During a special address, chipmaker Nvidia unveils the H200, its latest AI chip to assist with generative AI tasks. Never miss a deal again! See CNET’s browser extension 👉 bit.ly/39Ub3bv Subscribe to our channel: / @cnethighlights
So, under 10 years the average gamer will have GPU's that create real time Game creation environments. Today 40% of the planet actively plays video games. sounds like someone is going to make a ton of money.
Why are they using GPT3? Only the latest model matters... Likely just to make the improvement leaps look larger than they really are. Performance leaps matter for GPT4, GPT5 or Gemini. And l am guessing the compute needed here is either keeping pace with improvements, or outstripping it
Its not unrivaled for certain inference loads requiring more memory and bandwidth MI300x can challenge. H200 changes the memory technology from H100 while also enabling the sixth stack of previously disabled HBM while also increasing memory density. This means that aggregate full device capacity moves from a previous maximum of 80 GBs to 144 GBs, 141 GB of which are addressable by applications. Bandwidth comes in at a massive 4.8 TB/s, a 40+% increase over H100’s 3.35TB/s. By comparison, AMD’s MI300X has 192 GB of HBM3 memory for 5.2 TB/sec bandwidth, although we’ll hear more about that later this year at the December 6th AMD event.
When are standard servers going to get similar upgrades or alternative not everyone can rent a super computer........ They might need a whole different application to even utilize things with all this power now like just delivering band width even if ai was used to compress it something need to be made to host server even from home on pcs or on server racks...
4:10 when u compare a gpu directly to a cpu and the performance is less than 200x for optimized tasks its just embarrassing. Why even compare the gh200 to a xeon. Everyone can see the comparison is janky
None of this matters when it's only available for the top 1-2% of the population. Nvidia controlling the market demand by cutting supply to increase prices for the goal of making more profit. I watched all summer the 4090 available on nvidia's website, the moment news was announced about sanctions "out of stock" Then during the week day it was back in stock then right as soon as Halloween passed "out of stock" normal \ average western people have not seen a stable supply since TU104 12nm (aka RTX 2080 or Turing) nvidia's new slogan should be "Nvidia - always out of stock" At this point it's obvious Nvidia no longer wants to deal with "peasants" as we are now in the PC gaming world. They want top end GPU's no lower than $2000-$2500 USD mid $1000-$1500 and lower $500+ People making less than $100k per/year will struggle to justify $1000 for a GPU let alone $2000+
AMD MI300 Instinct still isn't out yet 😅 Its going to take a long time for someone to check NVIDIAs power. They are relentless. Was this new beast even needed.
H200 is somewhat disappointing. It only has 141G of HBM3. AMD’s MI300X has 192G of HBM3. Big LLM needs a lot of VRAM. MSFT, Meta, & Amazon will buy the MI300X.
“When IBM comes in [a market segment], we run away” ~Jensen Huang (paraphrased, in a recent interview speaking about going against the ARC GPU). He knows where his company can compete, and where they can’t. IBM has been putting metric shittons of money into QC for years.
Taiwan has leveraged unfair policies, including low-cost labor, land subsidies, tax incentives, low electricity and water prices, and a favorable interest rate environment. Additionally, it has benefited from institutional advantages such as low-cost universal education. Seizing the opportunity presented by the semiconductor industry's shift after the US-Japan semiconductor war in the 1980s, Taiwan has vigorously developed the semiconductor manufacturing and chip design industries. Currently, due to the advantages of unfair policies, many semiconductor companies in Taiwan have grown to the extent that they can compete with leading US companies. For example, MediaTek surpassed Qualcomm to become the world's top mobile chip processor company. This success is not solely attributed to MediaTek's capabilities but stems from the unfair competitive advantage. Given this situation, the United States cannot tolerate the continuation of such circumstances. It is time to counteract and dismantle Taiwan's semiconductor industry, including companies like TSMC and MediaTek.