Availability is starting to open up, and more people are getting their hands on Nvidia’s RTX 3000 series cards and AMD’s Radeon RX 6000 series cards.
As a business, the graphics card market has been a dumpster fire on and off for the last 4 years, as surges and drops in cryptocurrency lead to people picking up cards for non-gaming purposes, then reselling them as new generations come out. However, this is but a part of the story – one that largely involves the greed of the companies involved.
Nvidia has been a near-monopoly on gaming GPUs. Now, I have to say “gaming GPUs” and not “graphics” because Intel’s integrated graphics are still the number 1 graphics in use in PCs around the world, but if you get more serious about gaming, you probably have an Nvidia GPU or have had one over the last decade. Nvidia has generally been out in front on performance, has sold their featureset better and used exclusivity of features like DLSS as a cudgel to bring people in, and is less dogged by memes about hot and loud cards or poor drivers – no matter how anecdotal both of those points are. AMD’s brilliant comeback with their current RDNA2 lineup is one of the better stories of the last year, as Nvidia has desperately needed real competition and having a second very strong option in several price brackets is a good thing for gamers.
However, there has been a problem.
Of course, I’m talking about the silicon shortage. The difficulty of obtaining any new, exciting transistor-based product has been well-documented – from the PS5 to the Ryzen 5, everything is just hard to get. That is getting better now, as production matures and demand from crypto-miners starts to ease, but both AMD and Nvidia are riding the wave in a few disappointing ways.
Firstly, the obvious one to me – Nvidia’s lineup refresh, the 3080 Ti and 3070 Ti, both engage in a sort of bonkers price gouging for the relative performance gain over their non-Ti equivalents. The 3080 Ti is by far the worse, at $1200, a full $500 over the original MSRP of the 3080. The card offers 2 GB more VRAM, a faster VRAM solution through a wider bus, and more processing power – coming in around 10% faster for a 71% increase in price! The 3070 Ti isn’t nearly as bad, but it still asks for a sizeable chunk of change for a non-equivalent performance increase.
AMD’s lineup has generally been better in this regard, but they remain unable to keep cards in stock for long compared to even Nvidia, and they’ve leaned pretty hard on increased VRAM that largely isn’t useful today and is of questionable value into the future, even if having a 16 GB graphics card is theoretically cool. What’s worse is that as the product stack winds down into lower costs, they’ve made compromises compared to the cards they replace – the RX 6700 XT has less memory bandwidth than the 5700 XT it replaces, which it sort of makes up for in the Infinity Cache solution, which is useful but not as universally reliable as more raw bandwidth.
Both manufacturers, however, are failing the market in a larger way.
The majority of PC gamers don’t buy cards like 80 series Nvidia or 800 series AMD – they buy low-to-mid range cards. Nvidia’s biggest sellers in most lineups have been their 60-series cards, and AMD’s most popular generally their 600-series cards. These cards usually offer great performance, decent amounts of VRAM, lower power consumption with no transient spikes that will knock an unprepared power supply out, and lower price tags. In the past, these cards usually filled the $200 pricepoint, and they did so very, very well. Over time, both manufacturers have pushed the price of these cards up ever higher, with Nvidia starting at the 1060 to move the 60-series to $300+ by offering a Founder’s Edition version, before the RTX 2060 finally just flatly started at $300. They have released an RTX 3060 – good luck getting one, but it is released – and that card starts at $329, with the price the card materials from Nvidia are sold to manufacturers at making most 3060’s more expensive even assuming the market isn’t eating the supply alive.
AMD, meanwhile, has no 600-series card out as of yet, and even the RX 6700 XT, which replaces the price-jacked 5700 XT which was $400, comes in at a hefty $479. Third-party cards are generally available near this price point, which is an improvement over Nvidia, but that base price is $150 higher as-is, so…not that great anyways, eh? In the last generation, AMD’s RX 5600 XT was price-jacked to $279, which made it a bit hard to recommend, and with no announcement of 6600 XT’s made as of yet, we can only speculate what the price would even be.
To me, this is the biggest failing of this generation. Sure, if I am to be fair, if a silicon wafer costs you the same no matter what to manufacture, then you want to have your profit-leading products be what gets churned out, but at the same time, this feels like such a disappointment. Right now, if you want sub-$200 graphics, you have to buy a GeForce GTX 1650 lineup card, with prices around $160. That’s not a bad card at all, but at the same time, you are locking in nearly $200 on iffy tech – not to mention that the vanilla 1650 doesn’t even have the updated Turing hardware encoder for NVENC, meaning if you are a streamer, you’re losing out on a big quality upgrade (the 1650 Super does have the Turing encoder block for a $10 boost in MSRP). There’s no new product out that comes anywhere close to that pricing, and given the current issues, there likely won’t be for a while either.
And the messed-up thing is that even if you decide to give up and go with console gaming, bad news there – getting a PS5 or either Xbox Series console is still a stretch and a half, and given their higher relative value for scalping, they get hit a bit harder by scalpers too.
But I miss the days of my youth, where a great GPU could be had for around $200, and the only compromises you were really making were max resolution and some bells and whistles like coolers, card design, and the really system-intense old modes of anti-aliasing where the entire screen had to be sampled and adjusted, consuming a ton of GPU horsepower.
The Future Ain’t So Bright
Right now, the market is at an effective standstill as COVID impacts start to abate but other impacts hit. Taiwan was suffering from a historic drought until earlier this month, which impacts the ability of chip manufacturing done at TSMC, who is responsible for AMD’s entire product stack (including console SOCs) minus a couple older GPUs, CPUs, and the IO die on modern Ryzen, and most of Nvidia’s product stack except their newest RTX 3000 series GPUs. Crypto impact is beginning to shrink, as mining becomes harder as intended and Ethereum, the main currency driving the run on GPUs, has again made noise publicly about moving to a proof-of-stake model which would change the paradigm substantially. However, as the positives begin to help supply along, there are negatives.
Chief among these is that the market is still starved for cards, so gamers will still be fighting for GPUs for a while yet. AMD and Nvidia both have no incentive to reduce prices in the current state of things, so they’ll likely continue to milk customers with cards that are, to be fair, quite good, but remain overpriced. Here in the US, tariffs implemented by the prior presidential administration have yet to be lifted, and there are few public signals that they will lift anytime soon. This has tremendous price pressure, which is being passed on to consumers, as most graphics card manufacturers made price adjustments for this very reason in early 2021, such that the MSRP of most cards launched prior to that point is unlikely to be hit anytime soon.
Intel is getting ever-closer to launching their Xe discrete cards, labeled as “DG” cards. The first version, DG-1, is out already in prebuilt systems and a few other forms, and it is…fine, but nothing special. It basically takes Intel’s current Xe graphics from their newest CPUs and puts just that part on an add-in card. Thus, it isn’t spectacular for gaming, but future cards are promising to get much, much closer, with purpose-built silicon dies that are much larger, faster and larger memory pools, and better overall configurations. AMD is about to release a set of Zen 3-based Ryzen APUs for the desktop market that will allow you to buy a fast modern CPU with a pretty good GPU right inside, one that, in recent testing from Gamers Nexus, beats Intel’s Xe graphics by a reasonable margin. In a sad way, until the GPU market moves to a better place, these parts may actually be the best hope for new systems short of buying a whole prebuilt with a newer discrete graphics card in it.
As long as this remains the state of play, we’re unlikely to see either GPU manufacturer make any moves to lower-stack products for a while. AMD has been teasing the RX 6600 lineup since CES 2021 in January, although only in the most recent driver updates have references to such a card shown up. The GPU product used for that card is currently being used in AMD’s professional Radeon Pro W6600 card, and the same silicon is also being sold to Tesla for their new generation center-console systems in their cars. Meanwhile, Nvidia has had rumors of both an RTX 3050 Ti and 3050 in the works for a while, and such parts are in use in laptops currently. However, short of a certification submission to the Eurasian Economic Commission from a board partner in May, no mention of desktop versions of these GPUs has popped up at all.
Of course, while it is very not-great that these companies are holding these releases, it also is how things function. Do you make two Geforce RTX 3050 dies that generate maybe $200 in profit, or do you sell one RTX 3080 Ti die that generates 3-4 times that value?
It is how it is supposed to work, and yet, it remains disappointing.