new july batch (233010WD) of 1tb SN550 have massive downgrade, buyer beware

[appleache]

Summary

 

RN the news is exploding in Chinese media circle, the recent NEW batch of sn550 1tb seems to have only half of the speed outside SLC cache compare to the original, old can do 800mb/s while new one is just about or less than 400mb/s. the new batch come with firmware version 233010WD, and 潮玩客 has tried flash older drive with this firmware but failed, and new drive also cannot be forced downgrade to the older firmware starting with 21 or older, this suggest a change in hardware level(NAND). RN the NEW sn550 1TB now has out of SLC cache speed similar to 500gb ver of sn550, or 960GB SN350. There is no solid prove that the NAND new SN550 is QLC as WD themself dont give enough details on their NAND chip part number, but there is a listing for 1TB QLC version of SN350 on their product sheet with 100TBW, old tlc SN350 is 960GB 80TBW, for context the OG SN550 1TB is 600TBW. RN there is no saying if this batch is CHINA/Asia only or global, but NA usually have older stockpile.

 

 Quotes

Quote

Running out with TxBENCH, it seems that the new SN550 has a faster write speed than the old SN550 when there is SLC cache, but we have verified with CDM that the continuous write speed of the two is the same whether in QD1 or QD8. But after the cache runs out, the average write speed of the new version of SN550 is only 390MB/s, which is not even half of the original version, which is basically the same level as the SN350. - expreview.com translated

 

experview tested speed, middle is sandisk ultra which is basically the old sn550:

潮玩客's video：

 

 

old vs new, dashboard shows firmware is already up to date and cannot be forced updated even by changing filename

 

 

 

My thoughts

even with the speed downgrade SN550 1tb is still a ok DRAMless drive, however now it is no longer the god tier budget DRAMless drive due to it "new performance". now it just on a slightly better position compare to other $100 DRAMless. consider stuff like sn750 1tb is only $20 there is no point gambling for SN550, since there is a possibility that the actual TBW is now the same as SN350 or just a bit better. PLUS other SSD that has DRAM occasionally comes down $110 or less, like Silicon Power P34A80, no point buying a (more) lootboxed SSD.

also WD is selling a pcie4 ver of SN550 in same Amazon listing as SN750, named SN750 SE, very scummy move.

 

Sources

expreview.com https://www.expreview.com/80127.html

潮玩客 https://www.bilibili.com/video/BV1eA411A7Z1

supplemental source:

https://www.bilibili.com/video/BV18P4y1s7Rt

https://www.bilibili.com/video/BV15L411b7Ur

[Moonzy]

Why do storage companies have to be so scummy?

[WereCat]

Also beware of Crucial P2. 

They changed TLC to QLC. while keeping the same product name. 

 

edit 

Source 

https://www.tomshardware.com/features/crucial-p2-ssd-qlc-flash-swap-downgrade

[Fasterthannothing]

Unfortunately the problem is that chip shortages are causes performance cuts to products. I'm not defending these scummy companies in any way but it seems to me that every single storage company has taken the approach of not saying anything about changing their specs because they know darn well it will blow over. Someone show me an article written in the last week or two about drives that got their chips switched out 2 or 3 months ago? People have too short of a memory and bigger fish to fry so to speak (Gigabyte and their exploding power supply) to care about something 99% won't notice.

[Stahlmann]

I have no problem with them changing out some parts for the sake of keeping up production. But if they want to keep the name the same then they should also keep the performance the same. And if the performance is different, they should just call it "SN550 v2" or so. This way the consumer at least knows it's a different product.

[That Franc]

Thanks for the heads up, I used to recommend these drives to people.

[appleache]

7 hours ago, SlidewaysZ said:

Unfortunately the problem is that chip shortages are causes performance cuts to products. I'm not defending these scummy companies in any way but it seems to me that every single storage company has taken the approach of not saying anything about changing their specs because they know darn well it will blow over. Someone show me an article written in the last week or two about drives that got their chips switched out 2 or 3 months ago? People have too short of a memory and bigger fish to fry so to speak (Gigabyte and their exploding power supply) to care about something 99% won't notice.

If you just changing source of nand chip for about the same or just a bit slower performance, that is normal way back in the days. But here we are talking about potentially changing Tlc to qlc with 1/6 of tbw without disclosures, crucial p2 at least has quietly changed the tbw listing.

[StDragon]

...and this is why I stick with Samsung SSDs

[appleache]

12 hours ago, Stahlmann said:

I have no problem with them changing out some parts for the sake of keeping up production. But if they want to keep the name the same then they should also keep the performance the same. And if the performance is different, they should just call it "SN550 v2" or so. This way the consumer at least knows it's a different product.

most important thing is TBW is now potentially different, which is SSD's lifeblood

[SeriousDad69]

2 hours ago, StDragon said:

...and this is why I stick with Samsung SSDs

Yup, paying $20-40 extra to know what you're getting without digging through reviews and product sheets is worth it. Samsung makes it easy by just marking all of their garbage drives with QVO.

[lostcattears]

Western Digital have been getting scummier it has lately always been them in the news for bad marketing/hard drives/ ssd 

[ApolloFury]

Are Seagate FireCuda affected by this?

[Senzelian]

Did the pseudo SLC cache size change aswell?

 

 

[appleache]

15 hours ago, Senzelian said:

Did the pseudo SLC cache size change aswell?

 

 

Still 12gb as old

[appleache]

16 hours ago, ApolloFury said:

Are Seagate FireCuda affected by this?

Seagate Is not owned by wd KEK, nor using same platform 

[Senzelian]

2 minutes ago, appleache said:

Still 12gb as old

This line isn't at 12GB

[Zodiark1593]

So, are companies just slowly handing a monopoly to Samsung?

[Distinctly Average]

On 8/20/2021 at 12:04 PM, SlidewaysZ said:

Unfortunately the problem is that chip shortages are causes performance cuts to products. I'm not defending these scummy companies in any way but it seems to me that every single storage company has taken the approach of not saying anything about changing their specs because they know darn well it will blow over. Someone show me an article written in the last week or two about drives that got their chips switched out 2 or 3 months ago? People have too short of a memory and bigger fish to fry so to speak (Gigabyte and their exploding power supply) to care about something 99% won't notice.

The chip shortage is not the problem, it is the lies. We make purchase decisions based on specifications published both by the manufacturer then implicitly through samples sent to trusted reviewers. Then the manufacturer switches parts later one screwing both the customer and the trusted reviewers. Let’s be honest, the manufacturers are doing their s for profit and nothing else. I would be far more likely to buy their kit if they were up front about the changes giving the devices new part numbers and detailing the difference. 

[leadeater]

5 hours ago, James Evens said:

Don't like Samsung particularly well for different reasons but that makes them slowly the preferred option for SSDs above the garbage tier.

Why? Unless it's super unrelated. Samsung has been the best SSD vendor since well.. forever. I only buy Samsung SSDs and I know all the high end large capacity enterprise SATA options from storage vendors are exclusively Samsung. Probably NVMe as well but that would be speculation where as I was directly told by these vendors that for SATA it was the case.

 

Only area I can think of to dock Samsung points is price, something I'm personally fine with pay that extra.

 

1 hour ago, Zodiark1593 said:

So, are companies just slowly handing a monopoly to Samsung?

Well it already was. Samsung as of latest figures is 42% of NAND market, side note 54% of DRAM market, and the lowest it's ever been is 28% in Q4 2014, next quarter it was back to 33%. No other single company has managed to get above 20% since the start of 2014, it really is and has been Samsung vs the rest with early credit going to Toshiba/Kioxia early on then nose diving at the end of 2013 (RIP).

 

On 8/20/2021 at 8:25 PM, Moonzy said:

Why do storage companies have to be so scummy?

Well I guess for the most part most people shouldn't have noticed the difference. What's most likely happened is the number of NAND chips halved from using newer twice as dense ones. Like I really really disliking saying this but if you need an SSD with consistent sustained performance you need to buy one and pay for one that is designed for it.

 

On to why I think people are complaining, Chia. Bottom dollar buyers trying to make a win fall on Chia farm/mining and getting hit by the performance loss. Because these buyers are doing it for revenue purposes I don't have much sympathy for them, spend more on better SSDs. For everyone else while it sucks in a way how much are you actually impacted and personally I'd rather capacity size changes not change performance this much at all anyway so if the models below 1TB already had this performance then the 1TB in my view should have been the same anyway, or the lesser capacity ones matching that of the 1TB.

[leadeater]

53 minutes ago, James Evens said:

For the chia guys this sucks as they might look at different drives but they probably didn't cared at all about the brand.

They don't care about the brand but they do care about the performance as well as the capacity too. I said it because they would be the most likely to notice and China is where the big push in Chia started/is. What I meant was they are likely the source of the discovery.

 

53 minutes ago, James Evens said:

Also isn't QLC not a bad choice for chia if you can still get TLC for somewhat the same price with higher write endurance?

Yes but I also wouldn't be jumping on the QLC theory yet, it could well be just half the number of twice as dense TLC. WD early last year transitioned from BiCS4 to BiCS5 volume production, which comes in TLC and QLC, and each layer is far more dense. Instead of having a 16 layer TLC BiCS4 NAND chip it could be an 8 or 9 layer TLC BiCS5 chip that is slower due to the lower number of layers. That is why the 500GB TLC BiCS4 is half the speed.

 

However with the above said the performance doesn't quite match what a TLC BiCS5 should be doing with 8 (ish) layers. If 8 layer TLC BiCS4 was doing ~500MB/s then 8 layer TLC BiCS5 should be doing very roughly 40% more than that. So on a pure performance basis it's closer to QLC BiCS5 than it is to TLC BiCS5.

 

Really need NAND part number to know.

 

Edit:

Also there are BiCS4 TLC 512Gb dies so it could be half the number of BiCS4 dies in usage.

 

But if I had to bet money, QLC BiCS4 has been swapped in.

[ouroesa]

23 hours ago, leadeater said:

On to why I think people are complaining, Chia. Bottom dollar buyers trying to make a win fall on Chia farm/mining and getting hit by the performance loss. Because these buyers are doing it for revenue purposes I don't have much sympathy for them, spend more on better SSDs. For everyone else while it sucks in a way how much are you actually impacted and personally I'd rather capacity size changes not change performance this much at all anyway so if the models below 1TB already had this performance then the 1TB in my view should have been the same anyway, or the lesser capacity ones matching that of the 1TB.

Why I generally agree, I don't think this is the main reason for people complaining. People are complaining because they don't like being taken advantage of nor do they appreciate products being misrepresented. 

When purchasing something, we rely on relevant specifications and compare that with the price. So changing half of the equation will severely impact the price vs performance and thus a crucial metric most people use when buying storage (price/perf; price/capacity and reliability are generally what storage purchase options are valuated on).

 

Think of it this way -  if you buy a car with a V8 which can do, say 0-100 in 6s but they deliver a car with an inline 4 doing 0-100 in 10s. Yes both will get you from point A to point B and both will eventually get to 100km/h but you paid for the larger engine because you might use it to tow something or do pulls or whatever someday and now that option has been taken away yet you got no decrease in price nor notice. Any reasonable person would be livid.

[leadeater]

23 minutes ago, ouroesa said:

When purchasing something, we rely on relevant specifications and compare that with the price.

This is where we hit problems though, because the product specifications are only based on the SLC cache performance so as far as the product specification goes, as tested and represented by WD they have not actually changed. I've said for a long time now consumer SSD product specs are complete garbage and have zero use or relevance at all.

 

WD made no claims or promises, or published specifications on performance outside of SLC cache mode or anything to do with long duration performance.

 

Quote

Test Conditions: Performance is based on the CrystalDiskMark 7.0.0f benchmark using a 1000MB LBA range ASUS Z170A desktop with Intel® i7-6700K 4.0Ghz, 8GB 2133MHz DDR4. Windows 10 Pro 64-bit version 1903 using Microsoft
StorNVMe driver, secondary drive. Performance may vary based on host device. 1 MB = 1,000,000 bytes. IOPS = input/output operations per second..

https://documents.westerndigital.com/content/dam/doc-library/en_us/assets/public/western-digital/product/internal-drives/wd-blue-nvme-ssd/product-brief-wd-blue-sn550-nvme-ssd.pdf

 

Run this test on the new revision and you'll get results within margin of error of the published specifications.

 

Edit:

The problem like I mentioned is the SSD market itself and that it is acceptable that mere capacity increases alone is deemed perfectly ok to get twice the performance within the same exact product model. Like actually wtf is with this, why is this a thing and why do we accept it. Less dense NAND chips should be used on smaller capacity devices to maintain the performance across the same model and different models to denote different performance levels or enduance.

Edited August 23, 2021 by leadeater

[ouroesa]

16 minutes ago, leadeater said:

This is where we hit problems though, because the product specifications are only based on the SLC cache performance so as far as the product specification goes, as tested and represented by WD they have not actually changed. I've said for along time now consumer SSD product specs are complete garbage and have zero use or relevance at all.

 

WD made no claims or promises, or published specifications on performance outside of SLC cache mode or anything to do with long duration performance.

I see your point. However a decent portion of people also lookas at third party reviews which will require updating after this. They (WD) could argue that this is not their responsibility nor does it change their data, as you mentioned, but still a really scummy practise hiding behind small text/loopholes. 

 

I will steer clear of WD products for the foreseeable future due to business practises like these. Wont make an iota of difference to them but I've got a little pride left which I intend to keep in tact.

[leadeater]

15 minutes ago, ouroesa said:

They could argue that this is not their responsibility nor does it change their data, as you mentioned, but still a really scummy practise hiding behind small text/loopholes

As with my edit, the entire consumer SSD industry has been this way near as much the whole time. Product model numbers are and will continue to be useless while capacity differences can account for up to 100% difference in performance.

 

It's not like Enterprise/Datacenter is much better as well, in fact it can be even more difficult to navigate. Something as simple as a formfactor change from U.2 to M.2 of the same model and capacity can result in half the performance dropping off.

 

Reviews can often be of little help as you'll commonly only get a review of one capacity, two if you are lucky, and it's almost always the largest capacity one that get the review. They are of course very helpful to spot within generation revision changes but if you are not buying the reviewed capacity then good luck with that dice roll because you really do have no idea what you are actually getting, could be within 10% or 100%, have fun.

 

SSD/NAND industry, all smoke and mirrors 

[ouroesa]

3 minutes ago, leadeater said:

SSD/NAND industry, all smoke and mirrors 

Almost all industries are getting to this point. Marketing, how I loathe marketing. They 'stretch/bend' the truth so much that you cannot go with any info a company puts out there.

Imagine a world run by engineers...wait, no.