This Product is a Meme - Corsair MP600 Pro Hydro X

[AlexTheGreatish]

It's an SSD... that is water cooled. Why? Well the answer is a lot more interesting than we originally thought.

 

Buy Corsair MP600 PRO Hydro X Edition 2TB NVMe SSD

On Amazon (PAID LINK): https://geni.us/L6iQC

On Newegg (PAID LINK): https://geni.us/sVQalp

Purchases made through some store links may provide some compensation to Linus Media Group.

 

 

 

[Murasaki]

And I thought my MP500 was a bit toasty. NVMouch!

[rodion.zissou]

Just now, Murasaki said:

And I thought my MP500 was a bit toasty. NVMouch

NVMeek

 

[Nystemy]

Coming in really quickly here to inform why there is differences between keeping the Flash warm or cool.

When one writes to a Flash memory cell, then the device creates a fairly intense electric field across the insulation surrounding the floating gate making up the cell. This forces some electrons to get trapped in or ripped out of the floating cell _{(depending on what the manufacturer thinks is better, both technically works.)} But each time this is done, it adds a bit of damage to the insulation, making it less good at stopping current from flowing through it.

The Flash cell is in effect just a fancy low leakage capacitor. And the faster the charge leaks out from it, the faster it forgets. _{(And for DLC ("MLC"), TLC, and QLC cells, it forgets even faster due to the smaller charge differences it has to deal with from the get go.)}

 

But how does operating temperature play a part in this?
Well, the insulation simply has higher leakage when warm, and doesn't need to break down as hard when that current is forced through it.
But likewise, the warm temperature has the deficit that the insulation also leaks more current, so the cells forgets faster.
 

In short.
If you want to put the drive on a shelf for 10 years, then keep it cool.
If you want to use it as a scratch disk for a storage/computing server, then keep it warm.
And if you use it as something in-between these two extremes, then it is a bit more of a mix...

Edit:
I should also probably mention that Flash memory manufacturers are working on solutions on "partly" solving this whole debate by just integrating heaters in the Flash array itself, that rapidly _{(as in ms, for writes it doesn't matter, and reading then this whole thing isn't even important...)} heat up the local area _{(talking areas in the tens of µm² here.)}  to well above 80C that makes the cell happily be written to with minimal wear. And since the amount of heat put in is actually fairly tiny, it quickly spread out and doesn't really add much to the overall operating temperature. And one can then put more actual heatsinks on the flash chips without worrying about this whole debate. _{But this is currently in patent hell....}

 

[Master Disaster]

Gotta say, I think you guys missed the point and missed the testing methodology.

 

Nobody is going to buy this thing to run standalone, you'd buy it because you already have a full loop in your PC and adding in the SSD is the price of 2 pieces of tubing.

 

Linus pointing out the GPU heats up the position where the SSD sits is kinda moot since the GPU would normally be in the loop too and not throwing out as much heat.

 

Its not that your conclusion was wrong, watercooling an SSD is not necessary but you ignored an entire use case and the one that IMO makes the most sense for the product existing.

[WolframaticAlpha]

At this rate, we will have an overclocked water cooled water cooler in a few years.

[jones177]

I have a MP600(PCIe 4 with heatsink) and it benches at 50c. My MX500 dose the same bench at 36c so I can see the need for better cooling with these high performance SSDs.

 

1 minute ago, WolframaticAlpha said:

At this rate, we will have a water cooled water cooler in a few years.

I think refrigerated cases is the way to go since it will not only cool components but keep beverages cool as well.  

[Quackers101]

will this become more of a thing with direct storage?

BLAST YOUR TBW AWAY IN ONE CLICK, THEY HATE IT!
*click the link to find out more*

[Nystemy]

8 minutes ago, Quackers101 said:

will this become more of a thing with direct storage?

BLAST YOUR TBW AWAY IN ONE CLICK, THEY HATE IT!
*click the link to find out more*

Direct storage should under normal operation only read assets. This doesn't meaningfully impact write endurance.

Though, personally, I sometimes wonder how direct storage is effected by simple things like access latency differences between different types of MLC Flash.
Or how something as simple as PCIe package buffering impacts things, all though, each of these are usually fairly small, only a few µs at the worse times.

[BaidDSB]

How did they measure the drive life endurance?

 

how can i measure my SSD's time?

[Quackers101]

41 minutes ago, Nystemy said:

Direct storage should under normal operation only read assets. This doesn't meaningfully impact write endurance.

Though, personally, I sometimes wonder how direct storage is effected by simple things like access latency differences between different types of MLC Flash.
Or how something as simple as PCIe package buffering impacts things, all though, each of these are usually fairly small, only a few µs at the worse times.

Unless we go all Microsoft Flight Simulator with the use of direct storage? I guess one would be bound to other stuff before that.

[Junction Runner]

I want to see them use up the whole write endurance on the drive, as they said it wouldn't take long, it would be cool to see what happens once it does reach that end of life. 

[Nystemy]

2 minutes ago, Junction Runner said:

I want to see them use up the whole write endurance on the drive, as they said it wouldn't take long, it would be cool to see what happens once it does reach that end of life. 

In regards to an ideal flash storage array.
Then one would just expect the data retention time to progressively get worse and worse with every new write cycle.

Since the insulation surrounding the floating gate in the cell progressively wears down. _{(Though, I wouldn't be surprised if migration of the metal itself eventually wanders into the gap and makes the conductivity rapidly shoot up. so that it at some point might go from having a retention time in the order of months at room temp, to the next cycle suddenly having a couple of days. And potentially hours after that. Though, I can't say I have deeply studied the long term failure modes on the individual cell level all that extensively.)}

But it is fair to say that the TBW score of SSDs is more or less what the manufacturer "guarantees" that one can write and still have "acceptable" data retention time. Ie, still in the couple of years at room temperature.

Though, put an SSD in a warm case and those "couple of years" is suddenly a few months. Since the leakage roughly doubles for every 10 C increase in temperature.  And the insides of PC cases are typically a good 10-25 C warmer than ambient, unless one has a good case.

But theoretically, an SSD used as a scratch disk wouldn't require any long data retention times _{(since scratch disk applications typically don't store the data for more than a few hours)}, so it should work fine for a lot more than its rated TBW. Unless the drive manufacturer has put in a counter, or stores the firmware in the flash array itself. _{(ie, eventually it might forget its own software and become a brick...)}

[Streetguru]

[REDACTED]

[Egg-Roll]

SSDs are longevity?  No... Are they getting better? Yes, but I wouldn't exactly trust them with my data. I wouldn't trust the 1.xPb claims manufactures state.

I guess that's what you get when you only own Seagate, anything looks better...

 

Also this videos testing is ultimately useless, they failed to do a whole loop in a case to see how well it works. I guess doing proper testing is too much work as obviously if you leave something in an open air conditioned room heat rises blah blah blah, but by putting it into a proper case and loop that changes the airflow patterns.

 

Equally assuming this video wasn't thrown together with haste this week, unless you recorded it last week you had AC on with weather in the low to mid 20's and you somehow made it sound like that made a huge difference  (72F = 22C to those who use F), only way it would have is if someone forgot to turn on the AC for that area prior to filming.

 

2 minutes ago, Streetguru said:

Wait, why didn't you test just filling the little block with water?

The little block as in the block attached to the SSD? They did about half way threw:

 

[Streetguru]

5 minutes ago, Egg-Roll said:

 

I look away for 2 seconds...

[AlexTheGreatish]

6 hours ago, Nystemy said:

Coming in really quickly here to inform why there is differences between keeping the Flash warm or cool.

When one writes to a Flash memory cell, then the device creates a fairly intense electric field across the insulation surrounding the floating gate making up the cell. This forces some electrons to get trapped in or ripped out of the floating cell _{(depending on what the manufacturer thinks is better, both technically works.)} But each time this is done, it adds a bit of damage to the insulation, making it less good at stopping current from flowing through it.

The Flash cell is in effect just a fancy low leakage capacitor. And the faster the charge leaks out from it, the faster it forgets. _{(And for DLC ("MLC"), TLC, and QLC cells, it forgets even faster due to the smaller charge differences it has to deal with from the get go.)}

 

But how does operating temperature play a part in this?
Well, the insulation simply has higher leakage when warm, and doesn't need to break down as hard when that current is forced through it.
But likewise, the warm temperature has the deficit that the insulation also leaks more current, so the cells forgets faster.
 

In short.
If you want to put the drive on a shelf for 10 years, then keep it cool.
If you want to use it as a scratch disk for a storage/computing server, then keep it warm.
And if you use it as something in-between these two extremes, then it is a bit more of a mix...

Edit:
I should also probably mention that Flash memory manufacturers are working on solutions on "partly" solving this whole debate by just integrating heaters in the Flash array itself, that rapidly _{(as in ms, for writes it doesn't matter, and reading then this whole thing isn't even important...)} heat up the local area _{(talking areas in the tens of µm² here.)}  to well above 80C that makes the cell happily be written to with minimal wear. And since the amount of heat put in is actually fairly tiny, it quickly spread out and doesn't really add much to the overall operating temperature. And one can then put more actual heatsinks on the flash chips without worrying about this whole debate. _{But this is currently in patent hell....}

 

This is awesome info.  I'm going to make this a pinned comment on the video since this is the first explanation I've seen that makes any sense.

[Nystemy]

9 hours ago, AlexTheGreatish said:

This is awesome info.  I'm going to make this a pinned comment on the video since this is the first explanation I've seen that makes any sense.

Thanks.

I try to provide as useful and informative posts as I can. _{(even if they might get a bit complex and sometimes long winded, but I tend to put such excessively detailed information inside of a parentages like this. In short, the extra nerdy stuff...)}

And I can relate to how most explanations on the webb for "ideal" flash operating temperature being of a conflicting nature. I read semiconductor research papers instead for a reason, though a lot less "user friendly" text in such, not to mention plenty of graphs and explanatory images.

 

Here is for an example a paper on extending write endurance in Flash cells by using a heater chip baked into the flash stack up to heat it up sufficiently for the insulation to start restoring itself. Even if this have some other downsides that eventually means that the chip will fail regardless. But it is fairly interesting.
https://www.usenix.org/legacy/events/hotstorage11/tech/final_files/Wu.pdf

 

_{btw, I am not a "researcher" in the field. Just some random furry that likes studying computer architecture design for the past 15+ or so years and semiconductor device manufacturing for the past 10+ years. It is starting to add up to a university course all though much less narrowed in on a specific subject matter within said fields, and then I also design computer architectures and implement them in silicone as a hobby, all though currently only theoretically and in simulations.  But if I make a bunch of chips some day, I might send a couple your way....}

[Mr.Nelf]

On 6/30/2021 at 11:08 AM, WolframaticAlpha said:

At this rate, we will have an overclocked water cooled water cooler in a few years.

ROFL!

Yo dawg! We heard u like water coolin', so we put a water cooler on yo water cooler, so u can cool yo water while u coolin' yo water!