Crucial T700 Gen 5 SSD Throttles Down to HDD Performance Levels Without a Cooler

[btarunr]

Crucial T700, the company's flagship M.2 NVMe Gen 5 SSD, runs hot—like every other drive based on the Phison E26-series controller (such as the Corsair MP700). ComputerBase.de discovered what the drive does without some sort of cooling. The E26 controller has a Tjmax value of around 86°C, and what happens when it's reached depends on the drive in question. The Corsair MP700 can turn itself off to protect the controller—something that will definitely cause your machine to hang with a BSOD.

The Crucial T700, on the other hand, aggressively throttles down the controller in an attempt to lower temperatures. While the drive won't stop (and your machine won't hang), its performance drops to hard drive levels, with CrystalDiskMark (CDM) measurements pointing to around 101 MB/s (of course, with much lower access times than a HDD). Both Crucial and Corsair offer the drive with large heatsinks, and recommend users to use them. This should severely limit the adoption of Gen 5 NVMe SSDs among notebooks, where the notebook chassis has room for only bare drives. However, some OEMs specializing in larger high-end gaming notebooks and desktop-replacement workstations, can find ways to connect the drives to the notebook's main cooling system using flattened heatpipes. You can catch ComputerBase's review of the MP700 in the source link below.



View at TechPowerUp Main Site | Source

 

[Minus Infinity]

It's crucial you avoid this turkey.

 

[lemoncarbonate]

Why is everything getting ridiculously hotter?

Next gen CPU and GPU are getting hotter, now SSD.
I thought next gen should also mean better power efficiency.

 

[TheLostSwede]

It's a firmware issue, Phison is already working on a fix. It'll most likely still throttle without a heatsink, but not to this degree and there shouldn't be a risk of data loss as in some cases.

PCIe 5.0 SSDs Generate Errors, Shut Down Without Cooler: Fix On The Way

Phison has isolated the problem and is working on a solution.

www.tomshardware.com

 

[Aleksandar_038]

Why is everything getting ridiculously hotter?

Next gen CPU and GPU are getting hotter, now SSD.
I thought next gen should also mean better power efficiency.

That is a trend for quite some time... Remember how P III or AMD Athlon Thunderbird cooler looked like?

Now, we need AIO or more than 1 kg of aluminum for cooling of high-end CPU. That is generally wrong direction that HW development took many years ago, along with some basic physics kicking in now when we are to single-digit nm level...

 

[Dr. Dro]

"The Corsair MP700, ComputerBase notes, can behave turn itself off to protect the controller"

That doesn't sound cohesive to me

W1zz found the same in his review of the Corsair drive, that it suffered from thermal shutdowns. The Phison rep actually answered folks' questions, I liked that.

 

[Hyderz]

In my early 20s as I was building and upgrading pcs, it never occurred to me that storage could get hot I noticed this when I upgrading os, new games and I couldn’t figure out what was causing crashes and hitches. Opened the case and touch individual parts for heat like cpu ram and gpu but they were mostly cool to touch but the hdd was furnace hot… so I bought a fan pull in cool air to the hdd and the problems went away… this is when I started thinking about air case flow

 

[bug]

Why is everything getting ridiculously hotter?

Next gen CPU and GPU are getting hotter, now SSD.
I thought next gen should also mean better power efficiency.

Well, if you built the same thing on the next-gen process, it would surely be more efficient. But if you also want to double the number of transistors or communication speeds, temps can only go up. It's still more efficient per transistor or per Gb/s, it just output more heat overall.

 

[john_]

That is a trend for quite some time... Remember how P III or AMD Athlon Thunderbird cooler looked like?

Now, we need AIO or more than 1 kg of aluminum for cooling of high-end CPU. That is generally wrong direction that HW development took many years ago, along with some basic physics kicking in now when we are to single-digit nm level...

It might be the wrong direction, but it was also the direction consumers wanted. AMD played the efficiency game and Intel won by just hitting 300W and winning single thread and gaming benchmarks. Then Nvidia pushed over 300W and won again, with the difference in Nvidia's case being that in the end they didn't really needed that extra power, at least for 4000 series.

As for the SSDs. Do we really need 10GB/sec? PCIe 3.0 speeds are more than enough and only pros and gamers with zero patience really need more.

 

[robert3892]

I have a gen 5 Corsair MP700 M.2 SSD. I have a third party m.2 SSD cooler on it with a fan. Temps seem to fluctuate between 64C and 67C.

 

[Arco]

These SSDs suck. Don't buy until Q3-4 2023 at the earliest. Too much heat, not much real gain, and really expensive.

 

[Aleksandar_038]

It might be the wrong direction, but it was also the direction consumers wanted. AMD played the efficiency game and Intel won by just hitting 300W and winning single thread and gaming benchmarks. Then Nvidia pushed over 300W and won again, with the difference in Nvidia's case being that in the end they didn't really needed that extra power, at least for 4000 series.

As for the SSDs. Do we really need 10GB/sec? PCIe 3.0 speeds are more than enough and only pros and gamers with zero patience really need more.

Yes, true.

Also, instead of optimization of software, we are just increasing complexity to stupid levels and solve everything with MOAR RAM, MOAR MHz, MOAR everything. Completely unsustainable... But, yes, that is what we (in general, as consumers) wanted...

 

[bug]

It might be the wrong direction, but it was also the direction consumers wanted. AMD played the efficiency game and Intel won by just hitting 300W and winning single thread and gaming benchmarks. Then Nvidia pushed over 300W and won again, with the difference in Nvidia's case being that in the end they didn't really needed that extra power, at least for 4000 series.

As for the SSDs. Do we really need 10GB/sec? PCIe 3.0 speeds are more than enough and only pros and gamers with zero patience really need more.

Efficiency is still there. Since AthlonXP, IPC became an important metric, nobody was talking about IPC before that. This efficiency works well till some point. After that, you can push the architecture some more, but it will require more power for limited performance gains. That will never change, it's just how physics work. Intel played that card with their CPUs, AMD played it too with their GPUs back in Polaris days. Everybody does it. Imho the trick is reading enough reviews to gauge where the sweet spot lies. Then you can buy into the sweet spot or buy something slightly outside of it an underclock/undervolt to tune it to your preferences. I did just that with my 12600k, put some hard limits on power draw and now I have a fast, cool beast.

Edit: And just to cover all sides, Nvidia also pushed their cards too far with Fermi.

 

[watzupken]

These SSDs suck. Don't buy until Q3-4 2023 at the earliest. Too much heat, not much real gain, and really expensive.

It doesn't matter when you buy it because this is an inherent issue when the controller runs faster and the NAND pushed to run faster as well. "Faster" here however, generally don't bring any benefit for most users.

 

[mama]

I don't think the business case for gen 5 drives is there yet. Good Gen 4 drives are plenty fast and cool enough to work.

 

[bug]

I don't think the business case for gen 5 drives is there yet. Good Gen 4 drives are plenty fast and cool enough to work.

It could be if you're YouTube and need to read lots of videos fast. Full disclosure, I'm not YouTube

But yes, the bottleneck with SSDs isn't their sequential performance, it's the random reads. PCIe (any revision) will do little to help with that.

 

[TheDeeGee]

It's crucial you avoid this turkey.

Any Gen5 NVME will throttle without a heatsink.

 

[TheinsanegamerN]

Why is everything getting ridiculously hotter?

Next gen CPU and GPU are getting hotter, now SSD.
I thought next gen should also mean better power efficiency.

Thermal density is a bitch.

That is a trend for quite some time... Remember how P III or AMD Athlon Thunderbird cooler looked like?

Now, we need AIO or more than 1 kg of aluminum for cooling of high-end CPU. That is generally wrong direction that HW development took many years ago, along with some basic physics kicking in now when we are to single-digit nm level...

TBF you can still do that. I've run desktop pentiums with NO heatsink without issue. Those skylake gold designs, even at 4.0 GHz, just dont get warm. A core i3 can run on a dinky little heatsink, so can T series i5s.

 

[bug]

Any Gen5 NVME will throttle without a heatsink.

More precisely, any PCIe5 NVMe will throttle without a heatsink and good airflow

 

[InVasMani]

We'll probably see a node shrink resolve all or most of the larger issues with Gen 5. I think the quick transition from Gen 4 to Gen 5 probably hasn't allowed for as much maturity on the node side of things and this kind of the end result of all of that. Give it time and I'm sure it'll reach better maturity though this early hardware for Gen 5 is looking rather subpar.

 

[trparky]

It's crucial you avoid this turkey.

I see what you did there.

 

[phanbuey]

I would much rather just have slower consistent nvme speeds than something that throttles...

 

[skates]

Well, if you built the same thing on the next-gen process, it would surely be more efficient. But if you also want to double the number of transistors or communication speeds, temps can only go up. It's still more efficient per transistor or per Gb/s, it just output more heat overall.

Absolutely correct. People automatically assume next gen means companies have magically found a way around thermal density when packing more and more transistors into a chip. The fact of the matter is, go fast means more heat.

Also don't get the push back on putting heat sinks on SSD drives, although I'm sure the same was said about adding a heat sink to the CPU and then to the GPU, and then all over again when water cooling became a thing...the horror.

I started collecting SSD heat sinks a couple of years ago because I was intrigued about their effectiveness. They do work, some better than others and the one that works best right now is one with a full copper heat sink. I've also had mixed results putting thermal pads on the entire SSD, as is norm, vs. just the chips. What does help most is getting airflow on the heatsink. My gen4 980Pro doesn't go above 43C under heavy loads with ambient temps at 78 degrees and it does not have a fan attached to it, but it does have a full copper heatsink.

 

[bug]

We'll probably see a node shrink resolve all or most of the larger issues with Gen 5. I think the quick transition from Gen 4 to Gen 5 probably hasn't allowed for as much maturity on the node side of things and this kind of the end result of all of that. Give it time and I'm sure it'll reach better maturity though this early hardware for Gen 5 is looking rather subpar.

Then you're probably aren't up to speed on how SSDs work. Node shrinks are actually detrimental to NAND cells (less charge per cell, more chance to err).
How much improvement have we witnessed since PCIe 4 drives were introduced? Zero. For SSDs, going faster = going hotter. The same thing to do, imho, would be to put a cap on sequential transfer speeds and concentrate on improving random access instead.

 

[Snoop05]

Then you're probably aren't up to speed on how SSDs work. Node shrinks are actually detrimental to NAND cells (less charge per cell, more chance to err).
How much improvement have we witnessed since PCIe 4 drives were introduced? Zero. For SSDs, going faster = going hotter. The same thing to do, imho, would be to put a cap on sequential transfer speeds and concentrate on improving random access instead.

Thankfully most motherboards let you limit the pcie interface. (which for the most part should acheive the goal)

Also the drive throttling (especially if done gruadually, not suddenly) while still operating correctly is a good thing, way better than not throttling (or not enough) and getting R/W errors like the Corsair MP700.