This site may earn affiliate commissions from the links on this folio. Terms of use.

NVLink-Feature

For the by 12 months, Nvidia has ho-hum-walked its GP100 GPU — that's Pascal, just equipped with HBM2 and a full suite of technological features meant to entreatment to the HPC and scientific communities — into its loftier-end markets. Initially, the carte was only available with a passive cooler and was intended solely for use in loftier-end servers. The first Quadro card to use a Pascal GPU were the P5000 and P6000, just these were derived from GP102 and combined a Pascal-class graphics processor with GDDR5X, the retentivity technology Nvidia used to skillful effect with its GeForce 10xx GPU family. Now Nvidia has announced the Quadro GP100 — an ultra-high terminate GPU based on the total Pascal implementation, with HBM2 fastened as well.

QuickSpecs

Traditionally, Nvidia's highest-stop cards slot into the Tesla and Quadro production lines, which don't always follow the consumer cycle's refresh patterns. This was particularly true with Maxwell, which debuted in the Quadro family unit, just didn't replace all of Nvidia'southward Kepler-derived parts in the Tesla segmentation. In that location 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 than than enough horsepower. In fact, the Quadro GP100 offers less FLOPS functioning than the GP102-derived P6000. GP100 is a 3584:224:128 menu (core count:texture units:render outputs) every bit compared with GP102 at 3840:240:96.

Where GP100 distinguishes itself is in three areas. Commencement, it does offer 128 render outputs every bit opposed to 96 on GP102, though information technology'south past no means clear that any loftier-finish workstation workloads will benefit from this. Second, it offers an estimated 720GB/s of retention bandwidth, up essentially from the 432GB/south on GP102. Third, it has custom brackets that support NVLink connections, meaning you can hypothetically hook two Quadro GP100 GPUs together in the same automobile, with upwardly to 80GB/s (40GB/southward per link, with two link brackets).

NVLinkConnections

But pushing GP100 out to the Quadro family sets up some other potential standoff betwixt GP100 and GP102 — total retentivity chapters. GP100 tops out at 16GB, as all first-generation HBM2 cards are expected to practice, while GP102 offers up to 24GB of GDDR5X. 1 reason why the workstation and HPC markets have an appetite for additional RAM is because many applications must be able to load their unabridged working data sets into retention earlier they can render them. The degree to which this is true depends to at to the lowest degree some degree on the awarding and which version(southward) of CUDA it supports, but it'southward been a persistent limitation when we've previously done workstation testing or spoken to diverse 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 and so again, there's opportunity here, as well. When reviews surface of the new scrap, we'll get a look at the dissimilar strengths of HBM2 and maybe a clearer picture of where providing huge amounts of memory bandwidth tin can improve operation or reduce information technology. While the two cards are not strictly apples-to-apples, they're likely the closest look we'll ever get at the "aforementioned" chip running on two different retentivity interfaces. If the Quadro GP100 proves to take 128 ROPS, that will tilt fillrate-heavy tests in favor of the HBM2 chip. Merely this can exist ameliorated by workload option — applications that don't push 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, as well future loftier-stop FirePro cards based on that architecture. Nvidia has still to reveal details concerning whatsoever refreshed Pascal GPUs that might be arriving in 2022.