INTEL IS BACK?? Intel 250K PLUS & 270K PLUS Review vs 265K & 245K

Intel, the only chip maker who hasn’t gone all-in on the AI bubble that’s poised to pop, is actually offering some new consumer-facing CPUs! It’s a shock, I know, but they are real, and they are here. These are the 250K Plus, and 270K Plus, part of the Arrow Lake S Refresh. So, let’s dive in and see what is actually new with these, then we’ll compare performance to the existing parts, the 265K and 245K. First, what’s new?

Well since this is a ‘refresh’, there isn’t much architecturally different. I’m assured that the silicon is different, not just a ‘binned’ version of the older chips, but it’s still the same Lion Cove P cores and Skymont E cores. The big difference physically is that Intel have slid the stack down, so the 270K Plus gets the 285K’s 8 P cores and 16 E cores (up from 8/12 on the 265K), and the 250K Plus gets an extra 4 E cores for a total of 6P and 12E. The clock speeds generally stay the same though – a 0.1GHz upgrade on the boost frequency at most. Intel is claiming 15 percent higher gaming performance than the last gen – we’ll be testing that in this very video – in part from the hardware changes, but mostly from the new hit feature, the Intel Binary Optimization Tool, or as I am absolutely going to be shortening it to, iBOT. IBOT is incredibly important so I want to dive into what that is, but I’ll finish the new bits. Official RAM support is now up to 7200 MT/s, 4 rank CUDIMMs are being trialed, the Intel Platform Performance Package is being updated to give more performance with less work, and amazingly Intel have done the most sensible thing and called this what it is, a refresh. This is still a 200 series part – not 300 series – while still being clearly marked as newer and better. THIS is what we’ve wanted all along, so great job Intel!

Right, back to IBOT. Intel’s Binary Optimization Tool is, in Intel’s words anyway, a revolutionary step in real-time instruction per clock improvement. Genuinely though, this is actually a really cool technology that swaps out instructions on the fly to get better performance. Basically, Intel runs a given game or program in their lab, and records all the instructions and functions the game or program calls. If they work out that there’s a faster way to do the same operation, they store that change, so when you then enable that profile, their software can swap out those function calls on-the-fly. If this sounds remarkably similar to AMD’s reviled Anti-Lag+ which did something awfully similar, resulting in people getting perma-banned from games, that’s because it kind of is. It’s a little different, namely that this function call swapping isn’t Intel injecting code into the game or program, it’s doing anonymous function call switching on the fly and always returning the same result, so the game is non-the-wiser. Also, Intel says it’s working with developers and publishers to ensure this won’t cause any issues. Because Intel has to profile each game or program on each CPU, it only works with a handful of titles at launch (12, here’s the list), and it only works on the Plus chips. Apparently some of those subtle hardware changes make it possible, and while they are considering backporting it to the non-plus chips, they don’t have anything settled for sure so don’t hold your breath if you’re one of the five people that actually bought one for gaming. IBOT is purportedly good for an average of eight percent more performance (at 1080p, with a 5090), so I guess according to their own figures, the hardware changes give you three percent, while IBOT makes up the difference. 

Right, that’s what’s new, now let’s talk performance. Let’s take a really quick peek at the raw CPU performance – Cinebench has a pretty linear progression between the generations – I mean the extra E cores, plus some clock speed and IPC improvement makes sense – with a sizable shift on each part. We’re talking a 5000 point jump in multi-threaded, and 100 points in single thread. That’s decent. Blender is the same, impressive gains where the 250K gets worryingly close to the 265K, and the 270K really kinda runs away with it. Interestingly though, power is commensurate with that extra performance. The 270K drew nearly 60 watts more at peak (and steady-state) than the 265K, and the 250K drank exactly 30 watts more on average across the Gooseberry render. That’s quite a bit, and that’s all on the stock settings. I can’t imagine what unlocking the power limits in the BIOS might do for your power bill!

When it comes to gaming I’m doing things a little differently to the other reviewers who will inevitably have reviews of these chips up today, in part because I need to do something a little different to stand out, but mostly because the highest end GPU I have is either an RX 6900 XT or an RTX 3070, both of which would probably constitute a flawed test in most people’s eyes anyway, so instead I’m leaning into it. I’m using an RX 9060 XT (the 8GB one), because I want to know if, using a real GPU you can actually afford, these chips are actually any better, and worth splashing out for. So, are they? Well I tested at both 1080p and 1440p to find out. Let’s start with the non IBOT games – although I did have APO enabled (that’s Intel’s last-gen trick to eke out more performance offering game specific profiles) for all chips. So, let’s start with CS2. At 1080p there is actually a really surprising spread here. The 245K gives us 535 FPS on average, while the 250K gives 607. That’s a surprisingly large delta – not a perceptible one mind you, but still – and there’s a similar jump from 613 FPS to 685 FPS on the 265K to 270K. At 1440p, as expected, the gaps narrow, and the 250K actually takes a very narrow lead over the 265K. All of this performance delta is kind of moot though, but interesting to see some improvement.

GTA V (enhanced) has a slight improvement across the board for the Plus chips too – 5 FPS on the u5 and 13 FPS on the u7’s. Again that isn’t significant, but it’s interesting. At 1440p, surprisingly, there isn’t much difference to 1080p, with the 250K overtaking the 265K (slightly), and the 270K only dropping three FPS on average. That’s impressive, showing a clear bottleneck elsewhere.

Lastly for the non-BOT-ed games, Rainbow 6 Siege. At 1080p there may as well be no difference here. It’s within like three FPS at 430+ FPS on average. Your CPU doesn’t matter, at least on medium settings and with this GPU. Same for 1440p, so I’ll flash this up and if you want to pause to look you’re welcome to do so, but otherwise I’m moving on.

With Intel’s Binary Optimisation Tool enabled for the supported games (as well as APO), Cyberpunk shows literally no difference. In fact the IBOT results are one or two FPS LOWER, not higher. There’s also functionally no performance difference between the chips, even at 1440p, so this one is gonna be a fly-by. 

Hitman 3’s built in benchmark lets me split out the CPU and GPU data, so obviously we are looking at the CPU data here, and happily there are actually some differences to talk about! The overall spread is actually important, with the 245K and 265K sitting a way’s back from the Plus chips – IBOT or not! There’s a 30 FPS spread here, which is significant. The 250K doesn’t see any average FPS improvement, but it does see improved 1% lows, but the real star is the 270K which added 6 FPS on average (but no 1% lows growth). That’s not bad for a toggle – although remember that this is the CPU data from the built in benchmark, so not necessarily the be-all-end-all for your gaming experience. At 1440p the gaps close – the 245K still struggles, as does the 265K by comparison, but it’s a much smaller gap.

Lastly we have Shadow of the Tomb Raider, the supposedly golden-child of IBOT. Sadly, at least on this mid-tier GPU, it doesn’t matter much. They basically all offer 217 FPS average, save for the 245K which gets a paltry 213 instead – shocking I know. Still, these results look identical, as do the 1440p ones – worryingly so actually. It’s weird. Anyway, that’s Shadow of the Tomb Raider.

So, do these new CPUs make any difference on a mid-tier card? No, not really. There are a couple games that have a few little gains, but it’s hardly monumental. This is a refresh – Intel made that clear – and the true win here is that Intel is selling these upgraded versions of last years kit for the same as current pricing. The 250K Plus is going to be $199 – the same as the 245K – and the 270K Plus will cost $299, which is the same as the 265K. If you were going to buy either of those older chips, these are here for the same money, and offer more cores and some minor improvements. Why wouldn’t you buy these instead? Obviously if you were going to buy team red’s offerings, it’s unlikely these minor improvements would move the needle enough to get you to buy these instead, and let’s be honest, right now is a truly terrible time to build a new gaming PC, so this whole thing kinda feels like a moot point right now, but it’s nice to see a shift inside Intel towards being more consumer friendly (now they’re on the back foot anyway). Good job Intel, if, for nothing else other than NOT calling this the Core Ultra 300 series.