient time to flush onboard DRAM? Yes. Will the SSD last longer "in the field": Probably also yes, since it will not draw the 10 W all the time but only when active.

Sidenote: It really looks like a 12 V -> 3.3 Buck conversion to charge the cap which then powers the SSD. But I could be wrong, so it might be a that the supercap is charged by PCIe port directly (I doubt that) so the regulator on the left is a boost converter, which can then use "whatever voltage the supercap has" -> 3.3V (thus meaning, that the "10%" consideration in my conclusion will not hold and thus, the SSD can live longer).

 

[systemofapwne]

The device arrived a few days ago and I just put it into my system today. First things first: It works.

I traced the PCB and routes can tell you the following:

• All traces for PCIe x4 are there
• The card is powered from the 12V lane
• It uses a buck converter (chinese clone of an lm2596) to regulate the voltage from 12 V -> "around 4 V"
• After the regulated voltage, there are three diodes in parallel, each 1 A max current capability, so they can fulfill the 3 A current output of the regulator
• After the voltage drop due to the diodes, the resulting voltage from intially "around 4 V" drops to 3.3 V, which charges the 4 F supercap
• The 3.3 V level at the supercap directly goes to the NVMe SSD and powers it,
• A "power loss" test with an idle SSD resulted in several minutes of powering the SSD with acceptable voltage levels (~ 3-3.3 V). I herefore say, this backup solution is fully sufficient (assuming, the SSD flushes DRAM data to FLASH storage within that time).

So: Go for this solution, if you trust in consumer NVMe SSDs to flush their data within the time the supercap can power them. I do by now. If you are an enthusiasts who needs almost 100% data integrity, go for an enterprise SSD with inbuild capacitor (and raid that SSD).