So, What's Next...? The rate of advancement in gaming performance... (Part 2) Is PC gaming dying?!

A couple of years ago, I looked at the rate of advancement of GPUs over time. It's the only way of tracking how the user experiences technological advancement. Recently, we've seen a lot of movement in the industry of various commentators reaching the same conclusions I did, way back when... Or, maybe I'm a particularly pessimistic person and these thoughts are the logical conclusion that people will come to, given enough time and data.

But given that I was ahead of the curve, I feel I have a little more to give than even the current trends in discussion...

Recappping the past...

To briefly summarise where we are, let's look at how things have progressed since the last time I covered this topic and I can do so in just a single sentence: we're shipping fewer dedicated GPUs (dGPUs) than even in 2021.

AMD's marketshare on new GPUs has dropped, and that's not news... [Tom's Hardware]

Now, these charts are not a 100% fair comparison. Back in 2006-2010, iGPUs were not the predominant force they are in the modern world and laptops constitute a larger proportion of the market than was even dreamed about in that time period.

According to Canalys, laptop shipments comprise of 80% of the market. So, we should expect some level of drop to have occurred over this time period as laptops became the predominant force. The problem with this accounting is that it completely ignores direct-to-consumer data, meaning that the number of desktops made by consumers is missing from this dataset. The other aspect to consider is that dGPUs can be attached to *existing* desktops and so, this explains the difference in yearly dGPU shipments and "PC desktop" shipments from system integrators.

The total market in 2024 was around 68 million units, with desktops making 13.6 million of that number... [Canalys]

In case the graphs don't really tell the story well-enough, the problem is that PC gaming is stronger and more popular than ever. We're talking a ~40% increase in PC gamers since 2008. The existing PC gaming install base is huge and the stagnation we've seen in hardware is only adding to this increase as older systems remain relevant and do not require as many upgrades to be able to play a majority of games. This means that GPUs are not only being sold to new builds, they're sold to generations-old system owners, which means that the Total Addressable Market (TAM) is actually very large.

We've seen this disparity between supply and demand play out in real-life over the last half a year as both AMD and Nvidia stopped producing their last gen higher end GPUs, to scenes of completely empty online and physical stores for these products. It's not a country-specific phenomenon, it's worldwide. 

It's simple: There are not enough GPUs produced to service the market.

Overall, we're now shipping around 40% of dGPUs per year compared to 2006...

And that tells us that, despite laptop dominance of the full system (complete solution) market, the reduction in dGPU shipments is outpacing the increase in marketshare of all-in-one solutions like NUCs and laptops in the consumer space. 

If GPUs are almost instantly out of stock at inflated prices after a couple of months stoppage in supply, it means that the market was not anywhere close to being serviced. 

If new GPUs can launch and instantly have price hikes to multiple tens of percent above MSRP, it means that the market is not being serviced.

If GPU makers can consistently market a lower-end product as a higher-end one, then it means that the market is not being serviced.

If GPUs already on the market do not lose their value over their lifespan, then it means that the market is not being serviced.

As I said back then, the performance uplift of the 40 series was the smallest since I started trending (i.e. the GTX 9 series)...

As I mentioned earlier, there is stagnation in GPU hardware but the issue with fewer dGPUs being manufactured year-over-year in a market that is larger than ever is that we have attrition. The rate of release of new GPU generations has slowed to a historical crawl. That's expected based on the slowing of technological advancement in the lithography sector, meaning that new manufacturing process nodes are taking longer and are more expensive to reach maturity. Additionally, low-hanging fruit in terms of algorithms and hardware architecture design have already been plucked and, outside of a complete redesign in how computation is performed and data is transmitted, there are very few changes to make for marked improvements*.

 

*I say this but it seems RDNA4 has done just this and fixed a load of bugs which have resulted in some of the largest performance increases per unit of compute that I've ever seen...

All of this means that as GPUs in consumers hands age-out, die of malfunctions due to heat and silicon degradation, or accidents, and because they no longer meet the performance requirements for modern games (or the particular user's expectations/preference), the pool of "viable" dGPUs already in the market shrinks over time.

Given the size of the PC desktop gaming market, and the decreasing quantity of dGPUs shipped per year, it is not unreasonable to realise that, at some point, there will be an intersection as the demand outstrips the supply. This would cause a tipping point whereby runaway price inflation would not only occur at the point of sale but also in pricing structures from the manufacturers.

I believe we have already passed this point.

The high prices of GPUs are not due to increased prices in manufacturing*, they are not due to competition with cryptocurrency miners, nor AI farmers**, they are not solely due to scalping***. The RTX 40 series and RX 7000 series did not see price drops as the generation progressed. That leaves one reason - chronic undersupply...

*They have increased but not by any proportional amount close to what we see at point of sale.

**There are different products which cater to those markets.

***The base MSRP price is increased per tier of performance before we even get to the point of scalping.

I've never seen a product have such a consistent price - and it's not just this model... [CamelCamelCamel]

Expectations...

Simultaneously, with the lack of hardware and hardware uplifts per price point, we have a new paradigm shift in that good quality 1440p high refresh rate monitors are becoming affordable. This means that 1440p is now the base resolution, instead of 1080p. GPU performance at 1080p is good for graphics card comparisons but a large (and growing) proportion of PC gamers will be moving to 1440p as their expected GPU requirement. Additionally, 30 fps is no longer acceptable and 60 fps is a bare minimum. Gamers want and expect higher refresh rates to be supported by their GPUs, especially when they are costing so much to purchase.

This has two repercussions: a greater than generational uplift for each GPU tier in performance; and increased VRAM requirements.

Now, at the low end of the spectrum in the not so cheap $300-400 range, GPUs are worth less than they were and are at the point of being relegated to legacy gaming and low-end gaming. It used to be that the mainstream GPU would get you good performance at the mainstream resolution and medium to high quality settings. Yes, you can look at the Steam Hardware Survey but this data is highly affected by markets which are not focussed on the new hardware. Sure, existing gaming PCs have 1080p screens, but what are they going to buy when they upgrade? What are new PC builders going to purchase? 1080p? Or a nice cheap, WAY better 1440p IPS high refresh rate screen? I know what I chose... and if you're spending more than $300 on a GPU, the average person probably will, too! Hell, you can get them for around $200, now... That's cheaper than the crappy $300 mainstream GPU we're talking about!

The RTX 5060 is bringing consumers all the way back to prior eras when Nvidia wasn't giving enough performance to meet then-current demands...

I recently posted on Twitter that Nvidia's 70 class has seen a 130% performance uplift since the RTX 2070 to the RTX 5070. Meanwhile, the 60 class has seen an improvement of just 45%. If we saw that same level of improvement in the 60 class, the RTX 3080 would be the level of performance we should expect... somewhere slightly faster than an RTX 4070 12 GB. The RTX 5060 doesn't even meet a consistent RTX 3070 level of performance...

This brings us to the uncomfortable question: If there are no (or very little) generational performance uplifts in the price range that the majority of gamers buy, what do consumers and developers do?

Well, you get behaviour like this...

Confusing causation with choice is a big problem, here...

Playing at 1080p for many years, including to this day on my TV, I have plenty of reasons to want more than 8 GB VRAM. Graphically demanding titles require more, they did back in 2020 when I was playing Cyberpunk 2077 on my RTX 3070 with raytracing enabled. People aren't buying 8 GB graphics cards because they want to, they are doing it because there isn't an alternative.

So, coming back to the awkward question - consumers keep buying the graphics cards at the prices they can afford, and developers keep advancing games which require more VRAM because monitor and TV ownership is increasingly moving to 1440p and 4K resolutions. 1080p is the 900p of 2012. It's dead, on its way out - the bargain bin of resolutions. Look, Nintendo have chose 1080p as the handheld screen resolution for the Switch 2. You know, Nintendo, that company that likes to live on the forefront of technol-oh! What's that you say? They don't?!

Yes, in 2025, 1080p is now "handheld" resolution.

This is all bad for gaming, as a whole. There's no upside. We have almost complete stagnation - both in VRAM and in graphics performance in the largest segment of the market of hardware buyers.

Consequences...

All of this is leading to a very unhealthy gaming hardware market as well! Continued high prices on new parts, paired with the lack of performance uplift is keeping used PC parts high as well. That means there's no trickle-down effect from new gamers upgrading and so less affluent gamers can't game like they could in times past. 

The low supply means that AIBs suddenly have lower sales and revenue numbers, and this will necessitate higher prices on their side as they need more profit per unit sold in order to keep their employees in their jobs.

Developers will suffer more complaints and (potentially) lower sales as players can't get cards that perform well or cards that can even manage to play the game (e.g. is ray tracing is a requirement).

Reviewers will get fewer views on their content and so less revenue, potentially leading to fewer independent outlets being able to operate in the hardware review space. 

This also comes hot on the heels of the recent shenanigans Nvidia pulled with the RTX 5060 by denying reviewers the ability to give information to the consumers, meaning that we're likely to see negative effects on both reviewers, "review" sites which capitulated to Nvidia's demands, and users having less money in their pockets to purchase hardware that is worthwhile...

It's a pretty negative-feeling time in the industry (at least from my perspective).

There are some positives, though. Just like the Intel/AMD CPU wars, the stagnation in performance across Nvidia's entire stack is allowing AMD to catch up with their GPUs* but the stagnation at the bottom of the product stack is also giving Intel a chance to catch up and to release a decent product at a decent price.

Now, this is a pretty slim chance but it's a potential silver lining on the horizon...

*If only they had some competitive pricing to match!

Moving Forward...

I'd like to see both real competition from Intel at the low-end of the graphics card market, and better products from all three manufacturers at the low end. I'd prefer that they made better products across the stack but let's focus on where we need the most improvement...

However, until graphics cards are manufactured in enough supply to satisfy the market, things will never be fixed. We need a total output across all manufacturers of around 50 million units per year to correct things a little and the current trend is heading in the totally opposite direction.

Previously, I had been cautiously optimistic that Nvidia would push silicon back up a performance level if AMD brought good competition. Well, AMD did and Nvidia didn't... though AMD hasn't brought any price competition so there's no pressure for Nvidia, at all...

I don't see a bright future for PC gaming if we keep on this trajectory. I've advocated for developers to target the performance level of the RTX 3070 in prior years and I will continue to leave this advice unaltered for the time being because that's all the performance that most consumers can access...

Coming back to the article sub-title: Is PC gaming dying? No, it'll never truly die but we're looking at a contraction and it's not looking very healthy...