Tag: research

Apr 30, 2026

• ai-hardware
• chip-design
• diffusion-models
• scientific-ml
• physics-simulation
• equivariant-networks
• GNN
• architecture-proposal
• research

Apr 11, 2026

• ai-hardware
• energy-efficiency
• memory-wall
• number-formats
• analog-compute
• parallelism
• research
• part
• the
• industry
• MXFP
• BitNet
• analog
• horowitz
• HBM
• architecture
• how
• key
• interesting

Apr 11, 2026

• inference
• benchmarking
• asic
• architecture
• research
• why
• workload
• LLM
• diffusion
• MoE
• vision
• systolic
• SIMT
• In-Memory
• dataflow
• reconfigurable
• worked
• NVIDIA
• groq
• google
• comparison
• key
• proposed
• useful
• cost
• energy
• flexibility
• validation
• calibration
• publication

Apr 11, 2026

• inference
• optimization
• quantization
• cuda
• blackwell
• moe
• kv-cache
• synthesis
• research
• 1
• 2
• 3
• TrtLLMGen
• thunder
• 4
• 5
• 6
• 7
• baseline
• with
• 8
• 9
• 10

Apr 11, 2026

• kv-cache
• quantization
• spectral-methods
• inference-optimization
• llm-serving
• memory-compression
• research
• 1
• 2
• 3
• 4
• integration
• 5
• hardware
• verified
• what
• 6
• 7

Apr 11, 2026

• kv-cache
• quantization
• inference
• attention
• compression
• spectral-methods
• transformer-internals
• research
• executive
• 1
• 2
• participation
• the
• what
• 3
• 4
• task
• statistical
• distribution
• 5
• why
• connection
• 6
• KIVI
• loki
• KV-CoRE
• RoPE
• random
• Rate-Distortion
• 7
• memory
• throughput
• calibration
• compatibility
• 8
• tested
• layer
• Training-Time
• dynamic
• interaction
• theoretical
• 9

Apr 11, 2026

• nvidia
• tensorrt-llm
• flashinfer
• moe
• cuda
• blackwell
• sm100
• inference
• open-source
• mlperf
• research
• 1
• the
• where
• 2
• why
• 3
• 4
• what
• NVIDIA
• 5
• MLPerf
• InferenceX
• 6
• 7
• Short-Term
• Medium-Term
• 8
• 9

Apr 11, 2026

• robotics
• edge-inference
• diffusion-policy
• VLA
• hardware
• research
• the
• why
• what

Apr 05, 2026

• semiconductors
• ai-infra
• nvidia
• tsmc
• research
• analyst-note
• executive
• technical
• the
• CoWoS-L
• supply
• CoWoS
• HBM3e
• bill
• financial
• bull
• bear
• key
• sources
• interesting

Apr 05, 2026

• semiconductors
• foundry
• tsmc
• gaa
• research
• analyst-note
• executive
• technical
• FinFET
• N2
• the
• yield
• supply
• TSMC
• customer
• financial
• who
• bull
• bear
• key
• interesting

Apr 05, 2026

• apple
• ai-infra
• edge-ai
• semiconductors
• research
• analyst-note
• executive
• technical
• the
• silicon
• supply
• data
• financial
• PCC
• bull
• bear
• key
• sources
• interesting