New GeForce RTX 4080 Super, RTX 4070 Ti Super, RTX 4070 Super

Original Article ↗ Hacker News Discussion ↗

Nvidia’s new GeForce RTX 4080 Super, 4070 Ti Super and 4070 Super cards are launching this month with lower MSRPs than their predecessors and modest performance gains, prompting debate over whether they finally correct the weak price‑to‑performance of the original 40‑series lineup. Commenters focus on how these GPUs compare not just to last‑gen RTX 30‑series cards but to the 4090 and Nvidia’s lucrative enterprise products, highlighting constrained VRAM on consumer models and its impact on AI/ML workloads. The launches also revive broader concerns about GPU naming complexity, long‑term value for gamers, and how competition from AMD and future Intel cards might influence prices and capabilities.

Launch specs & positioning

  • RTX 4080 SUPER: Jan 31, $999; RTX 4070 Ti SUPER: Jan 24, $799; RTX 4070 SUPER: Jan 17, $599.
  • Nvidia marketing emphasizes comparisons vs 20‑ and 30‑series, not vs existing 40‑series, which several find suspicious or at least incomplete.
  • Some assume the new cards will effectively replace non‑SUPER counterparts, making originals obsolete.

Performance, pricing, and value

  • Nvidia claims 4070 SUPER can beat RTX 3090 with much lower power (220W vs 350W); commenters see this as plausible but also as an indictment of how weak/expensive parts of the 40‑series launch were.
  • Multiple comments say performance‑per‑dollar has stagnated or regressed vs 30‑series, especially outside 4080/4090.
  • 4070 Ti SUPER is viewed more positively: moves to a bigger chip (AD103), 16GB VRAM and 256‑bit bus, ~10% compute bump but ~42% memory bandwidth increase, close to 4080 performance at lower MSRP.
  • Skepticism that AIB partners will actually sell near MSRP; others think crypto crash and weak AI utility for these cards will limit scalping.

VRAM, AI/ML, and workstation use

  • 4080 SUPER and 4070 Ti SUPER have 16GB; 4070 SUPER has 12GB.
  • Several ML‑focused users dismiss them for training because “sweet spot” for current large models is described as ~24GB, with >40–80GB left to expensive enterprise SKUs.
  • Nvidia’s tight VRAM on consumer cards is widely seen as deliberate segmentation to protect high‑margin data center/workstation products.
  • Discussion of quantization and fine‑tuning notes that 8–12GB can work for many inference and LoRA cases, but full‑precision training of larger models still needs much more memory or multi‑GPU/cloud.
  • Jetson AGX Orin (64GB LPDDR5, ~$2k) is mentioned as an interesting but bandwidth‑limited alternative.

Gaming use and future-proofing

  • For pure gaming, many say existing 30‑series (e.g., 3080, 3080 laptop, 2060/2070/1080 Ti) still run most titles well; often drivers, not hardware, are cited as the pain point.
  • Debate over how quickly >8GB VRAM becomes necessary; consoles and upcoming titles suggest more memory will matter for 4+ year horizons.
  • Some gamers like that ML‑oriented demand might keep prices high; others hope it stays low so they can buy “just for gaming.”

Product stack, naming, and upgrade patterns

  • Broad frustration with Nvidia’s crowded naming: 4070 vs 4070 SUPER vs 4070 Ti vs 4070 Ti SUPER is seen as confusing; some wish for simple 4060–4090 lineup or “4075” style names.
  • Supers are characterized as mid‑cycle refreshes that move higher‑end dies down the stack.
  • Several argue marketing targets owners of 10/20/30‑series, not current 40‑series, and that most people stick to a price tier (xx60, xx70, xx80) and skip at least one generation.

Alternatives and competition

  • AMD Radeon 7600 XT (16GB) is highlighted as a decent, cheaper midrange option, good for budget ML and displacing some 4060 SKUs; others call it unexciting but welcome for 16GB at lower cost.
  • AMD is preferred by some for easier Linux gaming, though Nvidia is still seen as far ahead for AI software support, with AMD “catching up.”
  • Some hope release pressure will drop prices on AMD 7800 XT or older used cards.
  • Intel Battlemage rumors (16GB, 256‑bit, ~4080 performance for $450) are greeted with strong skepticism but curiosity.

Other technical and practical concerns

  • Lack of DisplayPort 2.1 on the new cards is criticized; defenders ask what non‑gaming use really needs 4K/144Hz HDR right now.
  • Concern over high power draw and connector issues persists; some want SUPER models mainly for updated 12V‑2x6 power plugs and worry about 4090 availability/overheating.
  • One user with an older low‑power Quadro laments the absence of modern passive or very low‑power options.

Community sentiment

  • Mixed overall: enthusiasm for the 4070 Ti SUPER as a strong value and for modest price corrections, but ongoing anger about GPU pricing, product sprawl, and perceived intentional VRAM stinginess.
  • Some argue social media has become reflexively anti‑Nvidia and pro‑AMD, leading to pessimistic expectations that this launch partially undercuts.
  • Others remain resolved to skip this cycle entirely and wait for the 5000‑series or for better deals in the used and midrange markets.