• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.
  • The forums have been upgraded with support for dark mode. By default it will follow the setting on your system/browser. You may override it by scrolling to the end of the page and clicking the gears icon.

E-cores still evolve. But is there a reason for it?

I don't think I said anything contrary to that. But I was thinking of Zen 5. Testing at Phoronix, with Linux and big CPUs and HPC and server applications, demonstrated a solid generational advantage over Zen 4. Benchmarking at TPU and other sites, with Windows and little CPUs and desktop apps and games, resulted in disappointment of most enthusiasts. Most blame Zen 5. I blame the Windows scheduler.
Or those apps arent bottlenecked by the 6000Mhz max memory controller/ IF fabric and the dinky IO die that has been the same since Zen 3.

When you have 8 memory channels and a totally different IO and fabric design, maybe you give the Core what it needs to really stretch it's legs?
 
I don't think I said anything contrary to that. But I was thinking of Zen 5. Testing at Phoronix, with Linux and big CPUs and HPC and server applications, demonstrated a solid generational advantage over Zen 4. Benchmarking at TPU and other sites, with Windows and little CPUs and desktop apps and games, resulted in disappointment of most enthusiasts. Most blame Zen 5. I blame the Windows scheduler.
The same workloads on Linux got small perf gains. It's more a matter of what kind of workloads you're testing on, and not really a windows scheduler issue in this case.
Point in case, the single CCD Zen 5 cpus should not face any scheduler issue.
Or those apps arent bottlenecked by the 6000Mhz max memory controller/ IF fabric and the dinky IO die that has been the same since Zen 3.

When you have 8 memory channels and a totally different IO and fabric design, maybe you give the Core what it needs to really stretch it's legs?
The uplift can be seen in both the Ryzen and Epyc CPUs, so it's not just a matter of memory bandwidth.
 
Or those apps arent bottlenecked by the 6000Mhz max memory controller/ IF fabric and the dinky IO die that has been the same since Zen 3.

When you have 8 memory channels and a totally different IO and fabric design, maybe you give the Core what it needs to really stretch it's legs?
12 channels actually - but they feed as many as 128 big cores or 192 small cores. But at this point, I realise I've gone far off the topic. Super E cores are knocking at the door, and it sounds like tick and tock both at once!
 
Last edited:
Is the uplift on those chips greater than the uplift on the consumer platform for their respective tasks?

Games are specifically bottlenecked on latency and memory bandwith much more so than core performance (3D cache and tight memory timings/overclocking results are a testament to this), I think the zen 5 core performs great on windows -- if you run st cinebench, cpu-z, etc. Application benchmarks that need more raw compute power seem to scale well.

Those that are bottlenecked on other things tend to show 0 benefit, and then there are apps that fall between those two extremes. Seems that leaving the IF fabric and memory controller stagnant is really what killed the gaming benefit on zen 5 on the consumer side -- for the most part anyway. I think the 3D cache will mask that quite well, even on the dual CCD models.

Windoze scheduler may have a role, for sure - but it's the same role it has for Zen 4; the recent tweaks to the kernel boosted performance but didn't change the dynamic between the two generations.
 
Last edited:
Stay on topics please! We NEED to focus on Hyper-Threading!
Yessir. My main point, which is skewed by some that want to make it about off topic instead of the reality....

The HT was cut to accommodate the power usage of the e-cores on the new Ultra chips as well as die space and the difference of running 2 threads per core vs e-cores.

As funny as it is to some, I'll try to be more clear about this each time I reply.
 
I believe it is C0, but wasn't there a later one with AVX-512 fused off? Mine is an original CPU with it still available.
Turns out, The 14100F (that I have) is an H0 stepping. Revision.

So that's Raptor lake strictly, no Alder Lake parts here (correct??)

(Ya, I know there's no e-cores. Off topic a touch too, my apologies, just wanted to recap it real quick cause I said I would)

Oh and yes, AVX 512 has been off since 12th gen. There where some early models like 12400F that does support AVX 512, but only a few batches worth. The one I have, does not for sure, have AVX 512. Yes, I have a 12400F too.... Some call this an addiction....

Screenshot 2024-10-23 164801.png
 
Last edited:
Turns out, The 14100F (that I have) is an H0 stepping. Revision.

So that's Raptor lake strictly, no Alder Lake parts here (correct??)

(Ya, I know there's no e-cores. Off topic a touch too, my apologies, just wanted to recap it real quick cause I said I would)

Oh and yes, AVX 512 has been off since 12th gen. There where some early models like 12400F that does support AVX 512, but only a few batches worth. The one I have, does not for sure, have AVX 512. Yes, I have a 12400F too.... Some call this an addiction....

View attachment 368681
I can turn on AVX512 on my 13700kf if i shut e cores off... not sure if it works properly tho -- any good way to test? Not sure if just a holdover setting from 12th gen.
 

E-cores still evolve. But is there a reason for it?​


You haven't seen anything yet. As the tech fat cows push on core density and balancing power efficiency whilst getting behind AI/DATA/etc (not necessarily data-first, but profit-first), we might end up with one underpowered 0.75 P-core and a 101-minion swarm of e-cores. Its all good under the hood as long as I can continue achieving my performance goals (90-120fps does me nicely) providing pricing doesn't skyrocket to unimaginable levels. Hopefully, the future remains bright for hardware enthusiasts (gamers!!) because the alternative with cloud gaming sucks.

Yeah thats right the world does evolve around us gamers. "We will not go quietly into the night! We will not vanish without a fight!....."

Given that AMD describes itself as a 'data-first company,' does that mean they'll be rolling out C-cores across the entire range of future consumer desktop CPUs? Or, is "c" exclusively confirmed for EPYC with no official confirmation for desktop chips for the time being?
 
E-Cores are great for low power and background tasks. Hyper-Treading is a old tech for P4 days that is a powerhog.
Is modern SMT even comparable with the original from Netburst era?
 
So that's Raptor lake strictly, no Alder Lake parts here (correct??)

View attachment 368681

No. This is an Alder Lake processor subject to Intel's shameless rebranding. You can easily detect an Alder Lake design by checking the P-core associated L3 size. 1.25 MB L3 = Alder Lake (Golden Cove P-core), 2 MB L3 = Raptor Lake (Raptor Cove P-core). Raptor Lake parts are B0 silicon. This is valid for both 13th and 14th Gen CPUs. They share the same stepping and exact same physical characteristics. The rest (C0 and H0) are all Alder Lake derivatives, regardless of whether they are codenamed Raptor Lake or not. It's the first time Intel - or any chipmaker to the best of my knowledge - issues a false codename to a chip.

The entire 14th Gen lineup is a mere marketing designation, they are not newer processors, do not contain newer technology, do not feature any sort of silicon revision, do not have any physical changes from their 13th Gen counterparts (which were a refresh of 12th at best to begin with) and have not underwent any sort of manufacturing-level improvement over its 13th Gen counterparts. They load the exact same microcode and have the exact same specification down to the precise transistor - the only change is the default clock speed table. Nada, zilch.

It's all marketing talk. They're just different SKU models, with the lower-end 14th Gen having no changes whatsoever from the original 12th Gen products from back in 2021. With the exception of the i7-14700K (previously undefined configuration, misdetected by early microcode as a Core i9 with 4 E-core clusters while in reality, the fourth cluster is fused off leading to an unsuccessful initialization attempt) - 14th Gen chips will boot unmodified on 13th Gen-enabled BIOSes, displaying at best funky v/f curve behavior due to the out of date microcode. Intel has truly reached its absolute lowest with this generation.

I can turn on AVX512 on my 13700kf if i shut e cores off... not sure if it works properly tho -- any good way to test? Not sure if just a holdover setting from 12th gen.

It was fused off on Raptor Lake. It is not possible to enable this even though the option in the BIOS exists. Alder Lake CPUs which have the old round Intel logo can enable AVX-512 as long as the E-cores are off. Newer 12th gen chips that have the square intel logo cannot enable AVX-512. No 13th or 14th Gen chip can do it. It's all fused off.

1729728274487.png
 
Last edited:
No. This is an Alder Lake processor subject to Intel's shameless rebranding. You can easily detect an Alder Lake design by checking the P-core associated L3 size. 1.25 MB L3 = Alder Lake (Golden Cove P-core), 2 MB L3 = Raptor Lake (Raptor Cove P-core). Raptor Lake parts are B0 silicon. This is valid for both 13th and 14th Gen CPUs. They share the same stepping and exact same physical characteristics. The rest (C0 and H0) are all Alder Lake derivatives, regardless of whether they are codenamed Raptor Lake or not. It's the first time Intel - or any chipmaker to the best of my knowledge - issues a false codename to a chip.

The entire 14th Gen lineup is a mere marketing designation, they are not newer processors, do not contain newer technology, do not feature any sort of silicon revision, do not have any physical changes from their 13th Gen counterparts (which were a refresh of 12th at best to begin with) and have not underwent any sort of manufacturing-level improvement over its 13th Gen counterparts. They load the exact same microcode and have the exact same specification down to the precise transistor - the only change is the default clock speed table. Nada, zilch.

It's all marketing talk. They're just different SKU models, with the lower-end 14th Gen having no changes whatsoever from the original 12th Gen products from back in 2021. With the exception of the i7-14700K (previously undefined configuration, misdetected by early microcode as a Core i9 with 4 E-core clusters while in reality, the fourth cluster is fused off leading to an unsuccessful initialization attempt) - 14th Gen chips will boot unmodified on 13th Gen-enabled BIOSes, displaying at best funky v/f curve behavior due to the out of date microcode. Intel has truly reached its absolute lowest with this generation.



It was fused off on Raptor Lake. It is not possible to enable this even though the option in the BIOS exists. Alder Lake CPUs which have the old round Intel logo can enable AVX-512 as long as the E-cores are off. Newer 12th gen chips that have the square intel logo cannot enable AVX-512. No 13th or 14th Gen chip can do it. It's all fused off.
The stepping/revision makes it a Raptor Lake despite the cache sizes. Because it has much better voltage to clock curves.
But you're right either way, if it's a refresh, why bother naming it differently.

Well, what happened with my 14th Gen on 13th Gen bios B660, Z690 and Z790 chipsets, yes happened on all the boards I tried.... Posting at 1.7v, yes V-core. Wasn't a funky V-curve, it was straight Fkn hot. Posted windows cause I didn't believe it and was idle 85c and probably throttled all the way through post. I am currently running the release micro code of Raptor Lake on TUF Z790. Excellent performance. Way better than the latter micro codes I installed on previous MSI 790 to be honest. I'll never update it from here. W10 on a fresh install is cool though. Treats all the E-cores like P-cores, and sometimes comes with interesting and undesired results lol.
 
The stepping/revision makes it a Raptor Lake despite the cache sizes. Because it has much better voltage to clock curves.
But you're right either way, if it's a refresh, why bother naming it differently.

Well, what happened with my 14th Gen on 13th Gen bios B660, Z690 and Z790 chipsets, yes happened on all the boards I tried.... Posting at 1.7v, yes V-core. Wasn't a funky V-curve, it was straight Fkn hot. Posted windows cause I didn't believe it and was idle 85c and probably throttled all the way through post. I am currently running the release micro code of Raptor Lake on TUF Z790. Excellent performance. Way better than the latter micro codes I installed on previous MSI 790 to be honest. I'll never update it from here. W10 on a fresh install is cool though. Treats all the E-cores like P-cores, and sometimes comes with interesting and undesired results lol.

All C0/H0 chips are Alder Lake chips, Intel just named them Raptor because they felt like it. It's sad but it's true. You'll find that the 12th Gen limitations like E-core clocks being bound to ring domain also apply on these "Raptor Lake" 14th Gen low-end chips, specifically because they're just old Alder Lake chips in disguise.

The 14100F is actually one of the funniest chips, it's 1:1 identical to the 13100F and 12100F, but you can usually buy two 12100F's for the price of one 14100F; each "generation" here added 100 MHz. I kid you not.
 
Low power cores are pretty weak in general, same with low power cores for phones as well. See Ian's latest video - it's mainly thrown in for cost reasons but there are cases where they're useful in low power devices.
Is modern SMT even comparable with the original from Netburst era?
It's not and it entirely depends on the arch. Some people think of HT as the same as it was when P4 launched which is frankly strange.

If an arch is efficiently designed with HT in mind, it's going to be more efficient in MT with SMT on while having negligible performance hit in ST. See this. AMD seems keen on optimizing HT even further and Zen 6 looks to continue that path. I see why they've gone the route of dual execution engines etc.

Now it's entirely possible to design an arch without HT, but there's no obvious right or wrong when it comes to HT vs no HT.
 
At some point, E-cores will become invisible to the operating system. the hardware circuit does the scheduling.
Source? If that's true then the only issue I have with E-cores is P2E latency.
 
E-cores are evolving because they are the basis of the next generation architecture when Intel goes to a unified core arch after Nova Lake. It's the P-Cores that a dead end.
 
It's the P-Cores that a dead end.
That's definitely a statement.

Would've loved to read the thought process behind that, preferrably sans baseless speculations.
 
The 14100F is actually one of the funniest chips, it's 1:1 identical to the 13100F and 12100F, but you can usually buy two 12100F's for the price of one 14100F; each "generation" here added 100 MHz. I kid you not.
Lies and slander. I can’t believe you would do Intel dirty like that. They also have increased the maximum power budget on the PL2. By whole 21 watts. Totally different chip, pls donut steal.

That's definitely a statement.

Would've loved to read the thought process behind that, preferrably sans baseless speculations.
I mean, I guess the logic there would be based on the rumors that the E-core team feels like they made massive strides in relatively short time and are annoyed at the P-core team for basically stagnating and not really delivering anything significant in a while. Though that WOULD fall under speculation.
 
E-core team feels
And I feel like weather on Mars reflects the value of USD a little bit too dishonestly.
I want facts, not feelings.
 
A future where P cores and E cores are unified under real-time scalable core architecture is more possible than what people might think.
 
And I feel like weather on Mars reflects the value of USD a little bit too dishonestly.
I want facts, not feelings.
There ARE no facts that one can use to prove or disprove the future evolution of CPU architectures unless one works at Intel or AMD or some leaks come out, which are still not factual until confirmed by vendors themselves.
You’ve been given an answer already. E-cores are a die-space efficient way to increase MT performance. That’s the path Intel chose. There are already E-core only Xeons because of that. It’s not a grand leap in logic to suggest that in the future, if E-cores are strong enough, Intel might abandon the entire P-core idea altogether. Or they may not. It’s all speculation. But asking “why are they developing E-cores” is the same as asking “why AMD develops ZenC” - it makes sense for their primary focus. Which is NOT, as some still seem to think, desktop DIY enthusiasts seething about low increases in gaming performance.
 

Intel's Arrow Lake chips aren't winning any awards for gaming performance but I think its new E-cores deserve a gold star

source: https://www.pcgamer.com/hardware/pr...-i-think-its-new-e-cores-deserve-a-gold-star/


P.S. Paradigm shift: 12c .LITTLE Atom Skymont E-cores beats 8C big Lion Cove P-cores !
source:
 
Last edited:

Intel's Arrow Lake chips aren't winning any awards for gaming performance but I think its new E-cores deserve a gold star

source: https://www.pcgamer.com/hardware/pr...-i-think-its-new-e-cores-deserve-a-gold-star/


P.S. Paradigm shift: 12c .LITTLE Atom Skymont E-cores beats 8C big Lion Cove P-cores !
source:

Nice but not gonna convince the people calling them dumb names, who don't understand the point of them.
 
Nice but not gonna convince the people calling them dumb names, who don't understand the point of them.
EH cores? E for Effort Cores? Effortless cores?
Maybe nicer names instead!
16 Elegant cores!!
Enchanting Cores!!
Euphoric Cores!! (My personal Favorite)
:lovetpu:
 
Back
Top