Private Internet Access Test Results

Our latest Private Internet Access test showed excellent nearby performance using WireGuard. PIA retained 95.4% of our VPN-off download speed in Sweden and 95.2% in Germany. UK performance remained strong at 82.7%, while US East and Singapore showed much larger long-distance throughput reductions. Across all five tested regions, our fixed-target stability test recorded 0% packet loss.

← All VPN testing results

PPG LAB TESTED · PAID PLAN

We tested Private Internet Access (PIA) on Windows using WireGuard across all five Private Proxy Guide core regions. Sweden and Germany both retained more than 95% of our VPN-off download speed, while the UK retained 82.7%. US East and Singapore showed much larger long-distance throughput reductions. All five regions recorded 0% packet loss in our fixed-target stability test.

Test date: August 27, 2026 Test ID: PPG-VPN-2026-08-34AE Plan: Paid App version: 3.7.2 Protocol: WireGuard Runner: v0.4.9
95.4%
Best speed retention
Sweden median download
17.5 ms
Sweden latency
+8.6 ms vs baseline
0%
Stability packet loss
Across all five regions
5/5
Geo consensus passed
All PPG core regions

What our Private Internet Access test showed

Private Internet Access delivered excellent nearby throughput in both Sweden and Germany. Sweden reached 88.71 Mbps against our 93.00 Mbps VPN-off baseline, corresponding to 95.39% download-speed retention. Germany was almost identical at 95.16%.

The UK also performed well, reaching 76.88 Mbps and 82.67% retention. Long-distance performance dropped substantially: US East retained 22.10% of baseline download throughput, while Singapore retained 12.76%.

Tunnel stability was strong across the complete batch. Every tested region recorded 0% packet loss and 0% detected latency spikes in the 50-sample fixed-target stability test.

Private Internet Access speed test results

Our VPN-off baseline was 93.00 Mbps download, 93.87 Mbps upload and 8.97 ms latency. We performed five VPN-off baseline runs and three speed-test runs per PIA region. Median values are used for reporting.

Download speed retained

Sweden95.4%
 
Germany95.2%
 
United Kingdom82.7%
 
US East22.1%
 
Singapore12.8%
 

Quick interpretation

  • Sweden: excellent near-baseline throughput.
  • Germany: almost identical to Sweden at over 95% retention.
  • UK: strong European performance at 82.7% retention.
  • US East: substantial long-distance throughput reduction.
  • Singapore: lowest throughput and highest latency in this batch.
Location Download Upload Speed retained
Sweden 88.71 Mbps 89.86 Mbps 95.39%
Germany 88.50 Mbps 89.62 Mbps 95.16%
United Kingdom 76.88 Mbps 75.59 Mbps 82.67%
US East 20.55 Mbps 20.40 Mbps 22.10%
Singapore 11.87 Mbps 11.78 Mbps 12.76%

Regional latency

Location Regional ping Added latency Ookla jitter
Sweden 17.54 ms +8.57 ms 0.276 ms
Germany 21.40 ms +12.43 ms 0.590 ms
United Kingdom 29.40 ms +20.42 ms 1.665 ms
US East 101.99 ms +93.02 ms 0.375 ms
Singapore 177.70 ms +168.72 ms 0.721 ms

Gaming latency and connection stability

We separate regional latency from tunnel stability. Regional latency comes from the locked Ookla endpoint used for each location. A separate 50-sample fixed-target ICMP test measures median latency, p95 latency, jitter, packet loss and latency spikes.

Important: these are VPN-network measurements from our Swedish test environment. They are not direct ping measurements against specific game servers.
Location Stability median p95 Packet loss
Sweden 18 ms 18 ms 0%
Germany 22 ms 22 ms 0%
United Kingdom 29 ms 31 ms 0%
US East 102 ms 103 ms 0%
Singapore 177 ms 178 ms 0%

Stability jitter and spikes

Location Stability jitter Spike rate Result
Sweden 0.490 ms 0% Stable
Germany 0.388 ms 0% Stable
United Kingdom 0.592 ms 0% Stable
US East 0.571 ms 0% Stable
Singapore 0.592 ms 0% Stable

The fixed-target results were consistently tight. Sweden and Germany had identical median and p95 values, while UK, US East and Singapore were only 1–2 ms apart between the median and p95 measurements. No latency spikes were detected in any of the five regional stability probes.

PIA server selection and exit-location validation

This PIA test used the provider’s live region list rather than assuming that hard-coded server identifiers were still valid. Before each regional benchmark, the runner resolved the requested PPG location against PIA’s current region IDs and verified the selected region before continuing.

PPG region PIA region selected Geo consensus Result
Sweden se-stockholm SE PASS CONSENSUS
United Kingdom uk-london GB PASS CONSENSUS
Germany de-frankfurt DE PASS CONSENSUS
US East us-new-york US PASS CONSENSUS
Singapore singapore SG PASS CONSENSUS
Why this matters: PIA exposes multiple locations and specialized variants for some countries. The test runner therefore verifies the actual PIA region selected rather than simply issuing a generic connection command and assuming the requested location was used.

All five regions ultimately passed multi-source location validation. Sweden and Germany each had one external geolocation source disagree with the other sources, but the remaining geolocation services and IPLeak evidence agreed with the selected country. These were therefore classified as PASS_CONSENSUS, not manual overrides.

DNS observations

IPLeak.net DNS evidence was collected for the VPN-off baseline and each PIA region. The VPN connections produced observable resolver IPs, but the VPN-off baseline repeatedly reported zero detected resolvers.

Location DNS status VPN resolvers observed Reason
Sweden INCONCLUSIVE 26 No usable VPN-off resolver baseline
United Kingdom INCONCLUSIVE 5 No usable VPN-off resolver baseline
Germany INCONCLUSIVE 26 No usable VPN-off resolver baseline
US East INCONCLUSIVE 1 No usable VPN-off resolver baseline
Singapore INCONCLUSIVE 9 No usable VPN-off resolver baseline
What this means: the DNS result is not evidence that Private Internet Access leaked DNS. IPLeak explicitly returned zero DNS resolvers during both VPN-off baseline attempts, so there was no valid baseline resolver set to compare against the VPN connections. We therefore publish the conservative status INCONCLUSIVE_BASELINE_NO_RESOLVERS_DETECTED rather than Pass or Fail.

Browser evidence

BrowserLeaks evidence collection was enabled for this PIA test. IP, DNS and WebRTC evidence was captured for the VPN-off baseline and all five VPN regions.

6 BrowserLeaks IP evidence sets: baseline plus five VPN regions.
6 BrowserLeaks DNS evidence sets: baseline plus five VPN regions.
6 BrowserLeaks WebRTC evidence sets: baseline plus five VPN regions.

Browser evidence is retained as supporting material. Capturing the pages does not automatically produce a security Pass or Fail unless the evidence supports that conclusion.

IPv4, IPv6 and security observations

  • IPv4 location validation: PASS CONSENSUS across all five PPG core regions.
  • PIA region selection: verified for all five locations using PIA’s live region list and selected-region state.
  • DNS: INCONCLUSIVE BASELINE. The VPN-off IPLeak baseline explicitly reported zero resolvers after retry.
  • IPv6: INCONCLUSIVE BASELINE. The IPv6 lookup hostnames could not be resolved during the baseline.
  • BrowserLeaks: IP, DNS and WebRTC evidence captured for baseline and all five VPN regions.
  • Kill switch: not tested in this batch.
IPv6 limitation: the baseline IPv6 requests failed at DNS resolution before a usable IPv6 baseline could be established. We therefore do not interpret this as either an IPv6 leak Pass or Fail for PIA.

Connection timing

PIA’s command-line adapter returned approximately 0.07 seconds after each connection command. These values measure provider-command return time rather than verified end-to-end tunnel establishment.

Diagnostic only: we do not claim that PIA establishes a complete VPN tunnel in 0.07 seconds and do not use these command-return figures to rank connection speed against other providers.

Test evidence

The PIA batch generated raw Ookla measurements, fixed-target stability evidence, multi-source IP geolocation data, IPLeak observations, BrowserLeaks captures and automatic desktop and Speedtest screenshots.

26/26 Automatic desktop and Speedtest screenshots completed successfully.
20 Ookla runs: 5 VPN-off baseline tests plus 15 PIA regional tests.
18 BrowserLeaks captures covering IP, DNS and WebRTC across six scopes.

How we tested Private Internet Access

  • Plan: Private Internet Access paid plan.
  • Application: PIA 3.7.2 for Windows.
  • Protocol: WireGuard.
  • Test location: Sweden.
  • Runner: Private Proxy Guide VPN Test Runner v0.4.9.
  • Baseline: 5 VPN-off Ookla runs.
  • Core PPG regions: Sweden, United Kingdom, Germany, US East and Singapore.
  • Core-region coverage: 5/5.
  • Speed testing: 3 runs per VPN region, reported using median values.
  • Regional Speedtest locking: enabled.
  • Post-connect stabilization: 20 seconds.
  • Gaming stability: 50 ICMP samples against the fixed PPG stability target.
  • PIA server selection: live region discovery and selected-region verification.
  • Location validation: multi-source geolocation consensus.
  • IPLeak.net DNS test: enabled.
  • BrowserLeaks IP/DNS/WebRTC evidence: enabled.
  • Automatic screenshots: enabled.
  • Kill switch: not tested.

Raw Private Internet Access speed-test runs

The summary values above use medians. All 20 individual speed-test measurements used in this batch are shown below.

Show all 20 individual speed-test runs

Baseline – VPN disconnected

Run Download Upload Ping
#1 93.10 Mbps 93.87 Mbps 8.889 ms
#2 93.23 Mbps 93.60 Mbps 8.972 ms
#3 93.00 Mbps 94.11 Mbps 8.875 ms
#4 92.91 Mbps 93.54 Mbps 9.191 ms
#5 92.96 Mbps 93.97 Mbps 9.013 ms

Locked Ookla endpoint: Telenor Denmark, Søborg.

Sweden

Run Download Upload Ping
#1 88.71 Mbps 89.94 Mbps 17.539 ms
#2 88.80 Mbps 89.86 Mbps 17.651 ms
#3 88.65 Mbps 89.48 Mbps 17.435 ms

Locked Ookla endpoint: AltusHost B.V., Stockholm.

United Kingdom

Run Download Upload Ping
#1 74.58 Mbps 75.63 Mbps 29.395 ms
#2 77.09 Mbps 75.59 Mbps 29.459 ms
#3 76.88 Mbps 75.24 Mbps 28.184 ms

Locked Ookla endpoint: Structured Communications, London.

Germany

Run Download Upload Ping
#1 83.12 Mbps 88.98 Mbps 21.712 ms
#2 88.50 Mbps 89.62 Mbps 21.286 ms
#3 88.57 Mbps 89.70 Mbps 21.404 ms

Locked Ookla endpoint: Clouvider Ltd, Frankfurt am Main.

US East

Run Download Upload Ping
#1 20.60 Mbps 20.42 Mbps 102.262 ms
#2 20.54 Mbps 20.36 Mbps 101.991 ms
#3 20.55 Mbps 20.40 Mbps 101.869 ms

Locked Ookla endpoint: Surfshark Ltd, New York, NY.

Singapore

Run Download Upload Ping
#1 11.87 Mbps 11.83 Mbps 177.435 ms
#2 11.86 Mbps 11.78 Mbps 177.696 ms
#3 11.87 Mbps 11.77 Mbps 177.868 ms

Locked Ookla endpoint: GSL Networks, Singapore.

Private Internet Access test conclusion

Private Internet Access delivered excellent European throughput in this standardized test batch. Sweden retained 95.39% of baseline download speed, Germany retained 95.16%, and the UK retained 82.67%.

The Germany result is particularly notable because two of its three runs exceeded 88 Mbps. The first run was lower at 83.12 Mbps, but our median methodology prevents either the fastest or slowest individual measurement from determining the published result.

Long-distance routing had a much larger effect. US East retained 22.10% of baseline throughput, while Singapore retained 12.76%. This is consistent with why we test geographically separate locations rather than assigning one generic speed figure to a VPN.

Stability was strong throughout the batch. Every region recorded 0% packet loss and 0% detected latency spikes, while median-to-p95 latency remained tightly grouped.

The PIA-specific region-selection logic also worked correctly in the completed batch. The runner selected and verified se-stockholm, uk-london, de-frankfurt, us-new-york and singapore, and all five regions subsequently passed multi-source exit-country consensus.

DNS and IPv6 remain inconclusive because valid VPN-off baselines could not be established for those specific checks. We therefore make no DNS-leak or IPv6-leak Pass/Fail claim from this batch. The dataset covers PIA’s paid Windows app using WireGuard and does not establish performance for OpenVPN, streaming-optimized servers, specific game servers or kill-switch behavior.

← Back to all VPN testing results