News Intel Arrow Lake refresh might not have a new NPU after all — latest reports indicate a clock speed bump only

I'm sure all three people who care about NPU performance will be very disappointed.
The NPU resides on the SOC Tile which is where the memory controller also lies. This means there was potential for ARL's biggest flaw to be addressed in some fashion. So while the NPU is meaningless to most that wasn't necessarily the only thing that would have changed as the SOC Tile would have had to have been re-engineered.
 
"Intel is still alive and doing exciting things we can all look forward to". "Alive"? At the moment "yes". But "exciting things we can look forward to"? The author is quite optimistic. When was the last time Intel released something that didn't disappoint? You might be getting "excited" to be disappointed. When it comes to current Intel, the "I'll believe it when I see it" very much applies. Until then, I'd lower your expectations to lessen the disappointment.
 
I was hoping for more i have a 265k now but was going to get 285k refresh but not sure now
I got a 265K instead of a 245K in part because the writing was on the wall and upgrade likelihood seemed low. Barring any foundational changes (and the radio silence from Intel puts me firmly in the "this is a very minor refresh" camp) or you needing the additional cores for a workload keeping what you have is probably the best course. There's a lot of levers for manual tuning, but if that's not your thing then 200S Boost is a good option for tightening up performance a bit.
 
Very boring, but a 3-5% clock uplift is something Arrow Lake craves.
Part of me thinks this is just Skylake all over again... The initial release was rather underwhelming, then it was overclocked to reach its (efficient) potential under a refresh (Kaby Lake). And it'll get refreshed and overclocked incessantly (52-core SKU ≈ 10900K), as newer architectures and processes get cancelled (20A, 18A)...
 
Part of me thinks this is just Skylake all over again... The initial release was rather underwhelming, then it was overclocked to reach its (efficient) potential under a refresh (Kaby Lake). And it'll get refreshed and overclocked incessantly (52-core SKU ≈ 10900K), as newer architectures and processes get cancelled (20A, 18A)...
Well that would be a disaster and contrary to planned releases that are rather soon now, like Panther Lake. It's late in the game to declare 18A dead for Panther Lake, since it's probably launching within the next 3 months or so.

Also, as they were using TSMC nodes for Arrow Lake, they could just use new ones. And maybe that will happen with different dies or chiplets using Intel or TSMC depending on profitability.
 
The initial release was rather underwhelming, then it was overclocked to reach its (efficient) potential under a refresh (Kaby Lake). And it'll get refreshed and overclocked incessantly (52-core SKU ≈ 10900K), as newer architectures and processes get cancelled (20A, 18A)...
The only reason this happened is that Intel used to tie architectures to process nodes. Even then they ended up backporting Sunny Cove to 14nm to limited success due to how large the die ended up being. Today the architecture team is just making their cores using industry standard tools which means they get produced on whatever node it makes sense to.