We believe that a new era of computing is approaching. Access to high powered computing, rather than ownership, will define the thin-client era, eliminating our constant upgrade cycles and giving everyone the power of a super computer while simply carrying a cheap device. To make this future a reality, a streaming protocol that makes the shared computer feel like it’s right in front of you is required. We are trying to build that streaming platform and want to make it available on every computer, so any device can connect to any powerful computer, either at home or in the cloud.
Today, Parsec frees gamers from their hardware, and makes it possible to play PC games from anywhere on any device. In our quest to shave milliseconds from our streaming technology, we’ve spent more than a year finding millisecond savings while riding a wave of technology innovation across the hardware sector. Today, Parsec runs with the h.264 codec, but we’ll soon make h.265 available for an expected 50% improvement in bandwidth and latency.
With that being said, we wanted to push our technology to the limit and see how fast Parsec could encode, transfer, decode, and render a frame. At 240 frames per second, Parsec is only two frames behind the server PC with VSync on. The total latency required for the entire Parsec pipeline on the LAN is roughly 4–8 milliseconds in this test. We believe that focusing on our technology, protocol, and streaming software will be a piece of the puzzle to redefining the relationship between computer ownership and consumers, and we hope our technology makes the viability of cloud gaming finally possible.
Please note: We do not recommend using Parsec at 240 FPS. We only test the software at 60 FPS, but the test we did demonstrates the capability of our software.
The Test — Pushing Parsec To The Limit
According to literature from Nvidia, the Pascal Architecture should be able to encode 400 frames per second using its h.264 hardware encoder at 1080p. That got us thinking, “how far can we push Parsec?”
We bought two 240hz monitors and a 1000 FPS camera to record the video above. The server has an i7 6700k and a GTX 1070. The client is a mini PC with a GTX 1050ti. Both machines are on the same network connected via ethernet with 1Gbps of bandwidth. FRAPS was running on the server and client to confirm 240 FPS. For the testing, we used Test UFO with the flicker setting. We had the settings in Parsec set to match the refresh rate of the server (240hz) and max the bandwidth at 100Mbps.
The paused image below shows that Parsec powered device (right) is only 2 frames behind the server (left). At 240 FPS, this requires a total lag of 4–8 milliseconds.
Although this isn’t meant for the real world, it demonstrates the optimizations we’ve made with our software. We’re not sure where the rest of the market is on this, but we’re really stoked that our software was able to keep up with these high frame rates and maintain such low latency between two machines.