![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
I don't trust these prices - it's in USD - and when I do the conversions for my country (Canada/CAD), they're less than what the current series (4070 Ti, 4080) go for.
I believe these prices would happen, atleast in US
And you shouldn't trust those prices anyway. Nvidia is notorious for changing prices right before launch. Retailers have to be on their toes.
16gb on the 4080 Super seems a little low tbh.
The next option is 32 is they clamshell the memory like they did for the 4060 16GB.
They can't do 24 like the 4090?
Not with a 256bit memory bus. Honestly, I was initially surprised the 4080 super didn’t use the same ad102 as the 4090, but almost certainly it would have ended up too close in performance to the 4090.
*90/*90Ti is the new Titan. I know NV insisted that the 3090 was not a Titan, but it clearly was. The Titan cards were always in a weird position in that they kinda made the corresponding Quadro cards look bad and I doubt NV wants any more of that.
I agree that GPU prices have gotten ridiculous but a $599 flagship card is never happening. Everything is more expensive now- tech, housing, food, vehicles, people's labor, list goes on. 1080Ti was $699 seven years ago, 980Ti was $649 nine years ago, 780Ti was $699 eleven years ago... Heck, the 8800 Ultra was $849 seventeen years ago. And wafer costs are through the roof so high-end GPUs cost more to produce than they ever have. Even if consumer attitudes drastically shift, I would be stunned to see a flagship NVidia GPU going for less than $1000. If the absolute most people were willing to pay for a GPU in the 2020's was $599, that card would be using at the most a <400mm^2 chip with a 256-bit bus.
The Titan name is dead to force education and non profit business over to the professional series GPUs. I put something with Gamerz RGB XXX BJ Edition in the title I have a half dozen auditors up my ass before year end.
That makes a lot of sense, I've heard stories from folks with gov / edu backgrounds how difficult it was to build their own workstations for research and production because anything with a slightly gamerey-sounding name would get flagged.
This is rumoured to be the top RDNA 4 card, to be released in H2 of this year.
Yeah I heard that too. AMD is probably waiting till their dual-GCD setup is realistic to productize (since it apparently didn't make the cut for RDNA3 despite early leaks/rumors)
That's why I placed a safety disclaimer haha
The memory thing is overblown. Is it good to have moAr memory? Sure. You could say the same for the "other" side with AMD folks talking about how their more memory is the best. Sure...it's better to have more memory but for me, personally, I'd rather have a more mature driver set, toolset, CUDA, G-SYNC, etc. (not even going to mention RT because it's a non-factor for me, for the most part).
It’s more due to limitations on TSMC’s advanced packaging capabilities. TSMCs demand on the advanced packaging and assembly nodes increased far faster than their ability to procure the hardware to keep up with it. So not only did they jack the prices but they shortened the windows too, so consumer viability of those nodes is impacted if they want to keep their higher margins. And with consumer spending expected to fall off a cliff I’d bet AMD and Nvidia are both looking to tighten things up in advance.
Yes and no. The 90/90 Ti is clearly also the rebrand of the 80 Ti tier and what it used to be with a price hike, but also sort of a hybrid between that and what used to be Titan with the 24GB. That said, historically, Titans had functionality that was gimped on GeForce cards and their own drivers to enable those workloads. That might not be the case anymore I'm not sure, but at release there were definitely things the RTX Titan still outperformed the 3090 on because of that. To just blanket say 90's are Titans....I don't know.
If we see an 80 Ti that would only be if NVIDIA has decided to postpone the 5000-series until Q1 2025 (versus the usual Q4 2024). Now would be the time to make that type of move.
Agreed...yet all they announced today with 80, was a 4080 Super which is just full AD103, so ~5% better than a 4080. Granted, with a price cut. Which is good since they were trying to sell it for 80 Ti prices.
It's not just the memory capacity, it's the bandwidth. 4070 Ti with a measly 192-bit bus definitely shows its deficiencies at 4k in terms of performance scaling, and as for capacity, as a 3080 Ti owner, I can tell you 12GB is becoming not enough for 4k RT in some titles. Increased L2 helps, sure, but it isn't the magic band-aid Nvidia (or AMD for that matter with their Infinity Cache) wants you to think.
That - I can buy. But there are no games that I play that require 12GB. I know because even though I have RTX 4090 at home - I have an RTX 4060 in my laptop. It's overblown. Future-proofing is a fool's errand.
Overblown, I agree. And in tech there really is no such thing as future-proofing. There is though, the expectation that a card doesn't age too poorly, and I was extremely disappointed that I hit VRAM issues with one title less than 12 months after I bought the card. Last time I ever had VRAM issues was literally in 2012 with a 2GB GTX 680 and modded Skyrim.
GeForce RTX 40 "SUPER" refresh performance projections, now incl. ray-tracing performance
Sad part is I look back when I did my build and the card shortages were happening and I had nothing, but snagged a AMD RX6800 for $900 + tax...but I still go deum! at the thought of paying $1k for a video card.....
Same hardware, cheaper cooler (no RGB), and charge an extra $100 for the convenience of not having to answer questions about purchasing it.
The other side of that could very well be that the limited GPU capabilities of the existing 7000 series could similarly see performance fall off with future titles even if they are really good now because of the extra memory, trying to future-proof anything GPU-related right now is a hard ask because things are changing too fast.
