Introduction, Specifications, and Packaging
Intel 3D NAND in an M.2 NVMe SSD!
Introduction:
It's been quite some time since we saw a true client SSD come out of Intel. The last client product to use their legendary 10-channel controller was the SSD 320 (launched in 2011), and even that product had its foot in the enterprise door as it was rated for both client and enterprise usage. The products that followed began life as enterprise parts and were later reworked for consumer usage. The big examples here are the SATA-based SSD 730 (which began life as the SSD DC S3500/3700), and the PCI/NVMe-based SSD 750 (which was born from the SSD DC P3700). The enterprise hardware had little support for reduced power states, which led Intel to market the 730 as a desktop enthusiast part. The 750 had a great NVMe controller, but the 18-channel design and high idle power draw meant no chance for an M.2 form factor version of the same. With the recent addition of low-cost 3D NAND to their production lines, Intel has now made began another push into the consumer space. Their main client SSD of their new line is the 600p, which we will be taking a look at today:
Read on for our full review of the Intel SSD 600p M.2 NVMe SSD!
I know what you're thinking. That's no Intel controller. Well, you'd be right, but this is certainly not the first time Intel has sourced controllers elsewhere. To help them bridge the SATA 6Gb/sec gap, their SSD 510 and SSD 520 used Marvell and SandForce controllers, respectively. Fortunately, Intel's custom firmware design and quality control helped them avoid the SandForce-related bugs seen in competing products. The Silicon Motion controller employed here clearly has some Intel touches as evidenced by their logo stamped right on the part as well as a unique part number.
Specifications (source):
Performance looks good overall, but we do note that write speeds appear more limited than what we're used to from an M.2 NVMe part. Also take care to consider that the 600p is a hybrid drive, so those 'up to' write speeds are likely based on typical short bursts that go to the SLC portion and are later paid out to the TLC area. The approximate SLC cache capacities are as follows:
- 128GB – 3GB
- 256GB – 6GB
- 512GB – 16GB
- 1TB – 32GB
* EDIT *
After this article went up, there was some confusion about the encryption capabilities of the 600p vs. the 6000p. Intel sent along this note to try and clear things up a bit:
To get eDrive capability in Intel SSD 6 Series, the user needs the Intel SSD Pro 6000p Series product, not the 600p. The process for setting up and getting eDrive to work on platform with the SSD requires some additional steps. Additional background:
- BitLocker is Windows encryption/data security application
- By default, it does host based encryption using CPU features(AES-NI) if available(this is sometimes just called software encryption)
- BitLocker will work with SSD hardware encryption and offload from CPU if drive supports MSFT feature called eDrive (aka eHDD)
Manage-bde command
- If BitLocker not using drive encryption, then “manage-bde –status” will show encryption capability of the encryption mechanism on the host side
- If BitLocker using drive encryption, then the status will show “Hardware Encryption”
* END EDIT *
Packaging:
Simple and compact packaging for this small M.2 2280 part.
Considering the low price of
Considering the low price of this ssd nvme vs normal sata ssd , could I put it in my motherboard z77-a ( via an pci express adapter ) ?
It might not seem like a serious question , but my i7 2600k is still going strong; especially considering the cost of a new system that supports PCI3 nvme natively .
Sorry for my bad English ;
Thank for the excellent job Allyn
There is an old article done
There is an old article done by PCPer regarding usage of PCI-E SSDs on old systems. Do check t out, if I remember correctly that Z77 might support PCI-E SSD (NVME) as storage drive but won’t boot off it.
Edit: here is the link to that article, I hope it helps you.
https://pcper.com/news/General-Tech/Intel-SSD-750-Series-PCIe-Compatibility-Tested
You may be able to get boot
You may be able to get boot functionality.
Here is a site that tells you how to add NVME boot modules to your bios:
http://www.win-raid.com/t871f16-Guide-How-to-get-full-NVMe-support-for-Intel-Chipset-systems-from-Series-up.html
Pretty slick.
Very very interesting thank
Very very interesting thank you for the replay
I just got my 600p 256GB
I just got my 600p 256GB yesterday, put into empty laptop and clean installed Win 10 Pro onto it. After updating fully(no other installs) i ran bitlocker but was only able to get software encryption. uefi only mode, csm/legacy off, secure boot on.
Also, the Intel SSD Toolbox cant read the SMART info for the drive.
Is the drive too new? Their FAQ said not to install Intel NVMe drivers but to use what came with Windows.
“manage-bde -status c:” shows “Encryption Method: XTS-AES 128” isntead of Hardware…
I got this drive because their plastered all over that active power usage is 100mW, but it was burning hot to touch while encrypting. The 4W listed on back sticker seems more the truth. Not sure point versus Samsung 950 Pro except Intel firmware and cheaper.
We likely need a new version
We likely need a new version of SSD Toolbox to read SMART.
For hardware encryption, you likely need the Pro 6000p.
It lists “AES 256-bit
It lists “AES 256-bit self-encryption” under its product brief. Half reason i got it. Besides the low power usage.
So sad. thanks for reply.
self-encrypting is enabled
self-encrypting is enabled with a HDD BIOS Password
For Bitlocker e-drive you need the pro version
What’s new in Windows 10:
What’s new in Windows 10: Warning: Self-Encrypting Hard Drives and Encrypted Hard Drives for Windows are not the same type of device. Encrypted Hard Drives for Windows require compliance for specific TCG protocols as well as IEEE 1667 compliance; Self-Encrypting Hard Drives do not have these requirements. It is important to confirm the device type is an Encrypted Hard Drive for Windows when planning for deployment.
Yeah, ‘self-encryption’
Yeah, 'self-encryption' implies that all data is stored to flash in an encrypted manner, which means that data is secure if the drive is locked with something like a BIOS password. This is different than what is needed to enable hardware encryption via OS (i.e. Bitlocker). Intel passed me a note further clarifying this, and I've appended it to the specs section of the first page in this review.
For others reading this
For others reading this tread. You need to change windows bitlocker default encryption method from 128 to 256 to support hardware encryption. It works on 6000p for sure but they almost same.
http://www.howtogeek.com/193649/how-to-make-bitlocker-use-256-bit-aes-encryption-instead-of-128-bit-aes/
Did you happen to measure
Did you happen to measure temperatures?
I didn’t break out the FLIR,
I didn't break out the FLIR, but the heat spreader on the controller was very warm to the touch during the continuous load benchmarks. It was barely warm during the mixed (more realistic) workload test. I couldn't get it to throttle with no direct airflow, but this is an open testbed. Things may be different in a cramped mobile housing, but again, it's unlikely to see that much heat production unless you're intentionally trying to heat it up (continuous workloads seen in benchmarks).
Some of those tests looked
Some of those tests looked really bad, do you think a firmware could fix some of it?
I’m convinced that firmware
I'm convinced that firmware can fix it, but Intel has to decide if it is something that actually needs fixing (since the issue doesn't show up in typical use, and this is a budget SSD after all). If they do, then it will still be a few months for the new firmware to get through their validation process, and then there's the whole issue of there not yet being NVMe firmware update utilities that are 'easy'…
This looks like a great drive
This looks like a great drive for regular users, top on my NVMe recommended list from now on. Thanks for the good perspective Allyn
Can I ask for some power figures? If anyone has the equipment and expertise it’s you
I did a few quick checks, and
I did a few quick checks, and it was roughly similar to a 950 Pro when active. Sustained heavy writes can likely cause it to throttle, but then we're into workloads that don't typically happen. Intel claims 100mW average in typical use (MobileMark).
On the sequential performance
On the sequential performance page it says “After ~6TB of file creation, the 600p’s SLC cache was saturated…” should this not be 6GB?
And in reality, a regular consumer looking at a 512GB drive which is now almost the same price as a mid range SATA3 SSD of the same capacity would surely have more than enough with a 16GB SLC cache? Apart from copying 16GB+ worth of files across >120MB/s drives, which in itself is rare enough, what is going to saturate that? This seems like the sweet spot right now.
Thanks – fixed. Totally agree
Thanks – fixed. Totally agree on the 512GB's 16GB cache. Also note that since that model has 2x the number of dies, its 'low' sustained speed should double to 300 MB/s, meaning you'd have to write that 16GB at an even faster rate to saturate that cache.
I can think of a consumer
I can think of a consumer workload that would saturate that. Steam download on the really fast connections you can get in some areas. Gigabit connections are appearing in various areas of the us and even multi gigabit like comcast’s 2gig service is out there. Not to mention whats going on in areas of the eu and asia with fiber.
2Gb fibre still wouldn’t
2Gb fibre still wouldn’t saturate the 512GB model’s continuous write rate, though – and with 1Gb fibre it wouldn’t even saturate the 256GB model. That’s also assuming a home network fast enough to deliver those rates to a single client, which doesn’t exist. So you’re talking about the edgiest of edge cases here.
I just installed the 512 GB
I just installed the 512 GB version and I have to say I’m a little disappointed. About my system, XPS 8900, i7-6700 with 32 GB of memory. I performed a clean install of W10, which installed in under 10 minutes (nice), but after getting everything setup I ran some benchmarks. The highest read speeds in getting is barely over 700 and about 650 seq. Advertising is 1800 read and I’m not even getting half that. Maybe time to return and go with sammy.
Not sure what benches you are
Not sure what benches you are running, but in ATTO this one hits 1.6GB/s, and it hit 700 MB/s in our mixed burst test, which measures that read speed during simultaneous writes.
I can confirm this, I just
I can confirm this, I just got installed mine yesterday and it’s working as advertised.
I’m coming from an older x58 SATAII platform. Faster storage was really the one of the only reasons I could justify the upgrade, this did no disappoint for the money!
Sounds like you’re hitting a
Sounds like you’re hitting a chipset limitation – probably not getting full bandwidth to the slot you’ve installed the drive into. If so, the Samsung won’t fix that.
Maybe, your PCIE bandwidth is
Maybe, your PCIE bandwidth is limiting the drive. I had a problem with my 950 Pro where my sound card was sharing PCIE lanes with the M.2 slot on the motherboard. I just had to move my sound card to a different slot that didn’t share lanes and it fixed the problem.
If you plugged the drive into
If you plugged the drive into the built in M.2 slot, you’re hitting some kind of hardware limitation of the slot. I bought a 950 pro and was disappointed with the speeds I was seeing until I saw this forum post and bought the recommended PCIE to NVME adapter. It was really lame installing a 950 and only getting 800MB/s speeds. http://en.community.dell.com/support-forums/desktop/f/3514/t/19649617?pi21953=10
I installed mine into the m.2
I installed mine into the m.2 slot and got just above 700, switched to a x4 pcie adapter and got the same. Updated the BIOS, which I thought I already did, and now I get more like 1400.
Does Intel or IMF enter joint
Does Intel or IMF enter joint venture with SMI? How is Marvell left out in the controller development?
Intel always tends to jointly
Intel always tends to jointly work with whichever third party controller maker they are currently using in their products. They have worked with Marvell in the past.
If one was to put two (or
If one was to put two (or more) of these drives in RAID 0 would that mean that it would be less likely you would run into the issue of saturating the cache?
Also would two of these drives out preform 1 Samsung 950? It seems like two lower capacity Intel drives are roughly the same price as 1 higher capacity Samsung.
So long as your motherboard
So long as your motherboard supports NVMe RAID, yes, generally the specs will all double (up to the point of saturating the DMI bandwidth). More info about all of that in my triple M.2 RAID review.
Hopefully Allyn can start
Hopefully Allyn can start measuring power draw on SSD. Been looking for hotswap bays for ssd and wondering if a single 5v is enough for 4 SSD on write cycles.
Hot swap bays for M.2 NVMe?
Hot swap bays for M.2 NVMe? Single 5V what? If you mean a single 5V line of a SATA power connector, that is more than sufficient for four SATA SSDs.
You dont seem to like the 850
You dont seem to like the 850 pro (my ssd) 🙁 instead only showing evo sata models.
The 850 EVO and PRO are
The 850 EVO and PRO are *very* similar in performance so long as the EVO is in a 500GB or higher capacity, which is why the 850 EVO 500GB is a staple in our results.
from looking at this review I
from looking at this review I see little that would intice me to buy one , looks like it gets wooped bad by most all the other drives or am I reading the data wrong ?
It’s about the use case, and
It's about the use case, and that this drive is nearly half the cost/GB of the competing units. It doesn't like sustained writes, but it's fine for anyone that won't be beating on the drive constantly.
The Intel 600p M.2 NVMe SSD
The Intel 600p M.2 NVMe SSD is designed to compete SATA SSD, and yes it does use TLC to increase capacity, but slower performance to keep the cost down.
It would’ve been nice to have
It would’ve been nice to have 2-bit MLC 3D NAND Flash Memory and it should come with heatsink to prevent thermal throttling; in other words, better performance w/ heatsink attached to SSD itself.
M.2 SSDs are not going to
M.2 SSDs are not going to come with heatsinks as it makes them too large for the physical spec (they wouldn't fit in a laptop, etc).
So I was tinkering with my
So I was tinkering with my Intel 600p nvme ssd tonight, running crystal disk mark for benchies. I noticed that when disabling write cache buffer flushing in device manager, on the ssd, it gave me me almost 300mb/s on my 4KQ31T1 scores in CDM. Now I have always read the dangers of disabling that setting for a drive. But is this something that nvme drives should have disabled? That is a pretty big boost to get.
Write cache buffer flushing…enabled or disabled for nvme ssds? And is there Intel NVME drivers I should be using with the 600p? IRST are for sata based drives correct?
I have read a few forum posts that disabling it is a good thing for nvme drives. Recent posts, nothing dated.
What do you think?
It’s more risky, as Windows
It's more risky, as Windows will cache writes in RAM, and if you have a lot of RAM, that's potentially a lot of unwritten data to the SSD in case of an OS crash or power loss. If you're ok with the risks, yeah, it will help a lot with random write speeds. One note though – you can't disable buffer flushing on some vendor-specific NVMe drivers (like Intel's NVMe driver, for example). Also note that the Intel NVMe driver is only for their enterprise / SSD 750 and not for the 600p, which is meant to use the Windows Inbox driver.
So when disabling WCBF, it
So when disabling WCBF, it uses system RAM as the buffer, instead of the buffer on the ssd? Even if I have a power loss or power off my system hard by pulling the plug, there is no difference right? Unless my drive itself has a separate power source? I don’t understand why disabling this is extra bad in case of a power loss. Blue screens will write what is in the cache buffer if this setting is enabled vs not being enabled (I assume) but as for power loss, it doesn’t matter correct? WCBF won’t matter in that case?
Buffer flushing disabled
Buffer flushing disabled means Windows will cache writes *in addition to* the SSDs own SLC cache. The SLC cache is persistent and will survive a power loss, but the non-flushed Windows buffer (RAM) will be lost during a crash / power loss.
Wouldnt disabling ssd or hard
Wouldnt disabling ssd or hard drive write cache take up more ressource .since ms write it in memory and hard drive or ssd write it in their own internal and the speed up would actually be false since its the indexing of the two that would speed up thing but write would slow down thing ?i am all for speed up tweak .but in a day and age of patreon .youtube and most importantly twich gamer streaming wouldnt write speed affect everything negatively . write tend to be slow no mather what .if your drive write to itself and your os write to ram . you gain performance in some place but wouldnt overall thing end up in a huge loss
Intel normally very
Intel normally very disappointed
I prefer the Samsung 512GB M.2
Plextor M8Pe 512GB
up-to 2300
Plextor M8Pe 512GB
up-to 2300 MB/s for reads and 1300 MB/sec (512GB) for writes.
http://www.guru3d.com/articles-pages/plextor-m8pe-512gb-m2-nvme-ssd-review,1.html
225 EURO
225 EURO
This is a slam dunk by Intel.
This is a slam dunk by Intel. I agree with updating the rating if they actually fix that write issue.
That said, this is a no brainer for 99% of users out there. Including the nerds reading this who think it’s not good enough for them. Only heavy duty 3D rendering or video editing would write 32GB at over 120MB/sec (for the 1TB). There’s no way the PC gamer nerd will need more than that. It’s far more than enough even for moderate video editing.
You’d have to get to some seriously professional use cases to find the limits of the 600P. Namely, if you’re not paid to do video editing or 3D content all day everyday- you don’t need anything better. Even if you think you do. And if you do need more, you’ll be stepping up to some serious stuff like the Intel 750 line, big old heatsink and all.
Just saying, real talk. Hopefully Intel does smooth out that usecase but I can see why they didn’t even bother testing for it. If they do fix it, more reviews will show it and Samsung’s 960 Pro lineup might be threatened by the low cost and frankly, overpowered performance of the 600P.
With a 1tb drive you should
With a 1tb drive you should have write of at least 300Mb/s or even 600 Mb/s to saturate the cash
Allyn said that the 512 has 2x the dies so it can handle about 300Mb/s in the comments. Assuming linear scaling for dies and speeds you should you will write at 600Mb/s to saturate the 32GB on the 1TB model (which currently sells at 33 (euro) cents per gig in my region)
Hi ALLYN,
Is this basically
Hi ALLYN,
Is this basically the same flash as Crucial MX300 ??
If so must be decent amount of OP going on………
Same flash, but not operating
Same flash, but not operating the same way. 600p has a fixed SLC cache portion of each die, while MX300 dynamically adjusts how much SLC is present on-the-fly.
I picked up the 256GB version
I picked up the 256GB version on newegg for $95 + free shipping.
My only question is whats the performance on a x2 slot?
Thanks for bringing this to
Thanks for bringing this to our attention, great pricing on some high end hardware!!!
I’m curious to CNote’s question…. Plus any suggestions on how to get this to work on am3+ motherboard (more specifically msi 970a-g46) w the following setup:
2x PCIe 2.0 x16 slots
PCI_E2 supports up to PCIe x16 speed (when PCI_E4 is empty) or PCIe x8 speed
(when PCI_E4 is installed)
PCI_E4 supports up to PCIe x8 speed
2x PCIe 2.0 x1 slots
2x PCI slots, support 3.3V/ 5V PCI bus Interface
Allyn I believe you stated in last podcast something about a more cost friendly pci express adapter for m.2 that was for 2 channels. Would that be a good option for us with the gen2 pcie? What was the name of that thing?
That was the Phison E8. AMD
That was the Phison E8. AMD support for NVMe boot is spotty, so definitely make sure anything you try has been tried elsewhere first (research forum posts, etc).
PCIe 3.0 x2 / PCIe 2.0 x4
PCIe 3.0 x2 / PCIe 2.0 x4 will both cap throughputs at ~1.5 GB/s. For the 600p, it's not going to be impacted very greatly since its fastest speeds are only using roughly half of its native interface in the first place.
Sweet thanks
Sweet thanks
Thanks for the info Allyn~!
Thanks for the info Allyn~!