Because you're controller also needs to be able to handle it. It has to detect the loss of power and be able to write the cache to the NAND before the capacitors are drained. This tech might not be present in every controller. You would also need some other components and PCB traces to achieve it. it might be only a $10 BOM increase but by the time it reaches e-tailers you might be talking about a $40 markup
Can you substantiate that claim? Every general purpose microcontroller can be programmed to do that, so it shouldn't be a problem for a dedicated chip. Could it be that the industry doesn't care about the optimal price to feature ration, and is only diligent when it comes to high margin products? It certainly would not be the first time something like that happens.
Just the other I had a capacitor blow up, ruining a 400$ machine. The capacitor was rated at 240 volts and was running 220 volts, which is known to fluctuate at least a good 10%. The capacitor was essentially a time bomb. The same capacitor rated at 400 volts costs only 1 cent more, and would have lasted much longer, but they didn't go for it.
Because they're charging for engineering and for the value of such a feature. The BOM cost has little to do with the retail price of any device these days.
For the same reason we don't have ECC memory in our desktops and laptops: it isn't necessary. Power loss protection capacitors are a novelty. Since most consumer SSD's spend 99.9% of their time at idle, in the event of a system crash or power outage, a chkdsk will resolve any file system errors that would have occurred during the in flight dump the drive did before power failure (crucial MX series based on Marvell controllers) or in the case of most other SSD's there wouldn't be anything in flight because no user data is cached to DRAM. Many SSD's now don't even HAVE a DRAM buffer!
Caps are only necessary on drives that are caching user data off-NAND, use the NAND as a write buffer, or are in a high-write environment such as databases. Everyone else gets by without it. I love the idea of them, and I've never seen an SSD320 or my own SSD730 become corrupt during a power outage (and my SSD730, against Intel's recommendation, is in my laptop :) but at the end of the day it's inconclusive if capacitors actually matter for a consumer application, which is why they are rarely ever offered in one. DRAM often just caches the indirection table, every time it's changed, the changes are updated to the copy in NAND and if the power fails DURING that copy, the previous copy of the indirection table, also in NAND, is kept as a backup because that backup isn't deleted until the copy from DRAM to NAND is marked complete and passes verification. Drives that don't have DRAM are virtually immune to indirection table corruption, they're just slower because they access it and update it in NAND so IOPS suffer while that traffic causes overhead, bogging down data traffic.
These are enterprise SSDs and power loss capacitors are FAR from a novelty in that arena. I'm not sure why you keep talking about consumer applications for enterprise SSDs. You made a lot of unsubstantiated claims and threw out some anecdotes, but you might as well have been talking about a completely different subject.
Whether it's labeled consumer, enthusiast, pro, datacenter or whatever the hell else, if data that is reported as flushed to the drive isn't actually there after power is lost and restored the product is flawed since it violates the standard. Imagine buying a car where the brakes only work 99.99% of the time on the cheaper models...
Because it is still naive and foolish to think that BOM is or should be the main driver in the cost of a high tech product.
As for why, its simple and obvious, and yet many people seem determined not to see it, so, I'll try a different way.
A good approach to optimize for a healthy profit margin AND a low BOM is to drive for economies of scale. Fixed costs can be amortized over more units and larger volumes also improve negotiating positions with suppliers.
A proven strategy for this is to design a product portfolio consisting of multiple product lines with multiple models, rather than a single product. Products across the portfolio share a lot of common designs, code and components. They are differentiated from each other by small differences in components, and larger differences in price points (and profit margins).
For some reason, this approach outrages a lot of people because they think they are being cheated out of the lowest price and maximum capability, not realizing that its quite likely that this approach means that they are getting more value at just about every price point than they would be if companies didn't engineer their product lines this way.
Instead of everyone winning though, they still would prefer a situation where everyone is worse off, themselves included. That seems to be the very definition of stupidity.
Speaking off micron , weren't we supposed to start seeing their 3D-Nand 2Q 2016 consumer products and maby enterprise 3D Xpoint products? Those press annoucement were early 2015? Are they already beeing used in products ? Only thing i find when i google is old press announcements.
Given that these are enterprise drives, typical use would have them in queue depths >0-1 most of the time, so they are probably benched with fairly high queue depths.
I hope this presages the arrival of consumer PCIe parts from Crucial, not that I'd mind the Micron M.2's, I just don't think they'll be readily available or affordable.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
21 Comments
Back to Article
ImSpartacus - Tuesday, April 12, 2016 - link
What are those yellow things on the first pictured ssd. Are those like capacitors or something?questionlp - Tuesday, April 12, 2016 - link
Correct, "power loss protection capacitors".ddriver - Tuesday, April 12, 2016 - link
Considering those capacitors cost less than 1$, one is inclined to ask why are such not found in consumer products, they would barely add to the BOM.QinX - Tuesday, April 12, 2016 - link
Because you're controller also needs to be able to handle it. It has to detect the loss of power and be able to write the cache to the NAND before the capacitors are drained. This tech might not be present in every controller. You would also need some other components and PCB traces to achieve it. it might be only a $10 BOM increase but by the time it reaches e-tailers you might be talking about a $40 markupddriver - Tuesday, April 12, 2016 - link
Can you substantiate that claim? Every general purpose microcontroller can be programmed to do that, so it shouldn't be a problem for a dedicated chip. Could it be that the industry doesn't care about the optimal price to feature ration, and is only diligent when it comes to high margin products? It certainly would not be the first time something like that happens.Just the other I had a capacitor blow up, ruining a 400$ machine. The capacitor was rated at 240 volts and was running 220 volts, which is known to fluctuate at least a good 10%. The capacitor was essentially a time bomb. The same capacitor rated at 400 volts costs only 1 cent more, and would have lasted much longer, but they didn't go for it.
jordanclock - Tuesday, April 12, 2016 - link
Because they're charging for engineering and for the value of such a feature. The BOM cost has little to do with the retail price of any device these days.Samus - Wednesday, April 13, 2016 - link
For the same reason we don't have ECC memory in our desktops and laptops: it isn't necessary. Power loss protection capacitors are a novelty. Since most consumer SSD's spend 99.9% of their time at idle, in the event of a system crash or power outage, a chkdsk will resolve any file system errors that would have occurred during the in flight dump the drive did before power failure (crucial MX series based on Marvell controllers) or in the case of most other SSD's there wouldn't be anything in flight because no user data is cached to DRAM. Many SSD's now don't even HAVE a DRAM buffer!Caps are only necessary on drives that are caching user data off-NAND, use the NAND as a write buffer, or are in a high-write environment such as databases. Everyone else gets by without it. I love the idea of them, and I've never seen an SSD320 or my own SSD730 become corrupt during a power outage (and my SSD730, against Intel's recommendation, is in my laptop :) but at the end of the day it's inconclusive if capacitors actually matter for a consumer application, which is why they are rarely ever offered in one. DRAM often just caches the indirection table, every time it's changed, the changes are updated to the copy in NAND and if the power fails DURING that copy, the previous copy of the indirection table, also in NAND, is kept as a backup because that backup isn't deleted until the copy from DRAM to NAND is marked complete and passes verification. Drives that don't have DRAM are virtually immune to indirection table corruption, they're just slower because they access it and update it in NAND so IOPS suffer while that traffic causes overhead, bogging down data traffic.
Samus - Wednesday, April 13, 2016 - link
*Meant to say uses DRAM as a write buffer, not NANDddriver - Wednesday, April 13, 2016 - link
Capacitors - novelty? LOL. Together with hot water and sliced bread...jordanclock - Wednesday, April 13, 2016 - link
These are enterprise SSDs and power loss capacitors are FAR from a novelty in that arena. I'm not sure why you keep talking about consumer applications for enterprise SSDs. You made a lot of unsubstantiated claims and threw out some anecdotes, but you might as well have been talking about a completely different subject.nils_ - Wednesday, April 20, 2016 - link
Whether it's labeled consumer, enthusiast, pro, datacenter or whatever the hell else, if data that is reported as flushed to the drive isn't actually there after power is lost and restored the product is flawed since it violates the standard. Imagine buying a car where the brakes only work 99.99% of the time on the cheaper models...easp - Tuesday, April 19, 2016 - link
Because it is still naive and foolish to think that BOM is or should be the main driver in the cost of a high tech product.As for why, its simple and obvious, and yet many people seem determined not to see it, so, I'll try a different way.
A good approach to optimize for a healthy profit margin AND a low BOM is to drive for economies of scale. Fixed costs can be amortized over more units and larger volumes also improve negotiating positions with suppliers.
A proven strategy for this is to design a product portfolio consisting of multiple product lines with multiple models, rather than a single product. Products across the portfolio share a lot of common designs, code and components. They are differentiated from each other by small differences in components, and larger differences in price points (and profit margins).
For some reason, this approach outrages a lot of people because they think they are being cheated out of the lowest price and maximum capability, not realizing that its quite likely that this approach means that they are getting more value at just about every price point than they would be if companies didn't engineer their product lines this way.
Instead of everyone winning though, they still would prefer a situation where everyone is worse off, themselves included. That seems to be the very definition of stupidity.
theeldest - Tuesday, April 12, 2016 - link
Yes, Surface Mount capacitors to provide power-loss protection.DanNeely - Tuesday, April 12, 2016 - link
That's exactly what they are/extide - Tuesday, April 12, 2016 - link
Yep, those are Tantalum Surface Mount Capacitorsplopke - Tuesday, April 12, 2016 - link
Speaking off micron , weren't we supposed to start seeing their 3D-Nand 2Q 2016 consumer products and maby enterprise 3D Xpoint products? Those press annoucement were early 2015? Are they already beeing used in products ? Only thing i find when i google is old press announcements.ant6n - Tuesday, April 12, 2016 - link
I thought there were supposed to have some big event today, there were rumors they'd make an announcement re 3d xpoint.vFunct - Tuesday, April 12, 2016 - link
What Queue depths were the Random Read & Write IOPS measured?Billy Tallis - Tuesday, April 12, 2016 - link
The spec sheet doesn't list queue depth, but does say that the figures are for steady-state performance.extide - Tuesday, April 12, 2016 - link
Given that these are enterprise drives, typical use would have them in queue depths >0-1 most of the time, so they are probably benched with fairly high queue depths.Jay77 - Wednesday, April 13, 2016 - link
I hope this presages the arrival of consumer PCIe parts from Crucial, not that I'd mind the Micron M.2's, I just don't think they'll be readily available or affordable.