• 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
Pascal at Nvidia GTC,, Japan

Quote: . . . the Pascal GPU will be manufactured in Taiwan Semiconductor Manufacturing Company (TSMC), using the brand new 16nm FinFET process. This process is much more than a simple number, since it marks the shift from planar, 2D transistors to the FinFET i.e. 3D transistors. This shift required that the engineers make lot of changes in the thought process, and should result in significant power savings.

... Pascal will bring support for up to 32GB of HBM2 memory. However, the actual products based on Pascal will launch with 16GB HBM2 memory, and more memory will depend solely on memory vendors such as SK.Hynix and Samsung. What is changing the most is bandwidth. Both the Kepler-based Tesla (K40) and Maxwell-based M4/M40 featured 12GB of GDDR5 and achieved up to 288GB/s of memory bandwidth. Those 16GB HBM SDRAM (packed in four 4GB HBM2 chips) will bring 1TB/s in bandwidth, while internally the GPU surpasses the 2TB/s barrier. ...

NVIDIA’s Marc Hamilton said: “Using 3D memory, not only the memory capacity will go up, the memory bandwidth will go up significantly. With a much faster GPU, and higher memory bandwidth, the existing interconnects in the server are just plain outdated. So, we had to develop our own interconnect called NVLink, five times faster than existing technology.”

Pascal will also be available in multi-GPU packaging, replacing the Tesla K80 (NVIDIA skipped Maxwell-gen dual-GPU Tesla). Combined figures are very interesting to compare – 24GB GDDR5 and 480GB/s bandwidth should be replaced with 32GB HBM2 and 2TB/s bandwidth, mutually connected through NVLink rather than PCIe. The NVLink will enable up to 80GB/s, which should replace PLX PCIe Gen3 bridge chips that can only support 16GB/s (8GB/s per GPU). This part should be ‘warm up’ for 2018 and the Volta architecture.
The just didn't say when
This looks like it will be a massive leap in performance. Very exciting. Hopefully we will see it in 2016.
Hopefully? It's *guaranteed* unless something big and bad happens. I'd say, by Q2 we will see the first Pascal release. I am hoping to be at GTC 2016 in San Jose in March where there will be a lot more revealed (I bet)
I say hopefully because 14/16nm was supposed to be here a long time ago and we have been waiting and waiting, stuck on 28nm for what seems like forever. Actually weren't they supposed to go to 20nm first? I'm guessing they completely scrapped that node because of the delays?
Both AMD and Nvidia decided not to go with 20nm, years ago. There were no delays. 20nm was too expensive, and the process wasn't mature for it (without spending too much money for either company).

We are heading straight from 28nm to 16nm FinFET. That is a double node shrink jump, also using 3D transistors. The chips have taped out some time ago. We will Pascal before Summer; maybe by Spring.
Well I'm glad I held up on upgrading then. These cards are really going to shake up the market. Everything that came before them will be completely obsolete.
Hold your horses a bit. The GP204 - small die - is likely to be first (GTX 1080). The Big die Titans and GTX 1080 Tis will probably come late in 2016.

Of course all of this is speculation... but the picture is beginning to fill in.
This is speculation, and a little stretch, but the poster is well known ...
-- according to a recent and now removed MS download of Nvidia drivers , expect a top to bottom refresh of Nvidia's lineup with Pascal as GeForce driver 358.66 adds Vulkan, Pascal and Volta support(!)
(Google translation in Italics)

Quote:Wohl 3 verschieden Compute Capabilities bei Pascal:
Well 3 different Compute Capabilities in Pascal

Die CUDA-DLL nennt auch gleich ein paar Codenamen:
The CUDA DLL also calls equal to a few codenames



Those are 5 new GPUs in Nvidia's lineup plus a Tegra GPU (GP10B).

In contrast, AMD is bringing out 2 new GPUs.

That's what happens when you have 80% of the marketshare - you go for it! AMD can't; they are broke and R&D funds and and money for fabbing chips have nearly dried up. Nvidia no doubt will compete with Intel who will have 10-bit HEVC HW decoders in Kaby lake next year - and Nvidia will have to add it to their low end also.
Nvidia is also planning to launch Volta in late 2017 on the same 16nm FinFET process.
Cool. I wonder if this means that a launch is imminent.
No. We see these kinds of breadcrumbs dropped usually about 4-6 monnths before launch. That doesn't mean that Nvidia won't launch some of their GPUs early. It depends on what they want to do.

Forum Jump:

Users browsing this thread: 1 Guest(s)