[Bug]: The inference speed of the whisper model under the v1 engine is much slower than v0

[qaz-t]

Your current environment

The output of python collect_env.py

Collecting environment information...
==============================
        System Info
==============================
OS                           : Ubuntu 22.04.5 LTS (x86_64)
GCC version                  : (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0
Clang version                : Could not collect
CMake version                : version 3.22.1
Libc version                 : glibc-2.35

==============================
       PyTorch Info
==============================
PyTorch version              : 2.8.0+cu128
Is debug build               : False
CUDA used to build PyTorch   : 12.8
ROCM used to build PyTorch   : N/A

==============================
      Python Environment
==============================
Python version               : 3.12.9 | packaged by Anaconda, Inc. | (main, Feb  6 2025, 18:56:27) [GCC 11.2.0] (64-bit runtime)
Python platform              : Linux-5.15.0-136-generic-x86_64-with-glibc2.35

==============================
       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 RTX A5000
GPU 1: NVIDIA RTX A5000

Nvidia driver version        : 535.230.02
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):                               80
On-line CPU(s) list:                  0-79
Vendor ID:                            GenuineIntel
Model name:                           Intel(R) Xeon(R) Gold 6138 CPU @ 2.00GHz
CPU family:                           6
Model:                                85
Thread(s) per core:                   2
Core(s) per socket:                   20
Socket(s):                            2
Stepping:                             4
CPU max MHz:                          3700.0000
CPU min MHz:                          1000.0000
BogoMIPS:                             4000.00
Flags:                                fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid dca sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch cpuid_fault epb cat_l3 cdp_l3 invpcid_single pti intel_ppin ssbd mba ibrs ibpb stibp tpr_shadow vnmi flexpriority ept vpid ept_ad fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm cqm mpx rdt_a avx512f avx512dq rdseed adx smap clflushopt clwb intel_pt avx512cd avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local dtherm ida arat pln pts pku ospke md_clear flush_l1d arch_capabilities
Virtualization:                       VT-x
L1d cache:                            1.3 MiB (40 instances)
L1i cache:                            1.3 MiB (40 instances)
L2 cache:                             40 MiB (40 instances)
L3 cache:                             55 MiB (2 instances)
NUMA node(s):                         2
NUMA node0 CPU(s):                    0-19,40-59
NUMA node1 CPU(s):                    20-39,60-79
Vulnerability Gather data sampling:   Mitigation; Microcode
Vulnerability Itlb multihit:          KVM: Mitigation: VMX disabled
Vulnerability L1tf:                   Mitigation; PTE Inversion; VMX conditional cache flushes, SMT vulnerable
Vulnerability Mds:                    Mitigation; Clear CPU buffers; SMT vulnerable
Vulnerability Meltdown:               Mitigation; PTI
Vulnerability Mmio stale data:        Mitigation; Clear CPU buffers; SMT vulnerable
Vulnerability Reg file data sampling: Not affected
Vulnerability Retbleed:               Mitigation; IBRS
Vulnerability Spec rstack overflow:   Not affected
Vulnerability Spec store bypass:      Mitigation; Speculative Store Bypass disabled via prctl and seccomp
Vulnerability Spectre v1:             Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:             Mitigation; IBRS; IBPB conditional; STIBP conditional; RSB filling; PBRSB-eIBRS Not affected; BHI Not affected
Vulnerability Srbds:                  Not affected
Vulnerability Tsx async abort:        Mitigation; Clear CPU buffers; SMT vulnerable

==============================
Versions of relevant libraries
==============================
[pip3] flashinfer-python==0.3.0
[pip3] numpy==2.2.6
[pip3] nvidia-cublas-cu12==12.8.4.1
[pip3] nvidia-cuda-cupti-cu12==12.8.90
[pip3] nvidia-cuda-nvrtc-cu12==12.8.93
[pip3] nvidia-cuda-runtime-cu12==12.8.90
[pip3] nvidia-cudnn-cu12==9.10.2.21
[pip3] nvidia-cudnn-frontend==1.14.1
[pip3] nvidia-cufft-cu12==11.3.3.83
[pip3] nvidia-cufile-cu12==1.13.1.3
[pip3] nvidia-curand-cu12==10.3.9.90
[pip3] nvidia-cusolver-cu12==11.7.3.90
[pip3] nvidia-cusparse-cu12==12.5.8.93
[pip3] nvidia-cusparselt-cu12==0.7.1
[pip3] nvidia-ml-py==13.580.82
[pip3] nvidia-nccl-cu12==2.27.3
[pip3] nvidia-nvjitlink-cu12==12.8.93
[pip3] nvidia-nvtx-cu12==12.8.90
[pip3] pynvml==13.0.1
[pip3] pyzmq==27.1.0
[pip3] torch==2.8.0+cu128
[pip3] torchaudio==2.8.0+cu128
[pip3] torchvision==0.23.0+cu128
[pip3] transformers==4.56.1
[pip3] triton==3.4.0
[conda] flashinfer-python         0.3.0                    pypi_0    pypi
[conda] numpy                     2.2.6                    pypi_0    pypi
[conda] nvidia-cublas-cu12        12.8.4.1                 pypi_0    pypi
[conda] nvidia-cuda-cupti-cu12    12.8.90                  pypi_0    pypi
[conda] nvidia-cuda-nvrtc-cu12    12.8.93                  pypi_0    pypi
[conda] nvidia-cuda-runtime-cu12  12.8.90                  pypi_0    pypi
[conda] nvidia-cudnn-cu12         9.10.2.21                pypi_0    pypi
[conda] nvidia-cudnn-frontend     1.14.1                   pypi_0    pypi
[conda] nvidia-cufft-cu12         11.3.3.83                pypi_0    pypi
[conda] nvidia-cufile-cu12        1.13.1.3                 pypi_0    pypi
[conda] nvidia-curand-cu12        10.3.9.90                pypi_0    pypi
[conda] nvidia-cusolver-cu12      11.7.3.90                pypi_0    pypi
[conda] nvidia-cusparse-cu12      12.5.8.93                pypi_0    pypi
[conda] nvidia-cusparselt-cu12    0.7.1                    pypi_0    pypi
[conda] nvidia-ml-py              13.580.82                pypi_0    pypi
[conda] nvidia-nccl-cu12          2.27.3                   pypi_0    pypi
[conda] nvidia-nvjitlink-cu12     12.8.93                  pypi_0    pypi
[conda] nvidia-nvtx-cu12          12.8.90                  pypi_0    pypi
[conda] pynvml                    13.0.1                   pypi_0    pypi
[conda] pyzmq                     27.1.0                   pypi_0    pypi
[conda] torch                     2.8.0+cu128              pypi_0    pypi
[conda] torchaudio                2.8.0+cu128              pypi_0    pypi
[conda] torchvision               0.23.0+cu128             pypi_0    pypi
[conda] transformers              4.56.1                   pypi_0    pypi
[conda] triton                    3.4.0                    pypi_0    pypi

==============================
         vLLM Info
==============================
ROCM Version                 : Could not collect
vLLM Version                 : 0.10.2
vLLM Build Flags:
  CUDA Archs: Not Set; ROCm: Disabled
GPU Topology:
        GPU0    GPU1    NIC0    CPU Affinity    NUMA Affinity   GPU NUMA ID
GPU0     X      SYS     NODE    0-19,40-59      0               N/A
GPU1    SYS      X      SYS     20-39,60-79     1               N/A
NIC0    NODE    SYS      X 

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

NIC Legend:

  NIC0: mlx5_0

==============================
     Environment Variables
==============================
NVIDIA_VISIBLE_DEVICES=GPU-4a2a584b-b49a-0fb2-e38f-811149821da0,GPU-1e9f3782-ccfc-6cc3-f18b-41c250f8f073
NVIDIA_REQUIRE_CUDA=cuda>=12.8 brand=unknown,driver>=470,driver<471 brand=grid,driver>=470,driver<471 brand=tesla,driver>=470,driver<471 brand=nvidia,driver>=470,driver<471 brand=quadro,driver>=470,driver<471 brand=quadrortx,driver>=470,driver<471 brand=nvidiartx,driver>=470,driver<471 brand=vapps,driver>=470,driver<471 brand=vpc,driver>=470,driver<471 brand=vcs,driver>=470,driver<471 brand=vws,driver>=470,driver<471 brand=cloudgaming,driver>=470,driver<471 brand=unknown,driver>=535,driver<536 brand=grid,driver>=535,driver<536 brand=tesla,driver>=535,driver<536 brand=nvidia,driver>=535,driver<536 brand=quadro,driver>=535,driver<536 brand=quadrortx,driver>=535,driver<536 brand=nvidiartx,driver>=535,driver<536 brand=vapps,driver>=535,driver<536 brand=vpc,driver>=535,driver<536 brand=vcs,driver>=535,driver<536 brand=vws,driver>=535,driver<536 brand=cloudgaming,driver>=535,driver<536 brand=unknown,driver>=550,driver<551 brand=grid,driver>=550,driver<551 brand=tesla,driver>=550,driver<551 brand=nvidia,driver>=550,driver<551 brand=quadro,driver>=550,driver<551 brand=quadrortx,driver>=550,driver<551 brand=nvidiartx,driver>=550,driver<551 brand=vapps,driver>=550,driver<551 brand=vpc,driver>=550,driver<551 brand=vcs,driver>=550,driver<551 brand=vws,driver>=550,driver<551 brand=cloudgaming,driver>=550,driver<551 brand=unknown,driver>=560,driver<561 brand=grid,driver>=560,driver<561 brand=tesla,driver>=560,driver<561 brand=nvidia,driver>=560,driver<561 brand=quadro,driver>=560,driver<561 brand=quadrortx,driver>=560,driver<561 brand=nvidiartx,driver>=560,driver<561 brand=vapps,driver>=560,driver<561 brand=vpc,driver>=560,driver<561 brand=vcs,driver>=560,driver<561 brand=vws,driver>=560,driver<561 brand=cloudgaming,driver>=560,driver<561 brand=unknown,driver>=565,driver<566 brand=grid,driver>=565,driver<566 brand=tesla,driver>=565,driver<566 brand=nvidia,driver>=565,driver<566 brand=quadro,driver>=565,driver<566 brand=quadrortx,driver>=565,driver<566 brand=nvidiartx,driver>=565,driver<566 brand=vapps,driver>=565,driver<566 brand=vpc,driver>=565,driver<566 brand=vcs,driver>=565,driver<566 brand=vws,driver>=565,driver<566 brand=cloudgaming,driver>=565,driver<566
NCCL_VERSION=2.25.1-1
NVIDIA_DRIVER_CAPABILITIES=compute,utility
NVIDIA_PRODUCT_NAME=CUDA
CUDA_VERSION=12.8.1
LD_LIBRARY_PATH=/usr/local/cuda/lib64
PYTORCH_NVML_BASED_CUDA_CHECK=1
TORCHINDUCTOR_COMPILE_THREADS=1
CUDA_MODULE_LOADING=LAZY

🐛 Describe the bug

I'm using vllm 0.10.2(cuda12.8.1), my device is nvidia A5000, and I use the following command to start server

VLLM_MAX_AUDIO_CLIP_FILESIZE_MB=100 \
vllm serve models/whisper-large-v3 \
  --served-model-name whisper-large-v3 \
  --tensor-parallel-size 1 \
  --gpu-memory-utilization 0.9 \
  --host 0.0.0.0 --port 21002

I test whisper-large-v3 and whisper-large-v3-turbo, both of them are super slow than v0 engine. The table below shows the approximate speeds（token/s displayed in vllm logs）
v0 is test on vllm 0.10.1.1-cuda12.8.1, v1 is test on vllm 0.10.2-cuda12.8.1

model	v0	v1
whisper-large-v3	60-70	4-5
whisper-large-v3-turbo	135	22-25

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.

[DarkLight1337]

cc @russellb @heheda12345

[JakubCerven]

can confirm same behavior on A10 + openai/whisper-large-v3
4 hour audio takes ~140s on V0 and ~190s on V1

[russellb]

Thanks for the report. Can you try with #24866 ?

[russellb]

I'm working on this. We merged #24866. I posted a follow-up that applies the same fix elsewhere: #24989. It's still not as fast as you'd expect with V1.

I'm still tracking down what the bigger issue is.

[russellb]

My test is with a bunch of short transcription requests. The bottleneck for me seems to be running the encoder, as it's only running the encoder on at most one request in each batch.

The bottleneck in the single very-long transcription would be different. The changes we've already made probably make a bigger difference there.

[Sugar-zsg]

My test is with a bunch of short transcription requests. The bottleneck for me seems to be running the encoder, as it's only running the encoder on at most one request in each batch.

The bottleneck in the single very-long transcription would be different. The changes we've already made probably make a bigger difference there.

@russellb I think the issue is that the v1 engine of Whisper does not enable CUDA Graph. Previously, I fixed the problem where CUDA Graph was not effective in v0 (#22630). However, based on profiling results, CUDA Graph still does not take effect in v1.

[mingjack352]

+1

[JakubCerven]

I'm working on this. We merged #24866. I posted a follow-up that applies the same fix elsewhere: #24989. It's still not as fast as you'd expect with V1.

I'm still tracking down what the bigger issue is.

I tested with #24866

4h of audio now takes ~100 seconds on V1 (much faster than V0)
also worth mentioning is that I am not transcribing one 4 hour segment but sending 480 ~30 seconds segments/requests

adding cuda graph and improving performance even further would be amazing

[russellb]

Thanks for the input, everyone! very helpful

[russellb]

Based on @JakubCerven 's last update, I'm hoping everyone sees results that are at least faster than V0. There are obviously still optimization opportunities, but I'm hoping "slower than V0" is fixed. If anyone sees otherwise when testing from main, please provide some details on the test scenario so we can reproduce it. I won't close yet to leave some time for more feedback.

We may want to file a different issue to track CUDA graph enablement for Whisper.