We searched far and wide for benchmarks that run under Windows RT. But as a result of Microsoft's strict usage guidelines, there simply aren't any cross-platform tests available for comparing the Surface to other ARM-based tablets.
Everything we know about how Tegra 3 fares against competing SoCs comes from comparing Android- and iOS-based devices. We've seen the same processor perform differently under different operating environments, though. Right now, browser-based tests are the only way to quantify the performance of a Windows RT-based device compared to other tablets.
Unfortunately, that's not even a perfect solution, since browser support varies based on operating system. Safari is the default on iOS-based devices, and we saw a number of examples of it outperforming its competition in Which Browser Should You Be Running On Your iPad And iPhone?. Meanwhile, the Jelly Bean update to Android makes Chrome the default on Google's Nexus 7. With the Surface, we’re dealing with IE 10, complicating the comparison further. Even though we can’t standardize to a single browser, using each device's default is more appropriate for measuring device performance anyway.
BrowsingBench was created by EEMBC, also known as the Embedded Microprocessor Benchmark Consortium. It's a non-profit organization tasked with finding ways to develop testing methodology, specifically for embedded hardware. We're been playing around with this tool in the lab, and we love it. While it's meant for testing "smartphones, netbooks, portable gaming devices, navigation devices, and IP set-top boxes," it's just as applicable for measuring browser performance in general.
Meanwhile, Microsoft’s Surface falls ~30% behind the iPad 3 and ~40% behind the iPad 2. That’s in sharp contrast to SunSpider, where IE10 took a commanding lead. The good news is that Surface manages to outpace Tegra 3-based Android tablets like the Nexus 7 and Transformer TF201 by ~20%.
Bear in mind that the performance measured in these three tests is only one aspect of using a tablet, and by no means does a discrepancy of 40% in a browsing metric mean you're going to see a corresponding experiential gap in the real world.