[Performance]: No signficant speedup from Wfp8Afp8 (vs Wbf16Abf16) in Llama-4 Scout

[kailashbuki]

Proposal to improve performance

No response

Report of performance regression

Dynamic FP8 quantization currently does not lead to significant speedup for Llama-4 Scout model in 8xH200 at smaller batch sizes. Both BF16 and FP8 checkpoints use fused Triton MoE kernel w/ tuned config. The table below reports p50 end-to-end latency in seconds using benchmark_latency.py script.

BS	input len	output len	meta-llama/Llama-4-Scout-17B-16E-Instruct	RedHatAI/Llama-4-Scout-17B-16E-Instruct-FP8-dynamic	Speedup
4	32	128	1.0608315134886652	1.0581926168175415	1.0x
8	32	128	1.1672972993692383	1.1296635085018352	1.03x
32	32	128	1.5841029789298773	1.4512220670003444	1.09x
8	10000	100	2.044864689465612	2.0477603524923325	0.099x

python benchmark_latency.py \
    --model meta-llama/Llama-4-Scout-17B-16E-Instruct \
    --max-model-len 131072 \
    --tensor-parallel-size 8 \
    --input-len $IL \
    --output-len $OL \
    --batch-size $BS

python benchmark_latency.py \
    --model RedHatAI/Llama-4-Scout-17B-16E-Instruct-FP8-dynamic \
    --max-model-len 131072 \
    --tensor-parallel-size 8 \
    --input-len $IL \
    --output-len $OL \
    --batch-size $BS

Misc discussion on performance

No response

Your current environment (if you think it is necessary)

The output of `python collect_env.py`

INFO 06-16 20:49:44 [__init__.py:243] Automatically detected platform cuda.
Collecting environment information...
==============================
        System Info
==============================
OS                           : Ubuntu 24.04.2 LTS (x86_64)
GCC version                  : (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0
Clang version                : Could not collect
CMake version                : version 3.28.3
Libc version                 : glibc-2.39

==============================
       PyTorch Info
==============================
PyTorch version              : 2.7.0+cu126
Is debug build               : False
CUDA used to build PyTorch   : 12.6
ROCM used to build PyTorch   : N/A

==============================
      Python Environment
==============================
Python version               : 3.12.3 (main, Feb  4 2025, 14:48:35) [GCC 13.3.0] (64-bit runtime)
Python platform              : Linux-6.8.0-1029-aws-x86_64-with-glibc2.39

==============================
       CUDA / GPU Info
==============================
Is CUDA available            : True
CUDA runtime version         : 12.8.93
CUDA_MODULE_LOADING set to   : LAZY
GPU models and configuration : 
GPU 0: NVIDIA H200
GPU 1: NVIDIA H200
GPU 2: NVIDIA H200
GPU 3: NVIDIA H200
GPU 4: NVIDIA H200
GPU 5: NVIDIA H200
GPU 6: NVIDIA H200
GPU 7: NVIDIA H200

Nvidia driver version        : 570.133.20
cuDNN version                : Could not collect
HIP runtime version          : N/A
MIOpen runtime version       : N/A
Is XNNPACK available         : True

==============================
          CPU Info
==============================
Architecture:                         x86_64
CPU op-mode(s):                       32-bit, 64-bit
Address sizes:                        46 bits physical, 48 bits virtual
Byte Order:                           Little Endian
CPU(s):                               192
On-line CPU(s) list:                  0-191
Vendor ID:                            GenuineIntel
Model name:                           Intel(R) Xeon(R) Platinum 8488C
CPU family:                           6
Model:                                143
Thread(s) per core:                   2
Core(s) per socket:                   48
Socket(s):                            2
Stepping:                             8
BogoMIPS:                             4800.00
Flags:                                fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc arch_perfmon rep_good nopl xtopology nonstop_tsc cpuid aperfmperf tsc_known_freq pni pclmulqdq monitor ssse3 fma cx16 pdcm pcid sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves avx_vnni avx512_bf16 wbnoinvd ida arat avx512vbmi umip pku ospke waitpkg avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg tme avx512_vpopcntdq rdpid cldemote movdiri movdir64b md_clear serialize amx_bf16 avx512_fp16 amx_tile amx_int8 flush_l1d arch_capabilities
Hypervisor vendor:                    KVM
Virtualization type:                  full
L1d cache:                            4.5 MiB (96 instances)
L1i cache:                            3 MiB (96 instances)
L2 cache:                             192 MiB (96 instances)
L3 cache:                             210 MiB (2 instances)
NUMA node(s):                         2
NUMA node0 CPU(s):                    0-47,96-143
NUMA node1 CPU(s):                    48-95,144-191
Vulnerability Gather data sampling:   Not affected
Vulnerability Itlb multihit:          Not affected
Vulnerability L1tf:                   Not affected
Vulnerability Mds:                    Not affected
Vulnerability Meltdown:               Not affected
Vulnerability Mmio stale data:        Not affected
Vulnerability Reg file data sampling: Not affected
Vulnerability Retbleed:               Not affected
Vulnerability Spec rstack overflow:   Not affected
Vulnerability Spec store bypass:      Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:             Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:             Mitigation; Enhanced / Automatic IBRS; IBPB conditional; RSB filling; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds:                  Not affected
Vulnerability Tsx async abort:        Not affected

==============================
Versions of relevant libraries
==============================
[pip3] numpy==1.26.4
[pip3] nvidia-cublas-cu12==12.6.4.1
[pip3] nvidia-cuda-cupti-cu12==12.6.80
[pip3] nvidia-cuda-nvrtc-cu12==12.6.77
[pip3] nvidia-cuda-runtime-cu12==12.6.77
[pip3] nvidia-cudnn-cu12==9.5.1.17
[pip3] nvidia-cufft-cu12==11.3.0.4
[pip3] nvidia-cufile-cu12==1.11.1.6
[pip3] nvidia-curand-cu12==10.3.7.77
[pip3] nvidia-cusolver-cu12==11.7.1.2
[pip3] nvidia-cusparse-cu12==12.5.4.2
[pip3] nvidia-cusparselt-cu12==0.6.3
[pip3] nvidia-ml-py==12.575.51
[pip3] nvidia-nccl-cu12==2.26.2
[pip3] nvidia-nvjitlink-cu12==12.6.85
[pip3] nvidia-nvtx-cu12==12.6.77
[pip3] pynvml==12.0.0
[pip3] pyzmq==26.4.0
[pip3] torch==2.7.0
[pip3] torch-tb-profiler==0.4.3
[pip3] torchaudio==2.7.0
[pip3] torchvision==0.22.0
[pip3] transformers==4.52.4
[pip3] triton==3.3.0
[conda] Could not collect

==============================
         vLLM Info
==============================
ROCM Version                 : Could not collect
Neuron SDK Version           : N/A
vLLM Version                 : 0.9.1.dev128+gf32fcd944 (git sha: f32fcd944)
vLLM Build Flags:
  CUDA Archs: Not Set; ROCm: Disabled; Neuron: Disabled
GPU Topology:
        GPU0    GPU1    GPU2    GPU3    GPU4    GPU5    GPU6    GPU7    CPU Affinity    NUMA Affinity   GPU NUMA ID
GPU0     X      NV18    NV18    NV18    NV18    NV18    NV18    NV18    0-47,96-143     0               N/A
GPU1    NV18     X      NV18    NV18    NV18    NV18    NV18    NV18    0-47,96-143     0               N/A
GPU2    NV18    NV18     X      NV18    NV18    NV18    NV18    NV18    0-47,96-143     0               N/A
GPU3    NV18    NV18    NV18     X      NV18    NV18    NV18    NV18    0-47,96-143     0               N/A
GPU4    NV18    NV18    NV18    NV18     X      NV18    NV18    NV18    48-95,144-191   1               N/A
GPU5    NV18    NV18    NV18    NV18    NV18     X      NV18    NV18    48-95,144-191   1               N/A
GPU6    NV18    NV18    NV18    NV18    NV18    NV18     X      NV18    48-95,144-191   1               N/A
GPU7    NV18    NV18    NV18    NV18    NV18    NV18    NV18     X      48-95,144-191   1               N/A

Legend:

  X    = Self
  SYS  = Connection traversing PCIe as well as the SMP interconnect between NUMA nodes (e.g., QPI/UPI)
  NODE = Connection traversing PCIe as well as the interconnect between PCIe Host Bridges within a NUMA node
  PHB  = Connection traversing PCIe as well as a PCIe Host Bridge (typically the CPU)
  PXB  = Connection traversing multiple PCIe bridges (without traversing the PCIe Host Bridge)
  PIX  = Connection traversing at most a single PCIe bridge
  NV#  = Connection traversing a bonded set of # NVLinks

==============================
     Environment Variables
==============================
LD_LIBRARY_PATH=/opt/amazon/efa/lib:/opt/amazon/openmpi/lib:/opt/amazon/ofi-nccl/lib:/usr/local/cuda-12.8/lib:/usr/local/cuda-12.8/lib64:/usr/local/cuda-12.8:/usr/local/cuda-12.8/targets/x86_64-linux/lib/:/usr/local/cuda-12.8/extras/CUPTI/lib64:/usr/local/lib:/usr/lib:/opt/amazon/efa/lib:/opt/amazon/openmpi/lib:/opt/amazon/ofi-nccl/lib:/usr/local/cuda-12.8/lib:/usr/local/cuda-12.8/lib64:/usr/local/cuda-12.8:/usr/local/cuda-12.8/targets/x86_64-linux/lib/:/usr/local/cuda-12.8/extras/CUPTI/lib64:/usr/local/lib:/usr/lib:/opt/amazon/efa/lib:/opt/amazon/openmpi/lib:/opt/amazon/ofi-nccl/lib:/usr/local/cuda-12.8/lib:/usr/local/cuda-12.8/lib64:/usr/local/cuda-12.8:/usr/local/cuda-12.8/targets/x86_64-linux/lib/:/usr/local/cuda-12.8/extras/CUPTI/lib64:/usr/local/lib:/usr/lib
NCCL_CUMEM_ENABLE=0
PYTORCH_NVML_BASED_CUDA_CHECK=1
TORCHINDUCTOR_COMPILE_THREADS=1
CUDA_MODULE_LOADING=LAZY

Before submitting a new issue...

• Make sure you already searched for relevant issues, and asked the chatbot living at the bottom right corner of the documentation page, which can answer lots of frequently asked questions.

[kailashbuki]

tagging @mgoin, @simon-mo based on commit history

[mgoin]

Hi @kailashbuki. Using such a high TP size as 8 (--tensor-parallel-size 8) for a relative small MoE does not seem efficient. I would recommend exploring enabling expert parallelism if you must keep a large TP size (--enable-expert-parallel)

If your logs report there not being any tuned configs found, I recommend generating tuned configs for your deployment using benchmark/kernels/benchmark_moe.py. See https://github.com/vllm-project/vllm/blob/main/vllm/model_executor/layers/fused_moe/configs/README

[kailashbuki]

Thank you, @mgoin! Those benchmarks were produced using tuned configs obtained from the tuning script you shared.

When we reduce the TP degree, end-to-end latency still does not improve as shown in the table below. The first row below is mainly as a reference from the table above.

TP Degree	BS	input len	output len	meta-llama/Llama-4-Scout-17B-16E-Instruct	RedHatAI/Llama-4-Scout-17B-16E-Instruct-FP8-dynamic	Speedup
8	8	10000	100	2.044864689465612	2.0477603524923325	0.099x
4	8	10000	100	3.381372621515766	3.248984132427722	1.04x
2	8	10000	100	5.404737142496742	4.885081650456414	1.1x

For top-1 expert MoE models like Llama-4, using EP seems like a waste unless we aim for a very high concurrency, no?