Wednesday, October 22nd 2025
Intel "Nova Lake" Could Arrive Without AVX10, APX, and AMX Support
Intel is still finalizing its next-generation "Nova Lake" specifications. However, the core microarchitecture has been done, the NPU design is finished and upgraded, and the instruction set is finalized. It could be the case that these consumer CPUs will again lack support for Intel's AVX10, APX, and AMX instruction set—designed to encompass 512-bit acceleration and fast vector/matrix multiplication for tasks like content creation, encoding/decoding, AI, and much more—remaining exclusive to Intel Xeon processors. According to the latest GCC compiler patch, the initial Nova Lake enablement patch does not include AVX10, AMX, or APX. This might suggest that support for these new x86 instructions could be missing from the next-generation CPU. Initially, Intel decided to disable AVX-512 for "Alder Lake" and "Raptor Lake" client-oriented CPU lineups, meaning that only server-grade Xeon CPUs got the accelerated 512-bit data paths.
This GCC patch seems to contradict the findings from August, when Intel introduced AVX10.2 support for "future Intel Core processors" in the oneDNN software library, likely targeting the Nova Lake series. We need to wait for more information from other enablement patches or direct confirmation from Intel regarding whether AVX10, APX, and AMX will be included in the 52-core Nova Lake SKUs. Enabling fast vector and matrix acceleration on 52 cores would be ideal for consumers, content creators, gamers, and workstation users alike. In comparison, AMD introduced full AVX-512 support with its "Zen 5" cores across its product range, providing a performance boost in optimized applications for both desktop and server CPUs. This marked the first time AMD did not emulate 512-bit AVX, which previously required splitting 512-bit data into two 256-bit units to be processed over two cycles.The latest patch states the following instructions are now supported:
Source:
Phoronix
This GCC patch seems to contradict the findings from August, when Intel introduced AVX10.2 support for "future Intel Core processors" in the oneDNN software library, likely targeting the Nova Lake series. We need to wait for more information from other enablement patches or direct confirmation from Intel regarding whether AVX10, APX, and AMX will be included in the 52-core Nova Lake SKUs. Enabling fast vector and matrix acceleration on 52 cores would be ideal for consumers, content creators, gamers, and workstation users alike. In comparison, AMD introduced full AVX-512 support with its "Zen 5" cores across its product range, providing a performance boost in optimized applications for both desktop and server CPUs. This marked the first time AMD did not emulate 512-bit AVX, which previously required splitting 512-bit data into two 256-bit units to be processed over two cycles.The latest patch states the following instructions are now supported:
Intel Nova Lake CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, AES, PREFETCHW, PCLMUL, RDRND, XSAVE, XSAVEC, XSAVES, XSAVEOPT, FSGSBASE, PTWRITE, RDPID, SGX, GFNI-SSE, CLWB, MOVDIRI, MOVDIR64B, WAITPKG, ADCX, AVX, AVX2, BMI, BMI2, F16C, FMA, LZCNT, PCONFIG, PKU, VAES, VPCLMULQDQ, SERIALIZE, HRESET, AVX-VNNI, UINTR, AVXIFMA, AVXVNNIINT8, AVXNECONVERT, CMPCCXADD, AVXVNNIINT16, SHA512, SM3, SM4 and PREFETCHI instruction set support.

61 Comments on Intel "Nova Lake" Could Arrive Without AVX10, APX, and AMX Support
But I agree that most consumers will not notice. Those running professional applications that need AVX512 will simply buy the Xeons that come in the servers they use to accelerate said tasks, the pool that NEED AVX 512, but have to run consumer desktop CPUs, is incredibly tiny.x3d is definitely better for gaming. In productivity though, Arrow lake is very competitive with the 9950x, and depending on your production suite Intel poses an advantage.
Given how much performance can be unlocked via fabric tuning, I think intel has the chance to make significant inroads with panther lake, and nova lake's core count looks absolutely ludicrous. 52 cores on a consumer chip, with 16 of those being P cores....
2026 is gonna be fun.
Ryzen are bound by only 2 memory channels.
Threadripper and Xeon are power hogs.
Intel's "various" lakes are instruction set limited.
and in the audio world, it seems almost no one takes advantage of gpu processing. (but this is a software issue)
I'll continue to propose a synthetic test that seems plausible, but since the result isn't pleasing, it must be bad:
www.techpowerup.com/forums/attachments/01-jpg.419336/
The performance ratio of the 10900K vs. 11900K is indicative, as is the predictable superior performance of the Ryzen 9000s.
It's ridiculous to consider the results of a game on RPCS3 interesting, where there's a 2fps difference across 10 processor models, and you can't even see the point in that little bit (for example, the 9700X over the 9950X, or the 7000s flying by).As it is clear that that case is not at all significant of the potential of AVX-512.
But still if the instructions are basically bench/rpcs3/professional workloads only. I think intel chips have bigger issues.
Intel apparently has two or three processor development teams that implement various instruction sets using different hardware techniques. Those independent designs then get stuck into the same package with a hope that software can sort things out. Maybe it even will.
Within the same power limits.
So even in those conditions the advantage over Intel is significant.
Why would you need to remove the power limits?
To get even better improvements? But if the thermals and power draw deteriorate greatly it's not worth just so that they leave the competition WAY behind.
It would be sort of like adopting the Raptor Lake approach, where Intel was juicing the living daylights out of the CPUs, in that case so that they could match not surpass the competition.Honestly I disagree, the truth is that when trying to surpass the same obstacles there has to be convergent evolution, just like in nature.
Of course not knowing beforehand what is the ideal method to surpass the obstacle leads to a design choice which might seem competent but only until you compare to the competition's approach which might turn out better.
Also, the multitude of obstacles presented by today's software probably makes optimizing for all of them difficult, in the sense that you can't have something that gets top performance in every workload.
Again giving examples from nature you can't have an animal that has all of the attributes maxed out, speed, agility, endurance/stamina, power and intelligence. There's always some deficiency present.
Maybe it's the same here, where optimizing for the tasks most often encountered is of course the wisest approach, having some tricks available for other rarer tasks is also welcome.
Not being completely handicapped in some workloads also helps by not skewing the geomean downwards. It's all a balancing act.On top of benchmark results which isolate the CPU, in real world use Intel has an advantage due to the iGPU which does contribute to video editing workloads. And now perhaps that NPU also helps with some low level stuff, which might not be negligible in some scenarios.
Specs verified: NO AVX10 for Intel Nova Lake, but you will get a new NPU
10/22/2025 10pm Europe time
reference: www.computerbase.de/news/prozessoren/dokumente-bestaetigen-kein-avx10-fuer-intel-nova-lake-dafuer-eine-neue-npu.94767/
I would be surprised if Computerbase.de is wrong on that.
Imagine taking the upper hand again against AMD. Of course no one should be surprised. Intel is selling "10 core CPUs" for example, with only 2 cores being P cores for years and their supporters are defending this marketing.
Fusing off broken silicon is called "binning" and it's been a thing for decades. If one of those segments is hot, they will, yes, intentionally make more parts for that segment. Somehow you've twisted that into a major negative point.Those E cores are now on par with zen 3 in performance. Those E cores also make intel competitive in production apps and they easily smack AMD in battery benchmarks for mobile parts. Lunar Lake would never be successful without efficiency cores.
Do you refer to ryzen 5000 parts as "e" cores now? Or is that only reserved for whining about Intel?Sandy bridge doesnt support Windows 11 because of Microsoft's garbage artificial product segmentation. If the game doesnt need more than 4 powerful threads, then sure, why not run it?
Those games support coffee lake i3s, which have comparable performance.
At least you are very enthusiastic about Zen 3 performance, 2020 performance that is selling and priced as brand new, 2025, performance.
I am not the one whining here. You are whining, because I have the audacity of describing things the way they are. Three types of cores, not equal cores. You on the other hand, you are hiding behind a couple of benchmarks that are optimized by Intel.
And of course we are talking about desktop CPUs. Do people who want fast Intel CPUs really want Low Powered Efficiency cores like those on Lunar Lake, or would they want more Performance cores? Feeling the need to talk about Lunar Lake's LPE cores, in a thread about desktop CPUs, I don't think it is really an argument in your favor.
Any way, I am not a fun of supporting misleading marketing. Sorry.
The other more traditional benchies like PhotoWorxx fare much better, this same system scores north of 76,000 MPixel/s. Another AIDA benchmark where you'll see Zen 5 leaps and bounds ahead of the Intel systems is CPU AES performance. The reference is crazy, above 710,000 MB/s - my best score so far was at around the 300,000 mark.
Zen 5 is an excellent core architecture, especially for SIMD heavy, vectorized operations. It's also the first processor architecture that does AVX-512 without elevating the power consumption by a factor of 10 practically instantly, even though it does increase the power draw rather significantly all the same. Intel hasn't enabled it in their client segment processors for that reason, the amount of current these 10 nm processors would draw would probably ignite the processor at the clock speed targets they were after.
There are still many strong points that make Arrow Lake valid, and even Raptor Lake to this day (especially considering the Golden/Raptor Cove architecture is nearly as old as Zen 3), but it seems that Linus Torvalds was wrong about AVX-512 after all.
www.techpowerup.com/269770/linus-torvalds-finds-avx-512-an-intel-gimmick-to-invent-and-win-at-benchmarks
While those aren't relevant for most games or any heavy application, their existence and market share is still a hindrance for the OS kernels, drivers, system libraries and all sorts of basic applications to move require higher ISA levels (unless they want to compile multiple versions), which in turn causes all of us to miss out on a lot of "free" performance. Just imagine if everything was compiled for x86-64-v3 or higher.That's not a fair assessment.
The real world performance of E-cores will vary a lot depending on the use case, as four of those share L2 cache.The latest versions wouldn't even install on Coffee Lake systems, but installing from an older installation image will work fine (you wouldn't get feature updates though…).
But I'm much more anticipating APX, as it should yield substantial gains across numerous workloads, and it just needs recompilation. I do wonder though, if there is some kind of screw-up if it's disabled in Nova Lake.
But I think anyone knows the whole mess of 10nm and mess of management by Intel of updating everything but the Core and GPU architectures in the 14nm node for consumer eesktop products until it was too late with Rocket Lake (10nm was already viable in mobile with Tiger Lake and soon will be ready after only some months in desktop with Alder Lake).