Thursday, September 14th 2023

Intel Core i9-14900KF Geekbenched in v6.2

OneRaichu has conducted a series of Geekbench 6.2 tests on an Intel Core i9-14900KF CPU, very likely a preview sample—his results have arrived for public viewing in the form of three new database entries. The hardware enthusiast (and sometimes leaker) is expected to produce a full review of said flagship Raptor Lake Refresh processor. His evaluation arrives roughly a week after leaked Intel Core i9-14900K processors appearing online, via benchmark results produced in Geekbench 6.1 and CPU-Z. The KF variant is missing an integrated GPU, while its K sibling is likely endowed with a bog standard Intel UHD Graphics 700-series iGPU.

The database entries reveal single-core scores starting at 3322 and going up to 3347 points. Multi-core scores span from 22895 through to 23051 points. A Geekbench 5 result is thrown in for good measure, with achievements of 2412 points in single-core, and 26972 points in multi-core performance. OneRaichu's test build utilized an ASRock Z790 Taichi motherboard and 32 GB of DDR5-7000 memory, his OS of choice appears to be Microsoft Windows 11 (non-Pro) 64-bit. VideoCardz has crunched the numbers: "preliminary benchmarks suggest that the Core i9-14900KF outperforms the i9-13900K by approximately 5-6% in multi-threaded tests and a minimum 12% improvement in single-core performance."
Their evaluation continues: "However, it's crucial to bear in mind that we lack detailed information regarding the test settings employed by OneRaichu during these evaluations, so these scores may not provide a complete and accurate representation of the processor's overall capabilities."

VideoCardz has kindly produced some comparison charts based on the latest Geekbench data:
Sources: Geekbench #1, Geekbench #2, Geekbench #3, VideoCardz
Add your own comment

13 Comments on Intel Core i9-14900KF Geekbenched in v6.2

#1
AnarchoPrimitiv
What would lead to a situation in which single core is 12% improved, but multicore is only 5%? I'm guessing power and heat limitations throttling the CPU on multicore workloads?
Posted on Reply
#2
AnotherReader
Geekbench is very variable. Scores for the 13900k range from 2842 to 3419. Let's wait for reputable reviewers to get their hands on it. A 12% increase in single threaded performance seems rather unlikely.
Posted on Reply
#3
Wales
none-GPU Die CPU have stronger turbo?!!
Posted on Reply
#5
Chrispy_
AnotherReaderGeekbench is very variable. Scores for the 13900k range from 2842 to 3419. Let's wait for reputable reviewers to get their hands on it. A 12% increase in single threaded performance seems rather unlikely.
Not if they clock the everliving sh*t out of it! :)

253W power budget on a TSMC 4nm is a very different beast to a 253W power budget on Intel 10nm 'Intel 7'. At stock, non-cheaty power limits, the 13900K is barely ever clocking to 5.8GHz, 5.5GHz is far more realistic but I wouldn't be surprised if TSMC 4nm can sustain 6GHz on several P-cores without turning the socket into hot slag.

That's a 9% clock improvement before any IPC gains are considered....
All I really care about is performance at ~100W. Very few people really need 250W CPUs
Posted on Reply
#6
chrcoluk
I think easily achievable if they have implemented the new IVR.
Posted on Reply
#7
AnotherReader
Chrispy_Not if they clock the everliving sh*t out of it! :)

253W power budget on a TSMC 4nm is a very different beast to a 253W power budget on Intel 10nm 'Intel 7'. At stock, non-cheaty power limits, the 13900K is barely ever clocking to 5.8GHz, 5.5GHz is far more realistic but I wouldn't be surprised if TSMC 4nm can sustain 6GHz on several P-cores without turning the socket into hot slag.

That's a 9% clock improvement before any IPC gains are considered....
All I really care about is performance at ~100W. Very few people really need 250W CPUs
I thought the 14900k was built using Intel 7, not TSMC's N4.
Posted on Reply
#8
Chrispy_
AnotherReaderI thought the 14900k was built using Intel 7, not TSMC's N4.
Possibly. I can't find anything conclusive, Meteor Lake is definitely TSMC, but if these are just Raptor-lake rebrands on Intel 7, where is the 12% performance gain coming from?

The official announcement and concrete specs are due at the Intel event on the 19th Sept, so not long to wait for answers.
Posted on Reply
#9
AnotherReader
Chrispy_Possibly. I can't find anything conclusive, Meteor Lake is definitely TSMC, but if these are just Raptor-lake rebrands on Intel 7, where is the 12% performance gain coming from?
I'm skeptical of the 12% performance estimate. The leaker didn't claim it; rather, it was Videocardz that estimated it, but 13900k scores are all over the place so we don't know if this 12% estimate is accurate. Besides, Meteor Lake will be using Intel 4 for the CPU chiplet.
Posted on Reply
#10
Chrispy_
chrcolukI think easily achievable if they have implemented the new IVR.
Not sure I follow, that's just going to provide better overclocking headroom at stratospheric DDR5 clocks with eye-watering espensive DDR5-8000+ kits, no?
Posted on Reply
#11
AnotherReader
Chrispy_Not sure I follow, that's just going to provide better overclocking headroom at stratospheric DDR5 clocks with eye-watering espensive DDR5-8000+ kits, no?
I suspect @chrcoluk is talking about Raptor's Lake disabled DLVR. However, the DLVR only saved power in low load situations with CPU currents of 40 Amps or lower. After 70 Amps, the benefits are almost non-existent according to Intel's patent. Assuming voltages of 1.4 V, that is a power draw of 98W which is much lower than the 14900K will sustain.

Posted on Reply
#12
chrcoluk
Chrispy_Not sure I follow, that's just going to provide better overclocking headroom at stratospheric DDR5 clocks with eye-watering espensive DDR5-8000+ kits, no?
It should reduce voltages across the v/f curve which means more performance within the TDP limit via less throttling. (Turns out this may not actually be the case, see the reply above mine, so wont up play this feature anymore).

For some reason people seem to think binning quality and voltages only apply if you want to o/c, its really important for stock performance as well, especially on these power hungry pre overclocked chips.

However I would expect voltage efficiency to boost multi core performance more than single core as single core load's are probably not hitting TDP.
AnotherReaderI suspect @chrcoluk is talking about Raptor's Lake disabled DLVR. However, the DLVR only saved power in low load situations with CPU currents of 40 Amps or lower. After 70 Amps, the benefits are almost non-existent according to Intel's patent. Assuming voltages of 1.4 V, that is a power draw of 98W which is much lower than the 14900K will sustain.

Thanks, wasnt aware of the specific design limitations of it.
Posted on Reply
#13
AnotherReader
chrcolukIt should reduce voltages across the v/f curve which means more performance within the TDP limit via less throttling. (Turns out this may not actually be the case, see the reply above mine, so wont up play this feature anymore).

For some reason people seem to think binning quality and voltages only apply if you want to o/c, its really important for stock performance as well, especially on these power hungry pre overclocked chips.

However I would expect voltage efficiency to boost multi core performance more than single core as single core load's are probably not hitting TDP.



Thanks, wasnt aware of the specific design limitations of it.
It still makes sense for laptops. If the 14th generation has enabled the DLVR, then it would be significantly better for laptops than the 13th generation.
Posted on Reply
Add your own comment
May 8th, 2024 16:49 EDT change timezone

New Forum Posts

Popular Reviews

Controversial News Posts