Catching Up: The "Story" So Far

Here's the deal: We wrote about Kingston's NAND switch in its V300 a few months ago, whereupon the synchronous NAND V300 was benchmarked against its asynchronous successor. The controversy here was spurred-on by Anandtech, whose writers had spotted user-conducted benchmarks of the V300 SSD on its own forums; the benchmarks consistently showed performance lower than what professional reviewers had seen a year prior. The reason, it was discovered, was a result of a supply-side change of the NAND Flash utilized in the device. Although Kingston shipped the V300 with its original BOM for an entire year, it recently switched from Toshiba's 19nm Toggle-Mode 2.0 NAND to a 20nm Micron alternative; this was entirely for cost and supply reasons, we're told by Kingston, but users have seen a performance impact from the switch.

If you're not sure the meaning of this, our very in-depth SSD architecture article will teach you all about NAND and how SSDs work. It is critical that all who make large claims for or against Kingston's device read the two above-linked articles for a proper understanding of what's going on here.

An image from our SSD Architecture post. It's not so simple.

The user backlash stemmed somewhat misguidedly from a performance standpoint, though it would have had much stronger basis as an ethics argument. In my published testing, we noted that many of the performance metrics were within (or adjacent to) Kingston's original specification; some of the tests, like mass file copy, did see a tremendous performance detriment when compared to the original synchronous NAND V300. It was my conclusion that the V300 was still an OK SSD for its target market -- a very affordable, low-end solution for mainstream and mobile users (read: not performing professional tasks or mass I/O). I stated that most target users would not notice the performance difference anyway, even though power users, professionals, and piratical types may spot the change. These users, though, fall outside of the targeted market and would do better to purchase a higher-endurance device rated for their use cases.

I later stated that I have stopped recommending the V300 in favor of Crucial's new MX100 and older M500, but this was entirely due to value; Crucial's offerings outperformed Kingston's and were within $10 -- a plain-and-simple choice, in my mind. This choice is purely from a cost effectiveness standpoint.

But I haven't called the V300 a bad drive -- I just don't think it competes in value, and we'll get into why in a moment. We've got another half of the story to recap first.

PNY's Controller Switch 

We've only published article content specific to PNY once before; they're a much smaller organization in comparison to the Kingstons, Corsairs, and Samsungs of the world, but they've still got a large Enterprise division and a growing consumer division.

PNY's Optima SSD line was recently found to use LSI's Gen2 SandForce controllers in some devices, as opposed to the Silicon Motion (SMI) controllers found in earlier models. Some users have perceived this as being of a malicious nature.

Since then, given all the publicity from multiple media outlets, reddit threads, and social media, hundreds of thousands have seen accusations of a "bait-and-switch" tactic on part of PNY and Kingston. Further inflammatory comments have falsely claimed that reviewers received "special" samples that were never shipped to the public. This is incorrect, as we'll find out below.

As someone who has worked on both the media side and manufacturing side of the industry, this all seemed like a non-story to me; still, it's apparently gotten everyone's attention. Let's break it down piece-by-piece.

Why We're Writing This Piece 

Frankly, frustration. This posting is spawned entirely out of frustration from snowballing levels of unresearched, oft-unfounded rage. Accusations that are based entirely in a fantasy world have further caused frustration for reviewers, especially when users make claims that we're provided with better samples than the market will ever receive. It's not that simple, and to make such claims is offensive to the effort that reviewers across the industry invest in research, almost suggesting that we're easily "fooled."

Here's where we set the record straight for the reality of the industry we are all a part of.

This Happens Every Day - Stop Calling it a "Bait-and-Switch" 

First, a bit of a timeline.

It is important to note that neither Kingston nor PNY listed the NAND type in their original product specification sheet; PNY did not even list its Optima's controller, leaving it to reviewers who'd opened the device to detail.

Kingston sent its V300 to reviewers more than a year ago now (4Q12 / 1Q13, from memory). As always, it received numerous reviews throughout the initial sample seeding cycle, then began shipping to market in short order. Consumers who purchased the device would find that their models performed almost identically to media samples, when tested properly. It wasn't until around December of 2013, a full year after its announcement, that changes were found in shipping retail models.

A very similar story is true for PNY.

So, then: Why did the companies change what's being sold? This was explained in great detail in my V300 investigation, but we'll recap briefly.

These SSD manufacturers are fabless. They do not fabricate their own NAND Flash dies or controllers. The world's major Flash manufacturers include Hynix, Toshiba, Micron in partnership with Intel, and Samsung; each of these companies operates or leases sometimes multi-billion dollar silicon fabrication facilities. The fabrication plants source the synthetic silicon crystal, slice it into wafers, and then cut dies out of the wafer. A manufacturer -- Kingston, PNY, Corsair -- then purchases the dies, buys a controller, and sticks them onto a custom design. You can see the assembly process here. This is slightly different for Samsung, who make their own NAND and controllers.

When the supply changes, when partnerships change, when new NAND is engineered and old NAND ceases production, existing products face a serious hurdle to continued manufacturing. These fabrication plants aren't going to continue production of an older Flash design when a newer, higher-yield alternative emerges; similarly, manufacturers have to keep building their product lest it sees an early EOL. Controllers don't iterate that frequently, so "just make a new product" isn't really an answer that works. LSI's (soon to be Seagate's) new SandForce Gen3 controller is the first consumer SandForce controller since 2009, and Samsung's have seen big gaps in development as well, with large thanks to interface limits. This is important to note because manufacturers don't refresh devices too frequently when there's a lull in technology; there's a reason the HyperX SSD and M500 were available for as long as they were.

RAM is in a similar situation. In the past two years, it has become cost ineffective to manufacture 1333MHz RAM in favor of the (now) high-yield 1600MHz, faster alternative. You'll still find 1333MHz RAM sold, but it's using stepped-down components or a different supply from its original production run. It is also often a significantly worse value as a result of the cessation of 1333MHz DRAM fabrication.

The point is that—as discussed in my V300 investigation—supply changes force manufacturing changes in favor of retaining a low-cost device in its target market. The V300 and Optima wanted to be affordable devices, and with changes from Toshiba's NAND production (for Kingston), a new NAND supplier had to be found to keep the 120GB V300 in a ~$70 range. Keeping the synchronous NAND would result in an SSD that could cost several-fold more for the consumer. The Optima was another simple supply issue: PNY couldn't get enough of one type of controller (either SandForce or SMI) and still keep up with demand, so they did what any responsible manufacturer would -- they combined the two sources to ensure steady outflow of devices.

Here's the thing: This happens every day in computer hardware. That doesn't make it good or right, it doesn't make every instance excusable, but this isn't news. Power supplies change small components as the distance from original release grows, RAM changes DRAM modules or bins down, even CPUs and GPUs see this activity. Companies are trying to keep up with demand in one sector or another -- high-end or low-end -- and they can make tweaks to ensure one product doesn't "run out" before the others.

Look at GPUs: Users have historically been able to "unlock" AMD GPUs for higher performance. This isn't a mistake -- AMD is not stupid. AMD shipped those unlockable devices at a lower bin to keep up with demand for, in that case, the 6950. They may have been less stable in some instances, but such is the risk of modifying a product. Similarly, AMD has shipped processors with locked CPU cores when the X4 proved unstable at times, creating the X3 that could be unlocked with the right motherboard. These aren't attempts to "screw the consumer" out of a better product, but efforts to increase yield (in the case of the X3) and supply.

Power supplies generally get a free pass, seeing as very few people understand them nor have the expensive means to adequately test a PSU. RAM is seen as a commodity and is generally ignored as a boring component, making it somewhat invisible to the user when supply changes happen.

Kingston didn't specify its NAND and PNY didn't specify its controller. Both devices reportedly perform within original minimum product specifications. The V300 suffers hard in some specific aspects of performance that make it more questionable than PNY's change, so the change certainly isn't completely free of fault or criticism. That said, as someone who has worked both sides of the industry, these changes don't constitute "scandalous news" to me.

To call something a "bait-and-switch" necessitates actual bait. There is an argument to be made against the way the companies conducted these changes, though, but we'll get to that in the final section. I'm not giving Kingston and PNY a completely free pass, but I do think that the internet's reaction has been largely out of touch with the reality of the industry.

I fear for what would happen if the same buyers boycotting Kingston and PNY knew to what extent other manufacturers have changed SSD components. It'd be entirely impossible to purchase an SSD without violating some sort of boycott. Point to any SSD that's more than a year old -- chances are, it's not an identical BOM to the original.

Reviewers Aren't Getting Better Stuff

Also from our SSD Architecture post.

Let's set something straight. Although we received the synchronous version of the Kingston V300 SSD and other outlets have received the SMI PNY Optima SSD, that doesn't mean Kingston & PNY are "making better products for reviewers." The expense of such an endeavor would be astronomical -- a low production run with higher-quality devices isn't as simple as flipping a bit in the NAND or controller. It requires completely new suppliers in some cases, and it's not cheap to just run a few hundred SSDs through Toshiba while making the rest with Micron; it is also likely that exclusivity or contracts were involved, making such accusations even more unlikely.

These companies -- both Kingston and PNY -- shipped their devices with the original review sample components for a year or more in some cases, landing thousands of identical components into consumer hands. Although some companies have treated reviewers differently, this is not one of those instances.

The Moral High Ground: A Discussion of Ethics

A NAND gate.

Here's where things get more cloudy.

The argument to be made against both companies is pretty straight-forward: Prospective buyers are referencing benchmarks performed by reviewers during the first run of the device (unless they look at our bench).

If a user is expecting a 512MB/s 4MB SEQ READ throughput and receives 222MB/s, or if they're expecting a 4K QD16 random performance of 52K IOPS and receive 17K, it's potentially a huge problem. The listed specifications only detail "4K Random R/W" and "Sequential R/W," without necessarily going into detail on QD or other methodology. In light of the original criticism from months ago, Kingston published its own benchmarks between the V300A and V300S to ensure users know precisely what they're getting. Our benchmarks, quite honestly, are more definitive -- but it was a solid effort that seems fairly in-line with my own testing.

The company thinks that it "did no wrong" in this change and has not issued an apology. PNY allegedly issued this statement to Tweaktown:

"As mentioned in the product description of Optima, 'The PNY Optima SSD line utilizes multiple qualified controllers to offer the best available solution to our customers.'

Yes we did ship some Optima SSD's with SandForce controllers, but only if they meet the minimum advertised performance levels (in most of the benchmark tests, LSI controllers outperform SMI controllers). The readers assumption that PNY has abandoned SMI controllers is wrong as we have been shipping mostly SMI controllers, but also utilizing LSI to fill in the gaps."

Neither company thinks they've done wrong, neither has apologized. Both cite the fact that the rest of the industry does this constantly and that the devices perform within spec. Further defense is employed that users who'd require high performance in specific areas of SSD operation (see: 4K @ QD16, something that never happens in normal consumer use) should be purchasing a drive better-suited for them, anyway; I tend to agree with that. Buying a $70 SSD for professional video encoding and rapid file compression doesn't really make sense.

Kingston told us a few months ago (again, in the V300 investigation) that they would not be making silent changes again in the future.

But there's still an ethics concern: Should companies make announcements every time they change a product? Given how irresponsibly the internet has responded to the changes thus far, I just don't know that it's the best move. Users have given these companies all the reason in the world to continue normal operation; announcing a change in the future could be met with backlash by uninformed, trigger-happy consumers who see every change as an attempt at screwing the customer. There are bad changes. There are ethics issues of silence pertaining to changes when reviews are already online. I'll leave it to you all to decide the ethics of this matter, but I will leave everyone with the same thing I tell my staff:

"Don't call a company good or bad; don't ever say we trust a company or find them to be of consistently high quality; don't judge the products by corporate actions. Analyze every product independently on its own merits, then review it independently."

- Steve "Lelldorianx" Burke.