NVIDIA's Future: RTX 5090 to 5060

In this video, I want to talk about the upcoming NVIDIA GeForce RTX 5090 graphics card as well as other RTX 50-series GPUs including the RTX 5060 and RTX 5070. I see way too many people overhyped by claims that the RTX 5090 will more than double the performance vs the RTX 4090. Let's talk about it!

Support my work by becoming the iVadim channel member and get special perks:

Timestamps:
0:00 - RTX 5090 hype
0:30 - RTX 50-series development
1:30 - Overhyped Performance
3:26 - How RTX 4090 got so powerful
3:39 - How RTX 5090 improves upon the RTX 4090
5:42 - Realistic RTX 5090 gaming performance
6:04 - NVIDIA hates gamers?
6:57 - Will RTX 5060 & 5070 suck?

Sources:

[theivadim]

What do you think about the future of NVIDIA gaming graphics cards?

[johnc]

RTX 5090 will have DLSS 4.0. You won't need games anymore, you just type in what type of game you want to play and the A.I. codes a game for you in real time.

[hensonk]

Expensive games with unreasonable hardware requirements is a problem

[chlorobyte_projects]

Well, hopefully they adopt GDDR7, in that case they could deliver actually usable memory for the 5060 and 5070 cards.

...until we find out they'll be on a 64 bit bus

[tomallan]

I see Nvidia trying to cut corners as much as possible. If the 5090 beats the 4090 by more than 15% in raw rasterization I will be surprised and for at least a 20% increase in price. Nvidia is leaving gaming GPUs in the rearview mirror. Yes they will keep putting out overpriced GPUs but unless AMD or even Intel can increase their rasterization and raytracing to Nvidia levels GPUs will not have 1060 gen over gen leaps in performance.

[jouniosmala]

5060 and 5070 will probably be something like30-50% more powerful than 40 series equivalent at similar prices. Nvidia going for GDDR7 means, a lot more bandwidth with the same narrow bus and a lot more memory because memory chips themselves contain more memory. And clock speed bump means higher performance without extra cost, the improved density from 3nm node should be a bit higher than the 25% price bump TSMC has.
Oh. 40 series went from dirt cheap manufacturing node to ultra expensive node and people were angry because they didn't grasp that Geforce both lowered margins and increased prices because of that. And didn't realize that the company average margin people mention currently is based more on their professional lineup than Geforce lineup.

[najeebshah.]

I dont think it'll be worse or better, it'll be in accordance to what amd can do, amds failure to compete repeatedly has nvidia holding back on letting go too much, just like when the gtx660 was turned into a gtx680 when amd failed to compete and the 680 returned as the first titan.

[NBWDOUGHBOY]

I think it's gonna be a less about the performance increase and more about the a good price decrease. If history repeats itself then 20 series was a bad gen for prices. 30 series was better. 40 was bad and now 50 is gonna be better with more VRAM but maybe a 25% increase in performance tops.

[coolstar]

I’m so glad I got a 3080 a few months after their launch. Perfect timing for one of them as the newer 60 and 70 series cards are still lacking in comparison and arn’t really selling all that many units.

[grospoulpe]

2 to 3 times better than previous gen was I heard about RDNA 2 -> RDNA 3 (2.7x, from Red Gaming Tech); I don't think I heard something that silly with Lovelace (besides 600-1000W GPU) or +50% per Watt (not really the case with RDNA 3, besides, maybe, the 7900XTX)

[Pandemonium]

Typically a doubled paper spec translates into a 20-25% gaming increase in modern AAA titles.

[Humanaut.]

Price to Performance matters.
I voted with my wallet and left gaming behind (6months now).
Only really really good deals might be able to bring me back at some distant point in the future.
Ive had enough of the duopoly.

[viktortheslickster]

If the 512 bit bus rumour is correct, it seems likely to me that Nvidia will be using GDDR6W. 512 bit buses using regular 32 bit memory modules have been avoided due to increased board complexity and cost, I see no reason why that should suddenly change

[reallycome]

Agree with your statements throughout!

[kaystephan]

3:36 Isn't lovelace based on the 4Nm process? At least that's what it's called. Transistor sizing in general is way more complex but I'm pretty sure the name of the used process is 4Nm, not 5.

Also: I talked to someone a while ago who works in a big Server company. These people get information years ahead because they plan multi million dollar Server systems. And according to her Blackwell will cost around 70% more in production while only getting around 30% more performance out of 3Nm compared to 4Nm. Now, of course this could still change over the years between her hearing this and it actually being released as a product, but based off of this I think a practical increase in performance of 40 to 50% might be realistic. At least if they want to keep power draw the same (and they kinda have to because you can't just keep scaling power draw up until you have a 1000+W GPU).

[sidvicious]

You are correct and it will be more of the same. Nvidia has no will to release anythi ng better for consumers.

I got lucky and found an open boxed 4080 for 915.00 and I will be sticking with that for the foreseeable future, gaming and video editing is covered.

[MMTTsALd]

I agree with you on the performance being 40%-60% at best too the 4090 was such a huge leap in performance it's hard to believe Nvidia will do that again especially when amds best card is 20% behind and an even wider gap in RT theres no incentive for Nvidia to give us the best card possible

[MrGrombie]

I subbed when I saw the computer case. I love a good sleeper build. You’ve earned my sub. 😂

[neti_neti_]

Intel i9 15900k and RTX5080 Will be great combination.

[ChowdongsFameCrew]

Hopefully 5090 has the same good coolers as 4090, so quiet 😊