News Intel Core i9-11900K Spotted at 5.3Ghz, Beating Zen 3 in Single-Core Performance

[Admin]

Intel's Core i9-11900K found running at 5.3Ghz, beating all of AMD's Zen 3 CPUs in single-core performance.

Intel Core i9-11900K Spotted at 5.3Ghz, Beating Zen 3 in Single-Core Performance : Read more

 

[Gomez Addams]

"If any of these benchmarks are representative of the Core i9-11900K's real-world performance (spoiler alert, these benchmarks generally are not),..."

This article should have started with that sentence fragment, although it would have rendered the rest of it moot, it really is. These benchmarks are generally not representative so BFD.

 

[JamesJones44]

I actually don't trust Geekbench 5 results at all. I get wildly different results per run and per OS. Even if you compare GB 4 to GB 5 you see very different results. PassMark is more stable for comparisons IMO.

 

[Makaveli]

"(Strangely, the higher-clocked Ryzen 9 5900X and 5950X scored lower than the 5800X.) "

Single CCD vs Dual CCD design.

There are other benchmarks that show the 5800X being faster than the higher core count models in certain apps.

 

[CerianK]

"Strangely, the higher-clocked Ryzen 9 5900X and 5950X scored lower than the 5800X."
Like Makaveli said, in PassMark, some of the single-thread sub-tests are sensitive to cache latency, and other factors, which degrade when the thread scheduler must span multiple chiplets.

 

[JayNor]

Intel has plentiful 14nm capacity, although I presume they are busy converting some of it to 10nm. Their q4 er slides state Rocket Lake is already shipping. Maybe this is a mistake, since I believe it was previously stated to arrive in March.

Intel is ramping several products at TSMC this year ... Mobileye eyeq5, Habana Gaudi, the eASIC 5G chips, Xe-HPG. Perhaps their schedules are contributing to the tight TSMC capacity.

 

[dtemple]

Takeaway here is that the top-tier LGA1200 11th gen chip outperforms AMD's second-lowest AM4 chip in single-core performance, by 6%. In a synthetic benchmark that Tom's themselves has discounted as being unindicative of real-world performance. Gotcha! So the only leverage Intel has at this point is if they can provide widespread availability before AMD does.

 

[ingtar33]

Geekbench 5 is a very suspect benchmark, with credible claims of built in hardware bias in the past (largely in favor of intel); and probably should never be quoted as a source of anything. That said I don't doubt Intel has binned their rocket-lake chips to extreme levels to create the 11900k so that they can claim they have the fastest chip, I also don't doubt that chip will be impossible to purchase for any money as it's yields at that binning will be too low to actually sell to the public making it a paper chip that a few thousand people will ever get their hands on.

 

[GoatGuy]

Sounds to me like Team Red and Team Blue are neck-and-neck. A few points here, a few points there, really it hardly makes a difference in the 'so-called real world.

However, it is also fair to say, that this really does seem to be a resounding endorsement for unfettered technological competition. Amazing what the boffins of digital madness are doing!!!

 

[TerryLaze]

Their q4 er slides state Rocket Lake is already shipping. Maybe this is a mistake, since I believe it was previously stated to arrive in March.

If they want to have plenty of product on the shelves in march they would have to start shipping long before that, especially now where there can be huge delays due to covid but even without it you have to build up some supply if you don't want to be a laughing stock like every other companies launch this year.

 

[TerryLaze]

That said I don't doubt Intel has binned their rocket-lake chips to extreme levels to create the 11900k so that they can claim they have the fastest chip, I also don't doubt that chip will be impossible to purchase for any money as it's yields at that binning will be too low to actually sell to the public making it a paper chip that a few thousand people will ever get their hands on.

5.3Ghz is what 14nm can do for years now and on 10core CPUs no less, there will be zero issue to get 5.3 on 8cores.
Especially since we are always talking about 5.3Ghz on 1 core.

 

[TerryLaze]

Takeaway here is that the top-tier LGA1200 11th gen chip outperforms AMD's second-lowest AM4 chip in single-core performance, by 6%. In a synthetic benchmark that Tom's themselves has discounted as being unindicative of real-world performance. Gotcha! So the only leverage Intel has at this point is if they can provide widespread availability before AMD does.

With an arch that intel would have released against zen1 and on 10nm on their original schedule and it does well against zen3 and still on 14nm.


Also why are you even trying to make this sound worse than it is.
The second lowest AM4 chip has the best performance of all the ZEN CPUs.
(For this test at least)

If we compare the best of Team Blue to Team Red, the 11900K is 6% faster than the 5800X -- and even less if we compare it to the 11700K.

 

[Darkbreeze]

Hardware detective APISAK spotted Intel's Rocket Lake Core i9-11900K running full blast at 5.3Ghz (single-core) in Geekbench 5 and PassMark. Surprisingly, the chip manages to beat all of AMD's Zen 3 parts in the single-threaded tests.

Yeah, but at what power envelope? 300w? Because the last two gens of Intel CPU have increasingly and rapidly been climbing the mountain when it comes to power consumption, while those Ryzen parts are at least moderately refined by comparison.

 

[ingtar33]

5.3Ghz is what 14nm can do for years now and on 10core CPUs no less, there will be zero issue to get 5.3 on 8cores.
Especially since we are always talking about 5.3Ghz on 1 core.

except the new architecture doesn't quite clock that well (source, buddy at intel). Listen, I'm not going to get into this argument; Intel pulled this same trick with their last top end chip, they clocked it to the moon, overcharged for it, and it was never really available to the public; then they released countless SKUs of the same chip only clocked slower and slower, because in order to get those high clocks they binned the chips so hard they had a pile of rejects which couldn't clock laying around to sell.

Guess what? Rocket Lake already has 4... That's right 4 versions of the 8c16t chip in the pipeline, and they're all identical except for progressively lower clock speeds. This is just like what AMD did with the 9000 series piledrivers. Intel is selling a factory overclocked and super binned chip as a flagship part, a part which will have extremely limited availability at extreme prices, so unavailable that it will nearly be a paper launch. Tacking the K on the chip is the height of cynicism as you'll need to pore LN2 on the chips to get them higher then they came from the factory. to claim these are "overclockable" is hilarious.

 

[Sippincider]

Yeah, but at what power envelope? 300w?

I was going to guess 500 (and only half kidding).

Any generation getting that hot on the desktop will need to be named "Lava Lake". 🌋

 

[spongiemaster]

5.3Ghz is what 14nm can do for years now and on 10core CPUs no less, there will be zero issue to get 5.3 on 8cores.
Especially since we are always talking about 5.3Ghz on 1 core.

With at least 30% more transistors per core, an 8 core Rocket Lake CPU will be larger than a 10 core Comet Lake CPU. The existence of the 10850 (not in the original product stack) indicates the 10900k was already pretty heavily binned. The 11900k is going to be even more heavily binned. Single core boosts may not be affected, but I would expect peak all core clocks to be lower than the 10900k.

 

[Darkbreeze]

I was going to guess 500 (and only half kidding).

Any generation getting that hot on the desktop will need to be named "Lava Lake". 🌋

Pretty sure this is the engineer responsible for the latest architecture.