Share on facebook
Share on twitter
Share on linkedin
Share on pinterest
Share on whatsapp

HTML0NVIDIA Hopper will be set to become an enormous GPU.

Based on the most recent information provided by @kopite7kimi about NVIDIA’s upcoming HPC accelerator, the GH100 will feature the most considerable die size. It is just a bit smaller than 1,000mm2. It means that the GPU could be as large as 20% bigger than GA100 built on Ampere technology.

If you buy something from a this link, myelectricsparks Media may earn a commission. See our Read More.

NVIDIA's GH100

Yesterday, we discussed the “Hopper” trademark dispute that NVIDIA is fighting against a firm with identical product names. Suppose NVIDIA’s ongoing negotiations with the business “Dish Network” will not be resolved within the next few weeks. In that case, NVIDIA might not be in a position of revealing the brand new GH100 GPU before GTC 2022, when it will be held in March.

Furthermore, it appears that the rumours about that Multi-Chip-Module (MCM) model of GH100 aren’t accurate. There are two leakers like @kopite7kimi and @greymon55, affirm that GH100 is, in fact, an all-one-die model. But, it is apparent to be a Hopper design for the MCM. It is believed to be a “GH102” GH102 GPU.

It is anticipated that the GH100 could also be an extremely robust processor. Kopite has previously said that the TDP could be as high as 1000W, which will require a powerful cooling system. However, Kopite does have its water-cooled data centre GPU series, known as HGX, and, as such, technically, it shouldn’t pose problems.