NVIDIA is rumored to start production of it’s Pascal architecture GPUs in Q1 of 2016, utilising the HBM2 standard & 16nm FinFET manufacturing process.
NVIDIA is rumored to start production of it’s Pascal architecture GPUs in Q1 of 2016, utilising the HBM2 standard & 16nm FinFET manufacturing process.
NVIDIA is rumored to start production of it’s Pascal architecture GPUs in the first half of 2016.
Korean site Digital Times reports that NVIDIA will be using the new HBM2 memory interface & 16nm FinFET manufacturing process for their ‘Pascal’ architecture GPU’s.
Samsung will be competing directly with SK Hynix, although Hynix have not started mass producing HBM2 chips, these will come around August 2016.
Consumer & enterprise market’s may start to see the final product(s) appear on shelves as early as the second half of 2016.
With Samsung announcing yesterday production of “the industry’s first 4-gigabyte (GB) DRAM package based on the second-generation High Bandwidth Memory (HBM2) interface”, the above rumour could prove to be correct.
Whilst Samsung have only just started manufacturing 4GB HBM2 chips, 8GB HBM2 chips should not be far off which will allow NVIDIA to utilise up to 32GB of VRAM off a single GPU.
NVIDIAs Pascal GPU’s are rumoured to also feature GDDR5X for the lower-tier models in the line-up. GDDR5X breaks the current 8Gbps performance barrier of GDDR5 by doubling the pre-fetch to 16 data words per access from 8. First generation GDDR5X chips are expected to product 10-12Gbps with the company expecting to surpass 14Gbps by the second generation.
GP100/200; mostly likely reserved for Pascal top-end GPU’s, NVIDIA will ensure these chips are the first to utilise the HBM2 technology which could equate to over 10 TFLOPS compute performance – a large jump over current GPU’s compute power.
Pascal will be used for NVIDIA’s GeForce, Quadro & Tesla series of performance graphics cards.
