Ripping through Threadrippers - Core i9-12900HK Beats Threadripper 1950X In Cinebench R20 Benchmark

[Lightwreather]

Summary

Yesterday Romanian tech review site Lab 501 put up one of the world's first reviews of Intel's new mobile Core i9-12900HK (Alder Lake) processor. Packing Intel's all-new hybrid core microarchitecture, the CPU has managed to beat AMD's desktop Ryzen Threadripper 1950X in Cinebench R20

 

Quotes

Quote

In practice, the chip performs every bit and what the specs say. Thanks to the extra core count of the E-cores, and the enhanced IPC of the Golden Cove cores, the Core i9-12900HK puts up an impressive Cinebench R20 score of 6,741 points -- higher than any mobile CPU we've seen to date and equal to that of Intel's desktop Core i5-12600K.

For reference, Intel's previous mobile flagship, the Core i9-11980HK, is nearly 1,000 points behind the new Alder Lake part, with a score of 5,772 points. Even further back is one of AMD's top-tier mobile processors, the Ryzen 9 5900HX, with a score of 5,229 points.

But most impressively, the Core i9-12900HK is the first mobile CPU to out-right beat Cinebench R20's reference Ryzen Threadripper 1950X score of 6,670 points (which you can find by installing the app for yourself). An impressive result, considering the Ryzen Threadripper 1950X is a 180W behemoth with 16 core Zen cores designed for desktops.

 

My thoughts

Welp, that's competition for ya. A mobile CPU competing with a full on desktop CPU (albeit a 5 year old one, but still). Suffice to say, I'm impressed. Whether this is worth it or not, only time will tell but so far, Intel is putting out some big numbers here. and tbh, I can't wait to see what sort of response AMD's mobile 6000 series will have against this. 

 

Sources

Tom's Hardware

Lab 501

[TrigrH]

A overclockable laptop 14 core beats a 5yr old 16core AMD cpu.......ok?

[Quackers101]

when older threadrippers be ripping, feeling the age man.

things move fast as fu** boiii, but like the others said, it not as big of a deal. Although it kind of shows how some things have changed and it should.

 

until we get 4+4 core stacked and 6GHz beating older and slower 64 core premium CPUs

[dizmo]

Why is this news? That AMD chip is pretty old at this point. Seems like a rather unfair comparison.

What do you mean if it's worth it or not? In comparison to the Threadripper, that you won't be buying because it's old, and probably won't be cross shopping, since they're not even really in the same product category?

[seon123]

Amazing, a brand new CPU beats a 5 year old CPU. Next up, the RTX 4080 is rumoured to outperform the R9 Fury X, a water cooled flagship AMD GPU

[Lightwreather]

1 minute ago, dizmo said:

Why is this news? That AMD chip is pretty old at this point. Seems like a rather unfair comparison.

Well, apparently, it's first mobile chip to beat it, so.......

2 minutes ago, dizmo said:

What do you mean if it's worth it or not?

I meant, in terms of it's power consumption and cost (ofc this'll be dependent on the specific laptop) and compared to other processors in this price bracket.

[dizmo]

1 minute ago, J-from-Nucleon said:

Well, apparently, it's first mobile chip to beat it, so.......

I meant, in terms of it's power consumption and cost (ofc this'll be dependent on the specific laptop) and compared to other processors in this price bracket.

It's a mobile chip. You don't buy the processor alone. So you'd have to look at complete systems, which again, doesn't really make sense when you compare it to an old desktop chip. One is also meant to be mobile, and the other sit on a desk.

 

I hope you don't get injured, I've never seen someone try and stretch like that.

[Jurrunio]

The Romanian site does mention the 12900HK running at 65W PL1 and 115W PL2, I wonder what kind of turbo duration they chose.. If it's 115W the whole time,  beating a 1950X which also has process node and PCIe controller disadvantage doesn't mean too much in terms of measuring Intel's progress.

[Alder Lake]

15 hours ago, TrigrH said:

A overclockable laptop 14 core beats a 5yr old 16core AMD cpu.......ok?

14 hours ago, dizmo said:

Why is this news? That AMD chip is pretty old at this point. Seems like a rather unfair comparison.

What do you mean if it's worth it or not? In comparison to the Threadripper, that you won't be buying because it's old, and probably won't be cross shopping, since they're not even really in the same product category?

14 hours ago, seon123 said:

Amazing, a brand new CPU beats a 5 year old CPU. Next up, the RTX 4080 is rumoured to outperform the R9 Fury X, a water cooled flagship AMD GPU

first up, remember that this is a mobile chip

 

second up, remember how people 1 month before Intel's Alder Lake launch were trying to make fun of them for "not being able to make a true 16 core CPU," and now we have Laptops with 20 threads beating workstation CPUs with 32 threads, I personally find this to be pretty cool, though obviously I find anything related to Intel crushing AMD to be quite cool

 

 

 

like he said, it's the first mobile chip to beat a threadripper, granted, it's a really old threadripper, but a threadripper regardless, more of a checkpoint than news

[leadeater]

2 hours ago, Alder Lake said:

first up, remember that this is a mobile chip

A mobile chip consuming more than many desktop chips, AMD and Intel included 

 

2 hours ago, Alder Lake said:

second up, remember how people 1 month before Intel's Alder Lake launch were trying to make fun of them for "not being able to make a true 16 core CPU," and now we have Laptops with 20 threads beating workstation CPUs with 32 threads, I personally find this to be pretty cool, though obviously I find anything related to Intel crushing AMD to be quite cool

Well if beating a 16 core desktop chip is impressive then why not compare it to say the 5950X, it's also a 16c/32t CPU.

 

5950X: 10,428 @ 119W with Perf/Watt of 87.63

5980HS: 4,546 @ 35W with Perf/Watt of 129.89

12900HK: 6,741 @ 115W with Perf/Watt of 58.62

11980HK: 5,875 @ 86W with Perf/Watt of 68.31

11980HK: 3,806 @ 45W with Perf/Watt of 84.58

 

Notice a problem here? Hint, look only at the Intel CPUs.

 

That said though, CB 23 MT:

M1 Max/Pro: 12,402 @ 34W with Perf/Watt of 364.76

5950X: 28,641 @ 119W with Perf/Watt of 240.68

5980HS: 12,844 @ 35W with Perf/Watt if 366.97

12900HK: 19,402 @ 115W with Perf/Watt of 168.71

11980HK: 13,977 @ 86W with Perf/Watt of 162.5

 

Previous older generation software issue sort of fixed, barely.

 

That leaves us with a laptop CPU (12900HK) with lower performance, equivalent power draw and worse power efficiency than a desktop CPU. Hell it doesn't even look that impressive compared to Intel's own last generation.

 

Of course that said CB is not at all good for this comparison and Alder Lake will have much better performance in games etc which is very poorly represented in CB scores.

 

If Alder Lake mobile is doing this badly against AMD's desktop CPUs, entirely outclassed in power efficiency for AMD's mobile CPUs, then how do you think it's going to turn out when Ryzen Mobile 6000 hits the actual market? Slightly better performance at the same power as well as products actually using the 6980HX/6900HX parts this time which can be configured to use a lot more power than the base 45W so Alder Lake will neither have the performance lead or the power efficiency lead at the highest performance, in CB.

 

There's a good reason the highest performance highest TDP 3080 Ti mobile GPUs are mostly paired with the upcoming Ryzen Mobile 6000 HX CPUs and not Alder Lake. Not that their aren't Alder Lake laptops with 3080 TI's mind you but there is a difference in both the TDP and the design target of those laptops.

 

It's ok though, Intel gets a participation award in the mobile CPU market. Well done Intel, you exist 

Seriously though Intel is MUCH more competitive at lower power with Alder Lake. It's a complete joke to be trying to talk up the 12900HK which is probably the worst CPU to represent Alder Lake Mobile

[leadeater]

2 hours ago, Jurrunio said:

If it's 115W the whole time

115W the whole time, official default spec is all core boost power of 115W.

 

https://ark.intel.com/content/www/us/en/ark/products/132215/intel-core-i912900hk-processor-24m-cache-up-to-5-00-ghz.html

[RejZoR]

News like this are always so hilarious. Wow, a brand new state of the art top of the line CPU outperforms a CPU that was released in late July 2017... It's about a stupid statement as Apple's claim every year that "this is our fastest processor yet". Yeah no shit, I'd be rather disappointed if it didn't beat all the old ones, to be honest...

 

I ran R20 for the lolz, no optimizations, just ran it with bunch of crap running on my system as it does every day. And this is the score... a year old 5800X almost catches 5 years old Ryzen 1950X with half as many cores. Just this alone is telling how much things have progressed when CPU from same segment (desktop) is almost as fast as old one and it has just half of the cores old one had. It's higher all core clocks, higher IPC, lower CCX latencies, newer more efficient fabrication process etc.

 

Another indication is comparing 5800X to 1700X. Same segment, same core count and 5800X almost doubles the performance if you just glance at the graphs.

 

People need to realize 5 years is a VERY long period in computing. So, yeah, it's not really surprising or news worthy if you understand things even tiny bit.

[porina]

4 hours ago, Jurrunio said:

beating a 1950X which also has process node and PCIe controller disadvantage doesn't mean too much in terms of measuring Intel's progress.

Cinebench through the years has never shown much impact outside of cores and clocks, so "PCIe controller" almost certainly had negligible impact. Cinebench R15 was the AMD fanboy favourite as it was an example of software that scales near perfectly with cores and clock, unlike most other software.

 

Note a possible further impact here is that R20 was used, which was the first to introduce use of AVX. Before Zen 2, Ryzen has pretty bad AVX performance, more comparable to Sandy Bridge era. Having said that, I can't recall the exact details of my testing before, but I didn't see that strong a benefit with AVX in Cinebench anyway, implying it wasn't using it in a way that really gets a boost from it.

[Sauron]

Damn, my 5W laptop cpu beats this 100W desktop pentium 4, AMAZING

[cj09beira]

2 hours ago, porina said:

Cinebench through the years has never shown much impact outside of cores and clocks, so "PCIe controller" almost certainly had negligible impact. Cinebench R15 was the AMD fanboy favourite as it was an example of software that scales near perfectly with cores and clock, unlike most other software.

 

Note a possible further impact here is that R20 was used, which was the first to introduce use of AVX. Before Zen 2, Ryzen has pretty bad AVX performance, more comparable to Sandy Bridge era. Having said that, I can't recall the exact details of my testing before, but I didn't see that strong a benefit with AVX in Cinebench anyway, implying it wasn't using it in a way that really gets a boost from it.

pretty sure when he said those "disadvantages" he meant that those would negatively affect perf/w, making the efficiency improvements less impressive.

[Jurrunio]

5 hours ago, leadeater said:

115W the whole time, official default spec is all core boost power of 115W.

 

https://ark.intel.com/content/www/us/en/ark/products/132215/intel-core-i912900hk-processor-24m-cache-up-to-5-00-ghz.html

Older gen Intel mobile CPUs dont stay at PL2 all the time, it lasts like 28 seconds at stock before downgrading to PL1. That's why I wonder what is the official number for 12900HK (cant find spec sheets mentioning this) and whether the test manually increased it.

 

3 hours ago, porina said:

Cinebench through the years has never shown much impact outside of cores and clocks, so "PCIe controller" almost certainly had negligible impact. Cinebench R15 was the AMD fanboy favourite as it was an example of software that scales near perfectly with cores and clock, unlike most other software.

 

Note a possible further impact here is that R20 was used, which was the first to introduce use of AVX. Before Zen 2, Ryzen has pretty bad AVX performance, more comparable to Sandy Bridge era. Having said that, I can't recall the exact details of my testing before, but I didn't see that strong a benefit with AVX in Cinebench anyway, implying it wasn't using it in a way that really gets a boost from it.

I'm talking about those parts burning extra power on the 1950X, so even if the cores have the same efficiency the 1950X would appear to use more power

[leadeater]

Just now, Jurrunio said:

Older gen Intel mobile CPUs dont stay at PL2 all the time, it lasts like 28 seconds at stock before downgrading to PL1. That's why I wonder what is the official number for 12900HK (cant find spec sheets mentioning this) and whether the test manually increased it.

12th Gen is different. Intel never listed a maximum turbo power spec, but with 12th Gen they do and it's an actual power. Like with the K parts on desktop the HK parts are also PL1 = PL2.

[porina]

1 minute ago, Jurrunio said:

I'm talking about those parts burning extra power on the 1950X, so even if the cores have the same efficiency the 1950X would appear to use more power

Doh! Missed that context. Looking back, it should have been obvious.

[Jurrunio]

2 minutes ago, leadeater said:

12th Gen is different. Intel never listed a maximum turbo power spec, but with 12th Gen they do and it's an actual power. Like with the K parts on desktop the HK parts are also PL1 = PL2.

then maybe the non K parts will have PL1 != PL2? After all they all have base and max power listed, if all CPUs ignore the base then why have that at all

[leadeater]

2 minutes ago, Jurrunio said:

then maybe the non K parts will have PL1 != PL2? After all they all have base and max power listed, if all CPUs ignore the base then why have that at all

Correct, if you don't buy the top end OC unlocked mobile CPU then power usage won't be anywhere near as high.

 

Edit:

There is only a single HK mobile CPU, the 12900HK.

[CTR640]

8 hours ago, TrigrH said:

A overclockable laptop 14 core beats a 5yr old 16core AMD cpu.......ok?

I know, right? Pathetic and cringy.

[leadeater]

Quote

During its stress test, Lab501 confirmed that the part boots up to 4.988 GHz while reaching 113W of power and 99°C

 

 

As above, looks like thermal limited PL2 and it's dynamic up to 115W if it's cooled enough. CPU package power is still 113W so close enough to 115W I won't bother to go back and change my numbers.

[Imbadatnames]

Would like to see how quick it is over multiple runs. Probably the reason they used R20 instead of 23 too 

[Kronoton]

So a 115W brand new CPU ("mobile" my.....) beat 5 year old one running at 180W in a short synthetic benchmark?

 

Impressive!!!

 

 

 

 

.... how little real progress was made.

[Salv8 (sam)]

14 hours ago, dizmo said:

Why is this news? That AMD chip is pretty old at this point. Seems like a rather unfair comparison.

because @J-from-Nucleon was paid by intel to post this in the news section so it becomes news.

we're unto to you pat, we know you read the tech news section of the forums, YOU CANNOT FOOL US PATRICK!