Saturday, November 21st 2026

28-Core Intel "Nova Lake-S" Appears in Testing Running at 5.2 GHz

Intel's single-compute die "Nova Lake-S" variant has been spotted with a 28-core configuration and a boost frequency of 5.2 GHz, according to the OpenBenchmark testing suite. During the tests, a processor named Intel 0000 appeared with the "NVL" identifier, suggesting that a "Nova Lake-S" SKU with this configuration is indeed present. The processor's identification as Intel 0000 implies that the rumored four-number naming scheme is real, as confirmed by the benchmarks. This naming scheme could easily become standard, moving away from the three-digit SKU naming we've seen in the past.

Interestingly, the submission also identifies the integrated graphics as Intel NVL-S, indicating that the engineering platform is being tested beyond just CPU workloads. The tests include various graphics and compute applications, such as Quake II RTX, Unigine benchmarks, Hashcat, and multiple OpenCL workloads. While the 5.2 GHz figure is promising, it should not be taken as confirmation of the processor's final boost frequency. Engineering samples can operate with different firmware, power limits, and clock configurations than retail products. The listing also does not disclose whether the reported frequency represents a sustained operating clock or a maximum observed value.
The system information lists 28 CPU cores, likely indicating a configuration with 8 P-Cores, 16 E-Cores, and 4 LPE-Cores, along with a reported operating frequency of 5.20 GHz. The motherboard is identified as an Intel "Nova Lake-S" reference platform, specifically an NVL-S UDIMM 1DPC CRB Fab2 configuration running an early BIOS build. The system was tested with 32 GB of memory under Ubuntu 26.04, suggesting that it could be Intel's internal engineering teams testing the system, or one of their OEM PC partners.
Source: OpenBenchmark.org
Add your own comment

46 Comments on 28-Core Intel "Nova Lake-S" Appears in Testing Running at 5.2 GHz

#26
demirael
tsurugi28 cores 1C1T means 28 threads. Zen6 also has 24 threads, the difference is that e core performance is much better than SMT.
What workload do you run that need 24 threads?
Zen 6 has 24 threads per CCD, and you can bet on there being multi-CCD SKUs. I have some loads that can benefit from as many threads as possible (video transcoding at crazy settings, SCSKiller for precalculating shaders) and it's easy to come up with loads if you have the capability. Hell, throw BOINC projects at it. On desktop you're kinda memory bandwidth limited in some cases though (like AI on CPU). Quad channel is the one thing I miss from X99.
Posted on Reply
#27
Makaveli
NathaaabMaybe you don't use cores, others do. NL high end SKUs will practically be HEDT, which is great.
But that isn't what I asked I said what workloads?

So I am expecting a response with the actual workloads?

Example I render

That response it not an actual answer to the question.
tsurugi28 cores 1C1T means 28 threads. Zen6 also has 24 threads, the difference is that e core performance is much better than SMT.
What workload do you run that need 24 threads?
I will be looking at the 12core Zen 6 X3D chip when its released and I want it for the additional 4 cores over my current 9800X3D.

Which will help with alot faster software compiles, Handbrake will also benefit as I use that alot for videos I create when I need to reduce the size, will also allow for better multitasking when I have the processor at full load.

Also want the additional instruct sets AMD is adding for AI workloads on Zen 6.
Posted on Reply
#28
jesdals
Well Intel or AMD - I just want to see some real products not hints of proberly performance - hopeing for some new stuff soon ;)
Posted on Reply
#29
Lord Romulus
NathaaabThe E cores are getting so big and fast they will replace P cores. Having both P cores and E cores was great for peak perf on the P cores and good multicore on the E cores.
I believe it would be smarter for the semiconductor industry to adopt this approach: P-cores for desktop CPUs, and E-cores for laptop CPUs—no mixing. My reasoning is as follows: traditional, legacy computing—represented by the PC—is built on a hardware architecture that is always plugged into a wall outlet. Mobile computing, by contrast, is a more recent development; it relies on batteries, which makes energy efficiency a top priority. Consequently—at least in the context of modern systems like Android—mobile computing has had the time to adapt and make better, more efficient use of heterogeneous cores. Traditional computing, meanwhile, has matured over decades using homogeneous cores; it even took some time to fully embrace multi-core architectures (consider the advent of the Cell processor, which divided the software development community at the time, though the concept is well-integrated today). This is just my personal opinion: if I were a chip manufacturer, I would design x86-based CPUs with either E-cores or P-cores, tailored to specific product segments—reserving the mix of both for ultra-portable devices.
Posted on Reply
#30
tsurugi
Darmok N Jaladone CPU type
OS treats CPU designs like SMT and dual CCD no differently from p/e core. These designs all provide non-equivalent thread types, while boosting CPU parallelism. The OS can identify and schedule these hybrid threads.
Darmok N JaladThe single 12C CCD Zen6 should benefit all workloads because there will be 48MB of L3 versus 32MB.


I don't really think so. According to tests, zen6 latency in 1-32MB increased by 17% compared to zen5. Overall though, adding more cache is a good thing.
MakaveliI will be looking at the 12core Zen 6 X3D chip when its released and I want it for the additional 4 cores over my current 9800X3D.
True. You can never have too much performance since demand is always there, and in theory nobody complains about a faster PC.
What I don't get is why people still complain about e core when intel gives 28 physical cores that seriously boost CPU performance, backed up by real-world tests showing clearly faster processing.
Posted on Reply
#31
Onasi
@Lord Romulus
Why? I mean, yes, you outlined your reasoning, but what’s the point for Intel to basically nerf their desktop CPUs for… no reason? The amount of P-cores that will take up the space occupied currently by E-cores will absolutely not be enough to overcome the resulting MT performance deficit. You are essentially asking them to make worse CPUs just because. The software and OSs have caught up to the heterogeneous approach for the most part and will only improve with time. The approach WORKS. Denying it at this point is just being contrarian. Whatever issues Arrow Lake had at launch weren’t due to its hybrid nature. The refresh fixed what was left. The 270K is basically a 9950X at 2/3rds of the price. It’s a good CPU. Taking away its MT chops just to gain two more “muh real” cores is not a good idea. End of story.
Posted on Reply
#32
MufasaV
tsurugiOS treats CPU designs like SMT and dual CCD no differently from p/e core. These designs all provide non-equivalent thread types, while boosting CPU parallelism. The OS can identify and schedule these hybrid threads.



I don't really think so. According to tests, zen6 latency in 1-32MB increased by 17% compared to zen5. Overall though, adding more cache is a good thing.

True. You can never have too much performance since demand is always there, and in theory nobody complains about a faster PC.
What I don't get is why people still complain about e core when intel gives 28 physical cores that seriously boost CPU performance, backed up by real-world tests showing clearly faster processing.
Maybe their reference point is still Gracemont, while Skymont has 30-60% IPC gain depending on workload over that
Posted on Reply
#33
Makaveli
tsurugiTrue. You can never have too much performance since demand is always there, and in theory nobody complains about a faster PC.
What I don't get is why people still complain about e core when intel gives 28 physical cores that seriously boost CPU performance, backed up by real-world tests showing clearly faster processing.
Alot of people still hate on ecores because they still think of them in 12th gen intel.

They have improved quite abit since their first offering.

However I think if intels released a P core only chip with like 12-16 cores with BLLC that will be the big winner for most people.
Posted on Reply
#34
_roman_
Maybe intel loves dated hardware and software ore the claims are a hoax

Those claims are valid when tehy use a monitor from year 2017? Old Kernel, dated file system, etc.

check the line "gelistet seit" geizhals.at/hp-elitedisplay-e243i-1fh49aa-a1664713.html
last listed price was in year 2022 for that monitor

support.hp.com/gb-en/product/product-specs/hp-elitedisplay-e243i-24-inch-monitor/model/17210319

the gcc points to july 2025. I do install gcc later than the 16 branch.

Sat Aug 9 12:29:52 2025 >>> sys-devel/gcc-15.2.0

the storage in question which intel uses wiht so much money is only pice 3.0
www.techpowerup.com/ssd-specs/kingston-a2000-250-gb.d282

Why not spoof all teh operating system parameters? That is not that hard! The usual software look only in a few locations.
Posted on Reply
#35
AVATARAT
NathaaabA 14900KS with good RAM and tweaks still beats a 9850X3D in many games, so you must be talking about something else than gaming.
So, Intel need 24/32 CPU at 6.2GHz to beat 8/16 AMD CPU - this sound so bad...
Posted on Reply
#36
JohH
Measuring E core IPC can be slightly misleading because they share L2.
When measured in 1T, it is fine. If there is contention it falls off more compared to cores with private L2. So even if Coyote Cove ends up with very similar IPC to Arctic Wolf it will likely be higher real world performance even at same clock rate.
Posted on Reply
#37
Razor512
Intel needs to make a version that is all P cores and have all cores boost to 5.2GHz at the same time. If needed, give it a 600+ watt TDP, since that level of power has become widely accepted in the desktop PC market.
Posted on Reply
#38
MoistOintment
Lord RomulusI believe it would be smarter for the semiconductor industry to adopt this approach: P-cores for desktop CPUs, and E-cores for laptop CPUs—no mixing. My reasoning is as follows: traditional, legacy computing—represented by the PC—is built on a hardware architecture that is always plugged into a wall outlet. Mobile computing, by contrast, is a more recent development; it relies on batteries, which makes energy efficiency a top priority. Consequently—at least in the context of modern systems like Android—mobile computing has had the time to adapt and make better, more efficient use of heterogeneous cores. Traditional computing, meanwhile, has matured over decades using homogeneous cores; it even took some time to fully embrace multi-core architectures (consider the advent of the Cell processor, which divided the software development community at the time, though the concept is well-integrated today). This is just my personal opinion: if I were a chip manufacturer, I would design x86-based CPUs with either E-cores or P-cores, tailored to specific product segments—reserving the mix of both for ultra-portable devices.
Everybody by this point is aware of PPA.

All CPU core designs are balancing Power, Performance, and Area. Pretty much every CPU vendor has realized that the best solution for client is to have 3 different cores that are all optimized for one of these 3 attributes instead of making one core that tries to do it all.

Why are E cores bad for desktop? You think 4 P cores would be a better use of Intel's die space than 16 E cores? For what purpose? 8P+16E is already losing. 12P would get slaughtered.
Razor512Intel needs to make a version that is all P cores and have all cores boost to 5.2GHz at the same time. If needed, give it a 600+ watt TDP, since that level of power has become widely accepted in the desktop PC market.
Intel's E cores are pretty much the only good design they have left.

You want to get rid of them to add more of their garbage tier P cores? Intel's P cores are massive cores designed to sacrifice efficiency and area to have amazing performance....and then don't have amazing performance.
Posted on Reply
#39
Frizz
With bLLC then it will all boil down to efficiency, smoothness and pricing. If only AMD could compete at NVIDIA's level for gaming too T.T
Posted on Reply
#40
MufasaV
JohHMeasuring E core IPC can be slightly misleading because they share L2.
When measured in 1T, it is fine. If there is contention it falls off more compared to cores with private L2. So even if Coyote Cove ends up with very similar IPC to Arctic Wolf it will likely be higher real world performance even at same clock rate.
It's not misleading when measuring actual workloads as long as you isolate the E cores from the P cores. You're not extrapolating potential performance based on TFLOPS or something. You are seeing how long it took to do a 7Zip compression work or Javascript compilation or whatever
Posted on Reply
#41
DSTBench
It is still all the same.
14900Ks a bit less cores but turbo at 6.2Ghz
Intel 285 a bit more cores, but single Turbo at 5.6 Ghz
Now a bit (true P ) more cores but at max of 5.1 Ghz

its all the same wheel rooling, no CPU is faster than the one made more than 12 years ago, as for general purpose

And thats cause an Intel Pentium 4 with pure cooper, single core, two threads runs at 5 to 6 ghz. I did it. Now its only a matter of cores, with O.S. putting more processes that are ususally not necessary)

With the exception of youcrude and some video codecs (interframe) , the majority of apps runs as sequential, step by step, like the intra frame prores codec, so.. that no advancements for consumer..
Posted on Reply
#42
Zodwraith
phintsZen 6 is about to have 12 P-cores on one CCX so good luck Intel.
Doesn't matter. Apps that benefit from low latency core speed like games still rarely use more than 4 cores. Take a 9600x and raise the clocks to match the 9700x and it runs neck and neck in gaming. The 9700x only gets the edge from more cache.

Once you get into parallel compute workloads like comp/decomp and compiles they want cores over inter-die latency. This is where Intel's e cores come in and curb stomp AMD.

When it comes down to core to core Arrow Lake already meets or beats Zen5. Disable the e cores and a 270k easily still matches and beats that 9700x. Re-enable those e cores and Intel curb stomps it in multicore. AMD's only lead is thanks to X3D's cache for games.

Intel's fear is if AMD really can crack 6ghz and stay there. They both sit comfortably at 5.4-5.5 right now, but Intel can't hold 6ghz yet. AMD aren't hitting 7ghz so stop listening to Morons Law Is Dead. Early leaks still show it capped around 5.7 under load.

AMD's fear is Intel increasing core count again with e cores on a SINGLE DIE. That single die 7 will most likely stomp the 9950x in multicore. AMD HAS to increase core count per CCD so it's competitive with the 7, let alone the dual die 52core 9. They can't even rely on X3D for the win with gamers and enthusiasts anymore. X3D adds 96mb of 3rd layer cache. Intel's LLC is going to add 144mb of 3rd layer cache. Then when you look at architecture Intel's 18a is showing to be psychotically efficient with Panther Lake. TSMC's 2nm can clock high, but it eats power to hit those speeds. Probably why engineering samples have been in the mid 5ghz range.

AMD has everything to lose right now. Intel is about to cancel their X3D gamer card. They're also about to stomp all over the 16 core and most likely any 24 core in multi. AMD doesn't have a fat OEM following like Intel does. If Nova Lake is even half as good as I'm expecting and ALL the top chips have gobs of cache without a super special X3D moniker I worry about AMD in the long run.
Posted on Reply
#43
kazuviking
Darmok N JaladThe single 12C CCD Zen6 should benefit all workloads because there will be 48MB of L3 versus 32MB
Other than games and 7 zip nothing scales with cache.
Posted on Reply
#44
Vayra86
NathaaabMaybe you don't use cores, others do. NL high end SKUs will practically be HEDT, which is great.
Right, that magical nonexistant space of a consumer mainstream desktop system that is still handicapped by 2ch RAM and still has a small city worth of cores to.... epeen with, because nine out of ten workloads are going to be heavily constrained by throughput anyway.

GL with that awesome money saving you do there lol. There is a point where moar is just comical, and we're in that space for a while now with these Intel CPUs.
kazuvikingOther than games and 7 zip nothing scales with cache.
Those also happen to be workloads that are either in realtime or mandatory to continue further work so its pretty neat that those scale on cache.
A rendering workload that isn't realtime, can last a few minutes longer just fine, as do most other consumer related workloads. The only time critical ones are the realtime ones.
Posted on Reply
#46
Vayra86
MoistOintmentIntel's E cores are pretty much the only good design they have left.

You want to get rid of them to add more of their garbage tier P cores? Intel's P cores are massive cores designed to sacrifice efficiency and area to have amazing performance....and then don't have amazing performance.
Intel is trying to keep the release train rolling while improving their cores, but realistically, they haven't changed a thing and E cores are getting larger now too, while carrying a few disadvantages... it doesn't really look like they have something here for the long term, though 'it works for now', at least.
Posted on Reply
Add your own comment
Nov 22nd, 2026 04:39 PST change timezone

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts