Intel Core i9-12900K E-Cores Only Performance

[W1zzard]

With Alder Lake, Intel is betting big on hybrid CPU core configurations. The Core i9-12900K has eight P(erformance) cores and eight E(fficient) cores. We were curious and tested the processor running the E-Cores only to see how well they perform against architectures like Zen 2, Zen 3, Skylake, Comet Lake, and Rocket Lake.

Show full review

 

[pokazene_maslo]

I'm interested if it's economical to mine some cryptocurrency (maybe Monero?) on E-cores. Maybe someone in discussion can share his findings.
Thanks for great review W1zzard!

 

[bug]

Within 5% of the full enchilada in 4k gaming... Just wow!

 

[ghazi]

Nothing but respect for W1zzard but I have to strongly disagree with a part of the conclusion.

"SuperPi is arguably one of the best tests for IPC in a math-intensive environment that's not too encumbered with streaming data. Here we see the single "Gracemont" E-core achieve a score on-par with the Ryzen 3 3300X. What this means is that "Gracemont" has an IPC roughly comparable to AMD's "Zen 2" in math-intensive, low-bandwidth workloads."

SuperPi evidently is still bogged down by legacy x87 instructions from the 1980s. You can see this clearly in that the 5600X in your own data is almost twice as fast as the 3600X -- on Zen 2 and older, AMD had a half-assed (really, quarter-assed) x87 implementation designed only to support it for legacy applications. The execution resources were not performant and were not meant to be as the world moved on to SSE and later AVX. Only with Zen 3 did AMD decide to change this, as programs heavy on x87 continued to be a pain point in benchmarks and for some enterprise use cases relying on old software. And again, the results speak for themselves.

Meaning, by the same token, you could say Gracemont's IPC is barely more than half of Zen 3. So to suggest that these E-cores have the same IPC as Zen 2 on the basis of a 40 year old instruction set extension that Zen 2 only tacked on for legacy compatibility is not fair at all, it's approaching a worst case scenario for Zen 2 for reasons that have nothing to do with being math-intensive and low-bandwidth.

Appreciate the review nonetheless, it's good to see the P-cores and E-cores separated for analysis.

 

[AteXIleR]

Only P-Cores on without HT with a bit of OC could be a way to go for gaming.

 

[Wirko]

If the P-cores were clocked down to ~2.5 GHz to achieve (more or less) performance parity with the E-cores at 3.9 GHz, what would their power consumption look like?

(I think I can guess @W1zzard 's answer: Stop giving me ideas!)

 

[Prima.Vera]

Thank you for the article, however I have to question the results for the i7-6700K CPU.
I mean....what's going on there, why so ridiculously crappy performance from this i7?
Surelly the CPU is old, but not that old for an i7.
Did you use a completely different system to test th 6700K one? With cheap RAMs?
The CPU is underperforming big time.

 

[MikeSnow]

Looking at the last chart, Cinebench energy usage, the "energy efficient" cores are in practice less efficient than the P-Cores, and less efficient than all the Zen 3 CPUs, when you actually need to get some work done.

And when you enable both the P-Cores and the E-Cores, the energy usage drops by just 3% compared to using just the P-Cores. Is this worth the hassle of having to deal with all the quirks of a hybrid architecture?

Even in the single threaded SuperPI energy test, the E-Cores were barely more efficient than the 5800X cores.

While the big/little architecture had some success in the mobile device market, I don't think Intel's implementation is worth the hassle for the time being.

That being said, if I was looking to build a new computer from scratch, I would consider Intel, since the P-Cores give the best gaming performance at the moment. But, even if I bought an Intel CPU, I would probably disable all the E-Cores, to make sure the thread scheduler doesn't mess up.

 

[bug]

Looking at the last chart, Cinebench energy usage, the "energy efficent" cores are in practice less efficient than the P-Cores, and less efficient than all the Zen 3 CPUs, when you actually need to get some work done.

And when enable both the P-Cores and the E-Cores, the overall power usage drops by just 3% compared to using just the P-Cores. Is this worth the hassle of having to deal with all the quirks of a hybrid architecture?

Even in the single threaded SuperPI energy test, the E-Cores were barely more efficient than the 5800X cores.

While the big/little architecture had some success in the mobile device market, I don't think Intel's implementation is worth the hassle for the time being.

That being said, if I was in a the market for a new computer, I would consider Intel, since the P-Cores give the best gaming performance at the moment. But, even if I bought an Intel CPU, I would probably disable all the E-Cores, to make sure the thread scheduler doesn't mess up.

Well, Cinebench is one workload that's not meant for the E cores.
If a workload loves speculative, out-of-order execution and your efficient core is efficient because it doesn't do out-of-order, it won't be very efficient handling that workload. (This is just an example, I don't know exactly what Cinebench loves and Gracemont can actually do out-of-order execution.)

 

[Richards]

E-cores will be good on handhelds... e-cores are zen 2 level ipc 3300x+i59400

 

[InVasMani]

Still wondering about overclocking E-Cores if you dial back P-Cores clock speed multipliers how far might you push the E-cores and what's it like performance and efficiency wise? I'd say they have there up's and down's, but are certainly here to stay.

 

[BorisDG]

Within 5% of the full enchilada in 4k gaming... Just wow!

You can easily rock 6700K for 1440p+ with modern GPU and you will be fine.

 

[BigBonedCartman]

When Intel is desperate they make up “gears” and “E Cores” it’s pretty pathetic how far they’ve fallen

 

[timta2]

When Intel is desperate they make up “gears” and “E Cores” it’s pretty pathetic how far they’ve fallen

Yeah, "fallen", right there next to AMD.

 

[Minus Infinity]

Not bad, and for gaming even with only E cores in general you don't lose much at 1440p and for many games the frame rates are still excellent.

Can't wait for Raptor Lake which doubles E-core counts and improves the E-cores IPC. How will the new Zen 4c cores compare? I think Zen 4c cores will easily beat tweaked Gracemont cores and Zen 4 cores are going to have IPC uplifts of 25-30% over Zen 3. While Zen 4c could be 10-30% weaker, even at 30% weaker they'd be Zen 3 level performance which is much stronger than Gracemont. Zen 5's trump card is it's a Big.little and the little cores are Zen 4c cores. Also AMD is said to be preparing Zen 4c cpu's with 32 cores.

 

[Garrus]

It's an interesting academic exercise, but you can't truly disable the P cores and you don't get any major efficiency boosts. I'd much rather have a dual core made out of two P cores than 8 E cores, anyday, as a processor. Actually the P cores are quite efficient at equally low clock speeds.

 

[Berfs1]

Just curious why is the 10700K taking more power than the 10900K? That doesn't really make sense...

 

[W1zzard]

Just curious why is the 10700K taking more power than the 10900K? That doesn't really make sense...

Probably voltage or binning or heat, it’s only in one test

 

[blanarahul]

If the P-cores were clocked down to ~2.5 GHz to achieve (more or less) performance parity with the E-cores at 3.9 GHz, what would their power consumption look like?

(I think I can guess @W1zzard 's answer: Stop giving me ideas!)

Yeah. The fact that each P core consumes only 4-5 more watts and achieves 50% more performance is quite incredible.

Edit - 5800X is quite competetive with the lower clocked 8P cores in terms of efficiency. Very interesting. 5800X is clocked out of it's sweet spot and 8P cores are within their sweet spot, yet they are matched in efficiency.

Another interesting fact is that the 5800X at full speed (~4.5 GHz) with HT is only 20% ahead compared to 3.9 GHz Golden Cove without HT. Golden Cove is the IPC king for now (not counting M1 Macs of course, those are in a class of their own).

 

[TheinsanegamerN]

If the P-cores were clocked down to ~2.5 GHz to achieve (more or less) performance parity with the E-cores at 3.9 GHz, what would their power consumption look like?

(I think I can guess @W1zzard 's answer: Stop giving me ideas!)

I'd love to see this. Previous generations' 35 and 65 watt parts were far more efficient then their 125 watt parts. What can the P cores accomplish at E core speeds?

 

[yeeeeman]

Intel has to do a bit more work on these e cores cause they are quite inefficient. They should have just used the skylake core instead.

@W1zzard : Can you pretty please test this CPU at various TDP levels? 15,25,35,65, 105,125,150? So that we can see it will do in the mobile space for example.

 

[Solid State Brain]

@W1zzard : Can you pretty please test this CPU at various TDP levels? 15,25,35,65, 105,125,150? So that we can see it will do in the mobile space for example.

I think this would be a very interesting test. Looking at stress test results many people say that these CPUs are inefficient; I think they only look so because they are pushed to the limit by default like an overclocker would do. Lowering the power limit to more reasonable levels could make the situation clearer.

 

[Mussels]

I look at alder lake in general as a high consumption mess
These E-cores only? THESE impress me.

An E-core based laptop for example, would be quite interesting to see


(To be clear, this is sarcasm. Even the E-cores are power hungry monsters)

 

[zlobby]

Haha! Basically, the power consumption of a beefy laptop, minus its performance.

 

[THANATOS]

Thanks for the review W1zzard. Much appreciated.

I would also love to see a test simulating ADL as 2+8 with 15W and ADL 6+8 at 35W TDP, so we know what we can expect in the mobile space from Alder Lake P.