With the announcement this week from Nvidia of it’s new GeForce RTX 4000 “Ada Lovelace” series of cards it’s safe to say the internet is not particularly happy about Team Green at the moment. Pretty much every tech YouTuber is raging. I will give some thoughts from an editor’s perspective and try to get to the bottom of whether the anger is fair and whether you should be in the market for these new cards.
For many years now, Nvidia graphics cards have been the best performing cards in post production, beating their only rival AMD by some margin. The fact that Apple has moved away in recent years from using AMD cards to using the integrated graphics in their M1 and M2 family of chips has pretty much closed the previously large performance gap between PC and Mac. But savvy Windows based editors have had only one choice – stick with Nvidia.
So why are people angry?
Firstly, the pricing of the announced 4090 and two 4080 variants is just so high. In the US the MRSPs are $1599 for the 4090 (launching October), $1199 for the 4080-16GB and $899 for the 4080-12GB model (both launching November, with the latter having no Founders Edition). In the UK it’s worse with seller Overclockers listing prices starting at £1679 for the 4090 (which comes out at a whopping $1860!). And it will be even worse in various countries around the world. Comparing this to the launch MRSP’s of the 3090 ($1499) and 3080 ($699), there is a small jump for the 4090, but a large one for both of the 4080 models, which will be the more popular cards.
What’s more, most commentators agree that the lower of the two 4080 variants is a 4070 in all but name – the actual core chip is the lesser AD104 as opposed to the AD103 of the 16GB version, it has 20% less CUDA cores and a reduced memory bus and bandwidth – and this makes the price jump harder to take (the 3070 MRSP is $499). People are pretty annoyed with this and I think fairly – it is very confusing for any customer who doesn’t look too closely at the stats.
There is also the fact that the affordable cards that most people want – the 4070, 4060, etc. won’t be seen until next year as has been long rumoured and confirmed by the 3000 series being listed on screen during the presentation, not something they normally do. Not to mention the news that one of Nvidia main board partners EVGA decided to pull the plug on selling Nvidia cards, citing the way Nvidia handles pricing and communication.
Why are Nvidia doing this?
I think there are a number of factors at play here. A few years back when Nvidia were planning these cards, the GFX market was blowing up, with miners and gamers snapping up all the cards as soon as they went on sale. Industry speculator Moore’s Law Is Dead claims that Nvidia designed the next series to go after peak performance even if they cost more to produce. We do know they moved the production onto TSMC’s 4/5nm process away from Samsung’s 8nm process and Tom’s Hardware have reported that it could cost twice as much per wafer, though equally that gets you a lot more transistors on the chip with a higher performance to boot. There have also been various chip shortages and price hikes as well as delays reported over the last year.
Now the market is quite different. Ethereum – one of the main cryptocurrencies for miners – moved last week to proof of stake, which you can read up on if you like, but basically it is no longer profitable mining it, which in theory will mean a flood of used GPUs on the second hard market. And multiple sources say that there is a massive stock of RTX 3000 cards still unsold which Nvidia and partners are desperate to sell.
Nvidia has to recoup their investment, for both series of cards, and the pricing and strategy reflect that. The 4080-12GB might normally be called a 4070, but the likely production cost and the performance jump looks to be so great over the 3070, that it would make sense to up the name and the price. As Nvidia’s Justin Walker has said, “The 4080 12GB is a really high performance GPU… it’s faster than a 3090 Ti, and we really think it’s deserving of an 80-class product.”
Is there anything to be excited about?
For sure. For starters, it looks like there could be a pretty decent generational performance jump here as long you don’t mind the increase in power consumption, with the 3090 listed at 450W (and even higher from board partners), and board thickness (I hope you have 4 slots free).
What catches my eye straight away is the dual 8th generation NVENC encoders. If you’re like me and send clients edits for review on most days, time spent exporting is a big deal.
Before I got my 2000 series Nvidia card, I swore by only using software encoding for export, even though it was slower. Pixel peeping showed it was better than hardware encoding for me. But it depends on your hardware. RTX 2000 brought a generational leap to the Gen7 NVENC encoder and it’s indistinguishable from a software export and much faster. I’m excited to see what Gen8 brings and even more whether NLEs can utilize both encoders.
Nvidia claims cutting export times by up to 50% and they support AV1 encode and decode which is a bit sad for Intel as that was their USP for the yet to launch Alchemist series.
All in all I’m pretty excited about upgrading as I skipped the last gen due to unavailability.
What about AMD?
November 3rd is the date we are waiting for now for the next gen cards from Team Red to see how they compete. In the world of Post Production, AMD’s Radeon cards have yet to impress, but will this be the year that changes?
— Scott Herkelman (@sherkelman) September 20, 2022
What about Intel?
It’s safe to say that Intel expected to bring their cards to market a long time ago, but they’ve certainly had plenty of challenges. As you can see in the diagram, they are only targeting the lower price points – and even though they are going to be left in the dust by Nvidia’s latest cards, they still might perform quite well for those on a budget. We should be seeing them properly released fairly soon.
Bookmark Puget System’s site for the most rigorous tests that are available, which should appear on or near the embargo date (the date when all the reviews can come out). Other channels like Gamers Nexus, LTT & EpoxVox should do some benchmarking too. It will be very interesting to compare the price/performance for the 4090 and 4080’s to the previous gen cards, especially the healthy second hand market. I suspect some of the previous gen cards like the RTX 3080 and RTX 3070 will be a killer deal. But who wants last gen?