Lion Cove is said to reach 5.7 GHz on 3nm node, Skymont clock rising to 4.7 GHz
Last week, we covered the clock speeds of Zen 5 Ryzen 9000 CPUs. Now another information leak uncovered what clock speeds possibly reached by the Arrow Lake processors from Intel. This is an important piece of the puzzle, as we already know what the IPC of their cores will be, but the clock speeds were a big unknown, given how the new wider architecture and TSMC’s 3nm node could have drastically lowered them. And with that, the performance.
In the end, it seems that there might be some regression in absolute clock speeds, but it’s not going to be anything particularly bad, especially when compared to the impact that Intel’s 4nm node had on Meteor Lake generation clock speeds. Information from leaker called OneRaichu, who occasionally shares information about future Intel processors, has appeared on Twitter. It should still be taken with a grain of salt, these parameters are not yet verified from other sources.
Clock speeds drop on the big cores…
According to OneRaichu, Arrow Lake processors should reach 5.7 GHz on the big cores (P-Cores). This would represent a 5% deficit compared to the clock speeds achieved by the Core i9-14900K in the current generation (6.0 GHz). Intel also has an even more powerful Core i9-14900KS model with a maximum boost of 6.2 GHz, but that’s a special edition SKU (and it’s possible that some such “KS” CPUs will be created from the Arrow Lake generation as well, and thus i9-14900KS should be compared to that potential special edition Arrow Lake model).
Read more: Intel Core i9-14900K: The most powerful for desktop (review)
The 5.7 GHz clock speed should probably apply to the most powerful model, which should be the Core Ultra 9 285K, the lower models on offer will have lower maximum clock speeds as usual. This 5.7 GHz should apply to single-threaded applications (but it is also possible that it will be available for two threads as well), of course we are talking about P-Cores with the Lion Cove architecture. With all the big cores active (the fastest Arrow Lake has 8), the maximum clock speed is said to be 5.4 GHz.
The Lion Cove cores are said by Intel to have an average 14% higher IPC than the previous core, but there’s a slight complication to that value: the comparison is calculated based on Meteor Lake processors, not Raptor Lake processors. Anyway, the combination of the 14% increase due to IPC and the 5% reduction in clock speed would give a single-threaded performance increase of just over 8%. But remember that the IPC of a core varies depending on the application you’re running, so somewhere the increase may be better, and in theory, the overal performance gain could even tip over into a decrease in some unlucky apps (if the extracted IPC gain is low enough to not be able to compensate for the clock regression).
… but and improvement in the “little” ones
It is often said that the main star and hidden trump card of Arrow Lake processors is the efficient (so-called LITTLE) Skymont E-Core. It is actually a very wide architecture (given how it has as many as 8 ALUs, which probably says everything) and except for the SIMD instructions it is very powerful, with a massively increased IPC compared to the previous E-Core. And it’s possible that this will be true with respect to the clock speeds as well. Apparently, the clock speeds of E-Cores will not drop, but on the contrary, they could increase and reach up to 4.7 GHz (up form 4.4 to 4.5 GHz on Gracemont in fastest Raptor Lake CPUs), although this will probably not work on all E-Cores at once. The all-core boost for E-Cores is supposed to be 4.6 GHz.
This is actually not that far from the P-Core clock speed. Given that performance per 1 MHz is also getting closer (Skymont is supposed to have an IPC on par with the big cores in the Raptor Lake processors), it might actually be worth asking whether Intel could actually abandon the “Cove” line of big cores that has roots in the architectures from Sandy Bridge through Skylake and so on, and replace the big cores with something based on the “Mont” line of cores in some of the next generation processors. The “Mont” are originally based on Atom processors, but Intel has been steadily upping their complexity and performance significantly over recent years. They’re also more efficient per the area needed on the chip, whereas the “Cove” cores have a very large die footprint (they are significantly larger than the AMD architectures from the Zen line of cores that compete with them).
Who knows, perhaps this could also solve the problems of Intel processors with high power consumption compared to competitors (although high power consumption is mainly a conscious choice by Intel, not something given, as the processors have a wide range of efficiency and performance points Intel could choose). So perhaps there is a story emerging in which the current big cores will reprise the role of the Pentium 4 and its Netburst architecture, and the former “little” E-Cores will take over the throne and become the new P-Cores, like the disruptive Conroe architecture did in 2006.
Reduced clock speeds increase the likelihood that Intel will only score a draw against Zen 5
Arrow Lake processors with their final specifications should be revealed in September at the Intel Innovation event. Their actual launch and date of market availability should be in October. That’s probably also when the reviews will come out and the point where we’ll finally see how it all really turns out.
However, if the clock speed figure is true, then together with the IPC figure we can already guess a lot about the performance. Single-threaded performance should definitely be above the level of Raptor Lake processors at this clock, and also higher than with Ryzen 7000 processors. But AMD may have a higher intergenerational increase in single-threaded performance with its own Zen 5 core (used in Ryzen 9000 CPUs), as it appears to suffer no regression in clock speed.
So it’s possible that Ryzen 9000, whose official maximum boost will also be at 5.7 GHz, may be ahead of Arrow Lake in single-threaded performance. That is, at least in applications where Zen 4 wasn’t performing behind Raptor Lake (the IPC of different cores varies application to application, as mentioned). In software where Raptor Lake used to be ahead of Zen 4, Arrow Lake might in turn be able to get a win against Zen 5.
So the situation could be very close and highly competitive, which is usually a good thing. The relative standing of the processors in multi-threaded performance are harder to estimate. In games, though, we’d probably give Zen 5 a better chance of taking the throne because it will be able to eventually rely on 3D V-Cache, which provides a significant improvement over and above the normal raw performance of CPU cores. The lower clock speed of Arrow Lake (compared to prior Intel CPU generation) will be a deficit and may ultimately prevent Intel from catching up with AMD’s V-Cache processors in games. This makes the Ryzen 7 9800X3D a likely favorite for the position of the most powerful gaming CPU for the next two years.
Source: OneRaichu, VideoCardz
English translation and edit by Jozef Dudáš
⠀
