The Uncertain Future of PC Gaming... (plus my opinion on Valve's Steam Deck)

 

Three things occurred today that kickstarted my thought processes and got me a little worried about the future of PC gaming. First off, I listened to a really good episode of the TECHonomics podcast. The second thing was that AMD pre-released FSR, along with some documentation. Finally, Valve released the Steam Deck for preorder.  Now I'm feeling very uncertain about the future of PC gaming...

Pricing out the competition...

I've had serveral conflicting thoughts about pricing over the last year or two. On the one hand, I expect to see price drops for GPUs over the coming months due to cryptocurrency losing value and because supply will finally meet demand. I also expected a big increase in GPU performance per [insert currency unit of your choice here] in 2022 because of the pending implementation of multi-chip GPU packaging, resulting in a lowering of price for each class of GPU.

On the other hand, I've been saying that AMD have been pushing up the cost of gaming for both CPUs and GPUs because they are not providing downward pressure on either Intel or Nvidia. This did come as somewhat of a surprise to me given Nvidia's history with the RTX 20 series and past Titans and other "high-end gamer" cards. The thing is, I was always looking at the systemic creep from the wrong end - AMD's relative performance and market share meant that they were more competitive in the low-to-mid range and that's where the price increases have mostly come from.

The cost of buying the graphics card that met the average performance for games released each year targetting "recommended" hardware requirements in games released that year... AMD has been pushing things higher...

Unfortunately, now that AMD are competing in the higher tiers of cards, they're taking that same philosophy with them and pushing up the price there too. The guys over on the TECHonomics podcast were talking today about the almost absurd manufacturer suggested pricing of the current generation of graphics cards being really high and how AMD should have priced them $100 lower.

Now, I'm maybe not as extreme as they are in their analysis. I would only suggest a $100 reduction in the RX 6800/RX 6800 XT and RX 6900 XT pricing for the same reasons they provide in their discussion. On the other hand, I think that the RX 6700 XT should be reduced by at least $50 (and this will probably apply to the RX 6600 XT as well given its rumoured $399 price point).

I can see that it's becoming more expensive to make GPUs for various different reasons (including shipping, substrate, and waver costs. However, it's not becoming that much more expensive to make them though I can totally see AMD's side of things where they can sell any graphics card that they put out there and they may as well make as much money on each individual card as they can because they're unable to grow marketshare due to their supply issues... 

Wait, what's that?

Oops...

Well, I guess people have a problem with buying over MSRP all of a sudden... or it could be releated to the fact that cryptocurrency has taken a bit of a dive and the performance of AMD's 6000 series cards was already pretty low for that function. Now supply of Nvidia cards is increasing, the pressure on AMD's stack is also being releaved and, quite frankly, normal economics are coming back into play because AMD does not have the same general consumer awareness or value as Nvidia does!! Thus, they cannot normally price their cards at the same level - even ignoring software support differences the two product lines have between them that make Nvidia the preferred choice for creators and/or streamers and/or graphics professionals.

So, AMD pushed their cards' prices higher to meet Nvidia's, their middlemen and partners inflated prices to match the crypto demand and now their partners in the ecosystem are finding that they bought a lot of overpriced stock that they can't get rid of even when they reduce the price to only 160% of the MSRP. This will push prices down eventually - companies will need to take a bath on this, only it's not clear which companies will. Personally, I think the retailers will end up losing out.

Unfortunately, though, this still does not bode well for graphics card pricing for the future generations where we're expecting am almost doubling of performance for both Nvidia and AMD but with possibly a doubling of prices. Which... is where we come to the rest of the TECHonomics discussion that a focus on the high end, higher profit margins and lack of competition will actually kill PC gaming ecosystem because no one will be able to afford GPUs.

I've spoken about this many times in various YouTube comments and on Twitter (no links, I'm lazy) but I really feel that there's more than just price increases to worry about here. If this comes to pass, there's a two pronged assault on PC gaming that will occur:

1. GPU prices will get higher, pricing the low to mid-range consumers out of the market.
2. Developers will be unable to bring new technologies to bear.

Now, that first point is obvious - expensive GPUs; the low-end of the market fails to progress past the heady days of 2016, when the GTX 1060 6GB and RX 480 8GB were released for $250.

The second point is what happens when the first comes to pass: developers cannot target increasingly new architectures, features and performance envelopes because there isn't enough of a market to support their development. This results in stagnation of games' ability to technically innovate. 

No, VRS, no fancy upscaling, no ray tracing.

Developers are potentially left in a desert of old hardware that cannot be pushed further. In an ironic twist, the consoles from Microsoft and Sony are both more powerful than most DIY gamers' rigs are.

This leads to the second item on today's list:

Behold, the saviour of gamers* everywhere... *though mostly those playing at 4K on high-end GPUs, the same as DLSS...

Leaving the past behind...

AMD, the defenders of all that is good and pure in gaming, finally released preview code and examples of FidelityFX Super Resolution (FSR) to the world, as well as some light documentation for people to peruse. However, during the process of familiarising myself with that documentation I came across this snippet:

Huh, matching performance, seems okay... but wait, does this mean a new 70 class card from AMD is no longer a 70 like it was in the old stack?

I actually was more interested in the time cost for implementing FSR across the different resolutions but then something caught my eye: Why is AMD comparing performance tiers against different generations of GPUs? And why is AMD saying that its current 70 series card is not equivalent in the performance tier to the prior 70 series card?

Now, here's the thing - there's nothing wrong with the GPU comparisons here: all these pairs of cards from Nvidia and AMD are similar in performance (actually, AMD wins out on each comparison by a small margin) but what confuses me here is that they are algining the performance of the cards with a specific performance tier in the market/product stack.

The problem with this is that the tier of cards moves around based on the stack in question. HD 7970? High end. HD 7870? A performance card. HD 7770? Mid-range. RX 480? High end. RX 470? Mid-range... and this is despite the fact that the RX 400 series didn't compete with Nvidia at the high end. The "70" or "7" always denoted a mid-range card. Okay, so historically that seems fine. The nomenclature even vaguely matches Nvidia's in terms of place in the product stack, if not in actual numerical performance.

Here's where things get suspect.

The RX 5700 XT was the highest end AMD card last generation. This generation, the RX 6700 XT is the mid-range card. However, that's purely a naming scheme sort of thing - the 5700 XT could have lain in the same position in the stack as the 6700 XT if AMD had released the higher-end cards that they had actually developed - but decided not to. The 5700 XT duked it out with the RTX 2060 Super and the RTX 2070. It is in the same position now as the 6700 XT is with regards to the RTX 3060 Ti and RTX 3070. It literally occupies the same spot.

So, why is AMD not classifying them in the same tier? Obviously, AMD haven't released their RX 6600 XT yet so they can't reference that. In this sense, I believe that AMD is tipping their hand and stating, in an oblique way, that the 6600 XT will have a performance target of the RX 5700 XT, as the RTX 3060 has the performance of the RX 5700 XT/RTX 2070/RTX 2060 Super.

But this belies the actual situation - performance moves with each generation. By moving the performance tier of the product stack upwards, AMD is trying to reposition its pricing structure higher without actually providing better performance per [currency unit]. This is bad because it stops consumers from being able to upgrade to the same performance tier that they would have, historically.

Further to this, we have the last piece of news:

This beastie is slightly shorter than two Sony Vita's in length, or if you prefer, quite a bit larger than the Nintendo Switch...

Futureproofing...

Valve's Steam Deck looks sleek and sexy. Build quality looks great and the overall design, while not perfect, is very, very good. The Steam Deck also doubles as an honest to goodness plug-in desktop computer - which is absolutely amazing and which is why (aside from Valve apparently taking a loss on each unit, even at the stated prices) the price per unit is relatively high. I doubt they get the discounts of huge scale that Sony and Microsoft do! 

What concerns me, though, is the internals.

I think Richard Leadbetter said it well in the Digital Foundry Eurogamer article covering the release:

"The fundamentals are there to ensure that games run - but how well are they going to run? We are in a time of cross-gen transition in the industry: if the titles of today run fine, what about the games of tomorrow?"

Really, for a handheld, the console is very impressive and it could serve a really nice hybrid function in terms of gaming/work/entertainment by being able to connect to monitors/TVs and be portable in a way that only the Aya Neo really is. However, I feel that 4C/8T is not ideal for a gaming system in the here and now and definitely lacklustre in the years ahead.

That's unfortunate because, as per my research, the recommended specs for gaming at 1080p are moving away from 4C/8T to 6C/12T and the data gathered so far this year are supporting that shift - a shift towards the average performance of a Ryzen 5 3600 in 2022.

Obviously, this year is not yet over so the data collection is incomplete but already we can see that the average and mode cores are 6 whereas the average threads are 10 but the mode threads are 8. This discrepancy reflects the legacy of Intel's lack of multithreading in their i5 series...

The Aya Neo itself is very similar in design and specs to the Deck, including the quantity and speed of the LPDDR memory, but is superior in two factors - it's a 6 core Zen 2 design and it comes in only 512GB and 1 TB NVMe capcacities... I feel like that's a great combination, despite being held back by the 8 compute units of the Vega architecture at 1.5 GHz in comparison to the (max.) 1.6 GHz of the 8 RDNA2 CUs in the Deck. Unfortunately, the Neo is inferior in all other regards because it's priced at double the cost of the Deck. 

In this light, it's clear that Valve have done the best job they can do with the money they have available. Trying to shoehorn in two cores/threads into this device probably would have just cost them more money, drained battery and increased heat output without really helping in terms of longevity. Especially because the Deck has a relatively low CPU clock frequency of 2.4 GHz (and boost, 3.5 GHz), well below that of the Ryzen 3300X and also of the Neo which tops out around 4.0 GHz. In fact, it'll be peforming well below the R5 3600 in single threaded applications due to its lower clock speeds. This isn't good in terms of gaming...

The 8 RDNA2 cores are also clocked relatively low so we don't really have a good handle on how well this unit will perform when combined with the LPDDR5 memory but it will likely be higher than the Ryzen 5 4500U that the Neo is based off of, which will likely put it around 60 fps at low to medium settings on most current and past games.

So where does this leave us?

Conclusion...

I'm feeling like the pricing structure of discrete hardware is going to push a lot of players away from PC gaming. At the same time, the amount of hardware available to play games on low to medium settings is just exploding, for not unreasonable amounts of money. That means that a lot of players are going to have lower performing systems, which means that developers will hard to keep their hardware requirements low in order to be able to court sales from those market segments.

From my perspective, this is bad because we were finally moving out of a period of 10 years of stagnation into a new era of innovation and technological advancement. It's unreasonable for us gamers to expect a 10 year-old PC to still play modern titles without a problem but if the userbase for newer equipment isn't there then developers can't move on either. 

Conversely, the Series X and PS5 are quite high powered devices and developers will certainly target them for the majority of games. This is looking more and more like a pending bifurcation in the gaming market - those that have very high-powered gaming systems and those that have very low powered gaming systems with very little remaining inbetween. While you may argue that was always the case, it was not usual that purchasers of *NEW* systems were only able to purchase low-end hardware without spending a lot of money. These consumers are getting hardware that is equivalent to years' old technology instead of the modern designs they should be able to purchase.

I believe that this combination of events has the potential to hold back the advancement of PC gaming against the strides the segment has made over the last years... and I'm not liking the possibility! I want new experiences, new games that were not possible on old hardware, if this trend continues, we won't get those experiences.