This site may earn chapter commissions from the links on this page. Terms of employ.

NVLink-Feature

For the by 12 months, Nvidia has boring-walked its GP100 GPU — that'due south Pascal, but equipped with HBM2 and a full suite of technological features meant to entreatment to the HPC and scientific communities — into its high-finish markets. Initially, the card was only bachelor with a passive cooler and was intended solely for use in high-end servers. The first Quadro card to apply a Pascal GPU were the P5000 and P6000, just these were derived from GP102 and combined a Pascal-class graphics processor with GDDR5X, the memory technology Nvidia used to proficient issue with its GeForce 10xx GPU family. Now Nvidia has announced the Quadro GP100 — an ultra-loftier cease GPU based on the full Pascal implementation, with HBM2 attached also.

QuickSpecs

Traditionally, Nvidia'south highest-cease cards slot into the Tesla and Quadro product lines, which don't always follow the consumer bicycle'south refresh patterns. This was specially true with Maxwell, which debuted in the Quadro family unit, but didn't replace all of Nvidia's Kepler-derived parts in the Tesla sectionalization. There were some questions about whether Nvidia would bring the GP100 GPU to Quadro at all, since the P5000 and P6000 shipped months ago and offered more enough horsepower. In fact, the Quadro GP100 offers less FLOPS performance than the GP102-derived P6000. GP100 is a 3584:224:128 card (core count:texture units:render outputs) as compared with GP102 at 3840:240:96.

Where GP100 distinguishes itself is in three areas. Kickoff, it does offer 128 render outputs as opposed to 96 on GP102, though information technology's past no means articulate that any high-finish workstation workloads will do good from this. Second, it offers an estimated 720GB/s of retention bandwidth, up substantially from the 432GB/s on GP102. Third, it has custom brackets that back up NVLink connections, significant you can hypothetically hook two Quadro GP100 GPUs together in the same machine, with up to 80GB/due south (40GB/s per link, with two link brackets).

NVLinkConnections

But pushing GP100 out to the Quadro family sets up another potential collision between GP100 and GP102 — total retentiveness chapters. GP100 tops out at 16GB, as all first-generation HBM2 cards are expected to do, while GP102 offers up to 24GB of GDDR5X. One reason why the workstation and HPC markets take an ambition for additional RAM is because many applications must be able to load their unabridged working data sets into retention before they tin can return them. The caste to which this is true depends to at least some degree on the awarding and which version(s) of CUDA information technology supports, but it'due south been a persistent limitation when we've previously done workstation testing or spoken to various vendors. Their differing capacities suggest that some markets will remain better-served by the GP102-derived P6000 with its 24GB of RAM than the 16GB on Quadro GP100.

QuadroProducts

But then again, there's opportunity here, too. When reviews surface of the new chip, we'll get a await at the unlike strengths of HBM2 and perhaps a clearer picture of where providing huge amounts of memory bandwidth tin amend functioning or reduce it. While the two cards are not strictly apples-to-apples, they're likely the closest await we'll ever go at the "same" chip running on ii dissimilar memory interfaces. If the Quadro GP100 proves to accept 128 ROPS, that volition tilt fillrate-heavy tests in favor of the HBM2 chip. But this tin can be ameliorated by workload pick — applications that don't button the 96 ROPS on P6000 won't push the 128 ROPS on GP100, either.

AMD has already stated that its upcoming Vega GPU refresh will use HBM2, also hereafter high-end FirePro cards based on that architecture. Nvidia has yet to reveal details concerning any refreshed Pascal GPUs that might be arriving in 2017.