Intel Core i9-13900 "Raptor Lake" Processor Gets a Preview

[AleksandarK]

Intel is preparing to launch its 13th generation of desktop processors codenamed Raptor Lake. Succeeding Alder Lake, the 13th gen design will implement up to eight P-cores with 16 E-cores manufactured on Intel's improved 7+ technology node. Today, we got a performance preview from SiSoftware that has collected SiSoftware Sandra database scores of Intel Core i9-13900 Raptor Lake-S processor. They present an overview of a few benchmarks. Firstly, the SoC features 36 MB of unified L3 cache versus 30 MB in Alder Lake. With DDR5 memory running up to 5600 MT/s and PCIe 5.0, the SoC features the latest IO and memory standards. The big P-cores now lack AVX-512 and feature 2 MB of L2 cache per core. We see 4 MB of L2 cache for a cluster of small E-cores. An exciting addition to E-cores is the AVX/AVX2 support, which is a first for Atom cores.

Regarding testing, the author has collected a few tests that seemed appropriate to compare to the equivalent Alder Lake model. Starting with ALU/FPU tests that benchmark basic arithmetic tasks, Raptor Lake delivered 33% to 50% improvement over Alder Lake. The Raptor Lake design achieved this with 3.7 GHz P-Core and 2.76 GHz E-Core frequency. In vectorized and SIMD tests, the 13th gen design showed only 5% to 8% improvement over the previous generation. For more benchmarks and accurate results, we have to wait for TechPowerUp's test, which will be coming on the release day.


View at TechPowerUp Main Site | Source

 

[Shatun_Bear]

About 5-8% better single threaded perf is what I expect. Multi-thread total performance uplift will be quite a bit higher because they'll compare the 12900K to the 13900K that has 8 more budget/half/nerfed cores.

Vs Zen 4 it will be quite even I predict, some wins some loses.

 

[ravenhold]

What is the TDP of Raptor Lake?

 

[aciDev]

About 5-8% better single threaded perf is what I expect. Multi-thread total performance uplift will be quite a bit higher because they'll compare the 12900K to the 13900K that has 8 more budget/half/nerfed cores.

Vs Zen 4 it will be quite even I predict, some wins some loses.

Don't forget AMD is going to push frequencies too

 

[ModEl4]

About 5-8% better single threaded perf is what I expect. Multi-thread total performance uplift will be quite a bit higher because they'll compare the 12900K to the 13900K that has 8 more budget/half/nerfed cores.

Vs Zen 4 it will be quite even I predict, some wins some loses.

https://www.sisoftware.co.uk/2022/06/10/intel-13th-gen-core-raptorlake-i9-13900-preview-benchmarks/

• In legacy ALU/FPU tests, RPL shows a large 33-50% improvement over ADL even at lower clocks which is tremendous.
• In heavy vectorised/SIMD tests, RPL sees only 5-8% improvement over ADL (at lower clocks) which is encouraging but perhaps not a surprise as the extra Atom cores are not going to provide much uplift. We are waiting for additional benchmark results to have a better understanding.
• The huge L2 combined caches (16+16 = 32MB) and the increased L3 (36MB unified) finally overtake/match AMD’s Zen3 – but now with Zen2-3D V-Cache and forthcoming Zen4 (1MB L2 per core vs. 512kB) it may still not be enough.

 

[Shatun_Bear]

Don't forget AMD is going to push frequencies too

Yeah, I think some are sleeping on Zen 4 performance because of that terrible '>15%' presentation. Max boost clock looks like it will be almost 800Mhz higher gen to gen which is unprecedented (4.9 vs 5.7). Those frequencies combined with the increased cache will make it killer for gaming (which people put too much stock in btw). Not sure if RPL can edge a win there or not.

 

[olymind1]

Games will like the extra cache a lot. It was clear when AMD doubled it from Zen+ -> Zen2, also when they doubled (the amount what a single core/thread can access within its CCX) with Zen3.

Cheap 6 core 13400 cpus will be awesome with great P/P ratio, not that 12400 or 5600 or even the 3600 are bad cpus, because they aren't.

Finally some progress after AMD and Intel sleeping through almost a decade!! Now bring this progress to the VGA market too!

 

[ARF]

Good... Unreleased future processor is compared to a 2020 old one that is not even the competitor's flagship.
Nice try

What is the TDP of Raptor Lake?

The same as Alder Lake, maybe. This is the same old broken 10nm intel process.

 

[Keullo-e]

This is the same old broken 10nm intel process.

Oh boy, wonder how many generations we'll see this time with the same node..

 

[ModEl4]

Yeah, I think some are sleeping on Zen 4 performance because of that terrible '>15%' presentation. Max boost clock looks like it will be almost 800Mhz higher gen to gen which is unprecedented (4.9 vs 5.7). Those frequencies combined with the increased cache will make it killer for gaming (which people put too much stock in btw). Not sure if RPL can edge a win there or not.

5950X, depending on the sample, had up to 5050MHz single core turbo frequency, so probably is something like 5.6-5.8GHz vs 5.05GHz (plus what IPC will give, AMD after the negativity that the 15% ST performance comment created clarified with 8-10% IPC) so best case, around 20-26% supposedly in total instead of 15%, but regarding L3 cache i have a feeling that we may have the same 32GB if you check the Zen4 based GENOA-X server info (Milan-X with 64 cores has 768MB L3 and it seems GENOA-X with 96 cores will be up to 1152MB (one explanation of the 1GB+) and possibly with minus some redundancy measure (88MB vs 96MB) giving us 1056MB L3 then)

 

[napata]

Good... Unreleased future processor is compared to a 2020 old one that is not even the competitor's flagship.
Nice try



The same as Alder Lake, maybe. This is the same old broken 10nm intel process.

TDPs are just artificial limits. A 12900k doesn't even really have a TDP.

Also not really sure how you can complain about Intel 7. More or less the same efficiency as TSMC's 7nm.

 

[ARF]

Err, TSMC has already begun risk or non-risk production of N3, N5 has been around for quite a while. So, yeah, intel is late.

 

[Shatun_Bear]

TDPs are just artificial limits. A 12900k doesn't even really have a TDP.

Also not really sure how you can complain about Intel 7. More or less the same efficiency as TSMC's 7nm.

It should have better efficiency than TSMC's 7nm though as that node is old at this point, we're already on 6nm about to move on to 5nm on desktop.

 

[InVasMani]

The AVX/AVX2 change on the E-cores is a nice surprise. That could really make a huge difference in a heap of different scenario's.

 

[ARF]

It should have better efficiency than TSMC's 7nm though as that node is old at this point, we're already on 6nm about to move on to 5nm on desktop.

I don't exactly understand what you meant. TSMC has several modifications of the N7 production lines:

7 nm process - Wikipedia

N6 is nothing but N7++.

 

[Fourstaff]

Performance looks decent enough. If Intel cannot compete on performance, they will compete on price.

 

[Chrispy_]

...and yet vendors are still trying to clear inventory of 10th gen desktop and mobile parts.

 

[InVasMani]

Price, efficiency, performance and features will determine how good it is as a whole. It's got plenty of room to separate itself from Zen 3 further even if it doesn't beat Zen 4 outright, but still might in area's.

 

[ARF]

Zen 4 with its integrated GPU units will be quite interesting (also to watch the reviews for performance metrics) - I wonder if they would make any effort to develop hardware acceleration in heterogenous computing direction, similar to intel's Quick Sync... GPU Accelerated Rendering and Hardware Encoding (adobe.com)

 

[Vayra86]

Yeah, I think some are sleeping on Zen 4 performance because of that terrible '>15%' presentation. Max boost clock looks like it will be almost 800Mhz higher gen to gen which is unprecedented (4.9 vs 5.7). Those frequencies combined with the increased cache will make it killer for gaming (which people put too much stock in btw). Not sure if RPL can edge a win there or not.

5.7 you say?
Got a source?

 

[ARF]

Yeah, I think some are sleeping on Zen 4 performance because of that terrible '>15%' presentation. Max boost clock looks like it will be almost 800Mhz higher gen to gen which is unprecedented (4.9 vs 5.7). Those frequencies combined with the increased cache will make it killer for gaming (which people put too much stock in btw). Not sure if RPL can edge a win there or not.

Core i9-12900KS boosts up to 5.5 GHz.
Ryzen 9 5950X boosts up to 4.9 GHz.
We have seen an ES of Zen 4 running at 5.21 GHz.

Latest Zen 4 leaks: 8-core Ryzen 7000X CPU pops up on OpenBenchmarking, estimated performance gains over Zen 3 revealed by MLID - NotebookCheck.net News

 

[defaultluser]

About 5-8% better single threaded perf is what I expect. Multi-thread total performance uplift will be quite a bit higher because they'll compare the 12900K to the 13900K that has 8 more budget/half/nerfed cores.

Vs Zen 4 it will be quite even I predict, some wins some loses.

It depends on how well that software scales above 16 cores. And at-most your speedup is going to be a lot lower than the 50% increase in core count. ( e-cores are worth about 60% a 2-thread p-core, so closer to 30% peak?)

 

[Shatun_Bear]

I don't exactly understand what you meant. TSMC has several modifications of the N7 production lines:

View attachment 250889
7 nm process - Wikipedia

N6 is nothing but N7++.

Point is you're comparing Intel's latest on Intel 7 (10nm) to Ryzen 5000 on TSMC 7 (7nm+).

Ryzen 7000 is on 5nm with the +25% ppw gains while Intel are stuck on Intel 7 with Raptor Lake. They'll be behind in efficiency again.

 

[chrcoluk]

...and yet vendors are still trying to clear inventory of 10th gen desktop and mobile parts.

Yeah a downside of rapid product releases, if I was a retailer I would be really paranoid about ordering too much as a new gen is always round the corner.

 

[ValenOne]

https://www.sisoftware.co.uk/2022/06/10/intel-13th-gen-core-raptorlake-i9-13900-preview-benchmarks/

• In legacy ALU/FPU tests, RPL shows a large 33-50% improvement over ADL even at lower clocks which is tremendous.
• In heavy vectorised/SIMD tests, RPL sees only 5-8% improvement over ADL (at lower clocks) which is encouraging but perhaps not a surprise as the extra Atom cores are not going to provide much uplift. We are waiting for additional benchmark results to have a better understanding.
• The huge L2 combined caches (16+16 = 32MB) and the increased L3 (36MB unified) finally overtake/match AMD’s Zen3 – but now with Zen2-3D V-Cache and forthcoming Zen4 (1MB L2 per core vs. 512kB) it may still not be enough.

Zen 4 has AVX 512 VNNI BFloat16 support.