Tuesday, January 3rd 2023
NVIDIA GeForce RTX 4070 Ti Launched at $799 with Performance Matching RTX 3090 Ti
NVIDIA today formally launched the GeForce RTX 4070 Ti "Ada" performance segment graphics card at a starting MSRP of USD $799. Based on the 4 nm "AD104" silicon, the RTX 4070 Ti is essentially the same product as the RTX 4080 12 GB, which NVIDIA decided to cancel from its original mid-November launch, toward a new one this CES, under a new model name. The card maxes out the silicon it's based on, featuring 7,680 CUDA cores, 60 RT cores, 240 Tensor cores, 240 TMUs, and 80 ROPs. It gets 12 GB of GDDR6X memory across a 192-bit wide memory interface, running at 21 Gbps (GDDR6X-effective). The card has a typical power rating of 285 W, and continues to use a 12VHPWR power connector, even on the custom-design products.
NVIDIA claims that the RTX 4070 Ti should enable maxed out AAA gaming with ray tracing at 1440p, while also being formidable at 4K Ultra HD in games that can take advantage of technologies such as DLSS 3 frame-generation, or even classic DLSS 2. The company claims that it offers performance comparable to the previous-generation flagship, the GeForce RTX 3090 Ti "Ampere," with a much higher performance/Watt rating. The RTX 4070 Ti doesn't appear to feature an NVIDIA Founders Edition model, and is a partner-driven launch, with custom-design cards dominating the scene. The RTX 4070 Ti will be available from January 5, 2023, but we'll have reviews for you before that!
NVIDIA claims that the RTX 4070 Ti should enable maxed out AAA gaming with ray tracing at 1440p, while also being formidable at 4K Ultra HD in games that can take advantage of technologies such as DLSS 3 frame-generation, or even classic DLSS 2. The company claims that it offers performance comparable to the previous-generation flagship, the GeForce RTX 3090 Ti "Ampere," with a much higher performance/Watt rating. The RTX 4070 Ti doesn't appear to feature an NVIDIA Founders Edition model, and is a partner-driven launch, with custom-design cards dominating the scene. The RTX 4070 Ti will be available from January 5, 2023, but we'll have reviews for you before that!
150 Comments on NVIDIA GeForce RTX 4070 Ti Launched at $799 with Performance Matching RTX 3090 Ti
The 4070 Ti is on TSMC 5 nm which costs ~$17000 per wafer.
edit: according to rumors.
"Foundry Sale Price Per Chip" 16/12NM - 331 USD
"Foundry Sale Price Per Chip" 5NM - 238 USD
??????????????????????????????????????
3070 ti cheapest are 790 usd(with vat), you can't get 3080 below 1060 usd (with vat and avg price for it is still over 1160 usd with vat), 3080 ti cheapest is 1670 usd(with VAT and avg 1850), 3090 with the best deal of the month 1965 usd (with vat but it's only one piece of card and after this somehow gone the next one is the avg min price is 2500 usd with vat GG Hungary, look how nice retailers we have), and there is no new 3090 ti.
Most of the 4080 tanked at 1560-1690 usd (with vat), 4090 cheapest at 2100 but avg 2600 usd (with vat).
7900 XT ohh boi price for it is 1170-1270 usd with VAT and the AIB design models are go for 1320-1450 usd, 7900 XTX 1530(only the standard)-1800(feel the premium, just feel it you can't see it).
So with that performance i can see it as an actual finally long last buy of a card for that price with 3090 performance, still it could be cheaper for a XX70 card like 600-650+VAT.I use a 970 and in 2021 wanted to upgrade for a 3070 ti or a 3080 HAHAHA lol, still using my 970.
The initial wafer cost for 14nm (12nm rebrand) was also around $16000 :D
Where are all the sub £300 cards!!??
2080TI 754 mm² die size -> 12nm 300mm wafer $4000 (yield of 80%) = 70 dies out of them fully 56 functional = $71 per die -> this die probably costed between $200 & $300 to produce at the start of N12 production
4080 379 mm² die size -> 5nm $17000 (yield of +90%) = 147 dies out of them fully 132 functional = $128 per die
As you can see new dies are more expensive to produce, but not groundbreakingly more expensive and prices will only get better once TSMC moves Apple's products to smaller node.
So, it needs to render at a lower resolution than 4K because it can't hack native 4K? How is that formidable, exactly?!
It can limp to 4K if you give it some crutches.
Btw a video card is more than just the chip. 2080 Ti has less memory, has GDDR6 memory instead of GDDR6X, it uses smaller heatsink (on the FE), copper and aluminum prices are up, pcb component prices basically doubled in the last 2 years, shipping is up, AIBs want more profit.
That's commonly called inflation.
And when I do buy, it'll be from AMD.
But take into account that 3080 Ti 3090 and all the halo products were discounted to under 1K. from the original 1199 to 1999. 3080 12 GB for as low as 750.
Clearly 48 MB L2$ and 192bit is as efficient as 384 bit 6MB. and the price being slashed from $999 3090 Ti 24GB, to a more reasonable 799 losing half of the bus and the memory, the same 40. Tflops.
The one to get is 4070 5888 Cuda version and if delivering 3070 Ti + 10% is as good as it gets, I'll take that. 3x faster than my 980Ti.
Turing expensive, Ampere affordable, Ada expensive, Whatever is next affordable...
AMD follows.