Corsair MP700 2 TB

[Metroid]

Too expensive x pcie 4.0 but when pcie 4.0 drives came in were very expensive too.

 

[AusWolf]

Too expensive x pcie 4.0 but when pcie 4.0 drives came in were very expensive too.

At least they didn't shut down due to thermal issues caused by normal use.

 

[Vya Domus]

I really don't understand what is the point of these, PCIe 5.0 drives provide effectively no real performance gain.

What are sequential read/writes actually useful for ?

 

[Mr. Perfect]

Is there a way to start testing these drives for DirectStorage performance?

I picked up a WD Blue SN570 a couple years back for my gaming PC because nothing was significantly beating it in your game load testing. Fast forward to today and it's still right there with the other drives for conventional game loading. At this rate, I'm not going to even bother upgrading from PCIe 3 unless DirectStorage starts showing some significant gains on newer gen drives.

 

[ymdhis]

It would be great if you could set it to permanent SLC mode with 660GB capacity and a lot higher endurance.

It only has 512GB SLC cache, the 660GB result comes from the cache constantly being written to the TLC portion. If the entire drive was just the SLC cache it would likely be something like 480GB at best due to over provisioning (unless the 512GB includes that too).

I'm sure there are some SLC only enterprise drives, for 10x the price.

 

[bug]

I really don't understand what is the point of these, PCIe 5.0 drives provide effectively no real performance gain.

What are sequential read/writes actually useful for ?

Sequential speeds have their uses: backing up, restoring, working with large images or video...
But there is very little reason to move beyond PCIe 4 without addressing thermal issues first. Without that, it's just marketing and a higher price tag.

 

[Psychoholic]

I really don't understand what is the point of these, PCIe 5.0 drives provide effectively no real performance gain.

What are sequential read/writes actually useful for ?

Agreed, the only way these sequential numbers would be useful is if you had two of these drives copying between them.

Otherwise, random 4k is what you actually "feel"
This drive did pretty well in random though, too.

 

[konga]

In what real-world scenario can you actually achieve 10GB/s? Or does this speed only exist in the realm of synthetic benchmarks?

 

[robert3892]

There is no fan, there is no heatsink. Our early sample included a heatsink with a fan, which Corsair decided to scrap for public release. The drive is sold without heatsink as detailed on Corsair's website.

Corsair released the MP 700 several weeks ago with a fan. It looks like they scrapped the fan idea. This is probably why they removed the MP700 from their website and replaced it weeks later minus the fan, I am using a third party fan and heatsink I bought from Aliexpress.

 

[AnarchoPrimitiv]

Anyone else think a "subjective" portion to SSD reviews should be added? What I mean is that they should take the drive being reviewed (only when it claims new or class leading performance figures), install the OS on it, and use it for 3 days and comment on whether the drive actually felt like it improved the overall user experience....better yet, it can be compared against an Intel Optane drive and the user experience provided (as that would be the ultimate in low latency). I think it would help a lot of potential customers, enthusiast and casual alike, to know whether an average human and actual perceive any improvement to the overall experience.

 

[Metroid]

I really don't understand what is the point of these, PCIe 5.0 drives provide effectively no real performance gain.

What are sequential read/writes actually useful for ?

The only thing I can think of is for virtual memory in case you use all your physical memory, other than that nothing.

 

[AusWolf]

Anyone else think a "subjective" portion to SSD reviews should be added? What I mean is that they should take the drive being reviewed (only when it claims new or class leading performance figures), install the OS on it, and use it for 3 days and comment on whether the drive actually felt like it improved the overall user experience....better yet, it can be compared against an Intel Optane drive and the user experience provided (as that would be the ultimate in low latency). I think it would help a lot of potential customers, enthusiast and casual alike, to know whether an average human and actual perceive any improvement to the overall experience.

In that case, my subjective portion would be that nothing improves user experience over a SATA SSD much, and nvme is only useful because it saves space in your chassis.

 

[Wye]

1. Bad math: Your CrystalDiskMark random 4KQ1T1 shows 93 MB/s, that is 23k IOPS, not 90k IOPS as you show on the next Random Access Performance chart.

2. Get some Intel drives tested before you declare a drive fastest. My 2 generations old 905P PCIe 3.0 wipes the floor with this one on random 4KQ1. High depth queues stats are meaningless for user workloads, they are only good for server workloads or publicity.

 

[chrcoluk]

Is the idle power usage in the review as well, 10w isnt too bad if its only whilst its under full load.

Read it, has high idle power as well compared to competition about an extra 2 watts over my 980 pro.

A drive that shuts down after some constant reads/writes (which is normal operation) is not a drive. If this is a trend with PCI-e 5.0, then I'll stay on 4.0, thanks.

Yeah seems the drive isnt ready, fair play in that the review still got published though.

Crazy they didnt provide a heatsink with this thing.

The only thing I can think of is for virtual memory in case you use all your physical memory, other than that nothing.

Yeah pagefile and hibernation are the two use cases where I have appreciated nvme speeds.

Although I now usually use sleep instead of hibernation and my pagefile is so small so it only pages enough to make sure all software runs but not enough to actually page a program out.

 

[W1zzard]

Your CrystalDiskMark random 4KQ1T1 shows 93 MB/s, that is 23k IOPS, not 90k IOPS as you show on the next Random Access Performance chart.

CDM is just used to provide that screenshot, people have been asking for it, because it gives them a baseline that they can reproduce with 0 testing skill.

The other tests on this page are performed with different software. 92k comes from the line charts at various queue depths. As explained in the subtitle for the bar chart, the number is weighted QD1: 75%, QD2: 20%, QD4: 5%, to approximate queue depths in actual real-life usage.

Get some Intel drives tested before you declare a drive fastest

Feel free to send me one, unfortunately they are all end-of-life. Still, congrats on having a 905P

Is the idle power usage in the review as well

First two charts on the power testing page. 2.8 W in desktop is super high, not a huge deal though. 1.0 W on mobile is, because it sucks your battery dry, other drives have 1/10th of that

 

[Dr. Dro]

Correct, the ASUS PCIe Gen 5 Adapter was used

W1zz, are you able to do a small follow-up regarding random and sequential performance with the SSD installed on the native Gen 4 slot of the motherboard? Or maybe once more Gen 5 drives come up for review, have this segment for all of them published as a review of its own?

I suspect quite a few people with high end motherboards that can't do internal Gen 5 NVMe would be interested in this as a potential upgrade (over it being a high-end, fast device, more so than it being Gen 5 capable), but given how close good Gen 4 drives such as the Kingston KC3000 come, it'd be interesting to see how it behaves.

The drive disappeared from Device Manager, accessing the drive letter returns an error.
Start Menu -> Restart: drive is missing after reboot. Press reset button: drive is missing. Shut down through power button or Start Menu, wait for computer to shut off, press power button, drive is back visible and works normally, no data lost

I've had this bug with a 512 GB ADATA/XPG S40G on the Crosshair VI Hero before. I only ever got it to stop with a BIOS/AGESA update. Exact same symptoms, drive missing after reboot, or sometimes it would show as a "Standard NVM Express Controller", code 43 on Task Manager. No data loss to report either.

I imagine there's still a lot of newness surrounding these Gen 5 drives. Motherboard or drive firmware/Windows NVMe drivers might not be up to snuff just yet, and may need some work.

 

[W1zzard]

I've had this bug

Just to clarify, "The drive disappeared from Device Manager", should be "Once it reached its thermal throttle temperature, the drive disappeared from Device Manager"

 

[Solaris17]

Once the drive got too hot and reached its thermal throttle point, the drive disappeared from Device Manager, accessing the drive letter returns an error.

jesus lmao. no way could I buy this. Great review though.

 

[Dr. Dro]

Just to clarify, "The drive disappeared from Device Manager", should be "Once it reached its thermal throttle temperature, the drive disappeared from Device Manager"

So, like an emergency thermal shutdown? That's terrible, imagine if that happened in your boot drive while you were actually running something important? Nope, big nope from me! Is going to need a firmware update asap.

 

[Ferrum Master]

I would like to see at what duration of uS is the peak power draw is seen. I know, I am a bad boy.

It is pretty hard find to the real latest specification of the M.2 thermal and power envelope. It gets way mixed with AIC and U2. I recall seeing some older Texas Instruments power management IC's and their limit is delivering only 3A of current to 3.3V rail.

Basically I can imagine some gimped board actually not capable of driving these things despite software features declaring maximum power states. I may be optimistic and think, that everything is accounted for, like burning sockets that's not always the case, considering these drives are still rare and the power envelope is really much higher versus the older gen M.2 drives.

 

[W1zzard]

I would like to see at what duration of uS is the peak power draw is seen.

those 10 W are not microseconds .. if you're doing sequential writes you're at 10.7 W _average_

http://imgur.com/e1rF9wN

Won't be an issue with those old power ICs, because they don't support Gen 5, in Gen 4 the drive writes at like 6-7 W

http://imgur.com/saO6Pyb

 

[Chris_Ramseyer]

Hello All,

Gen5 drives sold without a heatsink are meant to be used with the motherboard's existing M.2 cooler or an aftermarket cooler. In the review, Wiz stated the case is void of airflow and if you look at the picture of the system there is only exhaust fans, no intake fans that I can see. Without a heatsink, there needs to be some airflow. The specific number is 2 LFM, a very low number, but it is still some air movement. In most cases, the air moving from a CPU heatsink is enough to avoid the shutdown issue in this review. When using an AIO, there isn't air moving in that area and it gets even worse when using an add-in card in a PCIe slot that traps hot air in the area around the SSD. Gen5 SSDs will require some airflow and it has always been our recommendation to use at minimum a passive heatsink with some case airflow.

The JEDEC power limit for the M.2 slot has stayed at 11.55w for a long time and the E26 drives stay under that mark. The drives can move data at nearly 2x the rate as Gen4 SSDs, so it simply get heats up faster. 11w isn't a lot of heat to bleed off with a heatsink. As long as you have decent airflow (like intake fans at the front of the system blowing back across the motherboard). Right now, E26 is an enthusiast product and I don't know any enthusiasts not using intake fans. I'm not saying the review is wrong in methodology, it's just not an ideal setup for use with our controller at this time.

The Z690 chipset wasn't designed with Gen5 SSDs in mind. As someone already mentioned, the motherboard used in the review doesn't have a Gen5 M.2 slot (or a Gen5 M.2 heatsink). Many Z790 motherboards were designed with Gen5 SSDs in mind and even ship with robust passive M.2 heatsinks, some are even active (thanks ASRock!). ASRock even released aftermarket active heatsinks that integrate seamlessly on their lower-cost boards and sell them for less than $20. They are silent, but work incredibly well.

Z790 boards also significantly increase the ceiling for Gen5 SSD performance. Z690 motherboards don't allow the E26 drives to reach peak performance so the performance results are compressed down. The largest impact is on the all-important low queue depth random reads.

 

[bug]

@Chris_Ramseyer Appreciate the input, but I have a hard time understanding why the lack of a M.2 slot on the motherboard would be factor.
Also, do you figure a typical case with 1-2 fans would suffice if the room temperature gets closer to 30C?

 

[AusWolf]

Hello All,

Gen5 drives sold without a heatsink are meant to be used with the motherboard's existing M.2 cooler or an aftermarket cooler. In the review, Wiz stated the case is void of airflow and if you look at the picture of the system there is only exhaust fans, no intake fans that I can see. Without a heatsink, there needs to be some airflow. The specific number is 2 LFM, a very low number, but it is still some air movement. In most cases, the air moving from a CPU heatsink is enough to avoid the shutdown issue in this review. When using an AIO, there isn't air moving in that area and it gets even worse when using an add-in card in a PCIe slot that traps hot air in the area around the SSD. Gen5 SSDs will require some airflow and it has always been our recommendation to use at minimum a passive heatsink with some case airflow.

The JEDEC power limit for the M.2 slot has stayed at 11.55w for a long time and the E26 drives stay under that mark. The drives can move data at nearly 2x the rate as Gen4 SSDs, so it simply get heats up faster. 11w isn't a lot of heat to bleed off with a heatsink. As long as you have decent airflow (like intake fans at the front of the system blowing back across the motherboard). Right now, E26 is an enthusiast product and I don't know any enthusiasts not using intake fans. I'm not saying the review is wrong in methodology, it's just not an ideal setup for use with our controller at this time.

I would still say: avoid gen 5 SSDs without a heatsink like the plague at this point. Motherboard-supplied heat spreaders aren't really good. With no heatsink fins, they don't have the surface area needed to dissipate the heat, which just soaks in until the inevitable slowdown (or shutdown in this case) happens.

I also tend to say that a product should work as it is. If the SSD is supplied without a heat spreader, it means it should work perfectly even under a GPU or other add-in card. If it doesn't, it's a bad product. The box doesn't say "use with a heatsink" or "place only under a CPU cooler" anywhere.

 

[chrcoluk]

Hello All,

Gen5 drives sold without a heatsink are meant to be used with the motherboard's existing M.2 cooler or an aftermarket cooler. In the review, Wiz stated the case is void of airflow and if you look at the picture of the system there is only exhaust fans, no intake fans that I can see. Without a heatsink, there needs to be some airflow. The specific number is 2 LFM, a very low number, but it is still some air movement. In most cases, the air moving from a CPU heatsink is enough to avoid the shutdown issue in this review. When using an AIO, there isn't air moving in that area and it gets even worse when using an add-in card in a PCIe slot that traps hot air in the area around the SSD. Gen5 SSDs will require some airflow and it has always been our recommendation to use at minimum a passive heatsink with some case airflow.

The JEDEC power limit for the M.2 slot has stayed at 11.55w for a long time and the E26 drives stay under that mark. The drives can move data at nearly 2x the rate as Gen4 SSDs, so it simply get heats up faster. 11w isn't a lot of heat to bleed off with a heatsink. As long as you have decent airflow (like intake fans at the front of the system blowing back across the motherboard). Right now, E26 is an enthusiast product and I don't know any enthusiasts not using intake fans. I'm not saying the review is wrong in methodology, it's just not an ideal setup for use with our controller at this time.

The Z690 chipset wasn't designed with Gen5 SSDs in mind. As someone already mentioned, the motherboard used in the review doesn't have a Gen5 M.2 slot (or a Gen5 M.2 heatsink). Many Z790 motherboards were designed with Gen5 SSDs in mind and even ship with robust passive M.2 heatsinks, some are even active (thanks ASRock!). ASRock even released aftermarket active heatsinks that integrate seamlessly on their lower-cost boards and sell them for less than $20. They are silent, but work incredibly well.

Z790 boards also significantly increase the ceiling for Gen5 SSD performance. Z690 motherboards don't allow the E26 drives to reach peak performance so the performance results are compressed down. The largest impact is on the all-important low queue depth random reads.

We get that, but a drive shouldnt just thermally shutdown, throttling is much more sane behaviour.