The VPN Added 11 Milliseconds—and Made Xbox Cloud Gaming Feel Faster

The Speed Test Was Answering the Wrong Question

Xbox Cloud Gaming had expanded quickly by the time of this trip.

Microsoft reported in late 2025 that cloud-gaming hours among Game Pass subscribers had risen 45 percent year over year. The service had also expanded across more subscription plans and supported more than 1,000 owned games through its Stream Your Own Game library.

That was why I could play from a thin work laptop in a hotel room.

I owned the racing game, but it was not installed. I opened Xbox in a browser, connected the controller and launched the game from the cloud.

The official network requirements did not look demanding. Xbox recommends a connection of at least 10 Mbps, with some devices needing around 20 Mbps for the best experience, plus 5 GHz Wi-Fi where available.

My hotel connection exceeded those figures many times over.

But cloud gaming is not a video that can quietly buffer ahead. Every button press must reach a remote Xbox, affect the game and return as a new frame while the player still has time to react.

A large download number cannot hide a late steering command.Once I understood that, the 236 Mbps result stopped looking reassuring.The real question was not how much data the hotel could move.It was whether each input would return at a predictable time.

Disconnecting the VPN Lowered the Number, Not the Frustration

I had entered the hotel network through a large, established VPN provider.

It was a reasonable default. The company had years of public history, a large support operation, broad server coverage and an application I already trusted on work trips.

I had selected its automatic Canadian server before opening Xbox.To remove one possible source of delay, I disconnected it.The game reloaded through the hotel connection alone.

The average latency fell to about 41 milliseconds—roughly 14 milliseconds lower than it had been through the VPN.

For the first lap, the steering felt better.

Then the hotel route became busy. The picture blurred, the network indicator changed colour and several controller inputs arrived together.

Most corners responded normally.

Every twenty or thirty seconds, one arrived late enough to push the car off line.

That inconsistency was harder to play around than a small, steady delay. I could adapt to controls that were always slightly slower. I could not adapt when the same turn felt immediate on one lap and badly delayed on the next.

The direct connection had produced the lowest number of the evening.

It had also produced the most expensive mistakes.

So I returned to the VPN, not because a VPN automatically improves gaming, but because the hotel’s direct route had already failed the task.

The Large Server List Became a Series of Test Laps

I selected the provider’s nearest Toronto server manually.

The game settled around 55 milliseconds.

The number was higher, but the controls initially felt more consistent. I completed two clean laps.

Then the stream froze for a fraction of a second and rushed forward to catch up. The car moved from the centre of the track to the grass before I saw the frames in between.

I tried another Toronto server.

It opened at 51 milliseconds, then began jumping above 160 whenever the hotel Wi-Fi weakened.

A Montreal server was steadier, but its constant delay made quick corrections feel heavy. I could keep the car on the track only by steering before each corner rather than reacting to it.

All three servers connected successfully.

None gave me a dependable race.

The provider’s strength was choice, but cloud gaming made each choice expensive to evaluate. A web page reveals a bad route within seconds. A game reveals it after the player has launched a session and reached the moment where one delayed input matters.

Other travellers describe the same contradiction more simply: a slower connection can feel playable while a much faster hotel network becomes unusable once control delay begins jumping.

By the fourth restart, I was no longer comparing VPN against no VPN.I was choosing between unstable direct routing and repeated server experiments.Neither let me forget about the network.

The Hotspot Was Faster and Still Unreliable

I enabled the hotspot on my phone.

The 5G speed test exceeded 300 Mbps. For a moment, it looked like the cleanest solution.

The first lap was smooth.

Then the phone shifted between two indoor signal conditions. The picture softened, voice chat broke into fragments and the steering delay returned.

I moved the phone to the window.

The signal improved, but I now had a charging cable stretched across the desk and a phone balanced against the curtain. Every time I picked it up, the connection changed again.

The hotspot was faster than the hotel Wi-Fi.

It was not more dependable inside that room.

My friends had already completed the first event without me. The next race began in nine minutes.

At that point, the comparison finally became useful.

I did not need the route with the best speed test or the lowest momentary ping.

I needed controller timing that stayed predictable from the starting line to the finish.

Eleven More Milliseconds Removed the Lag I Could Feel

I returned the laptop to the hotel Wi-Fi and opened OnlydogVPN.

The smaller app did not begin by asking which Canadian city I wanted. I selected the preset for cloud gaming on a shared or unstable network and connected.

Then I relaunched the game.

The latency settled around 52 milliseconds—11 milliseconds higher than the lowest direct result.

I joined the next event.

At the first corner, the car turned when I moved the stick.

At the second, I corrected a slide without waiting for the picture to explain what had already happened.

Five minutes passed without the controls arriving in a burst. The latency moved within a narrow range instead of jumping between excellent and unusable.

Halfway through the race, the hotel Wi-Fi weakened.The picture briefly lost some detail.The car remained controllable.I finished third.

That was not a remarkable result for the group, but it was the first complete race of the evening. I had not changed servers, restarted the stream or blamed the controller.

The app added a small amount of baseline delay and removed the large interruptions that made the game feel slow. Its gaming preset selected the route around the task, while its HTTP/3-based connection recovered as the hotel network fluctuated.

I could not observe the hotel’s internal routing rules or identify the exact point behind every earlier spike. I could see the result on screen: the direct route produced the lowest isolated reading but unpredictable controls, the established provider required repeated server tests, and the smaller app kept the controller response consistent through the entire race.

The VPN icon stopped being part of the game.

The Hotel Login Expired During the Next Race

We started another event.

Near the final lap, the hotel’s captive portal expired and the laptop briefly lost its authenticated Wi-Fi session.

Earlier, that kind of interruption had ended the stream.

This time, the laptop moved to my phone’s hotspot while I reopened the hotel login page in the background. The connection recovered, voice chat returned and the game remained open.

I lost one position.

I did not lose the session.

After accepting the hotel terms, I moved back to Wi-Fi. The picture paused briefly, then continued.

That second race confirmed why the first had worked. The connection was not merely attached to one lucky server. It could recover when the network underneath it weakened or changed.

For ordinary streaming, a few seconds of buffered video can conceal a brief interruption.

A cloud game cannot pre-download the turn I have not made yet.

Every input must travel out and every new frame must return while the decision still matters.

That was why recovery felt more valuable than another 50 Mbps on the speed test.

More Bandwidth Had Never Been the Missing Ingredient

After the session, I compared the evening’s results.

The hotel Wi-Fi without a VPN produced the lowest average latency. It also produced the largest spikes.

The phone hotspot delivered the highest download speed. Its indoor signal changed too often to trust.

The major provider offered the largest server network. Finding a usable route required repeated launches and test laps.

The smaller app did not win every benchmark. Its average latency was higher than the best direct reading and similar to several routes from the larger provider.

The difference was consistency.

Once connected, the game responded in a way I could learn. A corner did not feel immediate on one lap and delayed on the next. The image could briefly lose sharpness without making the controller useless.

A speed test ends with one clean score.

A cloud-gaming connection has to deliver the same dependable round trip for every button press.

The Smaller Server List Was the Trade-Off

The service has fewer server locations, a shorter public history and fewer independent ratings than the established provider. Someone who needs a rare exit country or a large enterprise support operation may prefer the bigger network.

I needed one dependable route between a hotel room and an Xbox cloud session.

The direct connection gave me a low number without dependable timing. The large provider gave me many routes without a clear way to choose the one that would remain stable. The hotspot gave me raw speed without a reliable indoor signal.

The smaller app gave me two completed races and carried the second session through a network switch.

I had assumed any VPN would make Xbox Cloud Gaming slower because it added another step between the controller and the server.

The one that worked added 11 milliseconds to the number—and removed the delay I could actually feel.