Man, are they really making it difficult to buy any other brand of SSD. This drive is cheap and outperforms all the other low-end drives and they lead all the way up to the high end with the 950 PRO.
That's the power of their vertical integration. When everything is designed in-house, the margins are higher so they can sell lower.
Of course, you love Samsung SSD's until you have had to deal with their support. Which is needless to say, awful. I mean they still don't even acknowledge the 840 EVO's were outright defective (the firmware update did little to fix the problem, all it did was tell the drive to move static data around.) That level of response is just damning to an OEM, which is why some OEM's (such as HP) have dropped Samsung in favor of Micron. Apple would love to get away from Samsung storage products but...as you said, hard to ignore.
Another one who spreads FUD, 840 EVO is perfectly fine after the 2nd fix with no perormance degradation, I have one and don't see any need to change in the future. Yes, the fix is consuming 1 p/e cycle every 2 weeks but considering the reliability of todays SSDs (Samsung especially) the drive will be obselete long before its memory cells are going to die. Please stop continuing the same false myths about this drive and inform yourself better next time.
I ordered quite a few 840 evos where I work, and I've got to agree with vladx... *and* with Samus. Yes, performance is not generally an issue, but Samus is also right that the perf fix did not entirely work, and furthermore, Samsungs response on that issue has not been brilliant - I mean, there are considerably cheaper drives out there, so performance was definitely a selling factor, and if they can't deliver - yeah, that's a defect. We've also had one drive brick itself after a few weeks, and never had that issue with the crucial's (but the sample size is too low to be significant).
All else being equal, I would not pick Samsung drives again - but all else is not equal - Samsung drives are simply faster, and often cheaper too. And though their reputation for reliability has had a scratch, it's not a big one.
Basically what I'm saying is the 840 EVO isn't unreliable, and it is *partially fixed* but it should have been recalled.
Samsung is just too proud to do such a thing. The drive is not as fast as it was when the original (pre-"fix") firmware was released and numerous benchmarks back this up. The fix is a bandaid, not a fix. If Samsung has properly validated the drive it wouldn't have even come to market the way it is, the TLC technology just wasn't ready for prime time. Even Anand was shocked when the drive was confirmed to have TLC because the whole industry was years behind Samsung in releasing TLC-based NAND and now we know why.
That said, I have never seen an 840 EVO outright fail, but I have seen them, even with the latest firmware, have read performance on par with a hard drive.
"That said, I have never seen an 840 EVO outright fail, but I have seen them, even with the latest firmware, have read performance on par with a hard drive."
Sorry but I can't trust the validity of this claim since my main SSD is a 2 years old 840 EVO that has no performance issues after updating to the latest firmware and yes it's as fast as when it was new and almost full. So I hate to reiterate but please stop spreading FUD about this drive.
I have the 840EVO, too, 250GB for which I paid 193 USD about 2 years ago. It was dog slow after a while and I had to spend a lot of time to make it work properly again. I think Samsung should have said "here get a discount on the next gen" instead.
"However, endurance of the planar TLC-based drives is unsurprisingly somewhat lower compared to the 850 EVO (which uses 3D V-NAND)."
Please stop mentioning SSD endurance in any articles unless a normal consumer with average day-to-day usage has a reasonable chance of actually exhausting the endurance of the SSD during its usable lifespan (equal to the warranty period + a year or two).
SSD endurance hasn't been a problem yet for anyone but possibly the earliest adopters of SATA 2 SSDs. I have literally never heard of anyone saying anything resembling "Man, I gotta go buy another SSD because mine bricked itself because I wore out the NAND too much from excess usage." The reality of what ends up happening is that most SSDs (even TLC NAND based) will be able to read/write around a petabyte or more during its usable lifespan (except for some Intel branded ones which are programmed to brick itself after a preset amount of read/write operations), often giving many noticeable signs of death before it actually bricks itself.
And then the worrisome sheep who read tech articles and believe themselves to be more knowledgeable than others will debate on SSD endurance and bring it up in conversations as if it mattered to non-enterprise users.
The fact is, SSD manufacturers are aware of average SSD usage patterns and are aware of the typical endurance of their SSDs. They conveniently design product warranties to be an amount of time so that the sheer majority of users will have an expired warranty long before the SSD exhausts its NAND endurance.
tl;dr: Talk about warranty periods, not SSD endurance, because the former actually matters when the SSD stops functioning, whereas the latter will never be an issue (unless the user literally bought a consumer hard drive for enterprise-grade disk I/O based workloads) due to the warranty period not extending far enough.
More information only benefits us. I agree that the issue is sometimes overstated but consider this: If all the tech journalists / reviewers worldwide said "it's a non-issue" and never reported on it ever, what do you think would have happened? For consumer-level drives, manufacturers would have ceased to compete on (or care about) endurance, stopped publishing those numbers, and in some cases produced even-cheaper drives with crap endurance. The more drives die out of warranty (especially OEM boxes which do NOT carry the same warranty) the more drives they sell. The less space they have to reserve for endurance concerns, the more money they can squeeze. I mean why NOT keep them honest and mindful of endurance, even if, for most users, it's a non-issue?
You can just skip the parts where they talk about endurance. Or rant about it every article. That's good too.
Thank you for bringing this up, and I completely agree with your point.
Write endurance should be available somewhere on a spec table, but for consumer purposes, the amount of space that AT dedicates to it is completely ludicrous.
For the great majority of users, data retention is a much more valid concern than write endurance.
Endurance and data retention go hand in hand. JEDEC's consumer SSD spec states that drives must retain data for one year once the the endurance rating is exceeded.
It would be interesting to see this tested in reviews. I remember reading a PowerPoint presentation or a whitepaper that mentioned simulating years passing by baking the drive at certain temperatures for a short period of time.
Actually, with modern OSes running indexing in the background, antivirus doing their periodic stuff, automatic updates, snapshots and what not, the end user has little control on how much writing happens on their drives.
I have a three year (and a few months) old Samsung 840 Pro 128GB. It's at about 34TB of total writes right now. If it was a 750 Evo, it would be very close to the end of its rated life (35TB), and just past the end of its warranty as well. Given that the drive would be just about to exceed its rated lifetime right as the warranty ended, it would be way beyond that in two more years, and very possibly would have failed completely by then. At the very least, it would always be a concern until it did give up the ghost one day.
What are the "enterprise-grade disk I/O based workloads" I subjected the drive to on my personal PC at home?
Web browsing, mainly. I often leave a lot of browser tabs open (hundreds, sometimes), consuming a great deal of memory, and it would hit the page file pretty hard in those situations. The speed of the SSD allowed it to perform well enough to keep it from nearly grinding to a halt, but it has caused the total written to increase as fast as it has.
I'm glad Anandtech doesn't agree that certain bits of information that people are interested in don't matter enough to justify being mentioned in any of their articles. I'm glad to know that some of the lower-priced SSDs can't handle the kind of usage my at-home, completely non-enterprise workload throws at them. I spent a bit more than the bare minimum on my 840 Pro, but at a total-writes number that would have the 750 near the end of its rated life, my 840 Pro still has 83% of its rated life left, so I'm good for the better part of a decade to come. Maybe I'm not a normal consumer (and I would presume that few people who read Anandtech really are), but I'm certainly not doing anything enterprise-grade either.
"When Samsung released its 750 EVO lineup of SSDs based on planar TLC NAND flash memory earlier this year, it seemed like a big surprise, as the company was first to ramp up production of 3D NAND memory and to use it for SSDs."
I would expect that there's not TOO much commonality between 3D-NAND and planar NAND fab equipment. Given that fact, if you have the equipment in place, it makes sense to keep using it for as long as you absolutely can. Basically no different from Intel in the old days using n-2 fabs to manufacture Southbridges and n-1 fabs to manufacture Northbridges, or TSMC using their 40nm equipment to build low-end ARMs destined for things like smart scales or printers.
Exactly. It's a totally different process for planar vs. 3D nand, so whatever they've got that is planar is going to keep churning out planar chips. Expect to see them keep selling planar until those fabs are shut down or moved onto other products.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
28 Comments
Back to Article
SkiBum1207 - Wednesday, May 25, 2016 - link
Hey guys! I have a feeling that the title should be GB (rather than MB) :)Flunk - Wednesday, May 25, 2016 - link
Man, are they really making it difficult to buy any other brand of SSD. This drive is cheap and outperforms all the other low-end drives and they lead all the way up to the high end with the 950 PRO.RMSe17 - Wednesday, May 25, 2016 - link
If only 950PRO came in 1TB size :( Or how about they take that 2TB 850Pro, and put NVMe instead of AHCI?extide - Wednesday, May 25, 2016 - link
Because it is SATA. NVMe only works on PCIe.RMSe17 - Thursday, May 26, 2016 - link
I know, so how about they take that 2TB 850Pro and release the NVMe PCIe version...Samus - Wednesday, May 25, 2016 - link
That's the power of their vertical integration. When everything is designed in-house, the margins are higher so they can sell lower.Of course, you love Samsung SSD's until you have had to deal with their support. Which is needless to say, awful. I mean they still don't even acknowledge the 840 EVO's were outright defective (the firmware update did little to fix the problem, all it did was tell the drive to move static data around.) That level of response is just damning to an OEM, which is why some OEM's (such as HP) have dropped Samsung in favor of Micron. Apple would love to get away from Samsung storage products but...as you said, hard to ignore.
vladx - Thursday, May 26, 2016 - link
Another one who spreads FUD, 840 EVO is perfectly fine after the 2nd fix with no perormance degradation, I have one and don't see any need to change in the future. Yes, the fix is consuming 1 p/e cycle every 2 weeks but considering the reliability of todays SSDs (Samsung especially) the drive will be obselete long before its memory cells are going to die. Please stop continuing the same false myths about this drive and inform yourself better next time.Cellar Door - Thursday, May 26, 2016 - link
I completely agree with vladx - just a ton of misinformation this person keeps spreading.emn13 - Thursday, May 26, 2016 - link
I ordered quite a few 840 evos where I work, and I've got to agree with vladx... *and* with Samus. Yes, performance is not generally an issue, but Samus is also right that the perf fix did not entirely work, and furthermore, Samsungs response on that issue has not been brilliant - I mean, there are considerably cheaper drives out there, so performance was definitely a selling factor, and if they can't deliver - yeah, that's a defect. We've also had one drive brick itself after a few weeks, and never had that issue with the crucial's (but the sample size is too low to be significant).All else being equal, I would not pick Samsung drives again - but all else is not equal - Samsung drives are simply faster, and often cheaper too. And though their reputation for reliability has had a scratch, it's not a big one.
Samus - Thursday, May 26, 2016 - link
Basically what I'm saying is the 840 EVO isn't unreliable, and it is *partially fixed* but it should have been recalled.Samsung is just too proud to do such a thing. The drive is not as fast as it was when the original (pre-"fix") firmware was released and numerous benchmarks back this up. The fix is a bandaid, not a fix. If Samsung has properly validated the drive it wouldn't have even come to market the way it is, the TLC technology just wasn't ready for prime time. Even Anand was shocked when the drive was confirmed to have TLC because the whole industry was years behind Samsung in releasing TLC-based NAND and now we know why.
That said, I have never seen an 840 EVO outright fail, but I have seen them, even with the latest firmware, have read performance on par with a hard drive.
vladx - Friday, May 27, 2016 - link
"That said, I have never seen an 840 EVO outright fail, but I have seen them, even with the latest firmware, have read performance on par with a hard drive."Sorry but I can't trust the validity of this claim since my main SSD is a 2 years old 840 EVO that has no performance issues after updating to the latest firmware and yes it's as fast as when it was new and almost full. So I hate to reiterate but please stop spreading FUD about this drive.
Gadgety - Friday, May 27, 2016 - link
I have the 840EVO, too, 250GB for which I paid 193 USD about 2 years ago. It was dog slow after a while and I had to spend a lot of time to make it work properly again. I think Samsung should have said "here get a discount on the next gen" instead.eek2121 - Thursday, May 26, 2016 - link
You aren't very bright are you? The 840 EVO works perfectly fine. It's my main SSD in my system, and has been for a long time.Meteor2 - Friday, May 27, 2016 - link
Well that's the most rubbish comment I've seen in a long time.JoeyJoJo123 - Wednesday, May 25, 2016 - link
"However, endurance of the planar TLC-based drives is unsurprisingly somewhat lower compared to the 850 EVO (which uses 3D V-NAND)."Please stop mentioning SSD endurance in any articles unless a normal consumer with average day-to-day usage has a reasonable chance of actually exhausting the endurance of the SSD during its usable lifespan (equal to the warranty period + a year or two).
SSD endurance hasn't been a problem yet for anyone but possibly the earliest adopters of SATA 2 SSDs. I have literally never heard of anyone saying anything resembling "Man, I gotta go buy another SSD because mine bricked itself because I wore out the NAND too much from excess usage." The reality of what ends up happening is that most SSDs (even TLC NAND based) will be able to read/write around a petabyte or more during its usable lifespan (except for some Intel branded ones which are programmed to brick itself after a preset amount of read/write operations), often giving many noticeable signs of death before it actually bricks itself.
And then the worrisome sheep who read tech articles and believe themselves to be more knowledgeable than others will debate on SSD endurance and bring it up in conversations as if it mattered to non-enterprise users.
The fact is, SSD manufacturers are aware of average SSD usage patterns and are aware of the typical endurance of their SSDs. They conveniently design product warranties to be an amount of time so that the sheer majority of users will have an expired warranty long before the SSD exhausts its NAND endurance.
tl;dr:
Talk about warranty periods, not SSD endurance, because the former actually matters when the SSD stops functioning, whereas the latter will never be an issue (unless the user literally bought a consumer hard drive for enterprise-grade disk I/O based workloads) due to the warranty period not extending far enough.
Alexvrb - Wednesday, May 25, 2016 - link
More information only benefits us. I agree that the issue is sometimes overstated but consider this: If all the tech journalists / reviewers worldwide said "it's a non-issue" and never reported on it ever, what do you think would have happened? For consumer-level drives, manufacturers would have ceased to compete on (or care about) endurance, stopped publishing those numbers, and in some cases produced even-cheaper drives with crap endurance. The more drives die out of warranty (especially OEM boxes which do NOT carry the same warranty) the more drives they sell. The less space they have to reserve for endurance concerns, the more money they can squeeze. I mean why NOT keep them honest and mindful of endurance, even if, for most users, it's a non-issue?You can just skip the parts where they talk about endurance. Or rant about it every article. That's good too.
aryonoco - Wednesday, May 25, 2016 - link
Thank you for bringing this up, and I completely agree with your point.Write endurance should be available somewhere on a spec table, but for consumer purposes, the amount of space that AT dedicates to it is completely ludicrous.
For the great majority of users, data retention is a much more valid concern than write endurance.
Kristian Vättö - Thursday, May 26, 2016 - link
Endurance and data retention go hand in hand. JEDEC's consumer SSD spec states that drives must retain data for one year once the the endurance rating is exceeded.Quad5Ny - Monday, May 30, 2016 - link
It would be interesting to see this tested in reviews. I remember reading a PowerPoint presentation or a whitepaper that mentioned simulating years passing by baking the drive at certain temperatures for a short period of time.bug77 - Friday, May 27, 2016 - link
Actually, with modern OSes running indexing in the background, antivirus doing their periodic stuff, automatic updates, snapshots and what not, the end user has little control on how much writing happens on their drives.Ascaris - Wednesday, June 8, 2016 - link
I have a three year (and a few months) old Samsung 840 Pro 128GB. It's at about 34TB of total writes right now. If it was a 750 Evo, it would be very close to the end of its rated life (35TB), and just past the end of its warranty as well. Given that the drive would be just about to exceed its rated lifetime right as the warranty ended, it would be way beyond that in two more years, and very possibly would have failed completely by then. At the very least, it would always be a concern until it did give up the ghost one day.What are the "enterprise-grade disk I/O based workloads" I subjected the drive to on my personal PC at home?
Web browsing, mainly. I often leave a lot of browser tabs open (hundreds, sometimes), consuming a great deal of memory, and it would hit the page file pretty hard in those situations. The speed of the SSD allowed it to perform well enough to keep it from nearly grinding to a halt, but it has caused the total written to increase as fast as it has.
I'm glad Anandtech doesn't agree that certain bits of information that people are interested in don't matter enough to justify being mentioned in any of their articles. I'm glad to know that some of the lower-priced SSDs can't handle the kind of usage my at-home, completely non-enterprise workload throws at them. I spent a bit more than the bare minimum on my 840 Pro, but at a total-writes number that would have the 750 near the end of its rated life, my 840 Pro still has 83% of its rated life left, so I'm good for the better part of a decade to come. Maybe I'm not a normal consumer (and I would presume that few people who read Anandtech really are), but I'm certainly not doing anything enterprise-grade either.
joex4444 - Wednesday, May 25, 2016 - link
I like the part where you distinguish 10K IOPS random reads at 4K QD1 from 10K IOPS random reads at 4K QD1.name99 - Wednesday, May 25, 2016 - link
"When Samsung released its 750 EVO lineup of SSDs based on planar TLC NAND flash memory earlier this year, it seemed like a big surprise, as the company was first to ramp up production of 3D NAND memory and to use it for SSDs."I would expect that there's not TOO much commonality between 3D-NAND and planar NAND fab equipment. Given that fact, if you have the equipment in place, it makes sense to keep using it for as long as you absolutely can.
Basically no different from Intel in the old days using n-2 fabs to manufacture Southbridges and n-1 fabs to manufacture Northbridges, or TSMC using their 40nm equipment to build low-end ARMs destined for things like smart scales or printers.
saratoga4 - Wednesday, May 25, 2016 - link
Exactly. It's a totally different process for planar vs. 3D nand, so whatever they've got that is planar is going to keep churning out planar chips. Expect to see them keep selling planar until those fabs are shut down or moved onto other products.Ariknowsbest - Thursday, May 26, 2016 - link
Rather have the 850 Evo the price difference is too small to justify. If I would buy a TLC drive, I would buy the one with the best value for money.