Cognitive services speech onpremise 0.3.3 helm chart continuously crashing due to OOM Error

[vitas-pm]

I'm deploying the TTS on premise on a Kubernetes Cluster using the following values:

textToSpeech:
  enabled: true
  numberOfConcurrentRequest: 2
  optimizeForTurboMode: true
  image:
    registry: mcr.microsoft.com
    repository: azure-cognitive-services/speechservices/neural-text-to-speech
    tag: 2.8.0-amd64-de-de-conradneural
    pullSecrets:
      - mcr # Or an existing secret
    args:
      eula: accept
      billing: <my-endpoint>
      apikey: <my-api-key>
  service:
    autoScaler:
      maxAvailablePods: 4
    type: NodePort
  verification:
    enabled: false

speechToText:
  enabled: false

Shortly after the creation of the TTS Pods they start continuously crashing due to OOM Errors. Their calculated memory limit of 3GB is not sufficient as their average usage of memory seems to be around 3 to 4GB.
I can prevent the crashing by manually removing the limit of the deployment after installing the helm chart, but still this is not the desired behavior I assume.
Is there something I am missing or does the chart need an update?