Last week I gave a lightweight overview of what bitcoin is and how it works. ) Like I said, people can make money by certifying blocks for the blockchain. And to do this they need to try lots and lots of SHA256 hashes, looking for the magic nonce that will complete a block. The more computing power you have, the more hashes you can try per second. This improves your odds of winning the SHA256 lottery and netting yourself a few thousand dollars worth of bitcoins.
This works to keep bitcoin secure, but it has also created an intense and sudden demand for lots of computing power. At some point these bitcoin miners looked at consumer-grade graphics hardware and realized that they were an incredible deal in terms of cost vs. computing power. So they began buying them up and stuffing them into minimalist cases where the cards can sit and crunch numbers all day. They’re not even hooked up to monitors! They’re just crunching on SHA256 hashes!
(Note that when I say “bitcoin” I’m sort of doing this handwave-y inclusion of crypto currency in general. I’m talking about bitcoin because it’s the most recognizable and notorious of the currencies, but a lot of the things I talk about below apply more to the other currencies than to bitcoin specifically. But I don’t want to clutter this article up with all the asides and asterisks it would take to sort out the various strains of cryptobux. Just remember that this article is more concerned with graphics hardware and less concerned with crypto.)
The Graphics Hardware Market
All of this has pushed prices up. Normally a new graphics generation comes out, and then the price drifts slowly downward as the technology ages. Eventually a new card comes along to replace it, and the old one undergoes a price drop. Most of us budget-conscious gamers shop in this “recently dethroned” spot on the price curve, which is where you get the most bang for your buck. Any newer than that and you’re paying the hefty enthusiast markup. Any lower than that and the power falls off more quickly than the price.
Unfortunately bitcoin miners noticed this sweet spot too. So they bought up all the best cards. And when those ran out they bought all the next-best cards. (In terms of money vs. power.) And then the next-next best cards. The supply dried up just as demand spiked. And still the miners want more. They sell graphics cards in 6-packs now!
How bad is it? The GeForce GTX 1070 launched with a suggested price of $380. A year and a half later and you might expect the price to have dropped a little, but instead it’s at an insane $1,000!
And that’s assuming you can find them in stock.
Hoping For The Crash
The one thing PC Gamers can hope for is that the bottom will fall out of the bitcoin mining market. I realize it’s not nice to hope for the misfortune of another, but shortages have a way of making us turn on each other.
Everyone kind of has the sense that cryptocurrencies are going to take a tumble at some point, but nobody can agree on when it’ll happen or how far it’ll fall.
Note that I’m not saying this means cryptocurrency is “Doomed”, or a dumb idea, or that it’ll vanish entirely. There are reasons people created this stuff in the first place, and those reasons will continue to exist regardless of how the currencies perform on the open market.
I don’t understand all the reasons people are drawn to cryptobux, but here’s one example: Someone looks at the debt their country is racking up, and they decide (correctly or not) that this is unsustainable. Historically, when a country is on the brink of insolvency (or even over the brink) they begin printing money to satisfy their creditors. This fixes the short-term problem and allows the government to keep doing government stuff, at the expense of destroying the value of their citizen’s money. Someone worried about this kind of outcome might start looking for a way to buy something with their money before it becomes worthless. Historically, gold or other precious metals have been popular for this. But gold is prone to theft and it’s difficult to buy common goods with it. But cryptobux could let you transfer your money into something that won’t crash when your government turns on the printing press, and you can trade it over the internet. You can’t buy a lot of things with bitcoin yet, but all it would take is for someone like Amazon to look at all those billions tied up in bitcoin and decide they’d like to capture some of that value. If they (or another retailer of similar stature) began accepting bitcoin as payment, the entire landscape could change. Bitcoin might go up even more, which would further fuel the mining market and make everything that much worse. (For us PC enthusiasts, anyway.)
There are other reasons to be interested as well. Tax evasion, money laundering, privacy concerns, lack of trust in traditional global banks, and a host of other worries drove demand for this alternative money. It doesn’t matter if you think these people are wrong, stupid, crazy, or immoral. They believe this money is needed, and I don’t think those beliefs will change anytime soon.
However, it’s also true that a lot of people riding the cryptocurrency bandwagon aren’t here for the privacy, freedom, security, or whatever else people think bitcoin can give them. They’re here because they want to get rich. Their demand has inflated the price, which has drawn in more people, which has further inflated the price, and so on.
Which means that we’re in a bubble not unlike the Dot-Com bubble. In 1998, investors thought that the internet was the future of commerce. And they were basically right! But the rush of money attracted speculators, copycats, and get-rich-quick types who dumped their money into unwise ventures and produced a bubble. Certainly that’s at least somewhat true of bitcoin today. There are people buying Bitcoin solely for the reason that it’s been going up. They don’t care about alternative currencies, they’re just jumping on the bandwagon. Once it starts going down they’ll bail out, and the bubble will pop.
That might be a relief to us PC gamers who are terrified our old, obsolete cards are going to fail because we can’t afford to replace them in this market, but the truth is that I don’t think the markets will correct overnight. Our woes are not tied to the value of bitcoin, but the value of bitcoin mining. Even if bitcoin crashes tomorrow and is suddenly worth one-tenth what it is today, that’s not going to magically solve our problems. Maybe miners will stop building new machines and supply will gradually recover. But how low will the price have to go before the miners turn off the server farm and put their graphics cards on the used market?
(Then again, mining apparently puts an incredible stress on hardware, since it runs the GPU at full load 24/7. Most people don’t game 24/7. And even when they do, very few games really max out the hardware. So a card bought a year ago for crypto mining could easily have the equivalent of a decade and a half of wear and tear. Which means it might be near failure. If the used market were flooded with a bunch of unreliable, burned-out cards, then people may shy away from used and prefer new. Great. Another thing to speculate about.)
So bitcoin might continue to go up, or it might tank. And even if it does tank, it might not help the graphics card market. And even if it does, it will probably take some time. And even then, the market could be complicated by fraudsters and a glut of unreliable worn-out cards. In the long term this problem will sort itself out, but in the short term I don’t see any hope for people who need an upgrade.
So why doesn’t NVIDIA just raise production to meet demand? Isn’t that what companies are supposed to do?
Why Not Make More?
Like Apple, Qualcomm, AMD, and several other tech giants, NVIDIA is a “fabless semiconductor company”. They don’t have a facility for fabricating their own chips. Instead, they pay Taiwan Semiconductor Manufacturing Company (TSMC) to make the chips for them. These fabless companies schedule time on the TSMC fab facility, essentially renting a huge manufacturing complex. They compete with the likes of Apple for fab time, so there’s an upper limit on how many chips they can have made in a given year.
On top of this is some additional drama / complexity. Back in 2015, TSMC announced they were going to build a new “mega-fab” installation. That new facility isn’t online yetI’m basing this on the assumption that if it was up and running, there would have been an announcement.. This isn’t surprising. It takes years to construct, staff, and supply a fab facility, and I imagine it gets even harder if you’re building a “mega-fab”. Note that a fab facility isn’t just a magic box that makes microchips pop out. Each facility is geared towards a certain type of production. As technology improves, we figure out how to fabricate chips with smaller circuits. That’s what’s been driving this whole crazy Moore’s Law business for the last several decades. Back in 2007, circuits were around 100 nanometersHow small is a nanometer? One nanometer is to one meter as one meter is to THE SUN. (Give or take a little.). Every few years we refine the process and the circuits get a little smaller. 40nm. Then 28nm. 20nm. 14nm. Last year, TSMC began selling their first 10nm chips.
Every time the circuits get smaller, the power requirements change. You need to re-think how all that technology fits together if you want to get the most out of it. Which means you need to spend some time coming up with new circuit board layouts. (Or whatever it is the eggheads do.) The point is that there’s an overhead to moving to the next level of miniaturization. There’s R&D time, prototyping, and testing. You engineers need to spend months solving the new heat, power, and spatial problems. You don’t want to go through that every year.
If you’re a company like NVIDIA and you’re trying to maximize efficiency then you want to make the jump to the new fab process, wait for the technology to stabilize, spin up the manufacturing, then sell lots and lots. If you’re always making the jump to the latest technology then your hardware will always be in the rough “early shakeout” period. On the other hand, if you wait too long then your rivals will swoop in and began selling the New Hotness while you’re still stuck on the old technology. So there’s a bit of jousting and brinkmanship going on between companies as they try to minimize the cost of upgrades without leaving themselves too vulnerable.
What I’m getting at is that the whole system is really complicated and we can’t get more graphics cards by just cranking up the speed dial on the graphics card factory. Is there fab capacity available at TSMC? Is that capacity in the right size range for what we’re trying to make? Are the third-party guys ready to manufacture and sell the final product? This is a deeply complex system, and we can’t just plop down new factories like we’re playing SimCity and wait for the problem to resolve itself.
A Thought Experiment
For the sake of simplicity, let’s brush all of that aside. Let’s imagine an alternate world where NVIDIA makes their own chips in their own factories and builds their own cards and sells those cards directly to consumers. What does NVIDIA do about this graphics card shortage? Obviously they want to make as many as possible, as soon as possible. But how?
Assuming they’re already running the NVIDIA factories at capacity, they’re faced with a dangerous choice. They can spend billions to expand their facilities and hire and train a bunch more workers. It can take months or even years to get something like that up to speed. But then if the bitcoin bubble pops, all those bitcoin-mining graphics cards will end up dumped on the second-hand market. NVIDIA will have spent a fortune to expand just as the market is collapsing and prices are falling. They’ll have tons of new expenses and costs from the expanded production, but demand for their goods will be at an all-time low as the PC gamers feast on the now dirt-cheap used card market. A turn like that can put you out of business. Even if you survive, it can hobble you for years to come.
On the other hand, what if NVIDIA doesn’t expand and the mining market continues to grow? The graphics card shortage is really hurting PC gaming in the short term. We just had the launch of a half-step console generation. If the shortage continues, some of those PC gamers could migrate over to consoles. If we get a few more big-name titles like Wolfenstein II that launch with high system requirements and lots of driver problems, then then PC gaming public may be pulled towards getting consoles because it costs too much to keep their gaming PCs up-to-date.
That doesn’t just hurt your bottom line this year, that hurts it for several years. Fewer people on PC means fewer games get ported to PC. PC ports became an afterthought for a niche market, which means more sloppy ports and even more pressure to get a console. Sure, some of us will stay on the PC regardless, but a lot of people are more fickle / flexible in their gaming habits.
So now NVIDIA needs to figure out how to respond to this spike in demand, and to know how to respond they need to be able to predict what will happen with cryptobux bubble, which sort of makes them another speculator.
And of course, all of this becomes that much harder to think about if you consider that problem in the real world where it takes a long chain of companies to make the final product. NVIDIA probably really, really wants to make lots of new cards as fast as possible, but TSMC? They might not care at all. The GPU market is just part of their business and this bitcoin stuff isn’t really a serious concern for them. Somebody will need chips made in the future, and to them one company is as good as another.
So what is NVIDIA doing? Are they extending themselves to try to increase supply, or are they holding their breath and praying this whole things blows over as soon as possible?
My guesses aren’t any better than a coin flip, but I really hope this gets sorted out soon. I don’t care about internet funny-money, but I’m really due for a graphics upgrade and I can’t afford anything with prices where they are now.
 I’m basing this on the assumption that if it was up and running, there would have been an announcement.
 How small is a nanometer? One nanometer is to one meter as one meter is to THE SUN. (Give or take a little.)
The Strange Evolution of OpenGL
Sometimes software is engineered. Sometimes it grows organically. And sometimes it's thrown together seemingly at random over two decades.
Black Desert Online
This Korean title would be the greatest MMO ever made if not for the horrendous monetization system. And the embarrassing translation. And the terrible progression. And the developer's general apathy towards its western audience.
The Best of 2017
My picks for what was important, awesome, or worth talking about in 2017.
WAY back in 2005, I wrote about a D&D campaign I was running. The campaign is still there, in the bottom-most strata of the archives.
Steam Summer Blues
This mess of dross, confusion, and terrible UI design is the storefront the big publishers couldn't beat? Amazing.