Hello there and welcome to what's likely to be quite a small DF update. Lots going on, lots of press events happening and typically we can't really talk about them. Just to temper expectations, post SGF, a lot of the games shown there tend to go on a worldwide tour so those that didn't attend SGF (*raises hand*) get to see them.
But here's what's going on quickly!
Rich (ie me) is finally delivering the Bloodborne PS5 back-compat plus plus video tomorrow. Lots of surprises in this one and the screenshot above is basically the tip of the iceberg. I'm attending a press event or two this week.
Tom is handling one of the press events I was talking about, and is also taking point on a product photography day we are doing for an MSI sponsored video on ultra-wide gaming. Tom will also be taking point on the long-delayed Dark Souls 2 path tracing mod video.
John is actually doing the bulk of the work on the MSI video I've just talked about, but more importantly, he's going to be our appointed reviewer for Star Fox on Switch 2. Work continues in the background for the Batman Arkham Knight review.
Oliver will be returning to the two projects he's been working on for some time now - Switch 2 vs iPhone 17 Pro and PS4 vs PS5!
Will continues to power our website! The ARM coverage for their new ML stuff was particularly cool last week!
Alex doesn't have much news for you - he's currently off work on complete rest and we miss him!
I've been thinking a bunch recently about PC hardware balance and thought I'd share the following visualisation with you. It's about demonstrating absolute performance (final frame-rate if you like), along with CPU and GPU performance. This is Cyberpunk 2077 running on ultra RT at 4K resolution in DLSS performance mode on an RTX 5090 paired with a Ryzen 9 9950X3D.
What I'm trying to show here is that even with a 9950X3D, we can be CPU-limited. If we were using a 9950X non-3D, the Cherry Blossom Marketplace mesaured here would actually be CPU-limited with path tracing. The difference between the orange line and the green line essentially represents how much GPU performance you're leaving on the table.
It's quite dramatic - frame-rate differentials can be when you start moving beyond 60fps, so here's the same data graphed again in frame-time. I guess actually, it's still quite dramatic. In both cases, the green line represents maximum theoretical GPU performance, while the orange and blue lines represent actual output.
But essentially, I've been thinking a lot about responses to CPU reviews saying that we shouldn't bench at 1080p (ie removing the GPU as a bottleneck by running a 5090 at super low resolutions). I still think that methodology has merit but with selected test scenes and more plausible hardware combinations, we can still prove that faster CPUs can still under-feed the GPU even in more plausible gaming scenarios.
Anyway, I'll be coming back to that work at some point. In the meantime, any questions on this week's content, potential projects/topics... do let me know. Have a great week!