Enabling AVX/AVX2 in Kali Linux VM on VirtualBox (from Windows Host)

When running AVX-dependent binaries (like TensorFlow, LLMs, or h5 models) inside a VirtualBox VM, you may hit:

Illegal instruction (core dumped)

This guide walks through how to enable AVX and AVX2 passthrough to a Kali Linux VM hosted on Windows using VirtualBox.

Background

Modern machine learning and inference workloads often rely on CPU instructions like:

AVX / AVX2
FMA (Fused Multiply-Add)

By default, VirtualBox doesn’t expose these instructions unless:

VT-x is enabled on host and guest
Certain settings are correctly applied

Host Setup (Windows)

1. Check If VT-x Is Enabled in BIOS

Open Task Manager → Performance > CPU → Look for:

Virtualization: Enabled

If not, reboot your system and enable Intel Virtualization (VT-x) in BIOS/UEFI.

2. Disable Hyper-V & Conflicts

Open Command Prompt (Admin) and run:

bcdedit /set hypervisorlaunchtype off

Then, open PowerShell (Admin) and run:

Get-WindowsOptionalFeature -Online | Where-Object {$_.FeatureName -match "Hyper-V|Containers|VirtualMachinePlatform"} | Format-Table FeatureName, State

If VirtualMachinePlatform is enabled, disable it:

Disable-WindowsOptionalFeature -Online -FeatureName VirtualMachinePlatform

Then reboot.

VirtualBox Configuration

3. Use `VBoxManage` to Enable AVX

Open cmd and run:

"C:\Program Files\Oracle\VirtualBox\VBoxManage.exe" setextradata "kali-linux" VBoxInternal/CPUM/IsaExts/AVX 1
"C:\Program Files\Oracle\VirtualBox\VBoxManage.exe" setextradata "kali-linux" VBoxInternal/CPUM/IsaExts/AVX2 1

(Replace "kali-linux" with your VM name)

Also enable nested virtualization explicitly:

"C:\Program Files\Oracle\VirtualBox\VBoxManage.exe" modifyvm "kali-linux" --nested-hw-virt on

4. VirtualBox VM Settings

In VirtualBox GUI → kali-linux → Settings:

System > Motherboard:

✅ Enable EFI (special OSes only)
✅ Enable PAE/NX
✅ Chipset: ICH9 or PIIX3

System > Processor:

CPUs: 2 or more
✅ Enable Nested VT-x/AMD-V (should no longer be greyed out)

Acceleration:

✅ Enable VT-x/AMD-V
✅ Enable Nested Paging
Paravirtualization: KVM

Testing Inside Kali VM

Boot into Kali Linux and run:

grep -o -w 'avx\|avx2' /proc/cpuinfo | sort -u

✅ Expected output:

avx
avx2

Run your binary (e.g. TensorFlow or LLM inference):

python payload.py

Output should NOT crash, and you may see:

This TensorFlow binary is optimized to use available CPU instructions...
To enable the following instructions: AVX2 FMA...

Success

AVX and AVX2 instructions are now working in your Kali VM. You’re ready to run high-performance models or any AVX-tuned software from inside your virtual machine.

Optional: Re-enable WSL2 or Hyper-V

If you need WSL2 or Docker later, re-enable:

Enable-WindowsOptionalFeature -Online -FeatureName VirtualMachinePlatform