The Z9II was delayed. It’s unclear whether the delay is due to a new image sensor or a new EXPEED chip, but I thought I’d take a moment to explain what needs to be addressed with EXPEED.
Short version: processing pipeline bandwidth.
If you’re at all familiar with the progression of semiconductors, you know that each completely new generation is still following Moore’s law, which means transistors get smaller (smaller process size). Smaller can mean more, faster, or lower power. Engineers have to make a choice about how much of each benefit they choose. Once Apple got to some guaranteed battery charge level on their devices, they started balancing their Apple Silicon changes due process size reduction more towards faster, for instance.
EXPEED7 was a large step downward in process size from EXPEED6. A very dramatic change as it turns out, and most of that appears to have been gone towards faster. Much of the Z9 generation benefits derived from that.
However, it’s important to understand the internal structure of EXPEED, much like it is to understand how Apple Silicon’s internal structure is enabling things you don’t see on the Windows side yet.
EXPEED7 uses Arm processors and standard GPUs for its computational power. In that respect, EXPEED7 was close to but a bit behind an average smartphone’s processing power at the time it was introduced. However, EXPEED is more than just a set of processing cores. It also contains three specific additional known components. One is supplied by Socionext, which mostly resolves around external data transport (including CFexpress support). A second is supplied by Nikon, and centers mostly around Picture Controls and everything that ties into them (which includes raw processing). And the third is supplied by intoPIX, which is where most of the video processing is done as well as the High Efficiency raw capabilities.
From what I’ve seen in how Nikon has been able to update Z9 generation cameras, I’d say that the most likely issue that needs to be solved by EXPEED8 is internal processing pipeline bandwidth between the different IP modules. We know that the external bandwidth is high (e.g. recording to CFexpress reaches at least 800Mb/s). We know that the internal bandwidth for Picture Controls is high because the cameras have no problems generating 45mp files at 30 fps (essentially infinite buffer is possible).
When people talk about things they want in a Z9II, almost all of them probably aren’t possible with EXPEED7 because of internal bandwidth limitations. For instance, pre-capture raw: the issue here isn’t the data transfer into the intoPIX portion of the EXPEED7 chip, it’s the fact that you also need to touch the Picture Control portion of the chip, as well (JPEG previews need to be written into the file). I’m pretty sure it’s the pass-back-and-forth nature of what the data would need to do that’s made things like pre-capture raw and even proxy recording for raw video that’s put a limit on what Nikon can do with the Z9 generation.
When you add in things like examining the viewing stream for focus information and directing the lens and updating the viewfinder rendering, EXPEED7 is juggling a lot of balls that are using different parts of the internal components, and juggling is the right word here: you can’t control many balls simultaneously with one hand.
No doubt this will be addressed in EXPEED8: smaller process, more cores, better communication and bandwidth between cores, and so on.
Most of the user requests for something to change in the Z9II basically devolve into something that EXPEED7 would get overwhelmed by but EXPEED8 will likely address. In fact, outside of dynamic range, rolling shutter, and straight UI changes (e.g. changes to Save Menu Settings), almost every user request I’ve seen requires some work on the plumbing within EXPEED8 to accomplish correctly. Even things like applying user LUTs to video rendering in real time seem to fall into this category. So adding RED-things into EXPEED also comes into play for our next generation EXPEED.
I’m pretty sure that Nikon knows all the above (and more) and these items would have gone into the task list for creating EXPEED8. EXPEED, however, is at present a three-player dance (Nikon, Socionext, and intoPIX), which doesn’t play out as fast as a single player one, but Nikon’s done this dance before, so I’m confident that they’ll get things aligned again.
I believe that the current seeming delay is probably mostly outside of Nikon’s full control: everyone’s having difficulties getting new chips onto fabs and produced in quantity at the moment. The supply chain involved with semiconductors is still not fully recovered—and now is getting another hit from the Iran war—so the Really Big Players (Apple, Nvidia, etc.) are getting the highest priority, and they are paying for that priviledge.
We’ll see what Nikon was up to at some point, but at the moment I’m guessing that Nikon won’t have a production-level supply of EXPEED8 until very late this year. (If I’m wrong, and they already have EXPEED8 in hand—again, I don’t think they do—then it’s a delay with image sensor that’s slowed the Z9II release.)
As for the wait, I’ll remind you of 2011. Nikon had been gearing up to release some key products when the earthquake and tsunami hit Sendei and then later that year floods shut down the Thailand plant. This was an externally imposed set of delays that forced them to cancel at least one key product and gave them time to change another significantly before it was introduced. Nikon engineering doesn’t sit on their hands when they’re waiting for something to resolve that’s outside their control. It’s entirely possible that a delay in getting EXPEED8 production to volume might have also given Nikon the opportunity to change image sensor, too (or vice versa). So I wouldn’t get too worried about an extra year before the expected Z9 update is released.
Meanwhile, the current Z8 and Z9 are still near state-of-the-art cameras in almost every aspect. Personally, I’m feeling no need to jump to Canon or Sony while waiting, as for everything I might want changed on a Z9 I have an equal-sized list for the Canon R1 or Sony A1. Indeed, I still feel like the bigger issue for Nikon right now is that there some lens gaps that still need filling.
—————————
So what are the big user requests for Z9II generation? The following isn’t a complete list, but the ones I commonly see, ordered from most requested to less requested.
- Raw capture for pre-release.
- Improved autofocus, more modes, better subject recognition, better background rejection.
- More dynamic range.
- Better viewfinder. Sometimes coupled with better real-time viewfinder LUT/white balance rendering.
- Better settings saving capabilities.
- Proxy recording (record video to both slots).
- Higher frame rates (e.g. >20 fps).
- Higher flash sync. Related to better (or no) rolling shutter.
The remainder of requests tend to be isolated to a very specific use case or preference. A few of those are likely to come but have little true use impact (e.g. support CFe 4.0 directly for stills).