It’s a bad time to build a gaming PC. Although building a PC yourself can usually net you hundreds of dollars in savings, a continued rise in the cost of components makes it nearly impossible to put together a machine at a reasonable price. And the main reason why is the ongoing GPU shortage.
Rising prices aren’t anything new for PC components. GPU prices ebb and flow with demand and RAM prices rock up and down like a roller coaster, and those two components alone can represent a significant portion of your build price.
This time is different, though. A perfect storm of constrained supply, overwhelming demand, and market expansion would lock buyers out of finding a graphics card for more than a year. And although things are getting better, the problem is far from over.
To understand how we got here, here’s a history of the GPU shortage.
A dose of context
Before getting into the GPU shortage that’s currently going on, it’s important to look back. This isn’t a new problem for the GPU market. The difference today isn’t the shortage, but the fact that various factors have extended and worsened it.
The market experienced a similar shortage in 2017 with Nvidia’s GTX 10 series and AMD’s RX 500 series graphics cards. The shortage stemmed from increased demand due to rising Ethereum prices, which is an alternative to Bitcoin that is still contributing to the shortage today.
Graphics cards are especially vulnerable to shortages and price volatility compared to many other PC components.
Similarly, the Asia-Pacific region saw a shortage of GTX 980 and GTX 980 Ti graphics cards in 2015. The main culprit at the time was an issue with GPU supply chains, which was furthered by increased demand in the region. Prices rose about 15% in parts of the world, though they quickly returned to normal.
Graphics cards are especially vulnerable to shortages and price volatility compared to many other PC components. The two main GPU companies, Nvidia and AMD, don’t actually manufacture that many graphics cards. Instead, they provide the designs that add-in board partners (AIBs) license to build and sell their own versions.
This model does a lot of good, including offering users various card options for different types of computers. You could, for example, choose a liquid-cooled card for a high-end gaming PC or a slimmed-down card that doesn’t require external power for a stealthy build.
It carries some issues, however. Nvidia and AMD are free to set a target price for their graphics cards, but the final price depends on how much components, logistics, and taxes cost for AIBs. That makes the target price largely irrelevant, especially when the cost of components is high. This is important to keep in mind when looking at the GPU shortage happening today.
The few first-party cards that AMD and Nvidia make sell out immediately after launch. AIB cards come in after the fact and normalize supply, as well as provide more options to the market.
Given the past several launches, a GPU shortage was inevitable. The difference this time was that the PC gaming market was growing at an alarming rate, and a pandemic was about to shut people inside for over a year.
It’s a pandemic
On March 11, 2020, the World Health Organization (WHO) officially classified the coronavirus as a pandemic. California implemented the first mandatory stay-at-home order on March 19, and by May 31, 73% of U.S. counties had a stay-at-home order in place.
With an exodus from the office, the demand for computers grew. In 2020, the PC market grew by 13.1%, which is the largest growth the market had seen since 2010 when it grew by 13.7%.
Ryan Reith, an employee at market research firm IDC, said that the “growth centered around work from home and remote learning needs.” Reflecting on the boom, Reith said “the pandemic not only fueled PC market demand but also created opportunities that resulted in a market expansion.”
At the same time, the semiconductor industry started having some problems. Semiconductors are the building blocks of most modern electronics. In short, they can conduct manageable current through an electronic circuit, which is useful for everything from building a digital alarm clock to building a graphics card.
Semiconductor companies require long lead times, so companies like AMD and Nvidia usually request a certain allocation of manufacturing capacity months or even years in advance. The pandemic, as well as the rise in demand for computers, threw this delicate system out of balance.
In 2020, the PC market grew by 13.1%, which is the largest growth the market had seen since 2010 when it grew by 13.7%.
Decreased demand for cars in early 2020 caused many automakers to reduce their order size, only to try and increase their order size when the car market rebounded later in the year. Meanwhile, demand for Bluetooth and Wi-Fi chips skyrocketed, all while the tools essential for creating them decreased in supply.
Those are only two examples of issues the semiconductor industry faced in 2020. The important thing to remember for the GPU shortage is that semiconductor companies were operating at full capacity, all while the pandemic disrupted supply chains and grew the entire computing market.
Ampere arrives … for some
Nvidia launched the RTX 3080 on September 17, 2020, kicking off the Ampere range and the GPU shortage. Unsurprisingly, the cards sold out immediately only to be marked up by secondhand sellers. That’s not out of the ordinary for the graphics card market.
At the time, short-term shortages were blamed on increased demand. Nvidia CEO Jensen Huang said that “the 3080 and 3090 have a demand issue, not a supply issue” during a press event at the company’s GPU Technology Conference. Huang said that there wasn’t “a real problem to solve,” chalking the issue up to “just a phenomenon.”
That wasn’t the case, though. Nvidia chose the Samsung 8nm node for the Ampere range, and sources blamed the shortage on the poor yields of the node. Essentially, the manufacturing process produced fewer usable chips than expected. Nvidia later acknowledged the problem and changed its stance, pointing some of the blame at Samsung.
At the same time, DRAM prices started to rise due to limited supply, which is still affecting the market today. These memory modules are essential to multiple PC components, including graphics cards.
Between industry-wide issues, poor yields at Samsung, rising DRAM prices, and, of course, the pandemic, Nvidia had to fight an uphill battle to get cards built and shipped. These factors lead to a limited supply of Ampere cards at launch, which would have posed an issue regardless of the pandemic.
Because of the pandemic, the shortage came faster, lasted longer, and got more out of hand than any previous one.
Enter the scalpers
Scalpers were quick to capitalize on the limited supply and high demand. Almost immediately after launch, scalpers started sharing their winning strategy for buying up GPUs: Bots. One user reported grabbing 42 cards, while others said they scored 30 or more.
The pandemic helped fuel the success of bots. Instead of a physical launch, the supply of Ampere cards was limited to online retailers. By scanning multiple retailers and exploiting the checkout processes, bots were able to buy up cards faster than any human could. And without a way to buy a card in person, the available stock quickly dwindled.
Reports of bot use mainly came from social media posts, so it’s hard to pin down just how many cards scalpers claimed at launch. However, the power of hindsight confirms that the reports were at least somewhat accurate. In 2020, scalpers sold around 50,000 Nvidia RTX 30 series graphics cards, totaling $61.5 million in sales.
It was clear that bots were able to grab and, at least for a short period of time, control the market. Shortly after the launch of the RTX 3080, Nvidia announced it would delay the launch of the RTX 3070 to make “more cards available on launch day.” To stem the bleeding further, Nvidia partnered with Best Buy for exclusive distribution of Founders Edition cards in the U.S.
Given the pandemic, retailers like Best Buy resorted to selling the cards exclusively online. However, some retailers, such as Micro Center, moved to in-store sales shortly after launch. Even months later, hopeful buyers continued to camp out for hours at Micro Center locations around the country.
During this time, graphics card maker EVGA introduced a queue-based notification system for graphics cards to combat bots. This remains one of the most effective ways to buy a GPU at list price, providing buyers a window of eight hours to purchase a graphics card when their name comes up in the queue.
Despite these measures, bots grabbed much of the available stock, and, given the yield issues, that stock wasn’t immediately replenished. Online retailers like Newegg implemented anti-bot measures and order limits, but it was too late. The cards were already gone and it would be a slow process to get them back.
Under the thumb of a trade war
Across the Pacific, there were more problems brewing in the semiconductor industry. On September 26, 2020, the U.S. Commerce Department imposed restrictions on SMIC, China’s largest semiconductor company. That pushed U.S. business to chipmakers TSMC in Taiwan and Samsung in Korea, and they were already operating at full capacity.
The trade war between the U.S. and China began in 2018 when tariffs were imposed on solar panels and washing machines. A flurry of restrictions and tariffs came in the following years, but the restriction placed on SMIC came at the worst possible time.
It’s important to remember that the issues facing the semiconductor industry were ongoing at this point, and the restrictions on SMIC came only nine days after the RTX 3080 launched. Stock was already low before the restriction, and with big clients looking to buy up capacity, the GPU shortage would go from bad to worse.
AMD gets cocky with RDNA 2
AMD announced the RDNA 2-powered RX 6800 XT, 6800, and 6900 XT on October 28, 2020. Competing with Nvidia on performance and offering ray tracing for the first time on AMD hardware, the cards looked like an answer to the GPU shortage.
It looked like that was AMD’s view, too, as the company’s Chief Architect of Gaming Solutions and Marketing suggested it wouldn’t be a “paper launch” like Nvidia’s Ampere range.
It was, though, as all available units sold out within minutes of going live. Originally, AMD said it would meet the demand in the weeks following the launch. However, the cards remained out of stock at retailers and prices surged on the secondhand market, following the pattern that Nvidia’s Ampere cards experienced a few months earlier.
AMD faced some unique challenges compared to Nvidia, however. In addition to three new graphics cards, the company launched a series of new processors at the same time, as well as provided the framework for the Xbox Series X and PlayStation 5. And AMD chose TSMC as its chipmaker for all of them.
AMD entered the market as the GPU shortage started to come into focus. Multiple product launches drove up demand at semiconductor companies above their already high levels and the pandemic continued to disrupt the global economy. Although AMD’s cards sold for less on the secondhand market, they were still selling for much more than their suggested price.
Crypto booms
Going into the end of 2020, a boom in cryptocurrency prices would further extend the GPU shortage. Although the boom happened in December, demand from crypto miners has been constant throughout the shortage and the last several GPU launches.
It’s hard to tell just how many cards went to crypto-miners, but the number is high.
Shortly after launch, photos started circulating of mining rigs housing dozens of Nvidia and AMD graphics cards. Similarly, Nvidia’s specially designed mining cards drew in $155 million in revenue in the first few months after launching. This isn’t new for the GPU market either. Nvidia CEO Jensen Huang even recognized the issue publicly when a spike in Ethereum prices increased demand in 2018.
However, a massive increase in the value of Bitcoin and Ethereum in December 2020 drew the attention of industrialists to jump on the mining train. On December 1, 2020, Ethereum sold for around $590 per coin and Bitcoin sold for around $18,900 per coin. By the end of January, Ethereum was up $1,500 and Bitcoin was up to $35,000.
The rise caused more large-scale mining operations to pop up, like one set up by one of Russia’s largest oil producers, entirely powered by oil.
It’s hard to tell just how many cards went to crypto-miners, but the number is high. In the first part of 2021, while the crypto boom was continuing to happen, miners accounted for around 25% of all GPU sales. Like other factors contributing to the shortage, demand from crypto-miners isn’t solely responsible. Given the context, however, it continued to make GPUs extremely difficult to find.
Companies couldn’t ignore the shortage anymore, which is why rumors started circulating that Nvidia was reintroducing the RTX 2060 to combat demand. MSI eventually reintroduced the GT 730, a seven-year-old GPU, to provide an option for builders without integrated graphics, too.
Newegg also set up a raffle-style system called Shuffle, where the company would randomly select participants for a chance to buy high-demand products. The program is still going today, but Newegg often bundles cards with other PC components, bloating the price.
Out of the penalty box — tariffs
At the start of 2021, the GPU market continued to struggle as supply remained low and demand continued to rise. The aforementioned trade war with China was still in full swing, with tariffs taxing goods imported from the country. Originally, graphics cards were given an exception to the trade plan, but that expired at the beginning of 2021.
PC components aren’t normally subject to an import tax, but under the “list 3” trade action that taxes about $200 billion worth of products from China, graphics cards were subject to a 25% import tax.
Cards didn’t immediately jump 25% in price, and given the limited inventory at retailers, the list price of GPUs didn’t matter much. However, some graphics card makers passed the extra cost onto customers. For example, Asus said that it would update its MSRP to reflect “increases in cost for components, operating costs, and logistical activities plus a continuation of import tariffs.”
This is where AIBs come into play. They’re the main source of graphics cards on the market, especially after the initial launch window. Nvidia and AMD are free to set whatever price they want for their cards, but AIBs ultimately have to set a price based on the cost of components, logistics, and any taxes or fees they may be subject to.
At the end of 2020 and into the start of 2021, the price of AIB units grew to unprecedented levels. This is why graphics cards continue to sell for as much as twice their MSRP at retailers, matching the prices on the secondhand market. In other parts of the world, retailers were selling cards for more than triple MSRP.
Nvidia combats crypto miners
Toward the end of February, Nvidia came up with a new plan to combat the GPU shortage: Lite Hash Rate (LHR) graphics cards. Coinciding with the launch of the RTX 3060, Nvidia announced that it would halve the Ethereum hash rate — how much cryptocurrency the card can mine — through a driver.
At the same time, Nvidia announced Cryptocurrency Mining Processor (CMP) cards, which don’t have video outputs and are built specifically for mining. The two steps together were, according to Nvidia, meant to “ensure GeForce GPUs end up in the hands of gamers.”
But they didn’t work. Nvidia released a beta driver shortly after the launch of the RTX 3060 that disabled the mining limiter, and because the fix was only issued through software, users were able to download and distribute the driver to make the limiter obsolete.
When pressed for a similar response, AMD responded with “no.” Following the launch of the RX 6700 XT, an AMD rep said the company’s response was no response at all, saying “we will not be blocking any workload, not just mining for that matter.”
The beginning of 2021 was when the chip shortage that contributed to the GPU shortage started to become clear. Intel’s CEO warned that the chip shortage would last into 2022, and Nvidia said that it expected demand to exceed supply throughout 2021. It seemed the only way out of the shortage was to wait it out.
Nvidia wasn’t down and out, though. The company reintroduced the LHR concept to the rest of the Ampere range to deal with crypto-mining demand. This time, however, Nvidia didn’t solve the issue through a driver update. Instead, it introduced new GPU cores that would limit the Ethereum hash rate at a hardware level.
This was the first step from graphics card companies to get the market back on track, and in the months that followed, a number of factors would help improve GPU supply, if only by a bit.
A shift in strategy
In May, Nvidia launched the RTX 3080 Ti and RTX 3070 Ti, but the launch plans were a little different. Using its partnership with Best Buy, Nvidia released the RTX 3080 Ti exclusively in Best Buy stores, helping deal with bots and the chaotic checkout process that comes with new launches.
Meanwhile, the Ethereum Foundation said it would move from a proof-of-work model — like Bitcoin uses — to a proof-of-stake model. In addition to lowering the energy consumption of mining the coin, the shift would remove GPUs from the mining equation. In its quarterly filing with the Securities and Exchange Commission, Nvidia said the move could increase “aftermarket resales of our GPUs and may reduce demand for our new GPUs.”
On top of that, the prices for Bitcoin and Ethereum started to drop, and China, which held the lion’s share of Bitcoin miners, imposed new restrictions on crypto-mining. Although none of these steps ended the GPU shortage, they helped steer the market back toward normalcy.
And Nvidia was quick to capitalize. Following all of the shifts in the GPU market, a report circulated that Nvidia was taking the RTX 2060 back out of production to focus on production of the latest cards.
The current state of the GPU market
The GPU shortage isn’t over, but it’s in a much better spot. With tools like EVGA’s queue system, Newegg Shuffle, and in-store restocks at Micro Center, it’s possible to find a new graphics card in the U.S.
The situation is improving in other parts of the world, too. Recent months have brought an increase in GPU supply in European countries, driving down the price. In Germany, prices have dropped as much as 50% from their peak at the beginning of 2021.
Still, the GPU market is recovering — it hasn’t recovered. Tariffs and the increased cost of components have pushed AIBs to sell cards far above the marks Nvidia and AMD set. This is the biggest issue with the GPU market right now, especially as Nvidia pushes out updated LHR versions of its cards. For example, some updated RTX 3060 models — which should retail for $329 — are selling for close to $1,000 brand new.
Looking forward, all the GPU market needs is time. Scalpers have become less effective as more cards make it onto the market, and demand from crypto-miners has dropped thanks to falling prices and new regulations. Supply chain issues, which arguably kicked off this shortage, remain the main constraining factor, and it could be years before they’re worked out.
The biggest issue now is overcompensation. As Nvidia points out, it may need to place “non-cancellable inventory orders significantly in advance of our normal lead times, pay premiums or provide deposits to secure normal and incremental future supply.” Manufacturing capacity is valuable at the moment, and companies generally need to buy that capacity well in advance.
With the pandemic winding down in most parts of the world, the computing market expansion may wind down, too. That could lead to companies ordering parts they can’t reasonably sell, which could crash the GPU market.
In that case, at least hopeful buyers wouldn’t have to worry about overpaying for a graphics card.
Editors' Recommendations
Article From & Read More ( A complete, chronological history of the catastrophic GPU shortage - Digital Trends )https://ift.tt/3eQRKPZ
Technology
No comments:
Post a Comment