Intel retreats from AI race with NVIDIA

Intel abandons fight with NVIDIA and bets on new ‘revolutionary’ chip to dominate the AI ​​market!

According to a report from Wccftecha Intel decided to move away from the competition with NVIDIA with regards to computing power in IA and the large-scale AI model training market. The company is now exploring a less saturated segment of the AI ​​market, focusing on its new low-cost AI accelerator, the Gaudi 3.

Intel intends to position the Gaudi 3 as the product with the best value for money, although the Gaudi 3 “is not reaching” the latest GPU of NVIDIA in terms of performance, according to the words of Anil Nanduri, head of Intel’s AI acceleration office, during an interview with CRN.

Still, Nanduri points out that the accelerator chip Gaudi 3 is ideal for supporting cost-effective systems that run task-based and open source for companies.

On the other hand, as indicated in the report of the WccftechIntel claims that after the introduction of models LLM “focused on reasoning”, their line Gaudi 3 offers performance comparable to well-known AI accelerator H100 from NVIDIA, especially in inference workloads.

According to the report of CRNIntel also claims that the Gaudi 3 is about 9% faster than H100 not model Llama 3 and 80% more economical; already in the model Llama 2Gaudi 3 is 19% faster, with a price-performance difference of up to 2 times.

However, when evaluated in terms of floating point operations, the Gaudi 3 AI GPUs still lag behind the options of NVIDIAshowing that cutting-edge AI performance is not, at the moment, Intel’s strong point.

Therefore, according to the report of WccftechIntel does not plan to compete directly with Intel’s GPUs. NVIDIA. Additionally, a new competitor is emerging in the AI ​​computing sector: AMD.

As for Intel’s strategy, the company believes that smaller models of LLM will gain greater acceptance as the initial enthusiasm around AI and interest in data centers large-scale operations have declined, as indicated by the Wccftech.