Thursday, July 14th 2022
Intel "Raptor Lake" ES Posts 9.4% Higher Single-Core Performance Than "Alder Lake"
In what could be evidence of Intel pulling off a major generational IPC increase, Chinese PC enthusiast Extreme Player, with access to a Core i9-13900K engineering sample (ES), tested the chip on a handful synthetic tests, with the processor yielding significant performance gains over its predecessor, the i9-12900K. The most striking performance number has to be the CPU-Z Bench single-core test, which shows an impressive 9.41 percent increase over that of the i9-12900K.
The i9-13900K packs "Raptor Cove" performance cores, which Intel claims come with a generational IPC increase over the "Golden Cove" P-cores. The 9.4% performance increase could be a result of not just increased IPC, but also higher clock speeds (set at 5.50 GHz, the assumed maximum boost frequency of the retail processor). The multi-threaded CPU-Z Bench sees an incredible 46.34% performance increase. This stems from not just increased performance on the eight P-cores, but also the doubling in E-cores from 8 to 16. The E-core clusters also see a doubling in L2 cache sizes. The story repeats with Cinebench R23, with an incredible 13.53% single-thread performance increase, and a 40.25% multi-threaded performance increase.
Sources:
Extreme Player (bilibili), VideoCardz
The i9-13900K packs "Raptor Cove" performance cores, which Intel claims come with a generational IPC increase over the "Golden Cove" P-cores. The 9.4% performance increase could be a result of not just increased IPC, but also higher clock speeds (set at 5.50 GHz, the assumed maximum boost frequency of the retail processor). The multi-threaded CPU-Z Bench sees an incredible 46.34% performance increase. This stems from not just increased performance on the eight P-cores, but also the doubling in E-cores from 8 to 16. The E-core clusters also see a doubling in L2 cache sizes. The story repeats with Cinebench R23, with an incredible 13.53% single-thread performance increase, and a 40.25% multi-threaded performance increase.
25 Comments on Intel "Raptor Lake" ES Posts 9.4% Higher Single-Core Performance Than "Alder Lake"
The whole thing is not helped by this obsession with FPS and rubbish technologies like Ray Tracing pushed by some philistines. Since when we have to play at ultra, or dropping below 60fps a sacrilege.
Lest see how good Raptor overclclocks
Single thread Raptor @ 5.5Ghz is 892.2
My 5.6Ghz 12900k is 914.9
5.5GHz is a 5.6% increase in clock speed, ergo 5.6% higher performance from clock speed increase already.
The videocardz.com source lists a top boost of 5.7GHz, which works out to a 9.6% clock speed improvement.
Also to note, the 13900K (ES) hit a PL4 of 420W in the testing.
Allowing the processor to suck more power should allow it to boost higher, and more often, increasing absolute performance, but not IPC.
Indeed, look a the CPU-Z single thread testing.
At ISO clock speeds, they show a 0.25-0.3% improvement.
Now, to me, these tests are all worth a grain of salt as they are pre-release testing on ES chips.
I reserver all judgments for after the official release and multiple independent reviews.
Also if someone asked me now who is going to have the gaming performance crown (even if it was only 1% difference) taking account the IPC & frequency differences for Zen4 and the +14% in ST Cinebench R20 that 13900KF enjoys, I would say that chances are 13900KS to be on top vs 7950X!
Side question, where does your 5.6% and 9.6% clock speed percentages increase come from?
5.1GHz -> 5.5GHz = 7.84% more
5.1Ghz -> 5.7GHz = 11.76% more
WoW
If there was a problem it was pricing on most models because AMD asked for example $300 for a 5600X and Intel was asking $180 for a 12400F...
Nvidia's reference full AD102 GPU it seems will be below 600W according to AMD's prediction (maybe after discussions with partners of what TDP limits AMD can possibly target based on upcoming predetermined competition's designs) Also AMD is forecasting 700W for next-next gen GPUs in 2024 (let's hope 700W is for Nvidia otherwise except Deja Vu it will be mea culpa!
If 5.5Ghz is max, that means overclockers can get 6 Ghz.
And if Intel SR delay again in 2023, Intel will be like 2 generations behind in server products.
The sad thing, you have some fans of those so called reviewers, who go to great lengths to defend their favourite youtuber.
It might be all talk but AMD is still confident they will have the MT crown with Zen 4 and with clocks apparently able to hit single core 5.85GHz along with ~10% IPC uplift, other major architectural changes, ST will also be a lot stronger than Zen 3. And then we have Zen 4 with v-cache in H1 2023.
It would be too much to buy RL on a socket that is EOL. AM5 will be around until Zen 6 most likely. Trouble for Intel Meteor Lake is delayed until late 2023. So better to put RL in AL MB's.
I'm running my 12700k HT-less at 4.8 Ghz P, 4.1 Ring, 3.7 E, runs cool and quiet just fine at 1.16v, and performs more than "good enough" with a good RAM overclock.
Modern GPU or CPU too hot? As long as you're using a good PSU, then you should be able to underclock it a little (4-9%) and cut down on a lot of voltage - suddenly it draws 25-60% less power while delivering "more than good enough" performance.
Wanna go maximum fast? Sure, it does that too - but don't forget that it goes Cool & Quiet too :)
Single digit to 20%. No mention of what will get the highest increase. My guess would be server chips as the BOM is the highest there.
Wasn't competition meant to lower prices? So in addition to rising power consumption the hardware itself will also cost more. Plus better cooling needed to deal with the heat that further increases costs. Undervolting is great but most people run stock and don't overclock or undervolt.