Games

According to the article:

Irdeto has a plan, however: a program that will offer media outlets two versions of games to benchmark independently, with and without Denuvo Anti-Tamper, which he believes will prove "the performance is comparable, identical" between both. Apparently they hope to begin it within the next few months.

To me, this is a pretty good way of going about it. It gives people with professional and established testing methodology access to make the comparisons. Basically a "see for yourself" approach where its impact (or ideally, the lack of it) can be tested per-game.

One of the big problems historically with pre and post denuvo-removal comparisons is time. Casual benchmarkers run tests between older versions with DRM vs new versions without, which obviously have recieved a bunch of other optimizations. Some people don't even re-run the old tests, which brings their own system differences into the equation (e.g. BIOS and driver updates). Nor is testing between cracked / non-cracked versions perfect; DRM that's been circumvented is not necessarily prevented from running at all.

This is important to get right because, not only does actually good data take a LOT of work that casual benchmarkers wouldn't even think of, there aren't many definitive answers as to Denuvo's universal effect on games. Sometimes it appears to have had no effect, then sometimes it does, which may be related to different generations of it and how well the devs implement it on a per-game basis.

An example the article gives is [PCGamer] hiring Durante (a prolific PC modder and developer famous for their work on Dark Souls) to benchmark Final Fantasy 15 with and without, finding no definitive measurable difference in gameplay but a possible ~6.7% increase in load times.

The question I have though is whether Denuvo themselves offer these versions or if it's a program whereby every game with it is obligated to provide these versions to press. In the case of the former, Denuvo would likely have the power to only show off best-case scenarios, whatever the conditions for that may be. If the latter, we may get a positive effect whereby, if they don't already, Denuvo's contract with a dev requires that their implementation is done properly and meets their standards, otherwise they don't allow it. As a company, they do seem hung up on the perception of performance issues, so it wouldn't surprise me.

Putting on a tin-foil hat, if they're really scummy they could deliberately impose a performance limit on the non-Denuvo version. Say, something that is designed explicitly to mimic its runtime behaviour. That would be pretty insane though, not only is that tactic likely illegal, it wouldn't take much for it to become public. It would be a significant risk given how confident they are in its lack of performance impacts (which as we've said, appear to be backed by professional testing). If they're working to resolve their PR problems, it would be an utterly stupid thing to do.

Overall though, I look forward for what this means for testing going forward. Frustrations about DRM aside (I much prefer without for a number of reasons) from a technical level I'm interested in whether this approach will prove the performance impacts a myth or not. Either way, it'll give us more insight into how it actually works and what it's doing under the hood.