Hello there - I've got a few spare moments as various projects and tasks are juggled, so let's talk about yesterday's editorial meeting and what it means for the #content schedule ahead. First up though, we've received a lot of questions wondering how Alex is getting on and the news is that the priority here continues to focus on treatment and recovery, so I'm not expecting him to return for a while. We'll continue to support him throughout this process - we've just got to accept that managing his condition and working for us at the same time isn't going to work. He'll be back when the time is right. We're rooting for him and miss him a lot!
Let's talk about what's happening this week:
John has our Denshattack review, which I'm looking forward to sharing when embargo lifts. Beyond that, we're looking forward towards the end of the month when Halo: Campaign Evolved will quote-unquote drop. Batman Arkham Knight retro remains on the docket, but the recent release of an RT/voxelised version of Duke Nukem 3D is also potentially on the table for coverage!
Oliver is currently putting the finishing touches to his console coverage of Assassin's Creed Black Flag Resynced - which has caused some interesting issues, namely that while the game supports the new PSSR, his captures didn't seem to use it. All will be revealed in the #content. Also coming soon when the time is right, Oliver will also be looking into the Final Fantasy 7 Rebirth mod that addresses some of the game's lighting issues.
Tom is back from vacation and his 'secret' project still continues apace, albeit in the background as we await certain assets. In the meantime, he'll be taking point on our upcoming Splatoon Raiders Switch 2 review.
Will continues to power digitalfoundry.net and is currently working on resurrecting our automated CPU benchmark workflow. In the background, work is continuing on resurrecting the DF merch operation!
Rich (ie me) has a lot going on. I'm still working on getting our improved graphing tools ready for deployment to the wider team while transitioning our raw capture workflow to quality-based HEVC encoding. Turns out we can better intermediate codec quality (ProRes, CineForm) using HEVC while halving file size. Editing is faster too as Premiere uses hardware decoders. I'm also starting to spec out our revised CPU and GPU benchmark suite, which will require full automation.
As part of our sponsorship deal with Alienware, we're doing a video on the importance of X3D processors. This project has been invaluable in that I can attach a good deal of the revised CPU benchmark suite research to this specific video. In the process, I've been quietly blown away by how even at 4K, you can be CPU-limited. But I was not prepared for Space Marine 2. This is RTX 5090 at 4K DLSS performance mode with everything maxed (including RT), paired with a Ryzen 9 9950X. The idea here is to find CPU-limited games/scenarios where we would expect to see big "real world" improvements with the 9950X3D. Check this out:
PresentMon allows us to isolate how much time the CPU and GPU are taking to display the current frame. Here, it's about 10.3ms on the CPU side and 7ms on the GPU. So, converting that into frame-rate, it's about 97fps pumped out by the CPU, but about 142fps on the GPU (!). What's especially noticeable here is that... well... it's an entirely static scene with no enemy entities. In the "real world", you'd probably be ramping up the DLSS factor or using frame generation. Still, looks like rich potential here for a future CPU benchmark. If anyone wants us to look into other triple-A titles for CPU benchmarking do let me know.
Oh - and we're also looking to taking that PresentMon data into our custom visualisations as an overview of that data can be intriguing. Here's the 9950X paired with the RTX 5090 running at 4K resolution Psycho RT settings on DLSS performance mode. Consider the cyan line as "what the GPU is capable of" and the blue line as the ceiling in gameplay owing the CPU limit. This is from our Cherry Blossom Marketplace run:
Of course, you would want to be using RT Overdrive with a 5090, right? Here's where things get interesting because both CPU and GPU are stressed in this scenario at 4K DLSS Performance mode. However, where the yellow line is "higher" than the red line, the Ryzen 9 9950X is still CPU limited. In path tracing. Remarkable stuff. Now, would average percentage differences across the bench work? I don't think so, because the variance between CPU and GPU busy here changes very much according to context, something else to think about it in potential CPU benchmarks.
So, in previous updates I've often said "thinking about CPU testing" and this is the kind of stuff I've been thinking about. Benchmarking at 1080p and lowering the GPU requirement as much as possible remains - in my view - the best way to in giving you an idea of how much faster CPU A is than CPU B, but this kind of testing using settings appropriate to the rig may also have value too. At the very least I may head off the inevitable "Why are you testing at 1080p with an RTX 5090!!?!!" questions we get.
OK then, well, so much for the spare moments, time to get on with the tasks! As ever, feel free to suggest stuff and ask questions - I hope you have a great week!