Intel 600p Series SSD Review

[rhysiam]

Thanks for the review Chris, I've been wanting to see how these drives perform given the competitive pricing.

Are you sure the labels on your "Intel 600P vs SATA" series of graphs are correct? Because the 256GB 600P seems to massively outperform its 512GB bigger brother in most the tests (not all - notably), often with double the performance or better. Surely that's not accurate? If it is, can you explain why the 256GB would so utterly dominate the 512GB so often? Because that doesn't make any sense to me. Thanks!

 

[John Wittenberg]

Would you gents start testing boot times and game load times? These are real world metric's that end users want.

Yes, I know they'll essentially be the same as a good SATA III SSD, but most folks think that their computer will magically get super fast in those aspects - when in fact the computer does not since boot and game load times can't take advantage of NVMEs massive queue depths (so far at least).

 

[Guest]

SSD are still unreasonably expensive.

 

[logainofhades]

If I were buying an M.2 drive, I would probably settle on a Sandisk X400, or Crucial MX300, depending on price and capacity. Those high dollar drives are nice, but in real world usage, I doubt one would really notice much of a difference.

 

[bit_user]

How Intel's consumer SSDs expire once you surpass the endurance threshold is troubling. In an almost over-zealous move to protect user data, Intel instituted a feature on many of its existing SSDs that automatically switches it to a read-only mode once you surpass the endurance threshold (measured via the MWI SMART attribute). Surprisingly, the read-only state only lasts for a single boot cycle. After reboot, the SSD "locks" itself (which means you cannot access the data)

Wow. I didn't know that. When did they start doing this?

I have a 520 and 535. I thought I would be okay with a bit of overprovisioning, but it seems maybe not...

 

[apk24]

Wow. I didn't know that. When did they start doing this?

I have a 525 and 535. I thought I would be okay with a bit of overprovisioning, but it seems maybe not...

SSD Toolbox still says my 120GB Intel 520's estimated life remaining is 98%. This, despite the fact that it's rated for 20GBW/day for 5yrs (~36TBW) and I already have 22TBW according to SMART data, so...

This might be a thing for newer drives, but I'm betting my 520 will last until I get tired of it.

 

[PureBlackFire]

what motherboard was this tested in?

 

[bit_user]

Intel should not even use 'NVMe' anywhere on the product or packaging...would better keep user's initial expectations in check...everybody associates 'NVMe' with blistering speeds atm.

Sorry mate, I can't agree with this. NVMe is a technical standard. As long as they correctly implement the standard, then this is truth in marketing.

If people blindly treat it as shorthand for "fast", and buy without checking any benchmarks or reviews, then they get what they deserve.

SSD are still unreasonably expensive.

Not too expensive for the performance benefit you get. As long as you're using it for OS & Apps, they make sense. Just not for bulk data.

 

[Ricardotheanonymous]

The test was interesting.
Although, the story about endurance is ridiculous.
SSD's TBW comes from NAND's endurance.
So how can Chris think 128GB SSD and 512GB SSD has the same actual endurance?
Doesn't he know global wear leveling?
Those SSDs' endurance can't be the same.
This means that TBW is just what the company ensure and doesn't show the threshold.
Of course it can be used as a reference, but that's all.

If Intel refused the return because 73TB of data was written to the SSD, Chris is right.
But "72TB" or "TBW" is not written in Intel's waranty document.

 

[rhysiam]

The test was interesting.
Although, the story about endurance is ridiculous.
SSD's TBW comes from NAND's endurance.
So how can Chris think 128GB SSD and 512GB SSD has the same actual endurance?
Doesn't he know global wear leveling?
Those SSDs' endurance can't be the same.
This means that TBW is just what the company ensure and doesn't show the threshold.
Of course it can be used as a reference, but that's all.

If Intel refused the return because 73TB of data was written to the SSD, Chris is right.
But "72TB" or "TBW" is not written in Intel's waranty document.

How confident are you that your warranty would still be valid if you'd blown past the endurance rating? No mention of endurance in the warranty document might suggest you'd be fine, but I'm far, far from convinced. I'd want firm confirmation from a lawyer who works in this field before I'd be truly confident that Intel would honour a warranty on a drive that had exceeded it's "rated" writes. Chris is clearly sceptical that Intel would honour such a warranty, he says so on the last page.

Putting the warranty issue aside, obviously there's a gulf between actual NAND endurance and manufacturer's endurance ratings. But the big question that hasn't been answered (as far as I'm aware), is whether Intel has implemented the same the same enforced read-only/locking "feature" as they have on previous drives. The way I read the article, is that if Chris was 100% confident that Intel does not and will not ever lock the drive, then the endurance rating is much less of an issue. At this point however, we simply don't know whether that's the case. For me that's an absolute deal-breaker.

 

[Ricardotheanonymous]

How confident are you that your warranty would still be valid if you'd blown past the endurance rating? No mention of endurance in the warranty document might suggest you'd be fine, but I'm far, far from convinced. I'd want firm confirmation from a lawyer who works in this field before I'd be truly confident that Intel would honour a warranty on a drive that had exceeded it's "rated" writes. Chris is clearly sceptical that Intel would honour such a warranty, he says so on the last page.

Putting the warranty issue aside, obviously there's a gulf between actual NAND endurance and manufacturer's endurance ratings. But the big question that hasn't been answered (as far as I'm aware), is whether Intel has implemented the same the same enforced read-only/locking "feature" as they have on previous drives. The way I read the article, is that if Chris was 100% confident that Intel does not and will not ever lock the drive, then the endurance rating is much less of an issue. At this point however, we simply don't know whether that's the case. For me that's an absolute deal-breaker.

Sorry but I can't do anything for you if you have rejected from Intel's warranty.
I say I'm not sure, if you want the perfect accuracy.

I just wanted so say, in theroy, SSD with larger capacity have much endurance than less one.
And, TBW is refference for endurance, not the threshold of lifetime or warranty.
Chris had rated the endurance from TBW, but that's not accurate.
I think he should have ask about endurance of each capacity, instead of TBW.

BTW, from my experience, you don't have to be so nervous about read-only feature.
Something other than NAND will be defected before NAND comes to the end.
I'm using over 10 SSDs now and have blown up 3 or 4 SSDs before.
SSD suddenly disappear from desktop, even from BIOS, no data rescued.
Just back up your data everyday.

 

[rhysiam]

Sorry but I can't do anything for you if you have rejected from Intel's warranty.
I say I'm not sure, if you want the perfect accuracy.

I just wanted so say, in theroy, SSD with larger capacity have much endurance than less one.
And, TBW is refference for endurance, not the threshold of lifetime or warranty.
Chris had rated the endurance from TBW, but that's not accurate.
I think he should have ask about endurance of each capacity, instead of TBW.

BTW, from my experience, you don't have to be so nervous about read-only feature.
Something other than NAND will be defected before NAND comes to the end.
I'm using over 10 SSDs now and have blown up 3 or 4 SSDs before.
SSD suddenly disappear from desktop, even from BIOS, no data rescued.
Just back up your data everyday.

With respect, I think you're confusing the argument from the article. I don't believe the article is suggesting for a second that the NAND itself will be worn/unable to reliably store a charge at 72TBW. If you look on the first page it's specifically noted that this is only the second SSD ever tested which has the same endurance rating across the entire capacity range... I think we're all well aware of how ridiculous those endurance ratings are. The fundamental problem is that the drive *might* (we don't know for sure), lock itself after 72TBW, so who cares if the NAND is still good for loads of writes (which it almost certainly would be)... the drive is effectively a brick and you've got IMHO very little chance of a successful warranty claim. So not only does the 512GB model have the lowest endurance rating we've ever seen, it also *might* have a "feature" which effectively bricks the drive once that rating is reached... that's the problem, and one that absolutely should be raised in any half decent review.

I've expended many words on the forums over the last couple of years quashing fears about wearing out SSDs. But we've seen write cycles drop from 3-5000 in the early days, to 1000 more recently (still plenty IMO), to now just ~140 on the 512GB 600P. To put that in perspective, that's just 34 hours worth of writes at the SLC cache write speed. Again, if it was just warranty I could live with that, but a built in bricking feature... I'm out. I do agree that it's still unlikely that you'd reach that level of writes in the useful life of the drive, but it's definitely possible and absolutely hurts the value proposition of this drive.

 

[CRamseyer]

We have a busy schedule tonight with the 960 release (full details in an hour or so). I did want to reply though to the TBW not being in Intel's warranty.

http://ark.intel.com/compare/94917,94921,94924,94926

The ARK has the TBW limits listed. Yes, I do understand that flash more often than not outlives any warranty limitations but the "feature" is a brick wall. We have a statement from Intel and will have a news post as quickly as possible outlining their position.

For those asking, I still use an Asrock Z97 Extreme6 motherboard with direct-to-CPU M.2 PCIe lanes. I run a Z170 system for RAID Reports with two (or more) PCIe M.2 SSDs.

 

[Ricardotheanonymous]

With respect, I think you're confusing the argument from the article. I don't believe the article is suggesting for a second that the NAND itself will be worn/unable to reliably store a charge at 72TBW. If you look on the first page it's specifically noted that this is only the second SSD ever tested which has the same endurance rating across the entire capacity range... I think we're all well aware of how ridiculous those endurance ratings are. The fundamental problem is that the drive *might* (we don't know for sure), lock itself after 72TBW, so who cares if the NAND is still good for loads of writes (which it almost certainly would be)... the drive is effectively a brick and you've got IMHO very little chance of a successful warranty claim. So not only does the 512GB model have the lowest endurance rating we've ever seen, it also *might* have a "feature" which effectively bricks the drive once that rating is reached... that's the problem, and one that absolutely should be raised in any half decent review.

I've expended many words on the forums over the last couple of years quashing fears about wearing out SSDs. But we've seen write cycles drop from 3-5000 in the early days, to 1000 more recently (still plenty IMO), to now just ~140 on the 512GB 600P. To put that in perspective, that's just 34 hours worth of writes at the SLC cache write speed. Again, if it was just warranty I could live with that, but a built in bricking feature... I'm out. I do agree that it's still unlikely that you'd reach that level of writes in the useful life of the drive, but it's definitely possible and absolutely hurts the value proposition of this drive.

Thanks, now I've got your point, and sorry about some misunderstandings. The problem is that 600p may stop running over 72TB, however more data could be written. I agree about that, if it may happen.

But I don't believe it. The story means that Intel is spoiling their products. Theoretically, 1TB model has four times endurance than 256GB model. Intel won't let user to use it, and still give a 5 year warranty, and refuse RMA because of TBW? That doesn't make sense. Read-only feature is for fail safe. Why does Intel have to actuate the fail safe far before the threshold? If Intel doesn't want to warrant their product, they can simply shorten the warranty period, like 1 year. "TBW is not a threshold" is my idea, and that's much convincing for me.

In short, I believe Intel at this time.If you were right, I turn around.

If I add some about warranty, this is my logic.
1.Intel is basically responsible for their products.
2.But Intel can't do everything, so they made some limitation.
3.Thus, Intel is responsible for the product which is in limitation.
4.Intel hasn't define the limitation about TBW.
5.Thus, Intel is responsible for SSD written more than 72TB, even in case read-only feature did lock the SSD.

 

[rhysiam]

But I don't believe it. The story means that Intel is spoiling their products. Theoretically, 1TB model has four times endurance than 256GB model. Intel won't let user to use it, and still give a 5 year warranty, and refuse RMA because of TBW? That doesn't make sense.

Have you read the follow up posted here earlier? It's here: http://www.tomshardware.com/news/intel-ssd-endurance-600p-mwi,32731.html

Toms got confirmation from Intel which basically confirms the worst-case scenario. Drive will lock itself to read-only mode at 72TBW, irrespective of capacity. That's why we're all complaining about it.

As you say, it "doesn't make any sense" for the consumer. It absolutely makes sense if you're Intel and you to force your business and server consumers to buy the much more expensive (and higher margin) server grade gear. If you're a real sceptic, you could argue that it might also makes sense for Intel to build in a fixed point of failure into their products which would force customers to purchase new gear down the line. Whatever Intel's motives however, they're not for our benefit! Which I think is at the core of people's frustrations with this drive and this move from Intel.

 

[CRamseyer]

When I get back from South Korea we will look into the issue a little deeper. It may take two months to find what we are looking for but we will follow up with a definitive answer.

 

[Ricardotheanonymous]

Have you read the follow up posted here earlier? It's here: http://www.tomshardware.com/news/intel-ssd-endurance-600p-mwi,32731.html

Toms got confirmation from Intel which basically confirms the worst-case scenario. Drive will lock itself to read-only mode at 72TBW, irrespective of capacity. That's why we're all complaining about it.

As you say, it "doesn't make any sense" for the consumer. It absolutely makes sense if you're Intel and you to force your business and server consumers to buy the much more expensive (and higher margin) server grade gear. If you're a real sceptic, you could argue that it might also makes sense for Intel to build in a fixed point of failure into their products which would force customers to purchase new gear down the line. Whatever Intel's motives however, they're not for our benefit! Which I think is at the core of people's frustrations with this drive and this move from Intel.

Thanks again. But I couldn't find the text about Intel told that 72TB is the threshold. In the post you showed (I've missed it, thank you), Intel was talking about MWI, and didn't mention that it equals to TBW. I understand that there is read-only feature, but the trigger is yet unclear (MWI's detail is not clear). If MWI number decreases by 1 when 720GB is written inside the SSD, that's true. I hope Toms will clear this.

 

[bit_user]

I couldn't find the text about Intel told that 72TB is the threshold. In the post you showed (I've missed it, thank you), Intel was talking about MWI, and didn't mention that it equals to TBW. I understand that there is read-only feature, but the trigger is yet unclear (MWI's detail is not clear). If MWI number decreases by 1 when 720GB is written inside the SSD, that's true.

MWI and TBW are independent. So, MWI is effectively telling you how many actual block writes have happened, while TBW tells you how much data the host has sent to the SSD.

I think we can all understand and accept that MWI is the real issue. When a drive runs out of spare capacity, it makes sense to go read-only. This is because SSD sectors fail at a fairly predictable rate, and if you let the user keep writing data, it's like they're digging their own grave without even knowing it.

What is outrageous is if Intel puts the drive in read-only mode if TWB exceeds 72, even while MWI is still above zero. That's like an auto manufacturer disabling engine ignition once you've driven the number of miles for which it's warrantied. The car might have plenty of life left in it, if you drove it gently, but they're keeping you from using it because some aggressive drivers would've worn it out, by that point.

 

[Ricardotheanonymous]

MWI and TBW are independent.

What is outrageous is if Intel puts the drive in read-only mode if TWB exceeds 72, even while MWI is still above zero.

Absolutely. That's my point. MWI and TBW are different. And Intel only mentioned about MWI, not TBW. Then how come we think TBW (72TB) is the the trigger for the feature? I think MWI is. That (TBW is not the trigger) will make sense. (Although, we don't know much about MWI, that will be the next problem...)

 

[Bruce427]

I am wondering why ANYONE would buy a 256GB Intel 600P when you can purchase (in October) a 250GB Samsung 960 EVO for the same price?

Moreover, not only does the Samsung off about 28% greater endurance, but also handily outperforms the Intel 600P.

 

[wolflarson]

Erm I think you guys need to update this review. According to their website, the 512GB version has a 288 TBW. I don't know if I'm reading this wrong because last I checked it said 72 or they made some firmware changes that needs to be updated. I already bought a 512GB 600P drive so now I'm really confused. Do I need to update the firmware to get the new TWB or...?

http://ark.intel.com/products/94924/Intel-SSD-600p-Series-512GB-M_2-80mm-PCIe-3_0-x4-3D1-TLC

 

[CRamseyer]

That would figure. We'll reach out and see what they say. I have a 256GB drive writing data now to confirm what happens at 72 TB TBW. It just surpassed 25 TB.

 

[wolflarson]

That would figure. We'll reach out and see what they say. I have a 256GB drive writing data now to confirm what happens at 72 TB TBW. It just surpassed 25 TB.

Glad you're still keeping tabs on this. According to the updated specs, the 256GB version has a TBW of 144. Fingers crossed that this will hold true in your test. Just a quick question, Windows 10 should be able to see this drive in a clean install right? Or do I need to load up the NVMe drivers on an USB stick when I do a clean install? I'm asking because I had to return a 512GB drive for RMA because it flat out refused to show up at all. They checked it and sent a replacement drive within 3 days so I'm hoping that was just an isolated incident. It just crossed my mind today when I received the drive that perhaps it didn't show up because of a driver problem?

 

[Ricardotheanonymous]

That would figure. We'll reach out and see what they say. I have a 256GB drive writing data now to confirm what happens at 72 TB TBW. It just surpassed 25 TB.

Intel's decision to change the number is good. I'm looking forward to see the comment from them.
I think 72TB locking problem has been already negated on this article (http://www.tomshardware.com/news/intel-ssd-endurance-600p-mwi,32731.html). Although, it's still very interesting to make sure that nothing will happen when total written data passes 72TB.

 

[bit_user]

That would figure. We'll reach out and see what they say. I have a 256GB drive writing data now to confirm what happens at 72 TB TBW. It just surpassed 25 TB.

Excellent!

Given the new specs, will you try to go for 145 TB?