How about a 512-core video card?

EddieC 0 575 Views

Nvidia yesterday began shipping the GeForce GTX 480, with double the number of processor cores previously available at 512. The card is intended to to permit smooth playback and editing of 2D or 3D content, including animation, games and videos.

The GTX 480 supports Microsoft DirectX 11, Direct3D and , Nvidia's real-time physics engine and SDK that came along with its acquisition of Ageia in 2008. Now tailored for Nvidia's CUDA GPUs, PhysX off-loads physics processing tasks from a system's main CPU.

The board uses NVidia's latest graphics processing architecture--called Fermi after nuclear reactor inventor Enrico Fermi--which the company says will provide a platform not just for media producers and gamers, but for high-end modeling applications such as those used in meteorology, geology and medical research.

The GeForce GTX 480, which contains 1.5GB of dedicated graphics memory, is Nvidia's answer to AMD's Radeon 5870 product, which came out in September. Early indications are that GeForce keep pace with the Radeon in one-card tests, but in a two-card test reported in PC World, outperformed by a significant margin. The GTX 480 will cost you a PCI Express 2.0 x16 slot and about US$500.

Dani AI

Generated

Good point, — a couple of practical notes that matter more than the raw "core" number.

A GPU’s many “cores” are very different from CPU cores: they’re lightweight arithmetic lanes inside streaming multiprocessors and are scheduled in large groups (warps). Architectural details (how many cores per SM, instruction throughput, caches, and memory bandwidth) and how well software maps work to that parallel model usually dictate real performance far more than the headline core count. (docs.nvidia.com)

Real-world tradeoffs matter: that generation ran noticeably hotter, drew more system power and produced higher noise under load, so case airflow and a strong, reliable power supply are important. Reviews at launch measured high sustained temperatures and heavy system power draw; the board also requires multiple external PCIe power connectors and a substantial PSU headroom. Plan for those system-level requirements rather than buying on the spec sheet alone. (anandtech.com)

If the card is intended for GPGPU (CUDA) workloads, verify the card’s compute capability and which CUDA toolkit and drivers you need. Older architectures become legacy over time and toolchains or libraries can drop or limit support, so check the CUDA documentation and your application’s supported GPU list before committing. For multi‑GPU gaming or rendering, expect application-dependent scaling and to rely on driver maturity and benchmarks for the titles/apps you use. (docs.nvidia.com)

Quick checklist: confirm required PCIe power connectors and recommended PSU wattage; check physical card length and case airflow; look at independent benchmarks for your target apps (games, editor, renderer); prefer partner cards with improved coolers if noise/temps are a concern; and verify CUDA/toolkit and driver compatibility for compute workloads.

Be a part of the DaniWeb community

We're a friendly, industry-focused community of developers, IT pros, digital marketers, and technology enthusiasts meeting, networking, learning, and sharing knowledge.

Edit Preview