BETA
This is a BETA experience. You may opt-out by clicking here

More From Forbes

Edit Story

The Fastest Graphics Card In The World Will Be Unveiled In Two Days

Following
This article is more than 7 years old.

NVIDIA has a countdown timer running on its corporate website with the phrase and hashtag, “IT'S ALMOST TIME #ULTIMATEGEFORCE” prominently featured beneath.  It has been widely speculated that NVIDIA will be unveiling a new flagship desktop GPU at the Game Developers Conference, which will be taking place in San Francisco this coming week. GDC just so happens to align perfectly with NVIDIA’s countdown timer.

NVIDIA Has A Countdown Currently Running On Its Website.

In all likelihood, the rumor mill is correct. In a couple of days, expect NVIDIA.com to be updated and a GeForce GTX 1080 Ti to be unveiled – at least, that’s the safe bet. And if what we know of NVIDIA’s technology and current GPU line-up align with the GeForce GTX 1080 Ti speculation, expect it to be the most powerful graphics card on the market.

There are two products NVIDIA had already revealed in its Quadro graphics line that could be the approximate, pro-graphics equivalents of the GeForce GTX 1080 Ti, the Quadro P6000 and the recently-announced, but not yet available, Quadro GP100.

The Quadro P6000 (Image Source: NVIDIA)

The Quadro P6000 is built around NVIDIA’s GP102 GPU, like the current Pascal-based Titan X. But on the P6000 the GP102’s full complement of 3,840 CUDA cores are enabled – some are disabled on the Titan X. The Quadro P6000 is also outfitted with 24GB of GDDR5X memory, versus 12GB of the Titan X. Overall, the Quadro P6000 has got more CUDA cores than the Titan X, which results in higher peak throughput, and double the frame buffer memory.

Quadro P6000 Performance In 3DMark Time Spy. (Image Source: HotHardware.Com)

I tested the Quadro P6000 recently, and although it is designed for the professional workstation market, I ran a couple of gaming-related tests to see how it stacks up to NVIDIA’s current high-end GeForce line-up. The P6000 ended up being markedly faster than the Titan X.

The Quadro GP100 is a different sort of beast. It is built around NVIDIA’s GP100 GPU and features 16GB of bleeding-edge High Bandwidth 2 (HBM2) memory. The GP100 had previously been relegated to the custom server market and is at the heart of the Tesla P100. With the Quadro GP100 though, the GPU debuts in a traditional two-slot PCI Express add-in card form factor.  Due to the GP100’s reduced number of CUDA cores (3,584) and lower clocks, it doesn’t offer quite the same compute throughput as the P6000, but the GP100’s 4096-bit wide HBM2 memory interface, which runs at 1.4GHz, results in nearly double the memory bandwidth – about 720GB/s vs. 432GB/s.

NVIDIA Quadro GP100 Cards (Image Source: NVIDIA)

The rumored GeForce GTX 1080 Ti could be a consumer-class, tweaked version of either the Quadro P6000 or GP100. I’d lean towards a GP102-based card like the P6000, but with less memory – 12GB is likely.  If NVIDIA gooses up the GP102’s base and boost clocks, and keeps the memory humming at a similar speed, it should be the fastest GPU on the market and the most powerful GeForce card released to date, hence that #ULTIMATEGEFORCE hashtag.

Or I could be completely wrong. We’ll know in a couple of days.