Intel Core i9-12900HK Beats AMD Threadripper 1950X at Cinebench R20

[btarunr]

Armed with 6 "Golden Cove" P-cores, 8 "Gracemont" E-cores, high clock speeds, and clever enough power-management to fit inside a "halo-class" notebook, the new Intel Core i9-12900HK "Alder Lake" offers multi-threaded performance riving HEDT processors, beating AMD's first-gen Ryzen Threadripper 1950X 16-core/32-thread processor, according to an early performance review by Lab501. The 14-core/20-thread processor scores 6741 points, compared to the reference score of 6670 points for the 1950X. The processor ends up roughly 16.8% faster than the previous-generation i9-11980HK that's based on the 8-core/16-thread "Tiger Lake-H" silicon. Stress tests show that the chip can sustain boost frequencies of nearly 4.99 GHz on the P-cores, with 113 W package power draw, and core temperatures of 99°C.



View at TechPowerUp Main Site

 

[lola46]

lol thats 4 years old CPU

 

[watzupken]

The peak power draw have increased quite drastically over the years. With Comet Lake and Tiger Lake, I've seen reviews showing peak power at 90W, and that's with 8 performance cores running at full speed. With ADL, the Golden Cove CPU is certainly very power hungry when pushed hard since I don't expect the E-cores to guzzle that much power. Under light load, a laptop with ADL chips may last longer because of the E-cores, but under load, chances are that battery will run out very quickly, or the CPU may throttle quite a bit when running on battery.

 

[Totally]

Uh congrats, guess beating a now $350-580 cpu is something to brag about

 

[watzupken]

lol thats 4 years old CPU

This is true and the first gen Threadripper was manufactured on an older node. A better comparison will actually be their own Tiger Lake H processor, which I feel is alright for a generation improvement. Cinebench favours high core count, and this is where the E-cores will actually give the ADL processor a leg up in terms of multicore performance. My main concern is the longevity of these very hot chips. 99 degrees may not be the junction temp, but it is still extremely hot and likely bad for the chip and components around it in the longer term.

 

[eidairaman1]

Comparing an old unstrained cpu to a new already strained cpu is like comparing a v8 to an i4.

 

[Ruru]

"with 113 W package power draw, and core temperatures of 99°C"
I'm more worried about that as it's a MOBILE chip than it beating an old HEDT in synthetic benchmark..

 

[R0H1T]

Did Intel pay someone for this stupid leak comparison, I hope not

lol thats 4 years old CPU

5 years in a few months.

 

[Crackong]

Isn't CB not a "Real world benchmark" by the definition from Intel ?

Btw
Judging from its 99 degree C
This 12900HK is already stressed to its maximum
A quick search on HW bot says 1950x @4ghz results 7706 points in CB20

Maybe next time

Just checked the original post
The original post also tested plugged vs unplugged
The 12900HK scored an earth-shattering 2942 points while unplugged and eating 70+Watts
What's wrong?

 

[ElectrO]

Why is this news and why did the editor allow this?

 

[Lew Zealand]

Actually in CB R23 multiprocessor at the link, I see:

TR 1950x - 16315
i9-12900HK plugged - 16917
i9-12900HK unplugged - 6119

oof.

64% reduction in speed on battery, almost as good as a 7700K. Honestly, this looks broken somehow.

 

[Crackong]

64% reduction in speed on battery, almost as good as a 7700K. Honestly, this looks broken somehow.

Clearly the Error-cores strikes again.
The power profile is so limited so it decides to use E-cores primarily.

 

[Vya Domus]

Probably one of the most laughable comparisons ever. And keep in mind that the original Zen 1 core had half the AVX throughput.

 

[Broken Processor]

A very strange comparison.

 

[dj-electric]

I feel like this comparison is a bit misguided, and as you can see, birthed a bit of a wccftech exiled clown parade in the comments.

To put things in perspective - the score is similar to what you find on a 12600K or Core i9 10900K, and is slightly above the score of a 5800X in this test.
Extrapolating data is the hardest and most impossible mission for TPU users.

 

[CoolZone]

It's like comparing apples to oranges...no reason at all. Maybe the editor got bored due to lack of samples and started some side projects such as this one ))

 

[OC-Ghost]

65-115W, Why not buy one those laptops with desktop chips then..

 

[Dave65]

Why is this news and why did the editor allow this?

When bias rules nothing else matters.

 

[Prima.Vera]

If they compare it with 5 years old CPUs, I want it compared with my 3770K too. On single-thread. What is the difference I wonder?

 

[R0H1T]

What is the difference I wonder?

That's hard to extrapolate, you can probably guesstimate the numbers from AT by using older bench scores ~

CPU 2021 Benchmarks - Compare Products on AnandTech

CPU 2021 benchmarks: Compare two products side-by-side or see a cascading list of product ratings along with our annotations.

www.anandtech.com

Or use notebookcheck ~

https://www.notebookcheck.net/Mobile-Processors-Benchmark-List.2436.0.html

 

[1d10t]

1950X owner rejoice, now you can upgrade to 12900HK notebook.