SambaNova, Sonic AI

Skip to content

Home Discover Library Ask Sonic Projects

Use with Claude or ChatGPT

Home Discover Library Ask Sonic Projects

Use with Claude or ChatGPT

SambaNova, Sonic AI

Home/Library/SambaNova

S

SambaNova

Company · Tech

19

Mentions

3

Episodes

19

Claims

SambaNova, mentioned 19 times across podcast episodes and expert conversations analyzed by Sonic.

What SambaNova has said

Star Cloud plans to fly chips from SambaNova and Cerebras in its spacecraft.

bullish·Philip Johnston

SambaNova's "premium inference" technology enables running trillion-parameter models at full precision and high speeds.

The SambaNova SM50 chip delivers higher performance for premium inference than GPUs at a fraction of the cost and power consumption.

bullish·Rodrigo

Star Cloud plans to fly chips from SambaNova and Cerebras in its spacecraft.

Expert perspectivePhilip JohnstonJul 25

SambaNova's "premium inference" technology enables running trillion-parameter models at full precision and high speeds.

Official sourceRodrigoJul 25

The SambaNova SM50 chip delivers higher performance for premium inference than GPUs at a fraction of the cost and power consumption.

Official sourceRodrigoJul 25

SambaNova is shipping its new SM50 chip, a cloud-scale Reconfigurable Dataflow Unit (RDU), later this year.

Official sourceRodrigoJul 25

SambaNova's latest chip, the SM50, is designed for disaggregated inference, working alongside hardware from NVIDIA and Intel.

Official sourceRodrigo LiangJul 17

SambaNova runs large models at their original, full precision without quantization, which differentiates its performance.

Official sourceRodrigo LiangJul 17

The minimum deployment size for running a large model like DeepMind's 1.5 trillion parameter model on some providers' hardware is 10 to 20 racks, while SambaNova's minimum is a single rack.

Expert perspectiveRodrigo LiangJul 17

SambaNova's Generation 5 chip is scheduled to begin shipping later this year.

Official sourceRodrigo LiangJul 17

SambaNova has taped out 6 chips in the last 7 years and will tape out its 7th chip next year.

Official sourceRodrigo LiangJul 17

A single SambaNova rack can run a trillion-parameter model, a task that would require dozens of racks of competing hardware.

Expert perspectiveRodrigo LiangJul 17

A single 10-kilowatt SambaNova SN40 rack outperforms a 130-140 kilowatt rack of NVIDIA GPUs for inference tasks.

Expert perspectiveRodrigo LiangJul 17

Using NVIDIA GPUs for the prefill phase and SambaNova racks for the decode phase of inference can provide a 2 to 3 times throughput advantage at the same cost.

Expert perspectiveRodrigo LiangJul 8

A SambaNova rack consumes 10 kilowatts of power, whereas a typical GPU rack consumes 100 kilowatts.

Expert perspectiveRodrigo LiangJul 8

SambaNova's proprietary SM40 and SM50 technology is claimed to be 5 to 10 times faster on the decode phase of inference when compared to an NVIDIA GPU.

Expert perspectiveRodrigo LiangJul 8

SambaNova is one of a group of startups aiming to challenge NVIDIA by supplying AI infrastructure specifically for inference workloads.

Expert perspectiveEd LudlowJul 8

SambaNova's architecture uses N-1 generation HBM, which is in mature production, allowing it to run trillion-parameter models without being constrained by the supply of the newest HBM.

Expert perspectiveRodrigo LiangJul 8

SambaNova claims that using NVIDIA for the prefill phase of inference and a SambaNova rack for the decode phase can yield a 2 to 3 times throughput advantage over an all-NVIDIA setup.

Expert perspectiveRodrigo LiangJul 8

A SambaNova rack consumes 10 kilowatts of power, compared to 100 kilowatts for a typical GPU rack.

Expert perspectiveRodrigo LiangJul 8

SambaNova's SM40 and SM50 platforms are 5 to 10 times faster on the decode phase of inference compared to an NVIDIA GPU.

Expert perspectiveEd LudlowJul 8

Create a free account to see SambaNova's full intelligence report - every claim, the relationship network, and AI Q&A across all sources. No card needed.

Get started free

Back to Entities Entity Detail