Thursday, November 16th 2023
NVIDIA Halts Production of RTX 4070 Ti and RTX 4080 to Make Room for SUPER
NVIDIA has reportedly halted production of GeForce RTX 4070 Ti and RTX 4080 graphics cards, allowing the market to digest existing inventories of the two products, so it could create room in the channel for its upcoming GeForce RTX 40-series SUPER graphics cards. The company is expected to announce these cards around the 2024 International CES (January). The SUPER series is being designed to consolidate NVIDIA in the performance-enthusiast segments, and include three SKUs, namely the RTX 4080 SUPER, the RTX 4070 TI SUPER, and the RTX 4070 SUPER. According to the Board Channels post detailing this development, NVIDIA has stopped mass-production of the desktop RTX 4070 Ti and RTX 4080 GPUs, shipped its last orders for Q4-2023, each add-in card (AIC) partner has digested its inventories of these cards, and is awaiting arrival of subsequent new products.
Sources:
Board Channels, VideoCardz
85 Comments on NVIDIA Halts Production of RTX 4070 Ti and RTX 4080 to Make Room for SUPER
Otherwise its just going to be as unappealing as always. People with that sort of cash will still just rather spend a bit more for the full fat 4090.
4080 super for 1200, the current 4080 price point.
Not something I would care for, it would not even be a good deal, but it being Nvidia, I think not even that will happen.....
Next advert will be "Just buy" !
RTX 3000:
3090 Ti - 10 752 Cores
3090 - 10 496 Cores (-256)
3080 Ti - 10 240 Cores (-256)
3080_12 - 8960 Cores (-1280)
3080 - 8704 Cores (-256)
RTX 4090 - 16 384 Cores
RTX "4080 Super" - 10 240 Cores (-6144)
RTX 4080 - 9728 Cores (-512)
RTX 4070 Ti - 7680 Cores (-2048)
RTX 4070 - 5888 Cores (-1792)
See? There is a GAP on whole 4070 at top.
I see ZERO reasons to buy both: 4080 and "4080 Super". Jensen might use it by himself.
P.S. I got RTX 3080 Ti right now.
Why the heck is English the universal language anyway???
Historically SUPERS have fallen in between the non-Ti and Ti cards. Now they're... better? Worse? I don't know.
Why would you then not pay a little bit extra and get teh 4090? same reason why right now the 4080 is so unpopulair.
I will likely continue to direct people toward the 7900xt. While the price is currently bad; inflated to make Black Friday deals look good as Tech Jeebus himself has mentioned in an extremely very badly-timed video (if early October prices [~$700] had stayed people would have expected $650 on BF, with current pricing a sale at $700 [which I think it will be the permanent price pretty soon] will still look good). Past that, with this nVIDIA refresh launch, as much as it sucks for AMD, I have no clue how they get away with anything less than competing in price with the remnants of 4070 Ti non Ti only one suffix edition (~$650-700?) if 4070 Ti Ti (Super) is ~$800 or even $850. Likewise, while I feel ~$700/850 are what I consider 'fair/realistic' prices for the 7900xt/x, I have to imagine 7900xtx will have to drop to compete with said 4070 Titanium 16GB Super Edition, if perhaps only the Founders Edition (JFC, ferreal). At those prices (650/800, maybe even slightly less to catch peoples' attention?) those products will be a killer deal and would already be priced to compete with freakin' Blackwell. I mean seriously, a N31 might be a similar price to GB205. We can revisit the pros/cons of each companies products when that time comes, but I have to believe 7900xtx is going to look comparably pretty good in raw perf unless nVIDIA gives us a great performing card and/or it is less than $800. (<- Lol, no). Same thing for 7900xt and whatever a cut-down product might be.
I think the thing I will say to people is to think about how 4070 Ti may not even net you 50TF (on purpose, hence 66SMs), and then ask them what they expect of a PS6. A 7900xt will easily overclock to 60TF+ (make sure to use thebios repository;), heck even for a 7800xt ). As I've said before, those are very important (market segmentation/resolution) metrics if you do not account for nVIDIA's tensor (rt) cores. Which, you know, I don't because planned obsolescence and/or resolution/FR concessions. Even if you do, they are likely VERY close in that situation where compute RT can be used to it's utmost capability vs those cores. When that's not a factor, it's perf should conceivably mean one resolution bump higher in certain situations. That's real performance for less price, Kyle. Not the same joke as 4070 earlier Ti vs 7900xt, but still a similarish card (all things considered) for less money.I think $1000 with 4080 drop to ~$900. JMO. nVIDIA will never make the 7900xtx NOT look like a good deal bc greed, and like-wise AMD will never let AD103 look like a good deal bc they have to do that. It's amusing when (hot take, I guess) N31 is a better chip than AD103. 7900xtx is just a better setup than anything AD103 can muster. 7900xt will always be a better card than a cut-down version bc nVIDIA refuses to give that ideal perf at a lower price. Used to be ~20% better perf for ~10% cheaper, nowadays it more like 20% better perf for 30% going-on 40% cheaper, literally cheaper than one whole segment down with potential to perform similar if not better to an nvidia card in a certain tier, which is absolutely nuts.
All pricing is bad right now. The difference is AMD probably will fall to realistic price/perf land...There is literally no hope for expecting that from nVIDIA. If something is a good price, they screwed you somewhere. If it has decent performance, it will be overpriced. It's unfortunate, but it's true. That's why Ti Super 16GB is such a nVIDIA play. 16GB looks good. AD103 *sounds* good. Then you realize the compute just isn't there for the money they're going to ask you to spend. Maybe people won't see it in games tested out the gate, but rest assured, just like 4070/Ti of old, those concessions will come around to very-intently bite you.
That's the fine wine, you see. Not screwing people any way conceivingly possible so they'll need to upgrade quicker at a certain expected perf level. People think RT is a feature, it's actually just another thing they can segment with FOMO; something to fall back on when generations of cards end up performing approximately the same in actual raster/compute.
IIRC, the 780Ti was a more full-fat card than the OG Titan.