News PCIe 5.0 SSDs May Get Thicker and Require Wider M.2 Slots

InvalidError

Titan
Moderator
"Nobody in their right mind would buy a 5.0 drive to cripple it on an older motherboard."

Only for now. 3-4 years down the line, there won't be much of a price premium left to 5.0 over 4.0 just like there isn't much of a premium left to 4.0 over 3.0 today for the shrinking number of nearest equivalents available in both flavors.

BTW, just looked up WD's 1TB SN750 (I have the 500GB model in my PC) vs SN770 and the 4.0x4 SN770 is actually $5 cheaper at my local store.
 
Last edited:
However, it isn't a big deal if we look at it objectively since no one in the right mind would pay for a premium PCIe 5.0 SSD and cripple its performance on an older motherboard.

Wow, really? Not like people would ever get a PCIe 5.0 drive if the price difference between it and a 4.0 drive is small, or even not so small, so when they upgrade later they will already have PCIe 5.0.
 

USAFRet

Titan
Moderator
And with PCIe 5.0, maybe we'll see even less of the minimal difference between 3.0 and 4.0.

We are deep into diminishing returns.

5.0 drives will eventually become the norm (as did 3.0 and soon 4.0), but the rest of the system has a lot of control over how "fast" things are.

Sometime soon, someone will be wanting to forcefit a PCIe 5.0 drive into a i7-2600 system, and wondering why it isn't getting uber speed.
 

InvalidError

Titan
Moderator
And with PCIe 5.0, maybe we'll see even less of the minimal difference between 3.0 and 4.0.

We are deep into diminishing returns.
Mainly because mainstream has little to no use for it yet. If games ever evolve to pervasive use of DirectStorage, it may become essential to smooth gaming experience and mostly eliminating loading screens.
 

kjfatl

Reputable
Apr 15, 2020
216
157
4,760
Did I miss something? The headline of the article indicated that the SSD's would be getting thicker. Then the article talks about wider PCB, but the thickness of the PCB was not addressed. A thicker PCB would require a new type of connector and would likely be incompatible with current generation m.2 cards. A thicker CCA would improve signal quality etc..
 

Co BIY

Splendid
Did I miss something? The headline of the article indicated that the SSD's would be getting thicker. Then the article talks about wider PCB, but the thickness of the PCB was not addressed. A thicker PCB would require a new type of connector and would likely be incompatible with current generation m.2 cards. A thicker CCA would improve signal quality etc..

Agreed, I read it as you did. Although I assumed the thicker dimension would be the space between the card and MB to allow more heat sink on the bottom of the card.

But viewing the M.2 card vertically (like a supermodel) then wider does equal thicker.
 

InvalidError

Titan
Moderator
A thicker CCA would improve signal quality etc..
PCB thickness doesn't have much of an effect on signal integrity as long as it is thick enough to accommodate whatever number of ground and power planes are necessary to provide signal traces with continuous reference planes and uniform impedance. Also, since the NVMe controller is almost butting against the NVMe connector, there isn't much of a PCB between the card edge connector and controller to affect anything.

Agreed, I read it as you did. Although I assumed the thicker dimension would be the space between the card and MB to allow more heat sink on the bottom of the card.
Slapping a heatsink on the bottom is silly as it wouldn't have any airflow to get the heat off. The few studies I have seen about NAND vs temperature say NAND doesn't require cooling except for long-term off-line storage where you lose approximately half of the data retention time for every 10C above 25C. The controller is the only part that really requires cooling so it doesn't end up cooking itself and nearby NAND.
 

watzupken

Reputable
Mar 16, 2020
1,176
660
6,070
Wider, not thicker.

Currently, 2280
22mm wide, 80mm long

Proposed, 25110
25mm wide, 110mm long
It is actually both wider and thicker. The reason why it is thicker is because it is very clear that PCI-E 5.0 SSDs cannot sustain its performance without any heatsink on it. I suspect if a high end PCI-E 5.0 SSD is used in say a laptop, it will likely require some more elaborate cooling solution than what we see now with PCI-E 3.0 and 4.0 SSDs.
 

watzupken

Reputable
Mar 16, 2020
1,176
660
6,070
Mainly because mainstream has little to no use for it yet. If games ever evolve to pervasive use of DirectStorage, it may become essential to smooth gaming experience and mostly eliminating loading screens.
Perhaps. But will doubling down on transfer rate really help? Since SSDs hit retail market more than a decade ago, I've actually not seen much improvements other than a dramatic increase in transfer rate. I feel critical aspects like latency/ seek time have not made any tangible improvement. This situation reminds me of the great megapixel chase where every camera maker are just pushing out higher and higher megapixel cameras. However, the underlying problems like poor lens and sensor remains, resulting in the same crappy picture quality regardless of megapixels. In short, the product is only as good as the slowest/ weakest aspect of it.
 

InvalidError

Titan
Moderator
Perhaps. But will doubling down on transfer rate really help? Since SSDs hit retail market more than a decade ago, I've actually not seen much improvements other than a dramatic increase in transfer rate. I feel critical aspects like latency/ seek time have not made any tangible improvement.
There is actually a massive improvement going from HDDs to almost any SSD from access time going from ~10ms to 0.1-0.2ms, that is the main reason why boot time goes from 2-3mins to ~10 seconds by just swapping out the boot HDD for an SSD. I'd call that quite tangible.

Most of NVMe's potential won't be realized until software is written with NVMe's capabilities in mind and you'll have to wait for games showcasing DirectStorage for that. I'm sure that game engines being able to access resources from NVMe with microsecond-scale latency will open up quite a few possibilities.
 

g-unit1111

Titan
Moderator
Perhaps. But will doubling down on transfer rate really help? Since SSDs hit retail market more than a decade ago, I've actually not seen much improvements other than a dramatic increase in transfer rate. I feel critical aspects like latency/ seek time have not made any tangible improvement. This situation reminds me of the great megapixel chase where every camera maker are just pushing out higher and higher megapixel cameras. However, the underlying problems like poor lens and sensor remains, resulting in the same crappy picture quality regardless of megapixels. In short, the product is only as good as the slowest/ weakest aspect of it.

The average user probably couldn't tell the difference between something like a Samsung 980 Pro and an Intel 660P unless they're really looking for it. Enthusiasts would probably notice a ~1 - 2ms difference in load time, but you probably wouldn't know it either unless you were using bench marking software designed for that sort of thing. Is it worth upgrading? Depends on what you want to get out of the drive in terms of performance.
 

drajitsh

Distinguished
Sep 3, 2016
136
25
18,720
Why this bespoke form factor -- the E1S is already in use for Enterprise drives, and it is likely to completely replace M2 for PCIe5 servers. Also It has plenty of space for heatsinks for drives upto 25W. In addition the vertical orientation means that it will take up much less MB space. why go for a new form factor
 

InvalidError

Titan
Moderator
Why this bespoke form factor -- the E1S is already in use for Enterprise drives, and it is likely to completely replace M2 for PCIe5 servers.
Because consumer drives aren't going to adopt the backplane-based E1 form factor which requires a drive rack of some sort to support the SSDs. Not a problem for servers where motherboards and 1U/2U pizza boxes are tailor-made to fit the application, not so much in ATX-land where no means of supporting those exist. Motherboard manufacturers would either have to integrate the bulky E1 support directly on the motherboard, come up with an off-board connector to allow the drives to be installed and supported elsewhere, or a new ATX spec would need to be created to standardize E1 connector placement on motherboards and cases.