Monday, July 7th 2025

Intel Plans "Arrow Lake Refresh" for H2 2025 with Higher Clocks and Upgraded NPU

According to ZDNet Korea's sources familiar with the matter, Intel will release the "Arrow Lake Refresh" CPU family in the second half of this year. While no exact timeline is given, we can expect to hear more about these CPUs in the coming weeks, and for the rumor mill to intensify. The new refresh generation will introduce slight frequency adjustments, including an increased base clock and maximum boost clock, while also incorporating a faster NPU. A base clock increase, paired with a boost to maximum core clock, will give Arrow Lake Refresh a significant advantage in gaming performance. Keeping the same LGA1851 socket will extend the support for the current platform for one more generation too, until "Nova Lake-S" arrives.

Given that Intel attempted an IPO for a limited set of gamers, we could see a more widespread adoption of a similar approach. As a reminder, on the CPU side, IPO fine-tunes P-core and E-core frequencies, ring-bus speeds, the UPI interconnect, D2D links between tiles, and both PL1 and PL2 power limits. For RAM, IPO raises transfer rates and tightens timings, pushing modules beyond their factory XMP or EXPO profiles. If Arrow Lake Refresh ships with an IPO-equivalent out of the box, gamers will enjoy a significant performance increase, ready to offset some of the early disappointment with the base Arrow Lake. Additionally, Intel will place a new NPU on its IO tile. Using the NPU3 on base Arrow Lake, Intel was unable to meet the basic 40 TOPS requirement for Copilot+ AI PCs. And with NPU4, that TOPS requirement is met, allowing Intel to brand these new processors as AI PC ready.
ZDNet Korea also noted that Intel is preparing Nova Lake-S for the second half of next year, meaning that we are still months away from hearing more details about its platform. For now, we can only note that the Nova Lake-S, rumored to feature up to 52 cores, will be compatible with an LGA1954 socket and will include a mix of P, E, and LPE cores.
Source: ZDNet Korea
Add your own comment

78 Comments on Intel Plans "Arrow Lake Refresh" for H2 2025 with Higher Clocks and Upgraded NPU

#26
TheGuruStud
_roman_Some still advertise to buy an intel processor. I think I read it's cheaper. A Mainboard change is not an argument for the intel buyer. Buying new DRAM is an argument to avoid AMD. Therefore buy a new intel processor, a new intel mainboard and keep the old DDR4 DRAM.

Too bad AI is such worse to answer simple question. How many mainboard changes for intel platform vs amd platform for a given time-frame from year to year.
Intel STILL can't beat 5800X3D, so a lot of MBs lol. Mine Oced on X370 ties OCed 14900K (not that there's any headroom lul). Absolutely pathetic.
Posted on Reply
#27
JustBenching
TheGuruStudMine Oced on X370 ties OCed 14900K (not that there's any headroom lul). Absolutely pathetic.
No it doesn't. Not even close. This is just false.
Posted on Reply
#28
Tyler-98-W68
TheGuruStudIntel STILL can't beat 5800X3D, so a lot of MBs lol. Mine Oced on X370 ties OCed 14900K (not that there's any headroom lul). Absolutely pathetic.
a 245K does, and it absolutely demolishes a 5800x3d outside of gaming, cope harder amd fan boy
Posted on Reply
#29
JustBenching
Tyler-98-W68a 245K does, and it absolutely demolishes a 5800x3d outside of gaming, cope harder amd fan boy
I tested a friends 5800x 3d vs my 12900k, both oced, there is no comparison. I don't know where the heck do they get these weird ideas, the 3d chips are fast when stock cause they are not reliant on memory, when you start tinkering with memory then things kinda change.
Posted on Reply
#30
oxrufiioxo
JustBenchingI tested a friends 5800x 3d vs my 12900k, both oced, there is no comparison. I don't know where the heck do they get these weird ideas, the 3d chips are fast when stock cause they are not reliant on memory, when you start tinkering with memory then things kinda change.
I chuckled pretty hard when I seen that comment I was like wtf. The 265k beats it and cost about the same and destroys it outside of gaming for around the same price now at least in the states vs the last price you could get a 5800X3D for.



I am guessing the person making the comment only plays one game the 5800X3D is really good at and bases all cpu gaming performance on that specific game.
Posted on Reply
#31
Panther_Seraphin
JustBenchingI tested a friends 5800x 3d vs my 12900k, both oced, there is no comparison. I don't know where the heck do they get these weird ideas, the 3d chips are fast when stock cause they are not reliant on memory, when you start tinkering with memory then things kinda change.
DDR4 or DDR5 with that 12900k? Decent DDR5 is what makes Alderlake really shine in comparison but it was unbelievably expensive at time of launch and for a good while after.

5800X3D was amazing at time of release and was/is amazing for anyone still on AM4 now and doesnt have a decent chunk of change to do a full MB/RAM/CPU swap. Otherwise going to AM5/Waiting for intels next release would be my advice currently. Advising/advocating for AM4 parts now is silly.
Posted on Reply
#32
JustBenching
Panther_SeraphinDDR4 or DDR5 with that 12900k? Decent DDR5 is what makes Alderlake really shine in comparison but it was unbelievably expensive at time of launch and for a good while after.

5800X3D was amazing at time of release and was/is amazing for anyone still on AM4 now and doesnt have a decent chunk of change to do a full MB/RAM/CPU swap. Otherwise going to AM5/Waiting for intels next release would be my advice currently. Advising/advocating for AM4 parts now is silly.
DDR5.
Posted on Reply
#33
eldon_magi
For me, mini pc's and sunshine/moonlight streaming has changed the game completely. For my KDE Plasma desktop, on 3x 4K screens, i only need a mini pc running wayland, browsers, some bash terminals, and playing videos and music.

I bought an GMKtec K8 Plus which has a Ryzen 8845HS/780M, 96GB ddr5, and a 4TB SN850x. without ram and ssd it was 372 EUR. Could not be happier. Everything runs smooth, browser and video playback are fully hardware accelerated, not a single stutter anywhere, rock stable running 24/7.

I have a slider in KDE for 'silent, balanced and performance' mode, usually keep it at silent, sometimes compiling a kernel i can give it some power. For anything else silent is enough..

Have it mounted behind one screen with a vesa bracket, zero cables on the desk, and zero gadgets. Zero cables on the floor.

Other desktops are running on two servers in a separate room, in proxmox servers with pcie passthrough to GPU's. No need for dummy plugs, or anything connected to the GPU outputs using the EDID kernel trick. Stream those outputs 4K@60 with hevc over 2.5gbit ethernet, well tuned it is practically indistinguishable from a physically DP/HDMI connected GPU. I can stream windows, steamdeck, any linux GUI, or any app.

I will wait for nova lake, but even a core ultra 7 is interesting, as i could run a headless proxmox server on it, use cpu affinity to run a desktop system on the 12 e cores streaming from the igpu, and the 8 p cores running a gaming server streaming over a nvidia or AMD GPU. That's two very capable systems running on a single computer, with no need to make it run silent, no problem with heat, no big boxes or big cables. The mini pc just shows the video output, it's just an app in the taskbar.

below some pics of the K8 plus mounted behind screen..
Besides that all that's needed is keyboard and mouse. No heat, no noise.. it's heaven
Posted on Reply
#34
efikkan
Back in the day people used to welcome mid-generation refreshes.
Hopefully this refresh offers more than a tiny adjustment to clocks and voltages.

Despite all the FUD, Arrow Lake has been praised for how consistent the performance is, not to mention very good performance in Linux. The amount of mental gymnastics people go through to justify inferior products is astounding, and I sincerely hope that the amount who fall of this, is about equal to the amount of people who actually game in 720p on a RTX 4090/5090. (Especially those arguing for Zen 3 CPUs right now, which will not be snappy on a desktop by today's standards.)
Posted on Reply
#35
Tyler-98-W68
Since we got so off topic with stupid SFF's in this thread i'll just ramble on about my ITX Minisforum MB with R9 7945HX3D, because its fun....and I didn't put it in a SFF case.

Fun fact, my Ultra 5 225 games better than it does (and they both use about 100w of power under full CPU load)

www.3dmark.com/compare/spy/57264746/spy/56887902

And somehow the world's worst gaming CPU somehow matches/slightly beats a 7950X3D (lower power)
Posted on Reply
#36
kondamin
So will it get Battle mage as igpu like lunar lake?
Posted on Reply
#37
eldon_magi
I'm telling myself wait for nova lake before doing anything.. but arrow lake refresh this year.. If they upgrade the igpu to battlemage.. could be irresistable
Posted on Reply
#38
Outback Bronze
oxrufiioxoI was being sarcastic
I was well aware of that matey ;)

I just wanted to put my 2c in about that I will likely be skipping it. I'll be surprised if the refresh is any decent.
Posted on Reply
#39
oxrufiioxo
Outback BronzeI was well aware of that matey ;)

I just wanted to put my 2c in about that I will likely be skipping it. I'll be surprised if the refresh is any decent.
Maybe next gen gaming will get some love from both sides if it wasn't for X3D gaming on Zen5 would have been a letdown also.

Outside of gaming the 285k was super solid I just think the i5 is too compromised to matter for gaming anymore and that was their bread and butter for gaming for so long.

The problem and I've seen this locally is a 9600X performs about the same as the 285k in gaming and is like 1/3 the cost close enough that you wouldn't notice which you were using so anyone who only games the cpu is a waste.
efikkanBack in the day people used to welcome mid-generation refreshes.
Hopefully this refresh offers more than a tiny adjustment to clocks and voltages.

Despite all the FUD, Arrow Lake has been praised for how consistent the performance is, not to mention very good performance in Linux. The amount of mental gymnastics people go through to justify inferior products is astounding, and I sincerely hope that the amount who fall of this, is about equal to the amount of people who actually game in 720p on a RTX 4090/5090. (Especially those arguing for Zen 3 CPUs right now, which will not be snappy on a desktop by today's standards.)
It's the same thing in reverse from when Zen2 released the 3900X obliterated the 9900k at MT task and was like 5% slower in realistic gaming scenarios I had both so I am fully aware. You still had the Intel brigade saying Zen2 was trash etc.


Now it's the opposite which cracks me up a bit.

That's just fanboys in general if something is 1% slower or not chart topping at a specific workload it's trash when it isn't their preferred brand.
Posted on Reply
#40
TheGuruStud
JustBenchingNo it doesn't. Not even close. This is just false.
Take a look at stock 5800x3D gaming reviews and add 11% or whatever it is. Keep inhaling Pat's bottles.
Posted on Reply
#41
Tyler-98-W68
TheGuruStudTake a look at stock 5800x3D gaming reviews and add 11% or whatever it is. Keep inhaling Pat's bottles.
I mean I game at 720p all the time so obviously i'd choose a processor that is way older, and doesn't do anything else but game semi decent, when I could just buy a new platform and get better all around performance.
Posted on Reply
#42
ToxicTaZ
Arrow Lake Refresh... Usually I would say Refresh models to go to but Arrow Lake/Refresh LGA 1851is not a good platform my opinion. All because Intel 20A RibbonFET was broken. I felt the same way about LGA 1200 platform with broken PCIe 4.0

My upgrade path has been rock solid for 6 years now 9900KS (Coffee Lake Refresh) to my current 14900KS (Raptor Lake Refresh) on my MSI MEG Z790 ACE MAX motherboard with Bios 7D86vA7

My next RIG build will be based upon Intel Razer Lake Refresh (Intel 18A-PT) on second generation LGA 1954 Socket PCIe 5.0 Certified Motherboard.

From Raptor Lake to Razer Lake architecture huge level of performance.

Intel next good second generation nodes 18A-P & 18A-PT, I believe Intel dropped first generation 18A Node for 3rd party customers.

Cheers
Posted on Reply
#43
JustBenching
TheGuruStudTake a look at stock 5800x3D gaming reviews and add 11% or whatever it is. Keep inhaling Pat's bottles.
We can open a thread and test it if you want, no problem. Ill even run my 12900k stock. You said you are matching an oced 14900k - so clearly you must be curbstomping my 12900k. Wanna do it?
Posted on Reply
#44
_roman_
efikkanDespite all the FUD, Arrow Lake has been praised for how consistent the performance is, not to mention very good performance in Linux.
I still do not know what instructions those E-cores can do or not do.

I build my gentoo-sources linux kernels since 2006. Only proof available is my forums.gentoo.org forum user.

There are so many unneeded patches and fixes in the kernel sourches. I highly doubt that quote ... linux performance. For what optimisations? Which cpuflags?

You may check yourself all those changes first. Especially those which dates back 2-5 months ago for e.g. the ~amd64 kernel sources.

All that bloat which is needed because of those E-Cores. still not properly executed. Years after Years.

Gentoo compiles most packages from source code. My ryzen 7600x is enough yet.

I highly doubt INTEL has any support for the AVX512 instruction. Does intel now support those instruction? yes?

Where is AVX512?

I'm not sure if wikipedia is correct?
en.wikipedia.org/wiki/Arrow_Lake_(microprocessor)
->
www.intel.com/content/www/us/en/products/sku/241067/intel-core-ultra-5-processor-245k-24m-cache-up-to-5-20-ghz/specifications.html

check yourself how many instructions intel can not execute in comparison with a cheap entry processor from may 2023. (Hint: avx512_bf16 .... Any avx512.... is missing on intel side)
Sienna_Cichlid /home/roman # grep Ryzen /proc/cpuinfo |head -n 1; grep avx /proc/cpuinfo |head -n 1
model name : AMD Ryzen 5 7600X 6-Core Processor
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc rep_good amd_lbr_v2 nopl xtopology nonstop_tsc cpuid extd_apicid aperfmperf rapl pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_llc mwaitx cpb cat_l3 cdp_l3 hw_pstate ssbd mba perfmon_v2 ibrs ibpb stibp ibrs_enhanced vmmcall fsgsbase bmi1 avx2 smep bmi2 erms invpcid cqm rdt_a avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local avx512_bf16 clzero irperf xsaveerptr rdpru wbnoinvd cppc arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic vgif x2avic v_spec_ctrl vnmi avx512vbmi umip pku ospke avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid overflow_recov succor smca fsrm flush_l1d amd_lbr_pmc_freeze

A few days ago i checked some other cpu flag which intel did not even support and amd also removed later. FMA4 - nope / xop - nope

I state my statement from ages ago. Intel can not execute my code on those E-Cores. Fact. Just because windows 11 pro runs optimised code for i686 does not proof anything

further information of removed instruction
en.wikipedia.org/wiki/FMA_instruction_set
en.wikipedia.org/wiki/XOP_instruction_set

Nope for Intel desktop - as it seems
en.wikipedia.org/wiki/AVX-512

AVX512 is important
Posted on Reply
#45
JustBenching
Panther_Seraphin5800X3D was amazing at time of release and was/is amazing for anyone still on AM4 now and doesnt have a decent chunk of change to do a full MB/RAM/CPU swap. Otherwise going to AM5/Waiting for intels next release would be my advice currently. Advising/advocating for AM4 parts now is silly.
It was never really amazing. Today and on time of release it was always cheaper to just buy a new mobo and a new faster CPU than what the 5800x 3d cost on it's own. Not to mention that besides the gaming performance (which is pretty average for todays standards) the processing power is that of a calculator - even a 170$ i5 13600k just walks all over it. I always thought it's a CPU for people that are too bored to buy a new mobo and setup their PC (which is a valid reason to buy this, don't get me wrong) than it was really a good sensibly valued option.
Posted on Reply
#46
RootinTootinPootin
TheGuruStudIntel STILL can't beat 5800X3D, so a lot of MBs lol. Mine Oced on X370 ties OCed 14900K (not that there's any headroom lul). Absolutely pathetic.
I can smoke your 5800X3D with a 13600K (binned chip) and 8000MT's sticks on my Z790i Strix..before my son enjoyed my 1st 13900KS, he was rocking this chip and he has nothing bad to say, compared to his old Ryzen 9 5900X which was a stutter fest on COD during high motion game play..lol
_roman_AVX512 is important
Nope, on our daily normal lives, nope..AMD just went loolapallooza on this just to gain leverage vs Intel not adding this to their lineup..

I'd really say, you should try them both 1st before you'd throw out silly comments about needing this and that on the chip..I mean, I have them both now, I am just liking the ARL and RPL experience altogether without gaming hiccups..
Posted on Reply
#47
kondamin
RootinTootinPootinNope, on our daily normal lives, nope..AMD just went loolapallooza on this just to gain leverage vs Intel not adding this to their lineup..
it's not important for most, but it is important for those who need it. Intel basically forced people who need it on to their server platform or the competition.
They should have offered P core only raptor lake consumer chips with avx 512 still functional. and not make them run at 50Thz so they explode when plugged in
Posted on Reply
#48
RootinTootinPootin
kondaminit's not important for most, but it is important for those who need it. Intel basically forced people who need it on to their server platform or the competition.
They should have offered P core only raptor lake consumer chips with avx 512 still functional. and not make them run at 50Thz so they explode when plugged in
Give me an actual stat of those who bought who actually "needed" it and pretty much aware what it does. 7-zip users?? :roll:

personally I would get an Epyc on AM5 if I needed a cheap server platform, its cheaper unlike Ryzen..
Posted on Reply
#49
ToxicTaZ
_roman_I still do not know what instructions those E-cores can do or not do.

I build my gentoo-sources linux kernels since 2006. Only proof available is my forums.gentoo.org forum user.

There are so many unneeded patches and fixes in the kernel sourches. I highly doubt that quote ... linux performance. For what optimisations? Which cpuflags?

You may check yourself all those changes first. Especially those which dates back 2-5 months ago for e.g. the ~amd64 kernel sources.

All that bloat which is needed because of those E-Cores. still not properly executed. Years after Years.

Gentoo compiles most packages from source code. My ryzen 7600x is enough yet.

I highly doubt INTEL has any support for the AVX512 instruction. Does intel now support those instruction? yes?

Where is AVX512?

I'm not sure if wikipedia is correct?
en.wikipedia.org/wiki/Arrow_Lake_(microprocessor)
->
www.intel.com/content/www/us/en/products/sku/241067/intel-core-ultra-5-processor-245k-24m-cache-up-to-5-20-ghz/specifications.html

check yourself how many instructions intel can not execute in comparison with a cheap entry processor from may 2023. (Hint: avx512_bf16 .... Any avx512.... is missing on intel side)
Sienna_Cichlid /home/roman # grep Ryzen /proc/cpuinfo |head -n 1; grep avx /proc/cpuinfo |head -n 1
model name : AMD Ryzen 5 7600X 6-Core Processor
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc rep_good amd_lbr_v2 nopl xtopology nonstop_tsc cpuid extd_apicid aperfmperf rapl pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_llc mwaitx cpb cat_l3 cdp_l3 hw_pstate ssbd mba perfmon_v2 ibrs ibpb stibp ibrs_enhanced vmmcall fsgsbase bmi1 avx2 smep bmi2 erms invpcid cqm rdt_a avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local avx512_bf16 clzero irperf xsaveerptr rdpru wbnoinvd cppc arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic vgif x2avic v_spec_ctrl vnmi avx512vbmi umip pku ospke avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid overflow_recov succor smca fsrm flush_l1d amd_lbr_pmc_freeze

A few days ago i checked some other cpu flag which intel did not even support and amd also removed later. FMA4 - nope / xop - nope

I state my statement from ages ago. Intel can not execute my code on those E-Cores. Fact. Just because windows 11 pro runs optimised code for i686 does not proof anything

further information of removed instruction
en.wikipedia.org/wiki/FMA_instruction_set
en.wikipedia.org/wiki/XOP_instruction_set

Nope for Intel desktop - as it seems
en.wikipedia.org/wiki/AVX-512

AVX512 is important
Intel AVX512 is used in Their high-end Intel Xeon 6 processor family.

www.intel.com/content/www/us/en/products/docs/accelerator-engines/what-is-intel-avx-512.html

www.intel.com/content/www/us/en/products/docs/xeon-6-product-brief.html

Not sure if Intel is bring AVX512 to next-gen Intel Nova Lake & Razer Lake LGA 1954 Socket platform.

Intel Razer Lake is what I'm interested in most likely my next RIG build. Second generation 3D bLLC Cache.

Cheers
Posted on Reply
#50
LastDudeALive
agent_x007Can't Intel just figure out that all E-core CPUs for Ultra 3 (and lower) + all P-core CPUs for Ultra 5/7/9, are just fine ?
That wouldn't be good. Intel's biggest market is businesses, and that's the exact opposite of what businesses need. If you want a little micro-PC with an Ultra 3, you want 1or 2 P cores for single-threaded tasks so it feels snappy. And for the bigger workstations, a 4-core E-core cluster provide far more multithreading power than 1 P core that takes up the same space or power, so you actually want to load up on E-cores in the higher SKUs.
Posted on Reply
Add your own comment
Nov 17th, 2026 15:08 CST change timezone

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts