⋮    ⋮  

NVIDIA Rumored To Mass Produce Flagship Pascal GPUs With HBM2 In 1H 2016 – Availability in 2H 2016


Note: This news is tagged as a rumor and should be treated as such. There are some good points to believe that the flagship Pascal may arrive to GeForce audience later in the year but NVIDIA has had working chips at their labs since November of 2015, as was revealed at their GPU Technology Theater at SC15. NVIDIA has so far gave consumers no look at Pascal to date but they have previewed Pascal to the audiences at GTC 2015 in Japan and Korea several times before as seen in the slides below.

NVIDIA is rumored to launch their flagship GPUs based on the Pascal architecture in second half of 2016. In a report published by Korean site, Digital Times (via HWbattle), it is reported that NVIDIA will be making use of the next generation HBM2 standard on the flagship Pascal GPUs that feature the 16nm FinFET process aiming the consumer and enterprise markets.

NVIDIA Rumored To Bring Flagship Pascal GPUs in Market By 2H 2016

The rumor comes just a day after Samsung announced that their company have initiated mass production of their first, 4 GB HBM2 DRAM package. The company will be competing against SK Hynix who developed the memory standard in the first place in collaboration with AMD. SK Hynix deployed their first HBM package on AMD's Radeon R9 Fury series cards back in July which adopted a revolutionary package design.

Moving on, SK Hynix has yet to begin mass production of HBM2 DRAMs. SK Hynix is currently eyeing a production timeframe as early as August 2016. On the other hand, Samsung while having 4 GB HBM2 chips in mass production is further going to strengthen their HBM2 development by producing 8 GB HBM2 packages. These high capacity packages will be specifically aiming the flagship Pascal GPUs which will feature up to 32 GB of VRAM.

It is rumored that NVIDIA currently has a range of GPU samples based on their 16nm Pascal architecture that are being produced and tested internally. The NVIDIA lineup will range from solutions based on GDDR5X and HBM2 packages but it is the flagship which will have to wait till 2H 2016 to see the light of day since that is the time when the higher capacity HBM2 packages go in production. The current 4 GB HBM2 packages can go up to 16 GB VRAM with 1 TB/s bandwidth. The 8 GB HBM2 packages can go up to 32 GB VRAM with 1 TB/s of total bandwidth. NVIDIA had already revealed that their flagship GPUs aimed at HPC will be using 32 GB of VRAM that is only possible through higher capacity DRAM packages.

NVIDIA GP100/GP200 – The Flagship GPU, Powering Titans, Teslas and Quadros

NVIDIA GeForce Pascal

The heart of next generation supercomputers and High-Performance Computing platforms is without a doubt, the Pascal GP100/200 graphics chip. The NVIDIA GP100/200 chip will be the flagship GPU of the lineup and one which will determine the performance and efficiency of the new architecture. The Pascal GP100/200 has long been in the rumor mill and we still don’t have conclusive details on this monolithic chip. Being the successor to the GM200, the GP100/200 Pascal GPU will be built on the 16nm TSMC FinFET process node and feature up to a total 17 Billion transistors inside the package.

The GPU is going to pack a lot of performance for gamers and FP64 users since this chip will be powering some serious compute-oriented machines that demand double precision compute. Being the flagship of the lineup, NVIDIA will make their GP100/200 GPU their first graphics chip to support HBM2 memory with up to 1 TB/s of bandwidth and 32 GB VRAM. We know Pascal has a peak double precision performance rated at over 4 TFLOPs while the single precision compute performance is rated at over 10 TFLOPs. This will be by far the biggest leap in total available compute performance we have seen on any graphics card.

As for when it arrives, there’s a strong possibility that consumers won’t get the full GP100/200 first nor a cut down variant. The reason is due to high demand from the HPC market as they have to update the older Kepler based cards which are since being used as FP64 options as the Maxwell chips drove FP64 support away.

Powering the Tesla card first followed by GeForce and Quadro solutions, the card will be getting a range of products, without a doubt a new Titan offering and for $999 US which has been a consistent pricing for Titan graphics cards. The dual-chip cards are a totally different thing though (Titan Z). The GP100/200 chips will be available in a range of new packages such as the regular graphics (Add-In) boards and the new Mezzanine cards which were showed back at GTC 2015. Along with NVLINK support which is a new interconnect that NVIDIA is establishing with IBM and other partners such as CRAY, HP, DELL, TYAN, QCT and Bull, the connection would offer 80 – 200 GB/s access speeds between the several nodes integrated in HPC platforms.

NVIDIA GP100/GP200 Features:

  • Based on Pascal GPU Architecture.
  • 2x Performance per watt of Maxwell.
  • Launch rumored for 2H 2016.
  • Will support DirectX 12 feature level 12_1 and higher.
  • Successor to the GM200 GPU found in the GTX Titan X and GTX 980 Ti.
  • Built on the 16nm FinFET manufacturing process from TSMC.
  • Allegedly has a total of 17 billion transistors, more than twice that of GM200.
  • Taped out in June 2015.
  • Will feature four 4-Hi HBM2 stacks, for a total of 16GB of VRAM for the consumer variant and 32GB for the professional variant.
  • Features a 4096-bit memory bus interface.
  • Features NVLink and support for Mixed Precision FP16 compute tasks at twice the rate of FP32 and full FP64 support.

Both NVIDIA and AMD are currently testing their next generation GPU architectures codenamed Pascal and Polaris. Giving final polishes to their next generation GPUs, both companies expect to see a huge performance and efficiency gains from the latest FinFET process. NVIDIA will most likely have more detailed information on their Pascal GPUs progress at GTC 2016 in April where they are likely to showcase a range of GPUs for the GTX consumer markets. AMD on the other hand has already stated that they will be launching Polaris in desktop and mobility platforms in mid-2016 which marks the Summer of this year. Both vendors will see launches just months or even weeks away from one another.

NVIDIA Pascal and AMD Polaris – The FinFET GPUs:

Flagship GPUVega 10Navi 10NVIDIA GP100NVIDIA GV100
GPU Process14nm FinFET7nm FinFETTSMC 16nm FinFETTSMC 12nm FinFET
GPU Transistors15-18 BillionTBC15.3 Billion21.1 Billion
GPU Cores (Max)4096 SPsTBC3840 CUDA Cores5376 CUDA Cores
Peak FP32 Compute13.0 TFLOPsTBC12.0 TFLOPs>15.0 TFLOPs (Full Die)
Peak FP16 Compute25.0 TFLOPsTBC24.0 TFLOPs120 Tensor TFLOPs
Memory (Consumer Cards)HBM2HBM3GDDR5XGDDR6
Memory (Dual-Chip Professional/ HPC)HBM2HBM3HBM2HBM2
HBM2 Bandwidth484 GB/s (Frontier Edition)>1 TB/s?732 GB/s (Peak)900 GB/s
Graphics ArchitectureNext Compute Unit (Vega)Next Compute Unit (Navi)5th Gen Pascal CUDA6th Gen Volta CUDA
Successor of (GPU)Radeon RX 500 SeriesRadeon RX 600 SeriesGM200 (Maxwell)GP100 (Pascal)