Western Digital is trying to redefine the word “RPM”
arstechnica.com
arstechnica.com
The conclusion on page 4 of the above thread:
> Non-Pro RED = 5400rpm according to data sheet
> Pro Red = 7200rpm according to data sheet
> Internal both 7200rpm and 120Hz.
> Apparently all WD / HGST helium plates have real 7200rpm, no matter what it says.
Simply absurd, if true. Here, for example, is a screenshot of WD's WD100EFAX store page: https://imgur.com/HseW9Pa
No mention of an RPM "class". The spec sheet does refer to it as "5400 RPM Class", but with no further description of what the term means: https://media.flixcar.com/f360cdn/Western_Digital-3805661149...
What a world when you can't even trust the spec sheet. What's next, 10 watt class drives that pull 20 watts? 3-year class warranties?
Just like how internet speeds are advertised ("Up to 300 Mb/s").
If anyone wants to have an unfettered pipe then call up a Tier 1/2/3 ISP and ask for a commercial connection via a single-mode fibre, and be ready to pay quite a bit for it.
You might be surprised to find it surprisingly affordable. Cheaper than colocation, even.
If you get half a rack or less, that tends to be (quite a bit) cheaper.
Assuming you can easily get hooked up at a residential address. Perhaps if you live in an apartment/condo tower, then fibre may have already been run:
* https://www.fibrestream.ca/locations
Some random house, on some random street? Unlikely.
The community of Ammon, ID, built their own municipal fibre network because of this:
> It all began about seven years ago when the city government, suffering from horrible upload speeds, wanted a dedicated 100Mbps circuit between City Hall and the Public Works building, less than a mile apart. Qwest (now CenturyLink) didn't want the project, while Cable One offered to do it for a start-up fee of $80,000 plus $1,000 a month, Patterson said.
* https://arstechnica.com/information-technology/2016/06/what-...
At this year's (virtual) NLNOG there was presentation from a fellow who was fed up with slow Internet options in his rural Michigan county and so setup his own ISP by having fibre optic pulled to his house and then redistributing service to his neighbours:
Such rules are not within the American Overton window for the usual reasons.
[0] https://www.broadbandchoices.co.uk/ask-our-expert/broadband-...
I think the biggest reason they say "up to" is because a lot of poeple get a cheap $40 router that can't support that the advertised speed. Heck, I just moved into a new place that gives symmetric gigabit speeds. I had to replace my router to actually get one that can handle the advertised speeds (I had an APU1d4 with realtek ethernet ports that saturated at 300 Mb/s). My older laptop does not even have wired nor wireless speeds that can handle the connection.
So while I am no fan of a lot of ISPs, I completely understand why they say "up to", they don't want to deal with folks who say they don't get their advertised speeds while using a $200 laptop and a $40 router.
https://transition.fcc.gov/oet/mba/images/2018/chart1.png https://www.fcc.gov/reports-research/reports/measuring-broad...
Now, I absolutely would love to see it mandated by law that somewhere upfront in ISP info to consumers it's required to have a standardized SLA. A consumer-level SLA for sure, but still have it formalized and cover all of the basics. So not just absolute max bandwidth, but 90%/50%/whatever bandwidth, any guaranteed floor, 90% latency, any data caps, any throttling criteria, etc. All categories the same and measured in a mandated standardized way so universally comparable. I don't expect data center level SLAs at all, but there isn't any reason there can't be that level of clarity.
But residential/SOHO and such service inherently has more uncertainty and more compromises and capacity multipliers to meet lower prices. I think that's ok, just that it should be laid out and not buried in some fine print or support page in the disused lavatory defended by a leopard either.
When I was on ATT gigabit fiber in San Jose, it was hard to confirm the speeds, but they would vary a lot more. I think most of the time at peak I would only see 400 Mbps down, but closer to the line speed off-peak. Upstream was more consistent.
Sonic gigabit in SF. I’d say that’s within margin of error. I see 900s on the early AM. Plus my WiFi can’t go that fast so I’m ok with it.
Granted, before my current router I was only hitting just over 50% of the rated speed. But that turned out to be the router I was using, its CPU was not fast enough or something. With a new router and all wired connections between me and the 'net, I get the speeds I pay for.
Perhaps, with tongue in cheek, we could call this the p0 speed — as in it doesn’t make sense and it’s guaranteed that 0% of customers achieve it.
It's really really difficult to predict this statistically.
OTOH, there's a guaranteed minimum speed the ISP has to deliver.
Turns out that consumer protection laws are pretty useful.
Long ago an ISP sold me on VDSL for the office as "up to 55/3 Mbit/s". After install I notice they had deployed ADSL equipment with a theoretical maximum throughput of 12/1.3 Mbit/s. They claimed that this was ok and tried to hide behind the "up to" clause.
Only in the fine print do they disclose how many actual lumens.
And if you're buying products with Energy Star certification, they have to meet a specific output requirement to claim an equivalent wattage: https://www.energystar.gov/products/lighting_fans/light_bulb...
If you're buying no-name brands off Amazon with no certification, they may very well be lying to you. But "60W equivalent" is still a metric with a specific meaning, whether or not a particular shitty company is lying to you about it.
Maybe we need an independent "this spec sheet isn't full of bullshit" certification for hard drives.
I think Western Digital just made a year's worth of profits off of people who don't read spec sheets, and they have a plausible interpretation that would unfortunately will probably survive a court challenge. Western Digital has clearly taken the position that '5400 RPM' has always been a statement of disk I/O performance, never a statement about other characteristics, and so by saying '5400 RPM Class' they only mean the drive's I/O performance, not any of those other secondary characteristics that led to the uproar we're discussing.
Western Digital is in the wrong here — not for repricing 'cool and quiet' drives, but for communication — and should have just openly said:
"We're ending production of quiet-cool 5400 RPM drives in our consumer product lines. If cost is of most concern to your installation, our consumer products will continue to meet your needs. If heat and noise are of most concern to your installation, our professional products with WhisperDrive(tm) will continue to meet your needs."
EDIT: I also think WD should be publishing thermal characteristics, not just "watts used".
Exactly. The difference can be huge, even if the number of "dB" measured are the same, to the point that one would not recognize the noise of one running HD, whereas another could be extremely irritating.
There is an exact reason why: the curve of perception limits of our ears is not a horizontal line but a curve, where the higher power of the tone is needed when moving to the lower frequencies.
See: https://en.wikipedia.org/wiki/Equal-loudness_contour
Knowing that: 5400RPM == 90Hz; 7200RPM == 120Hz
The proper reporting of the noise levels would therefore be a noise histogram, or at least should include the frequency of the peak level and not be just a single number if one'd want to compare how irritating the noise is. And even a single histogram can miss the time component, as the noise with the main peak level at some frequency can also come in the waves, making it even more irritating, basically, the histogram of the histogram could also be needed to tell you what's going on.
Moreover, the said "equal-loudness contour" is not exactly the same for different listeners, and as a result different listeners could hear the same noise sources differently. E.g. my girlfriend and I hear the same noise sources (emitting different levels of noise over different frequencies) completely different.
In short, not having precise histograms and analysis simple "is it 5400 RPM" was a really useful information to know the expected noise characteristic of a HD. It's not an accident that these two became two different speeds with the expected use in different environments.
I've intentionally bought 5400RPM disks whenever it wasn't for some servers locked in the server rooms.
Edit (answering "floatingatoll"'s comment below): All the above should be an argument that: "“are they essentially silent from a few feet away”, which is often simple to answer given a drive to test" is not something that has the same answer for everybody. If I'm hearing the lower frequencies better than you, what is "essentially silent" for you can be very annoying for me, and that's why I'd need 5400RPM where you could tolerate 7200RPM. It's also an argument that the proper measurement could indeed be made to know what will be irritating for which person, but it's not common to be present even in the reviews, even less in the specs of the manufacturer. Which is again not surprising, as the frequency histograms sadly aren't common even in the reviews of the music equipment. Common consumers indeed prefer some simple "yes/no" verdicts ("is it quiet" enough), like the one "floatingatoll" below proposes. But such statements just aren't the answer usable for everyone -- that's anyway why the number for noise level is reported. But it's not enough. I know it firsthand, as I personally do hear better low level sounds than many people around me. What's "quiet" for them is "irritating noise" for me.
It's not regardless:
5400RPM == 90Hz
7200RPM == 120Hz
The human perception curve shows that the humans hear much better the sounds at 120Hz than at 90Hz,
https://en.wikipedia.org/wiki/Equal-loudness_contour
That means that for the same reported noise level of e.g. 30dB a 5400RPM drive is much more quiet for humans than a 7200RPM drive.
That is how the people recognized that these HDs aren't 5400RPM -- they sounded "too loud" so they took the sample of the noise and drawing the histogram they proved that the peak was at 120Hz.
As far as I know, nobody managed to produce a 7200RPM drive which does not have a peak at 120Hz. If you know some that doesn't have such characteristic, please write, everybody will celebrate that achievement.
Congratulations: Your idea is both technically possible and completely useless in practice.
For many years, the WD Green series didn't even specify the RPM.
I hate marketing...
In this case, the "5400 RPM" drive runs at 7200 RPM, but WD presumably makes it (either via software or hardware) perform worse in other ways to compensate, so that its performance is really only competitive with actual 5400 RPM drives.
That way, they can justify the lower price point without actually needing two separate manufacturing processes for each kind of drive (like how Intel saves on having to redesign each chip in a lineup).
[0] Of course it's a possibility that the analog front-end and ADC is still binned for performance or populated with different components, so you need a bit luck here if you want matched performance.
Why does it feel like big corporations are just getting more blatantly evil these days?
The 'confusion' as to whether the "K" prefix meant a factor of 10^3 or 2^10 goes back decades.
In 1961 a book says "it is more usual to express the speed in bits or kilobits (1,000 bits) per second"; an IBM 1410 manual says ""The 40K core array requires 40,000 valid five-position addresses from 0,000 to 39,999." In 1962 an ACM paper referred to a "4K IBM 1401" which had 4000 characters of storage, but in 1964 Amdahl wrote a paper that had 1K meaning 1024. Academics went back and forth a bit it seems.
* https://en.wikipedia.org/wiki/Timeline_of_binary_prefixes
For hard drives specifically, a 1974 example has Mbytes in the 10^6 sense. Further examples often have drives using the base-10 sense of the prefix (including the first 5 ¼ inch floppy), while memory chips using the base-2 sense of the prefix.
The powers-of-1024 nomenclature of kibi/mebi/etc was introduced in 1995, with IEEE mandating SI prefixes only be used for powers-of-1000 in 1997. The IEC follows suit a couple of years later.
That's easy to explain, as a 1401 is actually a decimal computer.
> In the 1950s, "1 kilobit" meant 1000 bits:[9][10]
Because hard drives, going back to the original (the IBM 350), have used decimal prefixes.
Given what this article is about, I don't know why you're giving HD manufacturers the benefit of the doubt.
None of those mitigating/confounding factors apply to RPM, so the current lies are more blatant.
(Though it does seem somewhat relevant to point out here that a "500GB" drive usually has an accessible capacity of 500,107,862,016 bytes, implying an effective definition of 1GB=1,000,215,724 bytes, which is neither 1024^3 bytes nor 10^9, nor 10^9 rounded up to the nearest whole (binary-sized) sector. And that effective GB definition varies with drive capacity.)
KB = Kilobyte, which from SI definition of kilo means 10^3.
KiB = Kibibyte, was formaly defined later by IEC in 1998 as 2^10
OSes (Windows/Mac/Linux), are the ones doing it wrong, using 2^10 internaly, but presenting using SI prefixes.
They got a new CEO this year, but the previous CEO had been HGST's CEO until WDC bought HGST and put him in charge. (He was also previously HGST's CFO, and before that, WDC's CFO.)
But I don't think a management change is particularly important here. Their competitors were caught doing the same BS with SMR drives. I think we're just seeing what happens when a lucrative duopoly gets sidelined by replacement technology. The consumer hard drive market has become a market for lemons, and only the datacenter customers get treated seriously anymore.
So it sounds like the reason behind it is that they can save money by eliminating the 5400rpm production line.
Easy fix though, smack a new 5400rpm label on those disks and fix the firmware to report 5400rpm, now they could get rid of the production line and retain the customers that weren't looking for 7200rpm disks.
Absolute garbage.
There's an order or two of magnitude less data and benchmarks on the Toshiba drives, but I just bought Toshiba 6x6TBs for a NAS build because I couldn't stomach buying WD or Seagate (or HGST which is just WD now).
EDIT: Forgot to say, the drives have been great. It's only been 6 months or so, but they've been performant, no bad sectors yet.
https://blocksandfiles.com/2020/04/16/toshiba-desktop-disk-d...
The only way to get good hard drives these days is to buy from the enterprise lines e.g. WD Gold or Seagate Exos. They're extremely expensive, but this is the only way to get the kind of quality that used to be standard a few years ago.
It's reasonable to expect companies to honour their spec sheets, and to not lie in their firmware.
Once they're caught doing otherwise it's important to hold them accountable, not throw our hands up and pay up anyway.
However, the difference was not in just branding which to me was surprising. In Elements the drive is held by rubber brackets which won’t fit in non-WD drives, but Easystore is using rubberized screws, which allowed me to use the case with older Seagates.
They do run hot in the external case. I was trying to run badblocks and when the temp became about 60C, I decided to run an external fan. I wonder how long do these drives last in these external cases.. In my NAS with fans, the highest it got was 49C
Some WD external drives will hang the boot process if the host computer probes the drive for bootability. These drives will even stop the host computer from booting if they're connected via USB, much less if they're directly connected.
Some USB external drives have the USB controller embedded on the drive PCB and don't provide an accessible SATA interface.
As of writing this comment, big external drives still come with 3.5" hard drives, with the only caveat being that some require you to disable the 3.3V rail on your SATA connector, either by cutting the wire or taping some pins on your HDD with kapton tape. Without the workaround some of these shucked drives will not power up properly when connected to a SATA power plug that also provides 3.3V.
In this case, it sounds (from the official response) as if WD is trying to “dumb down” the specs so that generally less informed buyers can compare them easily and make the right choices?! But this effort will fail because it hides information in scenarios where it certainly must cater to different needs — not everyone wants all the speed at a specific price just as not everyone wants the lowest power consumption at a specific price.
The discerning users who want to get the specs and then decide what trade offs are worth it for their use case must be able to get it, even if it’s on page 6 of the “technical specifications” document.
Which is total bullshit, because what "less informed buyers" even read spec sheets? The spec sheet is for people that want detailed technical information. Info for the average consumer should be on the retail box and store descriptions.
You'd think nobody would mind until you remember some people want lower power and noise.
Ideally these people would be buying based on those actual metrics, rather than the RPM as a proxy for them, if that's what they really wanted, but I don’t know if that’s on the data sheets.
>You cannot just run a drive at a different RPM (except for idling, different RPM while reading/writing requires different read head calibrations and glide height, etc.).
Power use and acoustics are listed in WD datasheets, but I guess they might not be well comparable to values from other manufacturers.
Given the fact that they lied about RPM on their datasheets, why should anyone believe other pieces of information on their datasheets are correct?
Personally the only metric I care about now on HDD is reliability.
The sketchy stuff shows up in cheap components with slim margins (budget AIO watercoolers, budget hard drives, budget PSUs budget/no-name ebay video cards).
That market is basically just Nvidia (80%+ market share and still dominating in performance/features).
I'm typing this on a machine with a fairly powerfull AMD GPU. The market for Nvidia is basically just Nvidia, everyone keeps using their special APIs. Their cards are not significantly more powerful than what AMD offers and their drivers suck because they're constantly using them for anti-competitive crap.
It's over 80% of GPU the market.
> Their cards are not significantly more powerful than what AMD offers
AMD can't compete with Nvidia's high end. Especially with the new 30XX series, which crush the performance of Nvidia's 20XX series.
> and their drivers suck because they're constantly using them for anti-competitive crap.
Their drivers have a lot of features that AMD doesn't, like G-Sync, DLSS, RTX and others.
Companies have been abusing the ability to force consumers into binding arbitration but this is another level. I'm not a lawyer but it doesn't seem to be clear if this would hold up in court; it is however pretty clearly a scare tactic, which is almost worse.
For the usual reasons, congress has refused to change the law to vacate the ruling.
For this drive, there is no explicit acceptance of the terms given and the terms are unavailable before not just purchase, but before opening of the box.
Thank you for the comment, I've edited mine to reflect the ambiguity.
There might be more power consumption, but in this case they are selling you a faster drive.
I kind of think this should be ok.
I wonder if mustang guys ordered a 6-cylinder mustang and got the v8 - would they complain about gas mileage?
There’s also heat, noise, and drive lifespan.
Do people order 6 cylinder mustangs because that’s what they want or because they can’t afford the 8?
Why wouldn't they? They were fraudulently sold a vehicle that didn't meet the claimed specs.
I don't know if they advertised it as a low power drive, but if they did it would be surprising in a negative sense.
I have a very standard commute where I drive boring and defensively. I enjoy cars as much as they get me from point A to B. There would be no benefit for me spending extra money despite the "better" engine.
Right now the high capacity, non-SSD hard drive market is a duopoly between WD and Seagate. Customers have little choice.
The small number of suppliers for flash memory might have something to do with this.
I'm amazed that there are people that would pick this up. I genuinely wouldn't notice a drive being a tiny bit louder/hotter
Exactly because of the issue in that video: if there are different RPMs they shake each other and cause problems.
I've never actually tested if this is true, but I've always made sure to honor it.
If in fact it is true, that this lie by Western digital is even worse.