The fallout from Nvidia's RTX 4080/4090 reveal last week has been intense. We saw the reveal of the RTX 4090 - a phenomenal achievement in performance - backed by the debut of pioneering new technology like DLSS 3 frame generation, the remarkable RTX Remix and a new version of Cyberpunk 2077 that is to all intents and purposes running with full path-tracing. The technology is there to move PC gaming onto a new level but reaction to the keynote has instead been dominated by RTX 4080 pricing. Two RTX 4080s of very different spec levels are coming - priced at $899 and $1199 respectively - and based on reaction to Nvidia's own benchmarks, there's a performance gulf between them and an 'fps per dollar' deficit compared to prior gen-on-gen leaps.
There's still much we don't know. Nvidia's numbers cover a mere handful of games - so the actual value of RTX 4080 in its dual guises can only really be ascertained via hands-on testing with a much wider array of titles, plus the take-up of DLSS 3. What is clear, however, is that the nature of an 80 series card has radically altered. Looking back to RTX 3080, it used a cut-down version of 3090 and 3090 Ti silicon, the GA102 processor. Both 4080 cards use different, smaller processors - AD103 and AD104 vs the much larger AD102 in the 4090 - and so the expectation is a significantly wider delta between 80 class and 90 class performance this time around. The differences in RTX 4080's compute power and memory bandwidth are remarkable, both against the 4090 and each other. With that in mind, it is difficult to understand why two quite different products both receive 4080 naming. Factoring out DLSS 3, only the RTX 4090 seems to offer a substantial gen-on-gen improvement, based on Nvidia's own numbers.
Part of the discussion in DF Direct Weekly this week covers the various reasons why this is happening. The backlash puts the blame squarely on Nvidia for over-pricing its products, for 'rebadging' a prospective RTX 4070 as an RTX 4080 in order to deliver an 80-class product for under a thousand dollars. However, last week, Nvidia boss Jensen Huang delivered a stark message: that cost reductions on performance, or the same performance for half the cost with a new generation are a thing of the past.
Backing up Nvidia's position, Xbox architect Andrew Goossen spelt this out to us in plain language two years ago - the cost-per-transistor is not reducing at speed anymore. Microsoft saw no route to cost-reducing Xbox Series X effectively, hence the launch of Xbox Series S. If existing tech cannot be cost-reduced, it stands to reason that more performance will hike up prices. Meanwhile, in recent times, the PlayStation 5 has increased in cost, despite using a smaller 6nm chip. Beyond all of that, we've not even factored in out-of-control inflation and its impact on GPU prices.
- 00:00:00 Introduction
- 00:21:59 News 02: DLSS 3.0 revealed!
- 00:42:19 News 03: RTX Remix modding tools impress
- 00:47:33 News 04: RT Overdrive patch teased for Cyberpunk 2077
- 00:52:38 News 05: Xbox DRM scaled back for disc games
- 00:58:48 News 06: NTSC/PAL switching for PS+ classic games introduced
- 01:02:05 News 07: New Star Ocean demo suffers performance woes
- 01:06:21 DF Supporter Q1: Given the leak of the Tegra T239, a likely Switch SoC, when is the new Switch going to be released?
- 01:12:32 DF Supporter Q2: How much difference does memory speed make in games?
- 01:14:07 DF Supporter Q3: Do you think Konami will go back to developing new big games?
- 01:16:51 DF Supporter Q4: What are your thoughts on the GTA6 leak?
- 01:22:49 DF Supporter Q5: Rich, did you imagine yourself doing this sort of work 20 years ago? And what’s one game you need to have a physical version of?
There may well be mitigating factors to explain the pricing but ultimately, the market will decide what to make of the RTX 4080. The reaction is such that there's undoubtedly a golden chance for AMD to make an impact with its upcoming RDNA 3 graphics line. If Nvidia is indeed over-pricing, AMD has the opportunity to significantly under-cut its opposition. If it doesn't meaningfully offer a much better deal, we should accept the probability that big silicon commands big prices going forward, just as Nvidia suggests. In its favour, AMD is moving to a multi-chiplet design, as opposed to the larger monolithic processors as found in RTX 4000 products. Using chiplets, elements like memory controllers can be entirely separate processors, fabricated on cheaper processes and potentially lowering cost. It's a strategy that has paid off handsomely for AMD in the CPU space - so Team Red may yet deliver. Right now though, the RTX 4090 will remain the preserve of the super-affluent, delivering far superior performance than RTX 3090, even beating the 3090 Ti into a cocked hat, by quite some margin. The message is clear though: the biggest gen-on-gen upgrade with Nvidia's new line of graphics hardware comes at a price: $1599, specifically.
As discussed by myself and the team in this week's show though, the RTX 4080 pricing storm has diverted attention away from some incredible stuff. Portal RTX is built on the new RTX Remix tool, effectively opening the door for many older games to receive full path-tracing 'remasters'. Portal itself is a brilliant example of what Remix can do: classic gameplay never ages, but now the title looks breathtaking. DLSS 3? We'll have much more on that soon, but with an effectively path-traced Cyberpunk 2077 running at perfectly playable frame-rates, the potential here is astonishing: PC can now scale beyond consoles in offering a totally transformed experience as opposed to the usual staples like unlocked frame-rates, ultrawide resolutions, tweaked settings etc. All of this has been overshadowed by RTX 4080 specs and pricing.
While the RTX 4000 reveal dominates our show - accounting for almost an hour of run-time - there's still more to discuss: John is happy with the recent easing off with Xbox DRM, but isn't so happy with the new Star Ocean demo's performance on PS5, nor its general showing on Xbox. And while it's great to see 60Hz support make its debut for European PlayStation 1 classics, it looks like the first title is fundamentally flawed. DF Supporter questions? Well, the Nvidia keynote dominates, but we also discuss the potential T239 processor in the new Switch, Ryzen's relationship with memory speed, the GTA6 mega-leak and Konami apparently getting back into games! And the best question of all? 20 years ago, could I have guessed what I would be doing today with Digital Foundry? Supporter Q+A is part and parcel of the DF Supporter Program and you can get involved - join us!
Will you support the Digital Foundry team?
Digital Foundry specialises in technical analysis of gaming hardware and software, using state-of-the-art capture systems and bespoke software to show you how well games and hardware run, visualising precisely what they're capable of. In order to show you what 4K gaming actually looks like we needed to build our own platform to supply high quality 4K video for offline viewing. So we did.
Our videos are multi-gigabyte files and we've chosen a high quality provider to ensure fast downloads. However, that bandwidth isn't free and so we charge a small monthly subscription fee of £4.50. We think it's a small price to pay for unlimited access to top-tier quality encodes of our content. Thank you.Support Digital Foundry