Nvidia packs Pascal into new Quadro graphics cards

[Scorpus]

Nvidia has today announced two new Quadro workstation cards built using the company's new Pascal GPUs. The Quadro P6000 and Quadro P5000 are similar to their consumer counterparts, but feature increased VRAM and target professional applications.

The Quadro P6000 is Nvidia's most powerful workstation card, featuring the same GP102 GPU used in the recently-announced Titan X. However, unlike the Titan X, the P6000 uses a fully-unlocked GP102 chip with 3,840 CUDA cores, providing a huge performance increase over the Maxwell-based Quadro M6000.

While clock speeds aren't known just yet, the P6000 does come with 24 GB of GDDR5X on a 384-bit bus, clocked at 9 Gbps. The card has a rated TDP of 250W, although double-precision performance is limited to just 1/32 of single-precision, similar to last-gen cards.

The Quadro P5000 is based on GP104, also seen in the GeForce GTX 1080. This card features 2,560 CUDA cores, and 16 GB of GDDR5X at 9 Gbps on a 256-bit bus. Like the GTX 1080, we're also looking at a 180W TDP, and the same double-precision performance limitation as the P6000.

Both new Pascal-based Quadro cards will be released in October, although prices aren't yet known. Considering the predecessors to the P6000 and P5000 cost $5,000 and $2,000 respectively, these cards won't come cheap.

Permalink to story.

https://www.techspot.com/news/65713-nvidia-packs-pascal-new-quadro-graphics-cards.html

 

[VitalyT]

Sucker-punch marketing...

nVidia has completely pulled the plug on HBM2, and not because it is unavailable - there is enough of it in production for the professional line of products, but because they want to milk buyers next year from the same chips, to announce a pseudo-new generation, with just HBM2 added.

For every engineer they've hired, they probably got two bean counters.

 

[mateau]

Lets put this into perspective.

Is it a BIG DEAL that P6000 offers 12 T-flops of single precision performance?

If so then how do you respond that 2 Radeon RX 480 will produce 11.1 T-FLOPS for $399.00?

And since DX12 can put those resources to work using Selective Multi-Adaptor then I would suggest to you that NVidia is wasting your money!!!

You know what Selective Multi-Adaptor is don't you? It could be described as Crossfire on steroids!!

Add 2 GPU's and with DX12 you virtually DOUBLE your performance (92% or so). Add three GPU's and you triple performance.

This is why NVidia dropped support for 3 and 4 way SLI. NVidia wants the consumer to spend LOTS of money for big fat expensive GPU's when all you need to do is add another small cheap GPU to your mobo!!

2 RX 480 outperforms 2 GTX 1080 and comes VERY close to the performance of P6000 in raw single precision T-FLOPS!!!!!! And how much is P6000?????

That is why AMD took the legs out from under the mid-price point market.

rtflol

 

[DAOWAce]

stuff about crossfire 480's

Multi-GPU tech is a poor answer to raw single card power. There's so many issues with it, and this is coming from someone who's run SLI for 2 generations over a period of 4-5 years.

Not to mention DX12 is still in its infancy and performance with multi-GPU natively is awful.

All I see here is the equivalent of a console fanboy.