• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Gen5 NVMe have active cooling

Page 3 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
BFG10K said:
I've called out Intel multiple times over their constant redefinition of TDP, and AMD trying to tell us 95C CPU temperature is somehow normal.

But at least CPU sockets and PCIe slots have room for cooling with big slow fans, and you can tame the parts with power caps and in many cases lose little performance.

OTOH NVMe is in the worst possible place for cooling. And they also do nothing but jack up sequential performance each gen which does nothing for the vast majority of users except create gum-stick furnaces.

Also the throttling is getting worse and worse. Gen 3.0 drives can still get 1GB/sec, but these new pieces of trash drop to 100 MB/sec. That's slower than my 2012 VelociRaptor.
Click to expand...

I completely agree with basically everything you're saying here, but the point you previously made isn't about this.

Even twenty years ago a CPU reviewer would have been laughed off the Internet for complaining that a CPU without any cooling system gets hot and throttles performance. The only valid exception was when Athlon XP CPUs without cooling toasted themselves rather than throttled performance, because an improperly installed HSF or a CPU HSF bracket breaking is plausible.
 
WelshBloke said:
Thats like saying an RTX4080 is no better than an Intel integrated gfx because minesweeper plays the same on both!
Click to expand...

Not really. You basically asserted that it's to be expected that an SSD will be "worn out" after ten years regardless of its usage scenario. I challenged that assertion. Then you dropped this straw man into the mix.
 
mikeymikec said:
Not really. You basically asserted that it's to be expected that an SSD will be "worn out" after ten years regardless of its usage scenario. I challenged that assertion. Then you dropped this straw man into the mix.
Click to expand...
You were talking about replacing a fan on a ten year old NVME SSD.
There's no guarantee that an m2 slot won't be obsolete by then.
 
WelshBloke said:
You were talking about replacing a fan on a ten year old NVME SSD.
There's no guarantee that an m2 slot won't be obsolete by then.
Click to expand...

i have to agree with this.
I don't think nvme's will even last 5yrs with how much capacity is required today vs what was required a few years ago.
Meaning that brand new 512GB SSD will be just about as good as a flash drive possibly several years from now.

I dont think speed really has a play in this after PCI-E 4.0 bandwidth.
I think we have another bottleneck somewhere else, like fetching, or in the actual I/O.
But definitely we'll see a shift in more capacity, as programmers are getting lazy as hell in optimization, and what took 512MB 10yrs ago, now takes 51.2GB to reproduce with today's programmers.
 
Yes. I think I used to be pretty happy with 10GB HDD and it had tons of games and movies for the time. The stuff kept me entertained for hours! Now just installing a single 100 GB game takes the fun out of enjoying it. It's like more and more hurdles have been created in the pursuit of graphical fidelity and realism but we can mostly still tell that it's a game. The games of the future may not even be downloadable. They will stream directly from Steam servers because the textures will be gigabytes in size and 10 gigabit internet will be common everywhere so games will have something like "100 GB temporary space" for minimum space requirement which will be filled with textures as and when needed. Offline AAA titles may cease to exist.
 
aigomorla said:
i have to agree with this.
I don't think nvme's will even last 5yrs with how much capacity is required today vs what was required a few years ago.
Meaning that brand new 512GB SSD will be just about as good as a flash drive possibly several years from now.

I dont think speed really has a play in this after PCI-E 4.0 bandwidth.
I think we have another bottleneck somewhere else, like fetching, or in the actual I/O.
But definitely we'll see a shift in more capacity, as programmers are getting lazy as hell in optimization, and what took 512MB 10yrs ago, now takes 51.2GB to reproduce with today's programmers.
Click to expand...
Think we'll ever see system ram and system storage converge?
If non volatile memory gets faster we could see laptops with what's effectively hundreds of TBs of system ram!
 
mikeymikec said:
Even twenty years ago a CPU reviewer would have been laughed off the Internet for complaining that a CPU without any cooling system gets hot and throttles performance.
Click to expand...
But CPUs/GPUs never made any claims to power usage or heat.

M2/NVMe did. It specifically claimed lower power usage, lower temperatures, and a smaller footprint compared to HDDs, advantages that no longer apply to some Gen5.0 parts.

Again, what's the end-game here? It's obvious die shrinks aren't helping the controllers because they're getting worse. So either it's mandatory liquid cooling because there's no room for CPU tower coolers, or we admit M2 is a failure.
 
BFG10K said:
But CPUs/GPUs never made any claims to power usage or heat.

M2/NVMe did. It specifically claimed lower power usage, lower temperatures, and a smaller footprint compared to HDDs, advantages that no longer apply to some Gen5.0 parts.

Again, what's the end-game here? It's obvious die shrinks aren't helping the controllers because they're getting worse. So either it's mandatory liquid cooling because there's no room for CPU tower coolers, or we admit M2 is a failure.
Click to expand...

It really does seem like PCIe 5.0 gen and higher needs something like U.2 or whatever will be its successor in the workstation/server world. Because these drives are not going to work in laptop form factors at all, and they're arguably too hot even for desktop with massive built-in M.2 heatsinks. M.2 form factor offers no value here.

That leaves OEMs in the lurch since they would mostly rather sell laptops/laptop hardware, which is why M.2 came to dominate over U.2 in desktops (that are relative afterthoughts at this point in the game). Without M.2, what do you do? Stick with PCIe 4.0 M.2 drives forever?
 
BFG10K said:
Here we are bois, water cooled NVMe:

www.tomshardware.com

Adata Details SSD with Self-Contained Liquid Cooling System: Up to 14GB/s

Extreme performance meets extreme capacity
www.tomshardware.com www.tomshardware.com

No way that'll fit alongside any tower CPU cooler, which leaves Intel/AMD stock coolers. The intersection between users of those and Gen5 drives is virtually zero.

We've reached a new level of lunacy, folks.
Click to expand...
It would fit on the Gen-Z.2 card on my ROG X670E Crosshair Extreme. It looks like a perfect solution for my setup. The fan would be at the bottom blowing air up through the cooler and into the exhaust fans in the top of my case.

Not a fan of the colors. All black would make it perfect.

1685624112807.png
 
This is the future of NVMe:

www.techpowerup.com

Innodisk at Computex 2023: Has the Right Idea About Gen 5 SSDs, to Make them AICs

Innodisk has the right idea about how to do PCIe Gen 5 NVMe SSDs—to ditch the M.2 form-factor, and make them PCIe add-in cards. This would remove the need for cartoonishly disproportionate cooling solutions with high-pitched 20 mm fans; and rather allow SSD designers to use cooling solutions...
www.techpowerup.com www.techpowerup.com

M.2 is a dead-end interface. The slot was never designed for bulky coolers with so much weight hanging from it. What's next, anti-sag brackets for SSDs? LOL.

SATA is perfectly functional and they should've just kept improving it. 2.5" drives are already very light heatsinks which can be put literally anywhere with double-sided tape, including by case intake fans.
 
Last edited:
BFG10K said:
This is the future of NVMe:

www.techpowerup.com

Innodisk at Computex 2023: Has the Right Idea About Gen 5 SSDs, to Make them AICs

Innodisk has the right idea about how to do PCIe Gen 5 NVMe SSDs—to ditch the M.2 form-factor, and make them PCIe add-in cards. This would remove the need for cartoonishly disproportionate cooling solutions with high-pitched 20 mm fans; and rather allow SSD designers to use cooling solutions...
www.techpowerup.com www.techpowerup.com

M.2 is a dead-end interface. The slot was never designed for bulky coolers with so much weight hanging from it. What's next, anti-sag brackets for SSDs? LOL.

SATA is perfectly functional and they should've just kept improving it. 2.5" drives are already very light heatsinks which can be put literally anywhere with double-sided tape, including by case intake fans.
Click to expand...

If the industry did shift to AICs, I guess most notebooks would remain on M.2 NVMe and take whatever passively-cooled solutions they can?
 
BFG10K said:
This is the future of NVMe:

www.techpowerup.com

Innodisk at Computex 2023: Has the Right Idea About Gen 5 SSDs, to Make them AICs

Innodisk has the right idea about how to do PCIe Gen 5 NVMe SSDs—to ditch the M.2 form-factor, and make them PCIe add-in cards. This would remove the need for cartoonishly disproportionate cooling solutions with high-pitched 20 mm fans; and rather allow SSD designers to use cooling solutions...
www.techpowerup.com www.techpowerup.com

M.2 is a dead-end interface. The slot was never designed for bulky coolers with so much weight hanging from it. What's next, anti-sag brackets for SSDs? LOL.

SATA is perfectly functional and they should've just kept improving it. 2.5" drives are already very light heatsinks which can be put literally anywhere with double-sided tape, including by case intake fans.
Click to expand...
M.2 is good for mobile.

The Phison e26 is built on 12nm, surely a shrink to 7 or 5 nm would reduced power consumption to tolerable levels?

At least I hope so, and that's why I'll wait for future generations before getting one.
 
mikeymikec said:
If the industry did shift to AICs, I guess most notebooks would remain on M.2 NVMe and take whatever passively-cooled solutions they can?
Click to expand...
The issue is one of volume. Why waste resources producing drives for desktops when the bulk of consumer PCs are now notebooks of one form or another? That's why we have M.2 slots on desktop motherboards in the first place, so that OEMs could produce products in a form factor that would fit desktop and mobile.

Of course that lead us to the absurdity of boutique M.2 drives with performance and power consumption so great that no notebook computer could hope to contain them.
 
biostud said:
The Phison e26 is built on 12nm, surely a shrink to 7 or 5 nm would reduced power consumption to tolerable levels?
Click to expand...
It would, but following the trend so far, they'd instead use it on Gen 6 which would more than offset any gains and would require even more cooling.

The latest Gen 3 Samsung Evo Plus has exactly the same thermal problems as my four year old version. There was no die-shrink that made it better. I also can't think of a single Gen 4 drive that was later re-released without a heatsink, and the bare drive no longer throttled.

The problem has been masked by more motherboards coming with heatsinks, so most people automatically use them and not bare drives. Also barely any SSD reviews test thermal throttling.

I didn't even know it was happening to me until I checked the temperature. Previously I assumed the dropping file copy speeds were just the SLC cache being exhausted.

Gen 5 has reached the tipping point where it needs active cooling not to throttle, so we can see rising thermals have been a problem all along.

It would be useful for BIOSes to allow controlling M.2 lanes just like PEG. I'd be happy to drop to 2x or 1x so I could run a bare drive without throttling. Sadly my MOBO only has a SATA / NVMe toggle.
 
Last edited:
Looks like a graphics card until you look closer.

videocardz.com

GALAX reveals HOF Extreme 50S PCIe Gen5 SSD, up to 12.4 GB/s read speed and 1500K IOPS - VideoCardz.com

GALAX HOF Extreme 50S, an even faster PCIe Gen5 SSDs GALAX reveals an updated PCIe Gen5 storage for high-end desktop and laptops. The HOF Extreme 50 was announced in December last year, and it was among the fastest solid-state storage for consumers at the time. Nonetheless, like with all new...
videocardz.com videocardz.com

In other news, PCIe 7 standard is out. Liquid nitrogen baths for such SSDs, I'll bet.

www.engadget.com

PCI Express 7 will be eight times faster than PCI Express 5

The PCI-SIG committee has unveiled the latest PCI Express 7.0 standard with blazing speeds, even as we wait for PCIe 5 devices to arrive.
www.engadget.com www.engadget.com
 
I still don't understand why they just move from 12nm to 7nm for the controller, it should save a lot energy, right?
 
biostud said:
I still don't understand why they just move from 12nm to 7nm for the controller, it should save a lot energy, right?
Click to expand...
Yes, but the companies think from a cost/business perspective.
They can probably produce 12nm controllers for 20% cheaper (or more). They will always lag the leading process tech by a few generations because of this.
The fact that Phison is fabless, and a smaller company, also mean they'll likely always lag samsung, micron, wd, seagate, etc.
Not just because they are fabless, but also less volume or negotiating power with the fabs (and likely more design time).
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top