Nvidia launches three new GeForce RTX 4060 GPUs, starting at $300

[Gastec]

I must be going slightly mad!
The RTX 4070/4070 Ti with 12 GB VRAM and the RTX 4060 Ti with 16 GB.
Nobody talks about that abomination. I must be definitely out of my mind!

 

[GettCouped]

I'm sure everyone has an opinion on this, but here is how I think Nvidia should have gone, making money while also making people happier:

4090 24gb $1599
4080 20gb $1099
4070 Ti 16gb $799
4070 16gb $599
4060 Ti 16gb $449
4060 12gb $330

Those prices are absolutely ridiculous. Sorry

 

[Hodor]

Just three new GPUs? That's gonna sell out fast. What about the 4th customer?

 

[tylerb1489]

My 1080 Ti will keep on trucking until the 5070 comes out and gives me 4080 performance with 16gb for 799$ bahahahaha.

 

[Gastec]

Just three new GPUs? That's gonna sell out fast. What about the 4th customer?

He lives in the 4th wall and his name is Emil.

 

[lripplinger]

I paid $250ish for the GTX 1060 when it came out. I thought that was an excellent bang for the buck. Still using it today. I might consider the RTX 4060, if it wasn't north of $300. That just seems like a rip off. I can't believe how the RTX 3060 is still selling for $330+. Ridiculous, NVIDIA. Guess we wait and see what AMD is doing with the lower priced RX7000 series cards.

 

[llamapanda]

Not content with trying to pull the wool over our eyes by using DLSS to make performance uplift claims, nVidia are now going to try to claim extra cache is a viable substitute for absolute memory bandwidth. Erm, no.

I'm not seeing anything here that would turn me away from any of AMD's 6xxx range at any of those price points.

 

[neeyik]

nVidia are now going to try to claim extra cache is a viable substitute for absolute memory bandwidth. Erm, no.

Apart from the fact that it does work, as evidenced by AMD's RDNA 2 and 3 performance, as well as that seen with other Ada products. Take the 4070 and the 3080 (10GB), for example -- at 4K, the Ampere card is just 1.5% faster, on average, than the Ada one. Yet it has a 2% advantage with FP32 throughput and texel fill rate, a 4% higher pixel fill rate, and 51% more bandwidth.

This isn't to say that local memory bandwidth isn't important but a large, low-latency cache hierarchy is by no means a poor substitute.

 

[rmcrys]

I think people are better off waiting and showing that these schemes won't work (or will they?)

I would rather buy a 20 or 30-series with big discounts or in second hand for a good price than buying expensive garbage. Nvidia is just seeing if it sticks and raises the (money) bar.

My card is a 3060ti 8 GB RAM and even if I had a 2060ti 8 GB RAM, I would rather wait for the next gen or that prices go down at least 30% and buy a 16 GB RAM card.

 

[Puiu]

Insane price for the 16GB version.

 

[llamapanda]

Apart from the fact that it does work, as evidenced by AMD's RDNA 2 and 3 performance, as well as that seen with other Ada products. Take the 4070 and the 3080 (10GB), for example -- at 4K, the Ampere card is just 1.5% faster, on average, than the Ada one. Yet it has a 2% advantage with FP32 throughput and texel fill rate, a 4% higher pixel fill rate, and 51% more bandwidth.

This isn't to say that local memory bandwidth isn't important but a large, low-latency cache hierarchy is by no means a poor substitute.

Valid points, but it brings a level of reliance on the drivers and games themselves to ensure efficient that avoids cache-misses forcing expensive fetches from main memory. We'll surely see some titles falling foul of this and posting unexpectedly poor benchmark results, even if the actual numbers are academic in practice and the general performance across the board makes it moot. Looking forward to finding out.

 

[Dr Roboto]

The underlying issue behind the complaints regarding price and performance is the perceived lack of focus from Nvidia on the gaming community. While GPUs were originally designed for gaming and 3D rendering purposes, they have found significant utility beyond gaming, such as in the fields of cryptocurrency, machine learning, and LLM (ChatGPT). Companies operating in these emerging areas are large-scale businesses that purchase GPUs in substantial quantities, often in the hundreds or even thousands (ChatGPT for instance utilized 10,000 GPUs for training). For these businesses, cost is not a primary concern, and they continue to purchase Nvidia cards even if individual gamers opt not to. In addition, the competition from AMD GPUs is currently limited, and Intel has yet to establish itself as a significant contender in the market. Although AMD produces good GPUs, they have not yet reached a level of competitiveness that can truly impact Nvidia's dominant position.

 

[Geralt]

I'm sure everyone has an opinion on this, but here is how I think Nvidia should have gone, making money while also making people happier:

4090 24gb $1599
4080 20gb $1099
4070 Ti 16gb $799
4070 16gb $599
4060 Ti 16gb $449
4060 12gb $330

Real prices:
4090 24gb $999
4080 20gb $699
4070 Ti 16gb $599
4070 16gb $499
4060 Ti 16gb $399
4060 12gb $299
The difference with the current Nvidia prices is theft. And don't come with the theory that prices are related to costs, because modern economists have already proved that prices are only related to people preferences and shortage.

 

[ScottSoapbox]

The difference with the current Nvidia prices is theft. And don't come with the theory that prices are related to costs, because modern economists have already proved that prices are only related to people preferences and shortage.

I believe the words you are looking for are supply and demand.

 

[Geralt]

I believe the words you are looking for are supply and demand.

Nope. People's preferences and shortage. If people don't want something, costs don't matter at all. Another factor is shortage. If a good is rare, its price increases. All that was already proved.

 

[Khamul]

4070ti 12GB $800
4070 12GB $600
4060ti 16GB $500
4060ti 8GB $400

WOW. So the 4060ti is just a 3070. But look at the VRAM **** show.

 

[ID10T]

I just buy a new card dependent on what I paid for my current card, if I paid more I keep it longer, in the end Nvidia only gets about the same from me.
Used to buy yearly, then every other now it’s every 3 and most likely will become every 4 quickly.

 

[ScottSoapbox]

Nope. People's preferences and shortage. If people don't want something, costs don't matter at all. Another factor is shortage. If a good is rare, its price increases. All that was already proved.

Yeah, the terms for what you are describing are supply and demand. Supply (rarity) and Demand (how much people want something) are commonly used in economics and business. The fact that you are unaware of this is inconsequential.