Troubleshooting
Fixing slow automation VMs on a Windows 11 host
Short answer
The usual cause is the Microsoft hypervisor layer under VMware, confirmed by the turtle icon in the status bar. Turn off Memory Integrity, disable Virtual Machine Platform and Windows Hypervisor Platform, run bcdedit /set hypervisorlaunchtype off, reboot, and rerun the same workload — with IT approval on a work machine. Guest memory is ruled out: 16 GB did not help, and working setups run smaller allocations.
On this page
- Why the host stays fast while the guest crawls
- The hypervisor check, step by step
- Working example: is 16 GB of guest RAM the constraint?
- Efficiency cores and power throttling on the Intel host
- Factory I/O's 3D load inside the guest
- Splitting Factory I/O onto the host
- When to switch hypervisors
- Questions asked
When the same VMware guest lags on two different high-end laptops — one with an AMD AI 350, one with an Intel Core Ultra 7 255H — while both hosts stay responsive, the shared suspect is the software path, not either processor. On a Windows 11 host, build 26200 in the reported case, Memory Integrity, Virtual Machine Platform, or Windows Hypervisor Platform loads the Microsoft hypervisor underneath VMware, and the guest pays for it. The turtle icon in VMware's status bar confirms the mode in seconds. Turn off Memory Integrity, disable those features, run bcdedit /set hypervisorlaunchtype off, reboot, and rerun the same TIA Portal workload — with IT approval on the work laptop. Guest memory is already ruled out: 16 GB in the guest changed nothing while others run the same software on less.
Why the host stays fast while the guest crawls #
A virtual machine runs its guest operating system on CPU time and memory handed over by a hypervisor. When the Windows hypervisor owns the processor's hardware virtualization, VMware stops talking to the CPU directly and executes on top of Microsoft's layer instead. Broadcom's VMware knowledge base documents guest performance problems after a host upgrade to Windows 11 24H2 with these security and virtualization features enabled, and VMware shows the turtle icon in its status bar exactly when this mode is active. The extra layer taxes virtualized guest work while native host applications keep full speed — which is why the host feels fine and the guest does not.
This workload adds a second suspect: Factory I/O renders a 3D scene that TIA Portal V18 and V15.1 never demand on their own, so graphics becomes its own bottleneck, independent of CPU and memory. A third lead — Windows parking vmware-vmx.exe on efficiency cores — is reported, not measured. It can only apply to the Intel host, because the AMD chip has no E-cores.
The hypervisor check, step by step #
- Look at VMware's status bar while the VM runs. Turtle icon present: VMware is on the Microsoft hypervisor layer; continue. No icon and the settings below already off: the hypervisor layer is unlikely — measure guest CPU and memory during the lag and work the later sections.
- In
Windows Security > Device security > Core isolation details, switch Memory integrity off. Microsoft documents that path. After switching, clear the reminder prompt rather than leaving it pending — VMware users report that step matters on Windows 11. - In Windows' optional features, turn off Virtual Machine Platform and Windows Hypervisor Platform, and Hyper-V if it is listed.
- From an elevated command prompt, run
bcdedit /set hypervisorlaunchtype off, then reboot. This stops the Windows hypervisor from loading at boot. - Rerun the identical workload — same TIA Portal project, same Factory I/O scene, same operations — and compare. Nothing else may change between the two runs.
Decision: icon gone and lag gone — the hypervisor layer was the cause; agree with IT whether the host keeps these features off or the guests move to a hypervisor that runs above them. Icon gone, lag unchanged — continue down this article. One safety note, once: this step set disables virtualization-based security on the host, which is why a work machine needs IT authorization and a restore of the approved configuration afterwards.
Working example: is 16 GB of guest RAM the constraint? #
Inputs, each with its source:
- Guest RAM assigned on the lagging VM: 16 GB, reported in the discussion; the lag was unchanged after assigning it.
- Working guest allocations for TIA Portal, from other users in the same discussion: 8 GB to 12 GB, with one light-use setup at 4 GB and 2 vCPUs. This 8-to-12 GB span is the working range used for every comparison below.
- A comparable working host: 32 GB RAM, an i7 laptop in the discussion running several VMs. The two lagging laptops' host RAM is unreported — read it in Task Manager before pushing the guest any higher.
Rule in words: if the guest already holds more memory than allocations that run the same software acceptably, memory is not the constraint, and adding RAM cannot fix the lag.
Arithmetic:
- 16 GB ÷ 12 GB = 1.33 — the guest sits a third, that is 4 GB, above the largest allocation reported to work.
- On the comparable 32 GB host, a guest at that top allocation leaves 20 GB for Windows and the remaining VMs.
Result: the guest sits 33% above the highest working allocation, and the lag never tracked memory at any point.
Decision, with its threshold: once guest RAM is at or above the top of the working range and the lag is unchanged, stop adding memory. Set the guest inside that range, leave the difference to the host, and spend the time on the hypervisor, CPU scheduling, and 3D checks. If host-side readings later show the machine itself near capacity, the decision becomes a bigger host, and sizing a virtualization server for Studio 5000 and FactoryTalk VMs works through that arithmetic for automation guests.
Efficiency cores and power throttling on the Intel host #
This check is for the work laptop, the Core Ultra 7 255H, only. The AMD host has no efficiency cores, so skip it there.
- The reported mechanism: Windows tags vmware-vmx.exe as a background process and parks it on efficiency cores or power-throttles it. Verify before acting — during a stall, the guest should be demanding CPU while vmware-vmx.exe in Task Manager sits well below the vCPUs you assigned. That mismatch is the signature; without it, move on.
- Fix it as a pair: set the host to the high-performance power plan, and from an admin prompt run
powercfg /powerthrottling disable /path "C:\Program Files (x86)\VMware\VMware Workstation\x64\vmware-vmx.exe"— after confirming that path matches your installation by checking where vmware-vmx.exe actually lives. - Retest the same workload. Responsiveness now tracks CPU demand: keep both settings. No change: leave the plan as it was and move on.
Factory I/O's 3D load inside the guest #
- Run the same project twice: once with Factory I/O closed, once with the scene running. Lag that appears only with the scene points at the graphics path.
- Enable 3D acceleration in the VM's display settings and rerun both conditions. Keep the setting if the scene measurably improves; restore it if nothing changes.
- Do not copy host graphics drivers into the guest. One user reports doing that under Hyper-V with good results; that is a Hyper-V practice, and this user also reports the VMware guest cannot use the host GPU directly in this setup.
Splitting Factory I/O onto the host #
- Isolation test: run Factory I/O on the host and TIA Portal in the guest, linked over the S7 protocol.
- That removes the 3D scene from the guest entirely. If the lag leaves with it, the guest's graphics path was the constraint — and the split is already a working configuration.
- Decision: keep the split, or keep both applications in one guest with 3D acceleration confirmed and accept the measured cost.
When to switch hypervisors #
If the turtle icon is gone, the power check is done, the workload is split, and the same VM still lags, the remaining variable is the hypervisor itself. Several engineers who hit this on Windows 11 report moving their TIA guests to Hyper-V and staying there — a reported result, not a benchmark. Treat the switch as the last step, and clear it with IT, because it changes the host's virtualization layer for everything else on the machine. One further report worth a single test: Windows 11 guests described as slower than Windows 10 guests in general, so a Windows 10 guest is one cheap comparison if your software versions support it.
Questions asked
What does the turtle icon in VMware's status bar actually confirm?
It confirms the execution mode: the Windows hypervisor is loaded underneath, so the VM is not running directly on the processor's hardware virtualization. It does not tell you how much overhead you are paying — that you measure by rerunning the same TIA Portal workload after Memory Integrity and the hypervisor features are switched off. If the icon is gone after the reboot, the guest is back on the direct path.
Is 16 GB too little RAM for a TIA Portal guest?
No. Field setups run TIA Portal guests acceptably on less than that, and one light-use configuration runs 4 GB with 2 vCPUs. At 16 GB the guest already holds a third more memory than the largest working allocation, and in the reported case assigning it changed nothing. At or above the top of the working range with the lag unchanged, memory is ruled out — set the guest inside that range and move on to the hypervisor, scheduling, and graphics checks.
Why does IT have to approve turning Memory Integrity off on a work laptop?
Memory Integrity is virtualization-based security: Windows checks kernel-mode code integrity inside a hypervisor-protected environment, which blocks a class of kernel-level attacks. Turning it off weakens the host's defenses, so on a company machine it is a security decision rather than a performance tweak, and policy may re-enable or block the change. Have IT authorize the test window and restore the approved configuration afterwards.
Will moving the VM to Hyper-V fix the lag?
It has worked for others, but that is a reported result, not a measured benchmark. Try it only after the hypervisor-layer fix, the power check, and workload isolation, because the switch costs a migration and changes the host's virtualization layer for everything else on the machine. Fix the Windows hypervisor state first: it is free and reversible.
Does 3D acceleration matter for TIA Portal itself?
The measurable 3D load is Factory I/O's scene rendering; TIA Portal's editors are 2D. Enable 3D acceleration on the guest when Factory I/O runs inside it and compare the same scene with the setting on and off. Expect the difference in the 3D view, not in ladder or function-block editing. If Factory I/O moves to the host, the setting stops mattering for this workload.