Nvidia was stolen ? A new AI chip is born that is ten times faster than a GPU

Mondo Technology Updated on 2024-02-24

Recently, a startup called Groq is in:Artificial intelligenceThe field caused a sensation. They're launching a brand new AIChipsThe LPU (Language Processing Unit) is known as ten times the speed of Nvidia's GPUs, but the cost is only one-tenth of it. What's innovative about Groq is that they have chosen a completely new system route through LPU'sReasoningThe engine provides the fastest for compute-intensive applications such as large AI modelsReasoningVelocity. It is worth mentioning that the groqChipsThe size of the 230MB is usedsramMemory, with a bandwidth of up to 80TBS, which gives it a clear advantage in terms of performance and speed. This article will unfold from the background of the establishment, technical characteristics and future development, etc., to bring readers a lecture about GroqTechnologyFeast.

Founded in 2016, Groq is a familyArtificial intelligenceSolution Inc. Eight of its founding team members are fromGoogleThey played an important role in the development of TPU. For example, Jona Thanross, the founder and CEO of Groq, designed and implemented the original TPUChipsand is responsible for 20% of the workload in the research and development of TPU. He also joined the GoogleX Rapid Assessment team for:GoogleThe parent company, Alphabet, designs and incubates new innovative projects. The whole team isArtificial intelligenceExperts in their field, their extensive experience and technical prowess provide a solid foundation for the development of Groq.

In addition to a rich background of experience, Groq works with:Googlerelationships also make them inArtificial intelligenceThere are more opportunities and resources in the field. This provides a bigger stage for GROQ to innovate, allowing them to take a different path and choose a different development path from GPU and CPU.

Groq has chosen a completely new system route, the Language Processing Unit (LPU). The LPU is a new type of end-to-end processing unit system with high speedReasoningcan provide the fastest for compute-intensive applications such as large AI modelsReasoningVelocity. LPUs have significant advantages in overcoming bottlenecks in large models, including computational density andMemory bandwidth

In terms of computing density, LPUs have more computing power than GPU CPUs, so they can reduce the computation time per word, resulting in faster generation of text sequences. GroqChipsIt is manufactured on a 14nm process and is equipped with a 230MB systemsrammemory, which allows the LPU to perform orders of magnitude better than the GPU on large models. In addition, Groq'sChipsIt also has an integer operation speed of 750tops and 188tflopsFloating-point arithmeticspeed, which provides a strong support for its high performance.

InMemory bandwidthaspects, groq'sChipsusedsramMemory bandwidthUp to 80TBS this makes:ChipsThe transfer of data is faster and more efficient. sramAs one of the fastest read/write storage devices at present, it has the characteristics of high energy efficiency and high accuracy. Especially after the development of in-memory logic technologysramIt is widely used in demanding fields such as CPU primary buffer and secondary buffer. GroqChipsin usesramAt the same time, it also pays attention to the balance of cost performance, so that it has achieved a good balance between high performance and high efficiency.

The launch of Groq has attracted a lot of attention to theirChipsConsidered subversiveArtificial intelligenceOne of the innovations in the field. The technical advantage of Groq lies not only in the increase in speed, but also in the reduction of costs. ChipsThe cost is only one-tenth of that of Nvidia's GPUs, which makes it easier for people to use high-performance onesChipsand thus pushArtificial intelligenceTechnological developments.

In terms of application areas, Groq'sChipsWith great prospects. First of all, it can be applied to the field of language processing to quickly generate text sequences and improveNatural language processingefficiency. Secondly, Groq'sChipsIt can be applied to large-scale data analysis andMachine learningdomains, providing more computing resources for data scientists and researchers. In addition, Groq'sChipsIt can also be applied to intelligent driving, medical diagnosis and other fields, providing more powerful computing and processing capabilities for the development of these fields. In short, groqChipsInArtificial intelligenceThe field has great application potential, which brings new opportunities for the development of the industry.

Groq as a familyArtificial intelligenceSolutions company, with its brand new AIChipsThe LPU has become the focus of the industry. TheChipsNot only is it ten times faster than Nvidia's GPU, but it also costs only one-tenth of it, making it in theArtificial intelligenceThe field has a huge competitive advantage. The team at Groq has a rich background, and their innovative thinking and technical prowess provide a solid foundation for the company's growth. The technical characteristics and application prospects of Groq also make it inArtificial intelligenceThe field has broad application prospects. In the future, we can expect Groq to be inArtificial intelligenceThe continuous development and breakthrough in the field contribute to the progress of the industry.

Related Pages