Question Raptor Lake - Official Thread

[Hulk]

Since we already have the first Raptor Lake leak I'm thinking it should have it's own thread.
What do we know so far?
From Anandtech's Intel Process Roadmap articles from July:

Built on Intel 7 with upgraded FinFET
10-15% PPW (performance-per-watt)
Last non-tiled consumer CPU as Meteor Lake will be tiled

I'm guessing this will be a minor update to ADL with just a few microarchitecture changes to the cores. The larger change will be the new process refinement allowing 8+16 at the top of the stack.

Will it work with current z690 motherboards? If yes then that could be a major selling point for people to move to ADL rather than wait.

 

[Hulk]

It require 170-180W for the 13900K to perform like a 7950X@100W.

Yes that is true. But I think it would be smarter to run the 13900K at 100W and just accept that it is 17% slower than the 7950X rather than run Raptor that high. It simply cannot compete with Zen 4 at those performance levels when it comes to efficiency.

 

[nicalandia]

13900K Linux Review with DDR5 this time(Will do a comparison with the DDR4)

Benchmarks - OpenBenchmarking.org

openbenchmarking.org

 

[Kaluan]

Don't like this character but apparently he have a sneakpeek of the 13900k out. Haven't watched the video myself

LOL, scrolled through his "catalogue"... the guy's a massive clown, worse clickbaits than LTT. The Dunning-Kruger in his tech talks/analysis is embarrassing.

BTW, does running P cores only on say the 13900K not = 12MB less of L3? Or 6MB/3MB less on 13600K/12600K and such?

 

[Kocicak]

Why do we have to see the clowns face twice already here?

BTW when are the real reviews coming out? It seems that 13700K was not sent out to reviewers, I wonder why.

 

[nicalandia]

BTW, does running P cores only on say the 13900K not = 12MB less of L3? Or 6MB/3MB less on 13600K/12600K and such?

The allotted amount of L3$ Per P Core remains the same at 3 MiB so 8 P cores have 24 MiB Of L3$(Unchanged from 12900K) But disabling the e cores will allow the 8P Cores to take advantage of the entire 36 MiB because the ring remains active

 

[uzzi38]

Alrighty, i have completed gathering the data for a tuned 7950x
4800MT/s memory with auto timings were used for all the runs.
SOC draw alone is around ~10w (important for the lower PPT runs)

PPT 250w, 160A TDC and 225A EDC limit = 40360 points (OBS: CPU only pulled 230w maximum)
View attachment 69276

PPT 210w, 140A TDC and 200A EDC limit = 40085 points (OBS: CPU only pulled 204w maximum)
View attachment 69277

PPT 160w, 115A TDC and 175A EDC limit = 39068 points
View attachment 69278

PPT 130w, 100A TDC and 160A EDC limit = 37344 points
View attachment 69279

PPT 100w, 80A TDC and 150A EDC limit = 34574 points
View attachment 69280

PPT 65w, 70A TDC and 130A EDC limit = 28387 points
View attachment 69281

PPT 50w, 65A TDC and 125A EDC limit = 22832 points
View attachment 69284

PPT 35w, 60A TDC and 120A EDC limit = 144667 points
View attachment 69282

PPT 25w, 40A TDC and 80A EDC limit = 4180 points (cores were running at a wooping 500mhz each while rendering cinebench lol)
View attachment 69283

At the lower end of the PPT scale the 7950x is limited by the chiplet design with SOC on a separate die drawing ~10w alone
At PPT limit 65w the cores had ~3.44w each
At PPT limit 50w the cores had ~2.5w each
At PPT limit 35w the cores had ~1.56w each
At PPT limit 25w the cores had ~0.93w each

A monolithic APU should be much better suited here..

*edit*
added 50w PPT run

When you say it's tuned, how so?

 

[Kocicak]

Tuned? I missed that, that makes the comparison somewhat less reliable, my data are from completely stock chip just capped at different power levels. And on a possibly non final bios.

 

[nicalandia]

Linux Benchmarks, 13900K DDR4 vs DDR5

Benchmarks - OpenBenchmarking.org

openbenchmarking.org

 

[Kocicak]

Linux Benchmarks, 13900K DDR4 vs DDR5

Benchmarks - OpenBenchmarking.org

openbenchmarking.org

There is twice as much DDR5 memory there, can it affect results?

 

[nicalandia]

There is twice as much DDR5 memory there, can it affect results?

Oh I am sure, but also the DDR5 Says @4Ghz(not sure if that is true) but that is all of the info we have on Linux performance from the 13900K

 

[inf64]

Linux Benchmarks, 13900K DDR4 vs DDR5

Benchmarks - OpenBenchmarking.org

openbenchmarking.org

Probably a newbish question, but is there a way to add another CPU as a comparison? Or use this result to compare it to some other runs on openbenchmarking website? I looked on the website but found no such option.

 

[Kaluan]

The allotted amount of L3$ Per P Core remains the same at 3 MiB so 8 P cores have 24 MiB Of L3$(Unchanged from 12900K) But disabling the e cores will allow the 8P Cores to take advantage of the entire 36 MiB because the ring remains active

Ah cool, cheers!

Makes that grifter's data look even more strange then I suppose.

Bit out of the loop, when's the RTL NDA lifting?

 

[nicalandia]

Probably a newbish question, but is there a way to add another CPU as a comparison?

Oh I am sure there is, I had to find it myself(just adding a , and then the ID for the run), Openbenchmarking is not the friendliest of sites.

 

[nicalandia]

13900K(DDR5) vs 7950X AVX-512(Default an Disabled) Performance(Using Intel's Optimized Libraries)

Benchmarks - OpenBenchmarking.org

openbenchmarking.org

Edit..
These benchmarks are me finding them and putting them together(Openbenchmarking). Default AVX-512 is Default 7950X and Disabled is 7950X with AVX-512 Disabled.

 

[Markfw]

13900K(DDR5) vs 7950X AVX-512(Default an Disabled) Performance(Using Intel's Optimized Libraries)


View attachment 69324

View attachment 69325

View attachment 69326

Benchmarks - OpenBenchmarking.org

openbenchmarking.org

why are they talking about avx-512 when that is NOT on the 13900k ?

 

[nicalandia]

why are they talking about avx-512 when that is NOT on the 13900k ?

From the source:

"Embree: Intel Embree is a collection of high-performance ray-tracing kernels for execution on CPUs and supporting instruction sets such as SSE, AVX, AVX2, and AVX-512. Embree also supports making use of the Intel SPMD"

So when the Benchmark does not find any AVX-512 it will use Other Advanced Vector Extensions SIMD instructions.

 

[Hitman928]

why are they talking about avx-512 when that is NOT on the 13900k ?

The with and without AVX-512 results are the 7950x.

 

[Markfw]

From the source:

"Embree: Intel Embree is a collection of high-performance ray-tracing kernels for execution on CPUs and supporting instruction sets such as SSE, AVX, AVX2, and AVX-512. Embree also supports making use of the Intel SPMD"

So when the Benchmark does not find any AVX-512 it will use Other Advanced Vector Extensions SIMD instructions.

well, its pretty stupid that we don't even know what it is using for the benchmark. How does the 7950x compare ????

 

[nicalandia]

well, its pretty stupid that we don't even know what it is using for the benchmark. How does the 7950x compare ????

I posted Pictures on the Original Posts. Are they not being Displayed?

The default 7950X is 40% higher than 13900K, but when the AVX-512 is disabled(on 7950X) the 13900K is 25% Higher.

 

[moinmoin]

The oneDNN results are instructive if only for the reason that there Intel clearly is not artificially disadvantaging AMD chips. Which I think is great news, for oneAPI's adoption potential and otherwise.

 

[inf64]

Even though Raptor Lake is clearly a massive improvement over 12th gen due to more threads and somewhat higher clocks (kudos to intel), it's just not as efficient or fast as Ryzen 7000. With Ryzen there is no thread director and no worry what is scheduled where and when, all cores are "P" cores and boost super high, guaranteeing more even performance scaling in MT workloads. IPC being neck even, and Ryzen supporting AVX512, while consuming much less power, makes it a superior platform (plus AM5 being at the beginning of its life cycle, unlike intel which is at its end). Still, in the mid range intel will sell a ton of 13600K and 13700K parts which will be a great success.

 

[Kocicak]

With Ryzen there is no thread director and no worry what is scheduled where and when, all cores are "P" cores and boost super high, guaranteeing more even performance scaling in MT workloads.

I observed how assigning work to the cores works in 13900K and as the thread count increases, it works like this:

1) P cores are getting 1 thread each
2) after P core has 1 thread each, E cores are starting to get work
3) P cores are starting to get work on their second threads only after E cores are highly utilised.

It seems it knows what to do with those cores... I would not worry too much about that.

 

[inf64]

I observed how assigning work to the cores works in 13900K and as the thread count increases, it works like this:

1) P cores are getting 1 thread each
2) after P core has 1 thread each, E cores are starting to get work
3) P cores are starting to get work on their second threads only after E cores are highly utilised.

It seems it knows what to do with those cores... I would not worry too much about that.

Yes that good to hear , but E cores have lower IPC, lower AVX throughput, lower clock. The benefit is lower power so it evens out, unless you push the E cores beyond its perf/watt sweetspot. Personally I don't like the hybrid approach as I want a straightforward performance scaling. Kudos to intel for making it work good, it's just not my preference.

 

[Kosusko]

As soon as I get 13900KF, I switch off seven big. P-cores and I will use the hybrid big.LITTLE processor only with 1C/1T Raptor Cove up to 5.8GHz and 16c/16T "Raptormont" up to 4.3GHz. I haven't needed a big cores for over twenty years. After twenty years, I'll only treat myself one big. P-core for "blink blink" single-thread performance.

P.S. please do not judgeme, it is my decision, my requirements and my money. Peace.

"Raptormont" is of course Gracemont

 

[nicalandia]

According to Geekbench the 13600 And 13400 Use the same Die(1.25 MiB L2$ For P cores and 2 MiB for e cores)

Founder Predator PO5-650 vs Founder Predator PO5-650 - Geekbench

browser.geekbench.com

The 13600K uses a different Die even if they have the same core/thread amount. The e cores on the 13600 are Gracemont with 2MiB per Cluster. The 13600K gets Raptormon cores with 4MiB per cluster.

ASUS System Product Name vs ASUS System Product Name - Geekbench

browser.geekbench.com