AMD Launches Radeon R9 380X, Fastest GPU Under $250 (hothardware.com)
MojoKid writes: Although AMD's mid-range GPU line-up has been relatively strong for a while now, the company is launching the new Radeon R9 380X today with the goal of taking down competing graphics cards like NVIDIA's popular GeForce GTX 960. The Radeon R9 380X has a fully-functional AMD Tonga GPU with all 32 compute units / 2048 shader processors enabled. AMD's reference specifications call for 970MHz+ engine clock with 4GB of 1425MHz GDDR5 memory (5.7 Gbps effective). Typical board power is 190W and cards require a pair of supplemental 6-pin power feeds. The vast majority of the Radeon R9 380X cards that will hit the market, however, will likely be custom models that are factory overlcocked and look nothing like AMD's reference design. The Radeon R9 380X, or more specifically the factory overclocked Sapphire Nitro R9 380X tested, performed significantly better than AMD's Radeon R9 285 or NVIDIA's GeForce GTX 960 across the board. The 380X, however, could not catch more powerful and more expensive cards like the GeForce GTX 970. Regardless, the Radeon R9 380X is easily the fastest graphics card on the market right now, under $250.
They're really drawing a fine line on this one.
Seeing as how the GTX 970 has broken under the $300 mark in the last few weeks, they're not doing much to sell me on the R9 380X.
Why doesn't AMD just shut down and leave all the work to Intel?
Because AMD is the only way we have to keep Intel from going full monopolist. The qualitative effect on the market from the difference between one and two makers of a particular product, such as CPUs that run a particular instruction set, is far greater than that between two and three.
I'm always surprised to see this mentality in tech oriented groups. AMD's contributions to GPU tech are huge, and very community friendly. They use and produce open source software, and actively support Khronos group efforts. Their tech is always non-proprietary, and works across even non-AMD devices. For development of any kind debugging information provided by the AMD gear is just plain more useful. As for CUDA - it is almost directly inferior to OpenCL. CUDA's prevalence is largely due to NVIDIA's attempts to jam it down every available throat. Hell, NVIDIA is going so far now as to con monitor manufactures into buying expensive chips to sit sidelong in variable refresh rate monitors to give the same functionality AMD cards provide for free.
Unfortunately AMD is trying to fight a war on three fronts - and losing all of them. It makes me sad to see them lagging behind when they are worth supporting on sheer principle alone. The performance and cost differences aren't nearly enough to make up for the business practices.