Specs of 125W Core Ultra 9, 7 and 5 processors of the 200 generation
As the October release of Arrow Lake CPUs with Intel’s first all-new CPU architecture since 2021 slowly approaches, the stream of information is starting to grow stronger. It seems that the preliminary specs of the 125W K-series models have been leaked, which will be the first to come out for the new LGA 1851 desktop platform, and also the fastest Arrow Lake CPU models. The first test of the cheaper Core Ultra 7 model has also surfaced.
This information comes from two sources. The first one is the leaker called OneRaichu on Twitter who also gave the first glimpse of the clock speeds of the top model of Arrow Lake. But now they also present the parameters of the remaining 125W Arrow Lake desktop CPUs. It is possible that parts of data might be preliminary and not completely final, though.
However, alongside this, the first test of the Core Ultra 7 265KF processor, which will be the lower model of this unlocked series, has also appeared in the Geekbench database. Thus, the parameters of this CPU are better documented. The sample in the database is apparently a qualification sample (it’s too early for volume retail units), which means it should have the same parameters as retail units.
K-models of Arrow Lake desktop processors: parameters
The highest Core Ultra 9 285K SKU is supposed to have eight Lion Cove cores with a maximum clock speed of 5.7 GHz, which we already reported last time. This clock speed will only be available for a small number of cores active at the same time. With all eight Lion Cove cores active, the maximum is expected to be 5.4 GHz. By comparison, the Core i9-14900K has a maximum boost of 6.0 GHz and 5.6 GHz when all cores are under load. So there is some regression in clock speed with the P-Cores. While the higher IPC (allegedly a 14% increase in performance per 1 MHz over the previous Meteor Lake generation) will make the processor a bit faster, some of that progress will be lost to the clock speed reduction.
On the other hand, the E-Cores, of which there are 16, will have higher clock speeds. While in the Core i9-14900K their maximum is 4.4 GHz, they will reach 4.6 GHz in the Arrow Lake K-series, including in this SKU. In addition, they are supposed to have massively higher performance per 1 MHz thanks to the new and significantly enhanced Skymont architecture.
The first preliminary benchmarks of this processor were shown by the leaker Jaykihn, we wrote about them here:
Read more: First Intel Arrow Lake benchmarks: Expectations turned inside-out?
The lower model Core Ultra 7 265K that was spotted in Geekbench has 8 Lion Cove cores and 12 Skymont cores, so it will be a similar configuration to the Core i7-14700K, but with fewer threads due to the elimination of HT on the big cores. The maximum boost clock speed is apparently 5.5 GHz (100 MHz lower than its predecessor) and the maximum all-core boost is 5.2 GHz (300 MHz lower). The E-Cores are also said to have a maximum clock speed of 4.6 GHz (which is 300 MHz higher than on the i7-14700K).
This processor will also have the Core Ultra 7 265KF version (which is actually the one that was leaked in Geekbench, instead of the basic 265K version), which will differ in its having the GPU deactivated and being sold for lower price. The top-of-the-line Core Ultra 9 won’t have such a KF version, so you won’t be able to use that to save money when buying the top SKU (this is a change from previous generations, where Core i9 “KFs” were/are commonly available).
In Geekbench 5.5.1 for Windows (note, this is not Geekbench 6), the Core Ultra 7 265KF scored 2252 points in the single-threaded test and 17,722 points in the multi-threaded test. However, since it’s more than two months until launch, these numbers may not be reliable as the firmware may not be fully optimized yet. Also, this is just one result, which may not be very representative.
And finally, the most affordable model should be the Core Ultra 5 245K (a Core Ultra 5 245KF is supposed to exist as well). This processor has 6 Lion Cove cores and 8 Skymont cores, the same as today’s unlocked Core i5s. The maximum clock speed of its P-Cores will be 5.2 GHz (100 MHz lower than the i5-14600K), but only 5.0 GHz under an all-core load (that’s -300 MHz compared to the i5-14600K). The E-Cores have a maximum clock speed of 4.6 GHz as well, which is +600 MHz compared to the LITTLE cores of the i5-14600K. So multi-threaded performance may improve by a lot.
These models (Core Ultra 200K series) will be released in October along with the Z890 chipset and the first LGA 1851 motherboards. Other cheaper models with 65W TDP (note that we’re only assuming TDP values for now, theoretically they could be changed) and locked multipliers, as well as cheaper boards with B860 and H810 chipsets, will come out in 2025, probably in January during CES 2025.
Read more: Intel Arrow Lake chipsets: Z890/B860/H810 specs and differences
The new generation of processors can’t come soon enough…
Intel desperately needs Arrow Lake processors, as the current offering of Raptor Lake desktop chips suffers from major silicon degradation issues causing loss of stability over time. Intel is planning a fix, but it’s not even sure if it will stop the process leading to the chip failure or just slow it down, and all the processors already in use have already been affected by this damage to varying degrees. The culprit, according to internal documents, is that the voltage management technology in these processors is inadequate for the high voltage operation that can occur on the LGA 1700 platform during boosting (and this could in theory be affecting HX-series laptop processors as well).
- Read more: Confirmed: Raptor Lake CPUs are degrading, Intel preparing fix
- Read more: More on Raptor Lake degradation: Chips getting damaged even when idle?
Confidence in the current 13th and 14th generation Intel Core desktop processors on the LGA 1700 platform has been thoroughly shaken by this, and it would be ideal if buyers interested in Intel’s offerings could get a new architecture and platform instead, where there will be at least some reason or hope that the voltage supply will be handled robustly enough to prevent such problems from happening. Although the previous kerfuffle probably serves as a warning that Intel isn’t a brand with which you can be 100% certain to never have any problems, as people sometimes imagine (it’s probably fair to say that no such thing as perfectly reliable brand exists).
Sources: VideoCardz (1, 2), Geekbench
English translation and edit by Jozef Dudáš
⠀
