Hacker News new | past | comments | ask | show | jobs | submit login
Silent changes to Western Digital’s budget SSD may lower speeds by up to 50% (arstechnica.com)
386 points by elsewhen 51 days ago | hide | past | favorite | 104 comments

This happens to video games also. Publishers give review copies with microtransactions not included, the reviewers review it, and then shortly after release they add their free-to-play garbage so they get reviews that don't take the microtransactions into account and then they add it later.

The also do it during ESRB submission to get around having to have a label that mentions in-game purchases. I'm surprised ESRB lets them get away with it, actually.

Jimquisition on it: https://youtu.be/9LeZ89_u2Gc

Article on it: https://www.tweaktown.com/news/66889/publishers-delaying-mic...

One of the latest events like this is Capcoms Resident Evil Village which launched with a DRM that added terrible stuttering to the PC version.

They only admitted to this and promised a fix when a cracked version came out that had the DRM removed and it didn’t have any stuttering [1].

[1] https://www.gamesradar.com/resident-evil-village-pc-patch-la...

The folks at Digital Foundry did a phenomenal job breaking down the specifics, as usual. I strongly suggest checking out their video if you're interested in the stuff:


Capcoms pc ports have always been… I remember my dmc3 that didn’t work at all (maybe a fan patch helped?), and the frame pacing in re4 on pc giving my wife and I migraines… or dmc5 deleting my save file… or re2 remake requiring 2 year old drivers to avoid crashes… I have friends who have issues with the street fighters on pc as well. That’s just off the top of my head.

I work on an upcoming free to play game, we will have microtransactions on cosmetics and possibly new character archetypes.

Microtransactions are not in the early builds of the game because we literally hadn't made the content for it yet, we were focusing on the game mechanics.

The store is a lot of work in a game, and the cosmetics themselves are mostly content work which is not hyper critical to have in place (unlike, say, matchmaking or gun mechanics).

I'm not saying it hasn't been done nefariously; just that: The game I'm making doesn't have microtransactions yet because we worked on other stuff.

But we're not lying about them being there, they have to be there otherwise we'll go bankrupt paying the server costs and failing to recoup R&D costs.

I'm not saying all microtransactions are bad necessarily. I've also worked on a couple of games that have had them, although we did have them on launch (and added more to it later).

But it is kind of a shitty move if you let reviewers review your game and let them believe there won't be microtransactions and then you add them later.

Now if your game is a free game to begin with, it's sort of expected you'll add something in there to make money at some point. But most of these big AAA games are $60-100+ to buy the game to begin with, so there's not an easy assumption to make. Also they almost always have it available a few weeks after launch, as if they were just waiting to get past ESRB and reviewers without mentioning those things.

Finally, I've worked on console games and handled ESRB submissions for our games. You're not supposed to leave anything out. We got dinged once for not mentioning that the card art on one of our cards had a character smoking a pipe on it, and another card where the anime girl had a slightly exposed nipple (we didn't even notice, it was a port of a game by another developer and the card art is busy enough you kind of have to be looking for it... if we had known, we would have told the developer to edit those out and the game wouldn't have had the 'smoking' and 'nudity' label on our rating). The video you submit is supposed to show everything in the game. And yet these companies aren't doing that so they can keep the in-game purchases label off their game.

If you're trying to hide something in your game from certain parties, you're being shitty and you should stop, is the moral of the story. If you're not you're probably fine.

For the ESRB rating for the game I worked on, here's the back of the box for proof: https://www.vgchartz.com/games/boxart/full_6178352AmericaBac...

You should be ensuring that all reviewers mention that this is an early access dev build with all paid content unlocked then, right?

I don't think people are talking about cosmetic micro-transaction though.

You should be ensuring that all reviewers mention that this is an early access dev build with all paid content unlocked then, right?

Sounds like since all the transactions are for cosmetics that haven't been made the content simply doesn't exist in review copies.

Bottom line is reviewers are reviewing one version of the game, and players are getting a sometimes significantly different version. That is anti-consumer behavior no matter how you cut it.

Some responsibility does also fall on reviewers for being aware and where appropriate mentioning this fact.

ESRB is industry controlled iirc, and microtransations/lootboxes are a huge cash flow for the largest games companies

Yeah, it's voluntary self-regulation that was establish to stave off laws being passed in the early 90s during the panic over Mortal Kombat, Doom, etc. Presumably if legislators start looking into microtransactions and rating tricks then the ESRB will start taking them much more seriously.

Et voilà. Conflict of interests all the way...

Phone manufacturers do this as well. Release a phone to reviewers with awesome performance and battery life (by aggressively killing background apps), then a month later, release a build that doesn't do those things so people don't complain about missed notifications, etc.

Hell, even Apple nuked the noise cancelling in the Airpods Pro right after launch [1], possibly for battery reasons or to prevent hearing damage lawsuits, but by all accounts they had jaw-dropping noise cancelling when the reviewers were writing their initial reviews.

[1] https://www.theverge.com/2020/1/17/21069953/apple-airpods-pr...

I see a market for a site called "The late review".

Not sure about consoles, but Steam solves that by ranking recent reviews and those submitted by players with more game time as more relevant.

IMO with games it's preplanned fraud for profit, With hardware it's due to pandemic induced chip shortage but I guess it's still a fraud when product states something and delivers another.

I was expecting that the silent feature reductions in the automobiles due to chip-shortage would seep into another hardware products sooner or later, In fact I was trying to validate whether there's a need to track 'What has been removed due to chip shortage'[1] and it seems like there's good reason to do it.

[1] https://needgap.com/problems/234-what-has-been-removed-due-t... (Disclaimer: I created this platform).

Sounds like the diesel scandal

What are those microtransactions doing? Crypto mining?

The reviewers get sent an account with everything enabled, and have a great time playing as Darth Vader Or Han Solo Or Whatever.

The general public buy the game, and find if you want to play as Darth Vader or Han Solo unlocking them costs $20 each.

Or playing for hundreds of hours per unlock

Or lootboxes, which guarantee you'll spend about $500 per unlock.

"Microtransaction" means small purchases. Such as buying a randomized pack of in-game unlocks for $1 each.


Why is this not fraud?

If you as a company source a device for reviewers and then change the specs in a way that you do not communicate, that is intent to defraud.

If you bought a corvette because all the car reviews said it went 0-60 in 5 seconds, and then you buy one and it takes 10 seconds to do the same thing, you’d have a pretty good argument - especially if everyone else experienced your same situation.

I’d be surprised if a class action suit or FTC investigation doesn’t happens within the next year.

The real question is what Western Digital says the speed for the drive is. I would guess that the changes don't touch the stated top speed, as the article seems to indicate that this only slows down write speeds after the cache is filled.

If Western Digital specified an average write speed, this might be affected. However, that average could technically not have changed depending on how they determined this average speed (for example, they could have determined this average by downloading and uploading smaller files).

I'd say that Western Digital isn't affecting the 0-60 time in ideal conditions, just decreasing the ability of the drive to attain that 0-60 time in less than ideal conditions.

> The real question is what Western Digital says the speed for the drive is.

They say "Up to 1,800 MB/s"

So if they decided to put in flash similar to a samsung 860 QVO, which drops down to 80 MB/s when the cache is full, that would still satisfy the spec.

There's a reason they state the specs using this kind of language. Anyone who experiences performance < 1,800 MB/s is getting what was promised. And anyone who sees performance > 1,800 MB/s is going to be happy.

Where it becomes absurd (and amusing, but still not fraud) is when the manufacturer claims "up to XYZ or more!". Any number is going to match that spec. So the statement says nothing.

Is this number correct? SSD with cache full can drop to speeds slower than spinning disk drives?

Reading those specs... I don't see any reason to buy it. It is worse than my current 1 TB WD Blacks (spinning disk, not SSD). And much more expensive too! I feel disappointed now... I was hoping SSDs would have been improved by now enough to make them good value for money for someoen that needs crazy amounts of storage.

The SSD will have a much better latency, and will in practice lead to shorter load times, even with slow QLC.

But if you need a lot of storage HDDs still have the better price per GB.

If you need crazy amounts of storage you are looking at enterprise SSDs which do not have this sort of problem, not consumer crap.

I think it's at least in part because everyone understands (or thinks they understand) 0-60 times for cars.

But when you start talking about multi-level caches single-level cell memory and the shingled magnetic recording debacle and get down to 02031 1T00 SanDisk Flash vs 60523 1T00 SanDisk Flash chips and continuous writing performance of more than 12 GB, eyes begin to glaze over. For most consumers, it's a black box, the implementation details are unknowable and irrelevant.

People would be ticked off if the Corvette had an undersized radiator that caused it to go into limp-home mode any time you took it to the track, we're starting into slightly esoteric "car guy" details. But enough people know what a radiator is and care about cooling system performance that car manufacturers listen. Very few consumers understand the workings of an SSD; manufacturers are under a lot less pressure as a result.

> eyes begin to glaze over

This argument is not especially helpful as a defense against a claim of fraud. It does nothing to explain why you'd give reviewers devices with different performance than the devices you sell.

Technically, you're correct, which is the best kind of correct. Yes, fraudulent activity is fraud. But what we really care about is whether it's noticed, acted on, and eventually prosecuted for.

Pragmatically, when you have enough complexity that explaining the fraud is difficult, it's still fraud but they're likely to get away with it.

I find it trivial to imagine someone (who's actually invested in the problem, and not just trying to make you go away) peering at you and asking "OK, so this one has 02031 etched on one of the components on the PCB, and this reviewer says it should be 60523, and you want a full refund because you consider that fraudulent? Does the box say which kind it's supposed to contain? No? Fine, I see that when you copy this enormous disk image, um, it looks like it's working? Sure, I can wait another 40 seconds. OK, yes, I see that the speed has dropped from about 2000 to about to 390. And you're saying they committed fraud because it's not supposed to drop? It is, but only to 660? Are you sure your disk isn't just filling up? Maybe you need to run Windows Update or defragment your hard drive...it just finished, was that delay really that bad? How often are you copying a file this big anyways?"

The point is that navigating the above conversation, much less winning it in a courtroom or having it so often that massive numbers of people boycott a manufacturer, is an enormous hurdle. Until it's more easily surmountable, it's immensely profitable for manufacturers to commit this sort of fraud.

> Pragmatically, when you have enough complexity that explaining the fraud is difficult, it's still fraud but they're likely to get away with it.

This is why you have subpoenas and discovery. It seems likely that an internal email chain exists discussing intent to deceive. Even for unsophisticated jurors, that's enough of a smoking gun to make it an open & shut case.

I think you're overstating it. People do recognize and hate when they're given a bum deal and it has nothing to do with getting into esoteric details.

You are going to get into the same debate had decades ago about hard drive size. Base 2 for techs vs base 10 for average consumers.

I hate to break it to you but this is EXACTLY what happens with a C7 generation Corvette. The C7 generation base Stingray, GrandSport and Z06 models do well enough to wow in magazine and vlog reviews but when owners got their cars and started taking them to track days they realized they started over heating after not many laps in under 30 mins.

They are obviously performance oriented machines but mostly designed for every day use on the street. They may have big brakes, great suspension and a pretty great engine but the cooling system included is not equipped to handle a track for more than 30 minutes. Within the first month of deliveries people were freaking out. After 30 mins the car would go into limp mode and you were lucky if you got on the track at all again that day.

But guess what? The reviewers never experienced this.

You don’t need to understand the workings of a car to get that 0-60 in 10s is worse than 0-60 in 5s.

Likewise, you don’t need to understand the workings of an SSD to get that 10s to copy 1GB of data is worse than 5s to copy 1GB of data.

Implementation details are irrelevant here.

But that's not the case here, it's still 5s to copy 1GB of data unless the cache fills up, which many people won't normally do.

So the parent post's analogy with an undersized radiator is still appropriate, you need to understand the whole system to know what its impact on you will be - you can get one good 0-60 run, then need to let the car cool down at idle for 15 minutes before you try it again.

> But when you start talking about multi-level caches...

No, this is not the point.

The average buyer knows very little about cars or buildings but is protected by regulation.

Consumers should not need to be expert to be protected.

I don't think there's any requirement that manufacturers guarantee results of third-party reviews for the entire run of a product. it probably just has to meet the specs published by the manufacturer. as long as you still hit that burst read number with some insane queue depth, everything's cool.

the corvette example doesn't really make sense because the manufacturer provides their own 0-60 number. if no one can hit the manufacturer's number in ideal conditions, that's a problem. now if the manufacturer updates their official specs to 0-60 in 10 seconds, that would be pretty weird, but not illegal.

I remember a similar situation with RAM a few years ago. there was some highly sought-after die that could be overclocked way outside the official specs. you could buy a couple different examples of the same SKU and they would have wildly different overclocking limits based on the underlying die. eventually that die wasn't produced anymore, so no one got the insane overclocks. people were really unhappy about that, but at the end of the day, they were relying on behavior that was outside the promised specification.

They seem emboldened by the event where they silently introduced SMR into NAS drives. That is, they didn't suffer any real consequences there, so now it's a free-for-all.

They must have lost a fair amount of sales?

I previously would never have considered buying Seagate, but since the debacle, when it came to replacing a failed drive in a ZFS array of WD Red CMR drives, I couldn't trust I'd get a drive that would actually resilver so I just replaced them all with Seagate IronWolf NAS drives.

The ops team at my work also avoids WD now, and I no longer buy WD SSDs or external drives because of this distrust.

> They must have lost a fair amount of sales?

(Anecdotally) Not from me. SMR or not, their easystore drives that you can shuck are cheap. Between buying a cheap drive and buying something else, I like to spend less money, so I'd choose the easystores.

Granted, I don't know how much has changed with the easystores since the SMR debacle since I've not really been in the hard drive market for a while.

AFAIK people who were pissed about the reds getting switched to SMR are disjoint from the people who shuck stuff. The latter group usually verifies the drives that they get, because you're not guaranteed to get a drive when shucking.

For anyone else unfamiliar: shucking is buying a USB harddrive purely for the drive inside, discarding the outer case/USB interface https://www.datahoards.com/whats-the-meaning-of-shucking-a-h...

> discarding the outer case/USB interface

I'm not really much of a shucker but I do it from time to time. That being said, I do save the SATA board and power supply (and USB) that I get when I finish the shuck. I haven't had to yet, but it possibly could be handy for connecting up an internal 3.5" drive externally, like if I want to quickly get some data off of it or something.

It is fraud, but laws against fraud are not generally enforced against large businesses in important industrial supply chains unless they consistently and repeatedly flaunt it.

I'd be surprised if there is a class action or FTC penalty.

One might argue that the changes are marginal and most users won't be affected by them. This change hurts write throughput for large write workloads, but these workloads are rare anyway.

If you change the specs, people expect your SKU to change. Don't call it the SN550 anymore. Call it the SN551.

There's a whole slew of internet benchmarks on the SN550 and how it performs on a wide variety of benchmarks. For WD to muddle their branding by calling this new, slower drive ALSO an SN550, weakens the brand and causes distrust in builders.

I remember when Linksys (and others) had wireless devices that worked really well in Linux. I bought 4 of them and got two different versions of the thing with radically different insides.

Someone at Linksys must have noticed that SKU X was selling really well and looked for ways to cut the cost by replacing the innards.

Eventually, online stores were putting the PCB revision number in the sale listings because people had to pick a specific rev.

You are right - it should be treated as fraud. It's a common business practice nowadays to bait and switch - first release good quality products, get good reviews for it and then use those good reviews to market your product. And after it has sold a certain quantity and established a brand value, start using cheaper and inferior quality components and extract more profit.

There's a very good reason Consumer Reports buys from normal retail channels.

No Consumer Reports like publication would survive for this stuff. They'd be so far behind the other reviewers, even if they had better quality data.

Ship the top tier hardware NVME drives to reviewers, wait a few months then start swapping out slower memory chips, controllers, go through multiple revisions while keeping the same name so when gamers look for reviews they see blinding speeds only to be duped. It's incredible what drive manufacturers are getting away with.

Linus Tech Tips recently did a video on this practice[1]. Almost every major component was changed but still sold as the same drive. In summary, performance was different though not necessarily in a major negative way. In this case it was an ADATA NVME drive.

[1] https://www.youtube.com/watch?v=K07sEM6y4Uc

>It's incredible what drive manufacturers are getting away with.

I think it's more incredible that we continue to elect governments that let them get away with it.

This doesn't need more regulation. Just stop trusting professional reviewers who are not bound by the existing false advertising laws that the manufacturers/retails already are. If you want to know how something performs, read the specs published by the manufacturer. If they're too shy to say how slow it is, then there's a reason for that!

Yes it does need more regulation. It is fraud and as a manufacturer you shouldn't be allowed to get away with deceptive shit like this.

It might be deceptive, but only because consumers somehow believe product names correspond to unchanging products. They obviously don't, across many industries. Cars keep the same names but change design every few years. Food keeps changing package size and recipe. It's everywhere! They didn't promise not to change it, and the products still conform to the published specs. If you're too lazy to read that from the horse's mouth and rely on outdated 3rd party reviewers, that's on you.

I've also seen this happen with laptops. They'll have a decent Samsung NVMe SSD when reviewers get their hands on the devices, but then after release, they switch it out for a cheaper one that performs much worse, particularly in longer workloads.

We only have 4 manufacturers of SSD drives who also making their own chips: Intel, Micron, Samsung, and SK Hynix. Here’s the table: https://en.wikipedia.org/wiki/List_of_solid-state_drive_manu...

I tend to buy Samsung and Intel, both working well in my devices.

> I tend to buy Samsung and Intel, both working well in my devices.

It's a generalisation, but be careful of using new Samsung consumer SATA drives (860, 870) if you're using Linux.

The Linux kernel has workarounds in place for the older ones (840, 850 series). But the workarounds for the newer models are only starting to be looked at this week:


Without those workarounds (eg disabling NCQ trim), there are data loss problems.

> disabling NCQ trim

Fixed support for queued trim was part of the marketing claims for the 860 series. Whatever's going on with the 860 and newer drives requires deeper investigation, because there's plenty of reason to believe that simply applying the old workaround is not accurately targeting the real problem.

The old workaround was to turn off support for NCQ (queued) trim, instead just using standard trim.

It doesn't seem to have any real downsides. :)

> It doesn't seem to have any real downsides

That depends greatly on what your software's policy for issuing Trim commands is. Partly because of the dearth of drives properly supporting queued trim, it is common for Linux distro to only run trim commands via cron job, rather than issuing them in real time during a workload.

Thanks, I had no idea about this. I've been using Samsung drives in all of my machines and servers and I run Linux exclusively. With a mixture of 850, 860 and 870's I must be at least partially affected.

Are 860s new? They are what's in my system and while I don't remember when I upgraded it was definitely pre-COVID...


smart says they have 631 days of power-on time, so maybe 2019?


I seem to remember having issues a while back but I haven't seen any since I upgraded my firmware to RVT04B6Q

k, I probably should have said "newer" (than 850 model) rather than "new". The 870 model though, is pretty new. :)

Thanks for the warning. Fortunately, I’m unaffected because running Windows.

I’m running Linux on top in WSL1 and VMWare.

Also running Linux on ARM devices which have neither SATA nor PCIx. When I need fast disks in them to compile my C++ code, I simply use a network-mounted drive to build, gigabit LAN is good.

I thought Western Digital had vertically integrated their SSDs after the purchase of SanDisk?

According to that Wikipedia list, Western Digital does manufacture their own flash, through a joint venture with Kioxia.

Also, Western Digital is looking at completely buying that Kioxia


Just read that Samsung changed the BOM on 970 evo plus. Switched to the controller of the 980 pro. Test reults are mixed. 4K QD1 is up ~50->82MB/s, serial writes after exhausting SLC cache (115GB) are nearly halved 1500->800MB/s.

That's not what happens when you switch to a newer, faster controller. Those performance changes are what happens when you switch to newer NAND flash that's manufactured with twice the per-die capacity, so you have half as many dies for a given drive capacity. (Assuming the newer NAND isn't split into twice as many planes per die to compensate.)

Edit: Looking into this further, TechPowerUp, Computerbase and others have mistaken K90UGY8J5B for K9DUGY8J5B. It's an easy mistake to make, and I've done exactly this before. But it completely explains the performance difference. The D that changed to a 0 signifies a switch from 16 dies per package down to 8 dies per package. The digits signifying the capacity of the package have stayed the same. The "B" at the end signifying the generation has also stayed the same, but Samsung didn't introduce 512Gbit TLC dies until a generation after they introduced 256Gbit TLC dies, so on the smaller dies the "B" means 92L and on the larger dies the "B" means 128L.

Thanks for the correction and the comprehensive information Billy, anandtech is my goto for hardware information, you guys rock. Forgot to mention that they changed the NAND.

> That's not what happens when you switch to a newer, faster controller.

Wasn't also caching strategy changed with the newer controller? Larger dynamic slc, slower folding speed?

I haven't done the math, but my gut feeling is that the relatively modest increase in SLC cache size could not on its own lead to that big a drop in post-cache sustained write speed except possibly where QLC is concerned, and especially not when mitigated by a generational increase in per-die performance.

Samsung's drives are still operating with SLC caches that are smaller than the theoretical limit, which means that even after the cache is full, the drive should be able to accept some amount of writes that bypass the cache and have near native TLC performance; flushing the cache can be deferred a bit longer until the drive is actually starting to run out of free blocks, as opposed to typical QLC drives with maximally-sized SLC caches that run out of free blocks at the same point where the cache runs out.

And Intel just sold their SSD business to SK Hynix, so soon there will only be 3!

This issue happened on WD (acquired SanDisk). So the "just buy NAND manufacturer's SSD" strategy don't work. Also you miss Kioxia!

Sadly my goto Crucial started doing it too, but they paired it with a huge price drop. Was the P2 NVME IIRC. ~70€ for 1TB.

Unbelievable, but is there a datasheet available and does the datasheet have read/write speed figures?

Edit: [1] The datasheet shows "up to" speeds, which means that it is the disk cache speed and not the NAND speed? Quite disappointing.

[1] https://documents.westerndigital.com/content/dam/doc-library...

Don't expect useful specs to be available for consumer drives. The datasheets have slowly been turned into marketing materials over the past 10 years. Only the devices targeted for business use have useful specs and even then, key details such as MTBF or similar can be missing.

Products that don't behave the same should really have different part numbers. In this example, perhaps they could call them SN550 and SN550A (B, C, etc.) or even SN551, SN552, ... if the majority of the design hasn't changed.

Yes, a great current example is consumer motherboards with Intel’s I225V network controller. The controller is unable to run at 2.5Gbps without a fix. The controller was fixed in revision 3, but a huge number of boards are shipping with revision 2. This is two years after the issue was identified, and there is no way to see which one you will get. Even boards manufactured this year have the old NIC, so I assume vendors are being shady to liquidate old NIC stock.

So you get sold a motherboard with a 2.5X NIC slowdown compared to advertised (it gets downgraded to 1Gbps). Update the figure on the box if you’re going to do that!

The most recent example of a "soft recall" would be Gigabyte's exploding PSUs. It makes you wonder how much retail outlets knew, because it is clear that Gigabyte was fully aware of the issue - and silently fixed their design.

This happens with products all the time. Strictly "shrink-flation" of consumer goods is the same thing.

One industrial vendor we once partnered with took a product P that had speed S and accuracy A, discontinued that product and replaced with product P1 which was speed S/10 and accuracy A paired with product P2 which was speed S and accuracy A/10. The original P products were precious if you could find them.

Silent change in this concept means there is no way to identify P2 from P, and both are sold under the same product name, product code, product specifications, and same product reviews. The later is the key aspect as selling products with the intent to trick the consumer is a defining aspect of fraud when done intentionally for profit purpose.

While shrinkflation often also intend to trick the consumer, the details are usually defined. The packet weight or volume number is changed, the weight cost is change (a list requirement for many products where I live), the number of items has changed. With silent changes however there is no such information available and the consumer just will have to hope that their P is not P2.

WD is on my personal blacklist for now. Don't forget last year when they lied and marketed SMR drives as NAS-compatible: https://www.tomshardware.com/news/wd-fesses-up-some-red-hdds...

I can forgive hardware failures. Packing a few trillion bits into a box and making them perfectly reliable is a hard job. What I absolutely cannot and will not forgive is Western Digital lying about their specifications. If I can't trust their words, damned if I'm gonna trust their products.

This is kind of a hyperbolic re-write of the linked article at Tom's Hardware[1], which is I guess originally sourced (by both articles) at a Chinese site called expreview. The linked article at Tom's Hardware has more information and more context that is, confusingly, left out of the top level article.

[1]: https://www.tomshardware.com/news/wd-blue-sn550-ssd-performa...

p.s is it possible to change what the link is to? it seems sort of pointless to have a discussion on an article that appears to be clickbait, in that it adds nothing substantive and in fact neglects to include a lot of information.

This was a problem with the Adata XPG SX8200 Pro, which LTT covered in this video[1].

[1] https://youtu.be/K07sEM6y4Uc

This happens a lot with flash in general, mostly because those who manufacture the end user product don't usually make the NAND (or don't always use their own).

I had the issue with "high end" USB sticks a while back during a review. The speeds given are completely irrelevant and always based on best case scenarios. And while a few manufacturers stick to the same formula for controller + NAND for the life of the product (say Sandisk), others just don't care and will silently switch both the controller and NAND with 0 indication on packaging (no rev., no nothing).

On SSDs I can't remember seeing that though, but then again, this is Western Digital we're talking about. A company best known for, let's say, being a very late entrant to the SSD market.

Edit : At least they (kinda) owned up to it after being caught, quoting the updated article : "For greater transparency going forward, if we make a change to an existing internal SSD, we commit to introducing a new model number whenever any related published specifications are impacted,"

Also, a warning on Western Digital”s SN750 SE series. These are PCIe 4.0 protocol ones that claim to outperform the regular SN750 drives. In reality they’re slower because they’re DRAM-less. Last I checked they were the same price, so be careful to not get the SE model.

How much of a difference does the DRAM cache make anyway, given that there is already a significantly larger disk cache in the system RAM, which is also closer to the CPU?

I think I'm done with WD from now on. The whole WD MyBooklive remote factory result fiasco annoyed me as well - for those impacted they offered a 40% discount on one of their new MyBookLive devices, as long as you send the old drive back (shipping paid by user).

We ought to have the equivalent of crowdsourced semantic versioning and CVEs for products and reviews. As a consumer, I should be able to know _easily_ if any breaking changes have happened that could render a specific set of reviews or benchmarks obsolete, and just as easily if a specific issue is resolved in future revisions.

Too often I've e.g. bought an Amazon product I've bought before, only to receive Revision 2 that doesn't fit my case any more, but that the company silently rewrote descriptions for just to keep their review score.

I understand why people want to keep their review scores, etc, and to some extent they _should_ keep them, but not at the product level, when the product changes.

Quick summary: The drive is 1TB with a 12 GB SLC cache that's capable of 2160 MBps write speeds. When you write more than 12GB in a short time to the drive you can run out of this cache and speeds drop to 390 MBps. The dive was previously tested as capable of 610 MBps in this scenario. (Read speeds are fast in every scenario.)

I bought a 1TB WD Blue M.2 SSD like the one on the picture and my computer has issues booting from it when all USB devices are plugged in (x299/10980XE). It looks like more people have issues booting from it as well, based on the feedback I saw on the seller site. How can one botch SSD that way?

Note that it's not just SSDs that can have completely different components on the same part. I've seen many USB wifi dongles with the same model number use completely different wifi chips internally.

On SSD market, silent change in same model and performance degrade is continuously happen. Any maker should change model number.

I'm super happy with my SN850 except those pesky leds that you cant turn off from Linux.

Not fair

Wanna buy a one legged duck? People who will want this are low end system builders who can add the "Has an SSD" who market to people have no idea how fast a 2 legged duck can run....

*fly. So it makes no difference until the duck lands :D

Maybe the duck has trouble taking off too? In effect they sell a crippled product for less $$ to the cheap system builders who market on price to people impressed by SSD, who know no better

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact