Intel Core Ultra 245K Boost + 8,000 MT/s RAM BENCHMARKED (vs 14600K)
This is Intel’s Core Ultra 245K – the newest i5 that everyone is talking about – and between this rather gaudy Corsair 8,000 MT/s CUDIMM RAM and the new “Core Ultra 200 Series Boost” feature, I thought I’d take a look and see what sort of performance uplift you can expect from both fast RAM and enabling the new ‘boost’ feature. First though, what is this “200S Boost” feature anyway? Well to be blunt, it’s Intel overclocking your CPU and hoping you thank them for it. This feature only applies to K-SKU chips, like this 245K, and Z890 motherboards, which is a bit of a punch in the value-proposition-gut for Intel here, considering for all intents and purposes most people should probably opt for a B or H series motherboard, much like with AMD Ryzen chips and their B series boards being the better choice for functionally everyone. This also doesn’t actually overclock the chip – at least in the way you might expect. This won’t lift your main clock speeds, instead it boosts the Fabric and Die-to-Die clock speeds from 2.6 GHz and 2.1GHz respectively to “up to 3.2GHz” on both – plus it allows support for 8,000 MT/s memory (CUDIMM or not) – with the key thing here being that enabling this explicitly named “overclocking profile” does not void your CPU’s warranty, unlike basically any other type of clock-changing overclocking. This also extends the RAM speed warranty from 6400MT/s to 8000MT/s. It’s literally just a single BIOS toggle, so let’s put it to the test and see what – if any – difference it makes to this 6 + 8 core CPU!
Starting with Cinebench R23 we see a pretty unusual result. As the single threaded performance increases, the multithreaded performance DECREASES… Weird, I know. What that means is the slightly slower single core performance on the 14600K – 2001 points versus up to 2098 points – ends up beating the 245K in every config (with 6000 MT/s RAM mind you) in multithreaded. Now it’s hardly a landslide victory – under 300 points out of 24,000 in it – but still. Interestingly while the Boost mode did increase the single threaded performance slightly, it decreased the multithreaded performance. This is a rather strange result to me, so let’s look at Blender which makes more sense. In the shorter BMW scene there’s only a spread of three seconds, with the 245K with 8,000MT/s RAM and Boost enabled taking the ever-so-slight victory there. Of course if we look at the longer Gooseberry render, well the story changes quite a lot doesn’t it? The 14600K runs away with the lead – a near ten percent lead over the 245K’s best offering. The Boost feature does make a difference, but so does the RAM. We drop 17 seconds just bumping up RAM speed, then a further 23 seconds enabling Boost. That comes out to 2.65 percent improvement with just RAM, 2.98 percent from just Boost, or 6.48 percent with both. One thing that the 200 series chips were pretty undisputedly better at than the 13th and 14th gen is power, and that’s true even with high speed RAM and Boost. The 245K sat at between 128 and 132 watts of constant full-load power draw, with peaks up to 155 watts, which is actually the 14600K’s stable power level. That peaked past 160 watts, although interestingly while the RAM does increase power draw slightly, so does Boost – but RAM does more, by a watt at most mind you.
As for gaming, testing with an RX 6900 XT, well starting with CS2 at 1080p on low settings we can see the 14600K still at the top of the charts, although the 245K does get awfully close with both high speed RAM and Boost enabled – just 4 FPS in it between them, and the 245K does have better 1% low performance too. To me it’s very clear the RAM is doing the heavy lifting here, as if we were just comparing RAM speed apples to apples between the 245K, Boost on, and the 14600K, there wouldn’t be much competition here. Admittedly CS2 running at 375 vs 410 FPS won’t exactly alter your gaming experience, but no matter how you look at it the 245K is at least a slight step back in performance, even with the flash RAM and the OC feature.
With Cyberpunk though, again at 1080p on Medium settings, you’ll find a different story. Here the 14600K is actually last, by 0.9 FPS but still last. What’s more interesting is that the Boost feature did functionally nothing for the performance. There’s a bit of instability on the 1% lows with 8000MT/s RAM and no Boost, but otherwise it’s functionally identical performance across the board, with only a noticeable difference in RAM speed, rather than boost.
Shadow of the Tomb Raider has the 14600K back up at the top, and comparing to the non-boost, matching RAM speed result from the 245K it’s a runaway win for the older chip. Boost does increase performance by 10 FPS – 4 percent higher – but RAM speed is what brings the 245K closer to the 14600K, offering 25 FPS more performance on its own, with functionally no improvement from Boost with the higher RAM speed – 1.3 FPS on average but 1.1 FPS lower 1% lows.
Rainbow 6 Siege – with its most recent Siege X update which annoyingly updated mid-testing making me redo a bunch of these runs – shows how the 245K really comes alive with that 8,000 MT/s RAM. Like, with matching 6,000 MT/s RAM the 245K – boost or not – is no match for the 14600K. Sure there isn’t likely much of a noticeable difference, but it’s hard to route for the newer chip when it’s actively slower than the last one – even with Boost – but with 8,000MT/s RAM it’s actively faster. Strangely boost actually hurt performance a decent bit, so that one’s a wash for Boost, but a strong indicator these chips need ultra-fast RAM to really see good performance.
Hitman 3 finally has a decently convincing result for the 245K regardless of Boost or RAM speed – the 14600K lags behind the 245K by anything from 7 to 14 FPS or up to 10 percent! Boost clearly doesn’t make much of a difference, and in the grand scheme of things RAM speed doesn’t seem to matter so much in this one. Sure, the 8,000 MT/s RAM with Boost is the fastest, but only by an FPS or two.
Starfield on low at 1080p is another one where Boost doesn’t seem to make a difference, but RAM speed definitely does. With matching RAM speeds, the 14600K has a paper victory over the 245K here. With the much faster RAM though, the 245K comes alive, and offers a noticeable performance difference of 10 FPS – which at these sorts of ranges is actually something you’re likely to feel and see. Boost though has no big effects here. It doesn’t hurt to enable it for sure, but don’t expect overclocked levels of performance.
Looking at the average of these games then, at 1080p, we see a pretty similar trend. The 14600K splits the two RAM speed options, with a pretty tight spread across the board. The 245K, with Boost and 8,000 MT/s RAM, is only 6.6 percent faster than the 14600K, and I have a sneaking suspicion that if the 14600K supported CUDIMMs (or Intel provided me with an 8,000 MT/s kit that wasn’t a clock driver kit) we’d see similar performance gains there too. If you opt to compare apples to apples with both chips running 6,000 MT/s RAM, the 14600K is the faster chip. It’s between 5 and 6.6 percent faster than the 245K, and somewhat as expected the Boost feature – which only boosts the interconnect speeds, not the clock speeds or multipliers – doesn’t give you a whole heap more performance. If you were interested, at 1440p as expected the layout is the same, but the gaps close slightly. The 245K with fast RAM and boost is 4.6 percent faster than the 14600K, and the 14600K is 4.5 percent faster than the 245K with the same RAM and no Boost.
So, Boost does help in some games, some of the time, but not by all that much. This isn’t going to be game-changing performance, as the biggest difference really is using faster RAM. At least this mode does allow for that faster RAM to be covered under warranty – that’s likely the biggest factor here really. The fact that this mode, which really only tweaks settings you could argue should have been set like that from the get-go, is only available to the most expensive K-SKU chips, on Z890 boards, that’s a bit of a kick in the teeth for anyone who did buy one of these Core Ultra 200S chips but opted for a better value proposition in a non-K chip or a B or H series board. For the five people this thing applies to, it is if nothing else a bit of free performance. It can’t hurt performance so you may as well enable it, but don’t expect too much. If you do opt for a Core Ultra 200S chip, you definitely want to consider ultra high speed RAM if you want to get the most out of them. 8,000 MT/s RAM like this Corsair Vengeance kit is actually a great shout. Being a CUDIMM it’s more stable and lower power, making it a good pairing for these chips. Whether or not this performance bump is enough to convince you to opt for Core Ultra for your next build, I can’t say, but I’d love to hear what you think in the comments below!
