Home Artists Posts Import Register

Content

Hi everyone, a huge amount of video uploads have completed and we can finally share our Spider-Man performance data in its entirety. The venerable GTX 1060 faces off against RX 580, the RTX 3080 10GB takes on the RX 6800 XT, while we have a somewhat less evenly matched face-off as the RTX 3060 trounces the RX 6600. 

There's quite an interesting methodology with these videos. Typically, we'd take the captured video then use an editing tool to synchronise clips and 'cut them out' for matching video streams. The cutting out procedure also cuts out the relevant performance data which is written on the fly in concert with the capture. Typically you find a common frame in the two streams then choose X number of frames from there.

For these streams with so much data, we're looking at just one capture for each GPU, no cutting out required. What we do here is to overlap two performance analyses, typically using scene cuts to synchronise the feeds. In this way you get to see a whole lot more data because you aren't 'cutting bits out' and therefore missing out in visualising some areas of performance. You can see how this version of the process works in this behind-the-scenes video: https://www.youtube.com/watch?v=h26HrHoTLBk

However, especially at 4K resolution, rendering out these head-to-heads is intense work, made all the more intense because the slow graphing renders are actually done twice - once per GPU, overlaid over one another. The advantage of the 'cut out' approach is that it's one frame graph render with two sets of inputs. My solution to this? Render out complete video streams overnight via batch processing. The unedited streams I rendered out typically took 2.5 hours each for circa 20 minutes of content... which says to me we need to heavily optimise the graph rendering. 

Anyway, here are all of the downloads:

GeForce RTX 3080 vs Radeon RX 6800 XT: https://www.digitalfoundry.net/bonus-material-marvels-spider-man-rtx-3080-vs-rx-6800-xt

GeForce RTX 3060 vs Radeon RX 6600: https://www.digitalfoundry.net/bonus-material-marvels-spider-man-rtx-3060-vs-rx-6600

GeForce GTX 1060 6GB vs Radeon RX 580 8GB: https://www.digitalfoundry.net/bonus-material-marvels-spider-man-gtx-1060-vs-rx-580

Files

Bonus Material: Marvel's Spider-Man RTX 3080 vs RX 6800 XT

A performance analysis using our FCAT-based frame-time capture tools, stacking up the Founders Edition 10GB version of the RTX 3080 up against the AMD Radeon RX 6800 XT.

Comments

Anonymous

Was just telling my son this morning that his Radeon 6600 was the semi-equivalent of the RTX 3060, but a litte less powerful.. Wow! Spider-man has some bad frametime issues on that card based on this video. Looks playable with some settings tweeks, though. I'm a little surprised that a console game targeting AMD hardware doesn't run better on the AMD desktop, but that just highlights some key differences between the platforms I suppose.