ADATA Unveils World’s First PCIe Gen5 SSD with Micro-Fan That’s Powered Directly Through the M.2 Slot

[Tsing]

The LEGEND 970 PRO, a new PCIe Gen5 x4 M.2 2280 SSD that delivers several improvements over the original, including being the first PCIe Gen5 SSD on the market to feature a micro-fan that is powered directly through the M.2 slot, is coming soon, ADATA has announced.

See full article...

 

[Grimlakin]

Ok I've never in my life seen a fan that needs a... 'charging cable' that wasn't hand held.

Are there NVME SSD's that run fans off of a battery that I'm missing out on?

Or was this just a translation error?

Regardless I would be suspect about putting something that could have physical resistance directly connected to my CPU and drawing additional power for a fan. Seems like a suspect combination to me.

 

[Brian_B]

Ok I've never in my life seen a fan that needs a... 'charging cable' that wasn't hand held.

I think it's an obvious translation error or some Gen Z'er who just things every cable is a charging cable because that's all they have ever seen in their life. Fans have always had a power cable of some sort - and most run out to either a 3 or 4 pin motherboard header, a molex/SATA cable, or they have an internal header on the card that they plug into.

So I guess ADATA is proud they have an internal header on their card.

 

[Zarathustra]

How loud is it?

 

[Endgame]

How long will it last? Even in a clean environment, most chipset fans only last a year. I’d guess this size fan will be dead in 3 months

 

[DrezKill]

Even in a clean environment, most chipset fans only last a year.

Oh gawd, is that really true? Back in the day they would last for years no problem.

 

[Zarathustra]

Oh gawd, is that really true? Back in the day they would last for years no problem.

I was about to refute this, but come to think of it, I can't remember the last time I actually checked on a Chipset fan. I think my Asus TRX40 board is the only one that has one, but I have no idea if it is still running. I did tweak it to lower the fan speed and make it quieter when I first got it. I wonder if that improves reliability.

But yeah, small fans generally have horrible reliability. Nothing innate about small fans, but they are just usually cheaply made by Chinese junk manufacturers.

If I need a small fan these days the only ones I trust are Noctua's 40mm units. 40mm isn't really that small, but it really is the smallest reliable fan I have found on the market.

It's difficult to tell here, but that's what I used when I needed a small fan on top of my two NVMe boot drives in my server:



Seen here before attaching heatsinks and fan (and almost everything else):



And here with the heatsinks on there:

 

[LazyGamer]

I like the idea of tackling the solution using what I assume is PCIe power from the M.2 slot for the fan - less cables more better.

But small fans have to spin fast to move air, and well, fast fans make more noise. Reliability and tuning are important here, I hope ADATA ships one of these to @Brent_Justice for a proper shellacking.

 

[Zarathustra]

I like the idea of tackling the solution using what I assume is PCIe power from the M.2 slot for the fan - less cables more better.

But small fans have to spin fast to move air, and well, fast fans make more noise. Reliability and tuning are important here, I hope ADATA ships one of these to @Brent_Justice for a proper shellacking.

Remember though, NVMe drives are not CPU's (or GPU's)

None of my NVMe drives have the capability of exceeding ~9.78w (though I only have Gen4 drives) but based on my poking around 15w is about as much as an m.2 port can supply, so that is probably the max heat that would ever need to be dissipated from a m.2 NVMe drive unless they start adding supplemental power connectors to them.

This is the output from smartctl on my SBC Mini PC I use for lunchtime browsing at work. It has a Sabrent Rocket 4 2TB in it (temporarily, until the drive I intend to use arrives)

Supported Power States
St Op     Max   Active     Idle   RL RT WL WT  Ent_Lat  Ex_Lat
 0 +     9.78W       -        -    0  0  0  0        0       0
 1 +     6.75W       -        -    1  1  1  1        0       0
 2 +     5.23W       -        -    2  2  2  2        0       0
 3 -   0.0490W       -        -    3  3  3  3     2000    2000
 4 -   0.0018W       -        -    4  4  4  4    25000   25000

I checked all the drives I have access to from here, and this one used the most power. (The Samsun 980's in the server max out at 8.49w)

Either way, the point I am trying to make is that it doesn't take much airflow if directly on the heatsink on the m.2 drive to make a rather huge difference in the amount of heat it can move away. A fan like this could probably sit at a surprisingly low rpm and still make a difference.

The fan in my server above is - according to the IPMI readout sitting at 480rpm and keeping those two 500GB 980 Pro's very happy at 38C. At that speed it wasn't audible to me when I set up the system on my desk and certainly isn't now inside the rack with all sorts of other things making noise. And that's with the ****tiest low profile heatsinks $3.50 can buy (the only ones I could find that would fit under those 16x cards) in an ~80F ambient closet (until I get my HVAC project up and running) inside a rack which is even hotter, inside a server case with 12x 7200rpm 3.5" hard drives, 14 other NVMe drives, and some hot NIC's, SAS controllers and a big 32C/64T server CPU. With the beefier heatsink on this unit, it probably won't need as much airflow...

...but - on the flip side - mine is a 40mm fan, and the smaller the fan, the higher the rpm will be needed. No idea how high they would actually have to rev it. It likely depends on how high temps they are comfortable with, as the higher the Delta-T between hot chip and cool ambient, the more efficient the heat flows from the hot to the cold side, and thus the slower the fan can spin.

I don't have any experience at all with Gen5 drives, but I know on my Gen4 drives the critical temps arent until 85C (Samsung) and 95C (Sabrent reference Phison design). I don't know at what temp they start throttling though. That probably happens below the critical temp.

 

[DrezKill]

But yeah, small fans generally have horrible reliability. Nothing innate about small fans, but they are just usually cheaply made by Chinese junk manufacturers.

I'm no stranger to that game. The northbridge fan on my nForce 4 board died within a year. I replaced it with a nice all-copper HSF meant for small GPUs, but the damage had already been done. The northbridge was toast. It didn't help that I had been running it overclocked (had to overclock the NB to overclock the CPU, which was an Opteron 165).

If I need a small fan these days the only ones I trust are Noctua's 40mm units. 40mm isn't really that small, but it really is the smallest reliable fan I have found on the market.

Completely agree with everything you said here.

 

[Endgame]

Oh gawd, is that really true? Back in the day they would last for years no problem.

From 2006 to 2012 I shared an office with the guy who did all the support for the high end developer workstations. They did custom builds until 2010 and then moved to HP Xeon systems. Constant stream of machines with grinding NB fans that the devs wanted fixed.

Before that, during college I worked at a local computer shop that sold 25-30 PC / month. We replaced so many fans on warranty it was crazy.

I can’t really say if they are still that bad, though. Maybe things have gotten better.

 

[LazyGamer]

I can’t really say if they are still that bad, though. Maybe things have gotten better.

I think the X570 chipset fans did alright, though AMD did put out an 'X570S' revision that didn't need one.

My personal experience has been with ITX boards though, which try to stuff in extensive power delivery components into a small space, and then don't tune the fans properly. These were the answer when I was wondering why the room an ITX build was in had started to faintly sound like a server room - that chipset fan was spinning in excess of 7,000RPM.

 

[Zarathustra]

All of that said, even if I did have a CPU/motherboard that supported Gen5, I wouldn't be buying a Gen5 SSD quite yet.

The Drives just aren't ready yet for the most part. It's just generic Phison rebrand after generic Phison rebrand.

The high sequential speeds are all good and well, but they are mostly useless unless you are repeatedly copying large files between similarly fast drives. I'm waiting for better 4k random reads.

There are a string of reviews out there touting the Corsair T705 as the "fastest SSD" but if you look at testing, the low queue depth 4k random reads are pretty abysmal.



68.71 MB/s is pathetic.

Not only are my Gen4 Samsung 990 Pro, Samsung 980 Pro and WD Black SN850x drives faster than this in the 80-90 MB/s range, but my Gen3 Intel Optane 905p is WAY faster than this 239MB/s

That's 3.5x faster in a six year old product.




That high RND4k read figure makes a system so much more responsive and faster feeling in most realistic cases. I am still using the 905p as my boot drive:



Even with its max sequential speed becnhed at 2700MB/s the system feels much faster than when I booted off of my 2TB Samsung 990 Pro which I benched at 7500 MB/s.

I wish I could have both higher sequential speeds and higher 4k random speeds, but unfortunately that doesn't seem to be an option.
Don't get me wrong, I'm not holding out for something to beat the Optane (that pretty much isn't going to happen unless I find a good deal on an Optane DC P5800X which pretty much isn't going to happen.) but I am likely going to wait when it comes to moving to Gen5 until something better than the garden variety Phison rebrands is available. Maybe Samsung or WD drives.

I also really wish the marketing people didn't keep pushing higher and higher sequential bandwidth. It is at the point where it is mostly useless. Shift the number. Start battling on QD1 4k random reads instead. THAT will make a huge difference.

We don't NEED Optane to drive improvements in 4k random speeds.

 

[Endgame]

Z nailed it. Capacity and sequential performance are not what I’m looking for. I want better 4k random performance, and the frustrating part is that the technology exists but no one is using it. Even if optane never comes back, low latency SLC would offer more than double the 4k random speed and would otherwise be tech all the drive providers are already using

 

[Zarathustra]

Z nailed it. Capacity and sequential performance are not what I’m looking for. I want better 4k random performance, and the frustrating part is that the technology exists but no one is using it. Even if optane never comes back, low latency SLC would offer more than double the 4k random speed and would otherwise be tech all the drive providers are already using

SLC is great, but compared to the TLC drives that are currently the norm, the storage density would be about a third.

Flash NAND certainly isn't 100% of the cost (controllers, boards, heatsinks, etc. Also have cost) but I would imagine that would still drive up the cost by ~3x.

I'd pay it. You might. But I think we are e light of a minority that there just isn't enough demand to warrant it. At least not until storage be Charles start considering the impact of other things than just high queue depth sequential performance.