Tuesday, September 2nd 2025

Intel "Arrow Lake Refresh" Core Ultra 7 365K CPU Surfaces

Intel may update its current CPU lineup with an "Arrow Lake Refresh" generation, which will potentially bring improved clock speeds and out-of-the-box software optimizations tailored to gaming and productivity. A new Geekbench entry listing shows what appears to be a Core Ultra 300 series family CPU, the Core Ultra 7 365K, carrying eight P-cores, and 12 E-Cores for a total of 20 cores and 20 threads running at 3.9 GHz base frequency. The CPU benchmark run was recorded on an MSI PRO Z890-A Wi-Fi motherboard, paired with 48 GB of DDR5 memory. In a Geekbench v6 single-core test result, the refreshed Core Ultra 7 365K achieved 2,140 points, while the multicore score showed 19,744 points total.

This places the refresh below the average score of the current Core Ultra 7 265K, which has the most recent single-core Geekbench score of 3,030 and a multicore score of 20,861 points. The lower perfromance of the new CPU is likely due to an engineering sample running the benchmark, as well as software not fully utilizing the CPU, which is expected to perfrom better than the current Arrow Lake equivalent Additonally, the software identifier in the benchmark is the same, showing "GenuineIntel Family 6 Model 198 Stepping 2," while the ID is different, carrying a number eight, instead of the regular Arrow Lake's number nine. Interestingly, all other CPU characteristics are the same, meaning that the L2 and L3 cache structures remain unchanged. We initially expected to see Arrow Lake refresh bring an improved NPU performance to classify it as a Copilot+ AI PC. However, later rumors indicated that the change will not arrive at the NPU. As Intel bets big on its 2026 "Nova Lake" family, we are left to wonder if Intel will launch this lineup at all and if it makes market sense to do it before Nova Lake arrives.
Source: BenchLeaks on X
Add your own comment

17 Comments on Intel "Arrow Lake Refresh" Core Ultra 7 365K CPU Surfaces

#1
kondamin
It should make the multi generation socket people happy.
As that would technically be a second gen 1800 chip, even if it isn’t.
Posted on Reply
#2
westlake
In my opinion, this is not real. As far as I know, GeekBench can still be beaten with a virtual machine. It makes no sense for the refresh model to be slower (5400 MHz vs. 5500 MHz). I don't think there has ever been a case in the past where the refresh operating frequency was lower. It has always been higher. It would be logical to think that this is an ES piece, but it is not. This is a retail model (C0662). If this is a real result, then it makes no sense for it to be clearly slower than the 265K. So, in this form, I cannot consider it to be real.
Posted on Reply
#3
ixi
westlakeIn my opinion, this is not real. As far as I know, GeekBench can still be beaten with a virtual machine. It makes no sense for the refresh model to be slower (5400 MHz vs. 5500 MHz). I don't think there has ever been a case in the past where the refresh operating frequency was lower. It has always been higher. It would be logical to think that this is an ES piece, but it is not. This is a retail model (C0662). If this is a real result, then it makes no sense for it to be clearly slower than the 265K. So, in this form, I cannot consider it to be real.
Could be better IPC and lower clock means less wattage required ( I do.hope so ), but knowing intel. They like wattage.
Posted on Reply
#4
N/A
Engineering samples are always slower, hence the 41.6 faster 265.
I'm waiting for a titan lake refresh, still a few sockets away.
Posted on Reply
#5
Hyderz
these cpu still dont have hyper thread? like it older sibling before this?
i wonder what performance bump we gonna see from arrow lake? if intel hopes to sell well or break even.. they should sell these at a lower price..
the current C U 9 285k is $599, the C U 9 385k should be at $529 or $519..
Posted on Reply
#6
kondamin
ixiCould be better IPC and lower clock means less wattage required ( I do.hope so ), but knowing intel. They like wattage.
IPC changes on a refresh are unlikely as that would require an architecture change
Posted on Reply
#7
nomdeplume
Arrow Lake was clearly there to reestablish OEM stability. No need to guess where the refresh is headed or who requested it.
Posted on Reply
#8
Squared
Core Ultra 365K? I sure hope not. All these new generation names for a mere refresh are exhausting. Call it Core Ultra 270K or Core 365K in keeping with convention.
Posted on Reply
#9
RootinTootinPootin
I can almost smell the AMD boi's popping here anytime soon to mock those numbers..Hope that's not a flop (it looks to me like one)
Posted on Reply
#11
ir_cow
I guess I will continue to wait for the X3D Intel version on the next socket....
Posted on Reply
#12
Hexage
arrow lake is just too slow in some regards. the entire architecture needs a overhaul.
Posted on Reply
#13
Squared
Hexagearrow lake is just too slow in some regards. the entire architecture needs a overhaul.
Arrow Lake and Zen 5 offer nearly identical gaming performance on desktops. Zen 5 with V-cache is a lot faster in games. Raptor Lake with higher clock speed is a little faster in games. If Intel could get Arrow Lake up to Raptor Lake's clock speed with a refresh, it'd do alright in games: worse than X3D chips but better than Raptor Lake and AMD's non-X3D chips. And outside of games Arrow Lake can keep up with any AMD offerings.
Posted on Reply
#14
oxidized
Am i the only one thinking this has to be a bad idea? A flawed cable, also made 90 degree like that will only made things worse
Posted on Reply
#15
Thatsbs
oxidizedAm i the only one thinking this has to be a bad idea? A flawed cable, also made 90 degree like that will only made things worse
I'm confused as to what your relating to? what cable?
Posted on Reply
#16
Thatsbs
nomdeplumeArrow Lake was clearly there to reestablish OEM stability. No need to guess where the refresh is headed or who requested it.
Do you think my socket (1851) has hope for a next gen of CPU (hopefully more than a refresh?) Or a complete new socket (1954) will actually be required for the likes of nova lake.. I'm a bit unsure how many pins on the grid are taken up? Is there still possibility that the 1851 socket could potentially support the nova lake? Or am i hoping too much?
Posted on Reply
#17
nomdeplume
ThatsbsDo you think my socket (1851) has hope for a next gen of CPU (hopefully more than a refresh?) Or a complete new socket (1954) will actually be required for the likes of nova lake.. I'm a bit unsure how many pins on the grid are taken up? Is there still possibility that the 1851 socket could potentially support the nova lake? Or am i hoping too much?
My comment referred to Intel's large contracts filling office and OEM machines. How they are delivered in years to come may separate in key ways from retail boxed processors. From the outside it certainly appears a lot of changes are taking place currently that might scrap a few rounds of plans or remove them outright from certain markets.

Upgrading within Intel socket historically offered very little. Gen 8+9 or gen 10+11 being good examples that are easy enough to compare against modern standards. Fair to good chance there won't be anything to hold your breath for. Basically, I have no idea what is going on at Intel or how it will shape future releases.
Posted on Reply
Add your own comment
Nov 16th, 2026 09:17 CST change timezone

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts