What’s the Fastest VPN?
Although everyone wants a fast VPN, value and privacy protections are far more critical in your choice of a VPN provider. We tend to de-emphasize speed in our reviews unless it is impressively good or abysmally bad. Still, readers are understandably concerned about the impact a VPN will have on their online experience. That’s why we test.
The results of all our testing to date are presented in the chart below. We’ll continue to update the chart, adding new results for products as we test them. Note that the table below shows the services sorted by their effect on download speeds, as does the list at the top of the story. You can click on the top headers to sort by a particular column you’re interested in.
How We Test VPNs
Before COVID-19, we at PCMag ran back-to-back speed tests on all the VPN products we planned to review over the next week or two. That’s no longer feasible.
Most of us at PCMag have been working from home since the onset of the COVID-19 pandemic in March 2020. Unfortunately, our homes don’t have the super-high-bandwidth connection speeds we enjoyed at PCMag Labs. Testing VPN speeds from our homes wouldn’t yield data comparable to previous years and would introduce all sorts of messy variables. For example: Did this VPN perform particularly badly because it’s a bad product or because someone else in the home was on a video call at the same time?
Rather than forgo VPN speed testing altogether, our solution has been to test VPNs in small batches and release the results as we go. This method allows us to reduce the risk to our analysts and still perform the same rigorous testing as always.
This rolling model also lets us provide fresher speed test results and gives us a chance to more easily update the results for different products throughout the year to see whether they still compete.
How We Test VPNs for Speed
When we are testing to find the fastest VPN, we use a custom tool provided by Ookla that uses the same technology as Speedtest.net. The only difference between our tool and the one online is that ours automates the test. There are some rare situations in which we use the online test as well.
(Editors’ Note: Ookla is owned by PCMag’s publisher, Ziff Davis.)
Ookla’s test measures three metrics: latency, download speed, and upload speed. Latency is a measurement of the time between when your computer sends a request and when it receives a response. It’s also called ping time. Many things can affect latency, such as the distance your data physically travels. Latency is measured in milliseconds, however, so even a seemingly large number for the increase may not be noticeable to the average user. Latency is important when playing video games over a VPN, as lower latency means a more responsive experience with less lag.
Download and upload speeds measure how much data is moved over your internet connection in a certain amount of time. Both are measured in megabits per second (Mbps). These are the metrics ISPs and wireless companies tout when they claim their connection speed is “fastest.”
When we test VPNs, we get a sense of the impact a service has on internet performance by finding the percentage change between using the VPN and not using the VPN for several speed measurements. We run 10 Ookla tests without a given VPN active and find the median of the results. We then do the same thing but with the VPN active.
By comparing the median results with and without the VPN, we find a percent change between the two. The bigger the change from the baseline performance, the more impact the VPN has on your internet speeds. We repeat this process for every VPN we test.
We have tested each of these services in as repeatable a manner as possible, but there are limitations inherent in our testing. The biggest issue is that our testing is conducted in New York City. If we performed the same tests on a similar network in a different part of the country, we’d almost certainly get different results.
Networks are fickle things and even small changes can greatly affect the results of a speed test, so while we have controlled for as many variables as we can, it’s hardly a sterile network environment. This is partly intentional; our tests need to be both controlled and representative of readers’ home environments.
We think of these tests as snapshots of performance that establish a replicable metric for measuring each service. The goal is to compare the services rather than to evaluate individual speeds. Your experience with these services will almost certainly vary from ours.
You can read a lot more about how we test VPNs and, importantly, how our testing has changed over the years in How We Test VPNs.
How We Interpret Our VPN Testing Data
We test dozens of VPNs and have to winnow down the glut of data this testing produces. Above all, our final list should be meaningful and easy to understand. We start with a list of every VPN we’ve tested that scored at or better than the median in at least one category (download, upload, and latency). We then drop all the services that only meet or exceed the median in one category, order the remaining products by download speed results, and take only the top 10 products.
We’ve used this approach for years because it gives products the chance to make the list by excelling in at least two out of the three categories. Most of the time, the results are all clustered, providing a well-rounded selection. However, it is possible for a product to make the list and still have subpar performance in one area, usually latency. We highlight this both as a word of caution and to demonstrate just how difficult it is to find a VPN provider that can truly deliver fast speeds.
Here’s the breakdown of all the products for each category:
- Scored better than the median in all three categories: AVG Secure VPN, ExpressVPN, HMA VPN, IPVanish VPN, Mozilla VPN, Mullvad VPN, NordVPN, Private Internet Access VPN, Proton VPN, and StrongVPN (all the products in our current list)
- Beat the median in two categories: IVPN
- Beat the median in one category: Avast Secureline, Bitdefender VPN, CyberGhost VPN, Hotspot Shield VPN, Opera VPN Pro, Surfshark VPN
The chart below breaks down the results from the 10 products, for easy comparison:
How Do We Deal With Unusual Testing Results?
For the first time in many years, we’re seeing a few products that produce Speedtest results that are better when the VPN is running than when the VPN is off. You might think that these would be our top picks, but they’re often not.
Let’s give them some credit first. Avast Secureline performed 6% better than the baseline result for download speeds. NordVPN and StrongVPN beat the baseline for latency, by 5.6% and 1.4%, respectively.
However, there are some important nuances to these results. We found that Avast Secureline VPN had a median download baseline of 816.38Mbps and an 865.49Mbps median download with the VPN active. That difference is significant and it persisted despite our rerunning these tests to confirm the results. But it’s not indicative of Avast Secureline’s overall performance because the app increased latency by a whopping 2,774.9% (3.75ms median baseline latency versus 107.91ms median VPN latency).
For NordVPN and StrongVPN, the improvements become minuscule when you look at the actual numbers rather than the percentages. NordVPN had a baseline median latency of 3.46ms and a VPN median latency of 3.26ms. StrongVPN’s edge was even slimmer: a baseline of 3.29ms compared to 3.24ms with the VPN running.
We don’t think it’s fair to expect these products to notch network speed improvements. We do think it’s fair to say that in some categories, these should have minimal impact on your surfing performance. If anything, the results seem indicative of improving VPN technology and infrastructure. At this point, we might sometimes be measuring the size of the pipe rather than the VPN’s performance.
VPN Cheating and Data Compression
Because we report how we test VPNs and use one of the most popular speed test tools available, there’s some obvious concern that a disreputable company might attempt to game the results. A nefarious VPN company could, perhaps, detect when an Ookla test is running and attempt to skew the results.
We spoke with engineers at Ookla when designing our tests, and they share our concerns about gaming their tests. They’ve told us that the company takes active measures to fool would-be cheaters. We defer to the experts at Ookla to prevent companies from juicing the results.
Because we make our methodologies public, it is also possible that VPN companies could tweak their networks to obtain better results. For example, a VPN company could endeavor to rent server space as close to our offices as possible, hoping that the shorter distances will yield better speed test results. There’s little we could do to prevent that. However, speed testing is only a small part of our reviews, and even an outstanding speed test result wouldn’t do much to raise the score of an otherwise mediocre product.
How to Speed Up Your VPN
There are a few ways you can offset the speed-reducing effects of using a VPN. First, choosing a service with many servers may (emphasis on may) make it more likely you’ll find one that isn’t crowded with other people all trying to use the same bandwidth. Some VPN services let you select specific servers and show you the current load on those servers.
Having many servers to choose from in different locations means you’re more likely to find one that’s physically close to you, shortening the distance your data must travel. This usually translates to lower latency and perhaps better overall performance. A few VPN clients include a ping test to help you zero in on the closest option, but it’s unusual for a VPN client to do anything other than automatically select the closest server.
A word of caution: A VPN with more servers is not necessarily better. Consider that a VPN service likely needs more servers to support more subscribers, and we don’t know how those servers are allocated among those subscribers. The size of a VPN server fleet may simply reflect a company’s popularity and not the quality of its service.
Split-tunneling is the generic term for when a VPN lets you define which apps send data through the VPN tunnel and which send data outside the tunnel. It lets you separate more sensitive activities, like web browsing or online banking, from more mundane but higher-bandwidth activities, like streaming music or playing video games. Split-tunneling is especially useful because Netflix blocks VPN use, as do other services. You can simply route these apps outside the VPN to avoid this problem.
Note that some VPN companies offer free versions that limit the number of servers available. We’d expect those servers available to free users to be crowded and therefore offer slower speeds. Proton VPN, for example, limits the number of servers available to free users but notably does not limit the amount of data a free subscriber can use.
Also important is the protocol the VPN service uses. OpenVPN and IKEv2 have long been the standard for the latest and greatest in VPN technology, but that is changing. More companies have begun deploying WireGuard, an open-source protocol designed for greater security, easier deployment, and potentially higher speeds.
The Fastest VPN Service Isn’t Always the Best VPN Service
Simply put, speed shouldn’t be your only consideration when shopping for a VPN. For one thing, your internet experience will almost certainly be faster without a VPN (our unusual outliers notwithstanding). For another, speeds depend so much on which server you use, where you are, what your network environment is like when you’re using the VPN, and so on. You might find that the service that’s as fast as lightning today is barely chugging along tomorrow.
Instead of speed, our VPN reviews stress value and trust. We prefer VPN services with a good geographic distribution of servers. VPNs that are easy to set up and use for first-timers and include a well-made local client go a long way toward getting PCMag’s endorsement. We also pay close attention to the efforts each VPN takes to protect its customers and their data. And, of course, price is a major issue. The average price among VPNs we’ve tested is around $10 per month. If a VPN is charging more, it had better offer something compelling.
It’s important to remember that while VPNs are an easy way to improve your privacy, they do not protect against every ill. We highly recommend that readers install a standalone tracker blocker and that readers make use of the advanced privacy tools available in most browsers. We encourage you to use a password manager to create unique and complex passwords for every login, install Antivirus software, and enable multi-factor authentication wherever it is available.
VPN speed will always matter to some extent. Just remember that there are many other factors to consider when selecting a VPN service.