Building the Steam Controller

Building the Steam Controller

When we first started designing hardware at Valve, we decided we wanted to try and do the manufacturing as well. To achieve our goal of a flexible controller, we felt it was important to have a similar amount of flexibility in our manufacturing process, and that meant looking into automated assembly lines. It turns out that most consumer hardware of this kind still has humans involved in stages throughout manufacturing, but we kind of went overboard, and built one of the largest fully automated assembly lines in the US. Our film crew recently put together a video of that assembly line, showcasing exactly why robots are awesome.

Here’s that showcase, where you can watch controllers being built entirely by robots. We aren’t crazy though, so humans are still on hand to keep the robots from becoming sentient.

Advertisements

bunniestudios.com: On MicroSD Problems

bunniestudios.com: On MicroSD Problems

First, the date code on the irregular card is uninitialized. Dates are counted as the offset from 00/2000 in the CID field, so a value of 00/2000 means they didn’t bother to assign a date (for what it’s worth, in the year 2000, 2GB microSD cards also didn’t exist). Also, the serial number is very low — 0x960 is decimal 2,400. Other cards in the irregular batch also had similarly very low serial numbers, in the hundreds to thousands range. The chance of me “just happening” to get the very first microSD cards out of a factory is pretty remote. The serial number of the normal card, for example, is 0x9C62CAE6, or decimal 2,623,720,166 — a much more feasible serial number for a popular product like a microSD card. Very low serial numbers, like very low MAC ID addresses, are a hallmark of the “ghost shift”, i.e. the shift that happens very late at night when a rogue worker enters the factory and runs the production machine off the books. Significantly, ghost shifts are often run using marginal material that would normally be disposed of but were intercepted on the way to the grinder. As a result, the markings and characteristics of the material often look absolutely authentic, because the ghost material is a product of the same line as genuine material.

Furthermore, the manufacturer’s ID is 0x41 (ASCII ‘A’), which I don’t recognize (supposedly the SD group assigns all the MIDs but I don’t see a public list of them anywhere). The OEMID is also 0x3432, which is suspiciously ASCII ’42’ (one more than the hex value for the manufacturer ID). These hex/ascii confusions are possible signs that someone who didn’t appreciate the meaning of these fields was running a ghost shift making these cards.

Armed with this evidence, we confronted Kingston — both the distributor in China as well as the US sales rep. First, we wanted to know if these were real cards, and second, if they were real cards, why were the serialization codes irregular? After some time, the Kingston guys came back to us and swore these cards were authentic, not fakes, but at least they reversed their position on not offering an exchange on the cards — they took back the programmed cards and exchanged them for new ones, no further questions asked.

However, they never answered as to why their card ID numbers were irregular. While I know chumby is a small fry customer compared to the Nokias of the world, I think it’s still important that they answer basic questions about their quality control process even to the small fry. I had an issue once with an old version of a Quintic part being accidentally shipped to me, and once I could prove the issue to them, I received world-class customer service from Quintic, a full explanation, and an immediate and full exchange of the parts at their cost. That was exemplary service, and I commend and strongly recommend Quintic for it. Kingston, on the other hand, did not set an example to follow.

Normally, at this point, I would simply disqualify Kingston as a vendor, but I’m more persistent than that. It’s disconcerting that a high-profile, established brand would stand behind such irregular components. Who is to say SanDisk or Samsung wouldn’t do the same? Price erosion has been brutal on all the FLASH vendors, and as small fry I might be repeatedly taken advantage of as a sink for marginal material to improve the FLASH vendor’s bottom lines. Given the relatively high cost of these components, I needed to develop some simple guidelines for IQC (incoming quality control) inspection to accept or reject shipments from memory vendors, so I decided to do more digging to try and find ground truth.

The first thing I had to do was collect a lot of samples. The key is to attempt to collect both regular and irregular cards in the wild, so I went to the SEG / Hua Qian Bei district and wandered around the gray markets there. I bought about ten memory cards total from small vendors, at prices varying from 30-50 RMB ($4.40 – $7.30), most of them priced toward 30 RMB. The process of shopping for irregular cards itself was interesting. In talking to a couple dozen vendors, you learn a few things. First, Kingston as a brand is weak in China for microSD cards. Sandisk has done a lot more marketing in the microSD space, and as a result, it’s much easier to find Sandisk cards on the open market. The quality of the grey-market Sandisk cards are also typically more consistent. Second, the small vendors are entirely brazen about selling you well-crafted fakes. Typically, the bare cards are just sitting loose in trays in the display case; once you agree on the price and commit to buying the card, the vendor will toss the loose card into a “real” Kingston retail package, and then miraculously pull out a certificate, complete with hologram, serial numbers, and a kingston.com URL you can visit to validate your purchase, and slap it on the back of the retail package right in front of your eyes. Hey, it’s just like new! … I suppose the typical buyer in those markets is not an end user, but someone who is looking to make a quick buck reselling these cards at a hefty markup in a more reputable retail outlet.

One vendor in particular interested me; it was literally a mom, pop and one young child sitting in a small stall of the mobile phone market, and they were busily slapping dozens of non-Kingston marked cards into Kingston retail packaging. They had no desire to sell to me, but I was persistent; this card interested me in particular because it also had the broken “D” logo but no Kingston marking.

Significantly, Kingston is revealed as simply a vendor that re-marks other people’s chips in its own packaging. Every Kingston card surprisingly had a Sandisk/Toshiba memory chip inside, and the only variance or “value add” that could be found is in the selection of the controller chip. Oddly enough, of all the vendors, Kingston quoted with the best lead times and pricing — better than SanDisk or Samsung, despite the competition making all their own silicon and thereby having a lower inherent cost structure. This tells me that Kingston must be crushed when it comes to margin, which may explain why irregular cards are finding their way into their supply chain. Kingston is also probably more willing to talk to smaller accounts like me because as a channel brand they can’t compete against OEMs like Sandisk or Samsung for the biggest contracts from the likes of Nokia or RIMM. Effectively, Kingston is just a channel trader and is probably seen by SanDisk/Toshiba as a demand buffer for their production output. I also wouldn’t be surprised if SanDisk/Toshiba was selling Kingston “A-” grade parts, i.e., parts with slightly more defective sectors, but otherwise perfectly serviceable. As a result, Kingston plays a significant and important role in stabilizing microSD card prices and improving fab margins, but at some risk to their own brand image.

Overall, the MicroSD card market is a fascinating one, a discussion perhaps worth a blog post on its own. I’d like to point out to casual readers that the spot price of MicroSD cards is nearly identical to the spot price of the very same NAND FLASH chips used on the inside. In other words, the extra controller IC inside the microSD card is sold to you “for free”. The economics that drive this are fascinating, but in a nutshell, my suspicion is that incorporating the controller into the package and having it test, manage and mark bad blocks more than offsets the cost of testing each memory chip individually. A full bad block scan can take a long time on a large FLASH IC, and chip testers cost millions of dollars. Therefore, the amortized cost per chip for test alone can be comparable to the cost of silicon itself.

In the end, I’d have to say that both SanDisk and Samsung look like they might be superior wholesale vendors to Kingston for memory cards due to their more direct control of their respective supply chains. Unfortunately, you can’t buy Samsung-branded microSD cards on the retail market, as far as I know — Samsung only sells their cards to wholesalers who then rebrand and/or resell the card, and like Kingston these non-OEM brands may blend their vendors so it’s hard to say if you’re getting the best card or simply a usable card.

www.Visual6502.org: Visual Transistor-level Simulation of the 6502 CPU

www.Visual6502.org: Visual Transistor-level Simulation of the 6502 CPU
The first of our projects is aimed at the classic MOS 6502 processor. It’s similar to work carried out for the Intel 4004 35th anniversary project, though we’ve taken a different approach to modeling and studying the chip. In the summer of 2009, working from a single 6502, we exposed the silicon die, photographed its surface at high resolution and also photographed its substrate. Using these two highly detailed aligned photographs, we created vector polygon models of each of the chip’s physical components – about 20,000 of them in total for the 6502. These components form circuits in a few simple ways according to how they contact each other, so by intersecting our polygons, we were able to create a complete digital model and transistor-level simulation of the chip. This model is very accurate and can run classic 6502 programs, including Atari games. By rendering our polygons with colors corresponding to their ‘high’ or ‘low’ logic state, we can show, visually, exactly how the ch

AnandTech

AnandTech
I started AnandTech as a hobby on April 26, 1997. Back then it was called Anand’s Hardware Tech Page, and it was hosted on a now-defunct free hosting service called Geocities. I was 14 at the time and simply wanted to share what I knew, which admittedly wasn’t much, with others on the web. In those days PCs were very expensive and you could often save a good amount of money buying components and building your own. We have our roots in reviewing PC components and technologies.

Adapteva

Adapteva
Adapteva Launches $99 Parallella Open Computing Platform

We successfully launched the Parallella project based on the Epiphany architecture through Kickstarter and will be delivering the first prototype platforms by the end of December.

raspberrypi.org

raspberrypi.org
The Raspberry Pi is a credit-card sized computer that plugs into your TV and a keyboard. It’s a capable little PC which can be used for many of the things that your desktop PC does, like spreadsheets, word-processing and games. It also plays high-definition video. We want to see it being used by kids all over the world to learn programming. Can I buy shares in the Raspberry Pi Foundation? We’re a charity, so you can’t buy shares in the company. If you want to support us, we’d love you to buy one.