News Intel's Core i9-14900K Shows Increased Multi-Threaded Performance in Leaked Benchmarks

[Admin]

Intel's Core i9-14900K pulls ahead of Core i9-13900K in Early CPU-Z benchmark.

Intel's Core i9-14900K Shows Increased Multi-Threaded Performance in Leaked Benchmarks : Read more

 

[Makaveli]

how much TDP increase for this performance?

 

[Alvar "Miles" Udell]

Not only does the assumed Core i9-14900K outperform its direct predecessor by almost 10% in single-threaded workloads given its higher default frequency, but it also beats Core i9-13900KS, which formally has the same 6.0 GHz maximum boost clock, by 3.5%.

So, we're looking at a potential very marginal 3-4% increase in performance per clock, while retaining the same core configuration. That's something review sites should slam Intel (and AMD, they did it too) for, not praise them for.

 

[RichardtST]

10%? I mean, why bother? We're still down roughly 30% for the year due to security bugs. Intel would have been far better off spending that effort into making a bigger jump a couple years down the road and building up a nice following on the current gen platform as it gets cheaper and cheaper. Waste of time and money, sadly.

 

[Amdlova]

how much TDP increase for this performance?

Same TDP of 360mm Rad with delta fans.

 

[Amdlova]

10% IPC increase and I think better IMC maybe Run DDR5 8000 out of the box

 

[Brian D Smith]

@RichardSt - If you stand still...you get passed. Give thanks to AMD (?). You can't simply go from point 'a' to point 'd' like some would prefer. You learn upon the way HOW to get to 'd'.

Anyway, the 'name of the game' is has become EFFICIENCY and less heat...not raw numbers and it is going to be MUCH more difficult to do those things as opposed to pumping more and more electricity into a CPU to make it go faster. That's 'dead end' thinking in a world where components have to get crammed together in a confined space.

 

[Lucky_SLS]

13900k - 949 points. Should be 901 as shown in the screenshot.

The performance jump is 8.5% why round it off and say 10%?

That's incorrect.

 

[bit_user]

Intel's Core i9-14900K Shows 10% Increase in Single-Threaded Performance in Leaked Benchmarks​

Wrong!

Performance of Intel's Core i9 in CPU-Z Built-In Benchmark​

	Single Thread	Multi Thread
Core i9-14900K	978	18,118.5
Core i9-13900KS	945.5	17,197
Core i9-13900K	978	17,167.5

Not only does the assumed Core i9-14900K outperform its direct predecessor by almost 10% in single-threaded workloads given its higher default frequency

OMG? Using what kind of math?

Here's the relative speedup between the i9-14900K and the others:

Performance of Intel's Core i9 in CPU-Z Built-In Benchmark​

	Single Thread (vs. i9-14900K)	Multi Thread (vs i9-14900K)
Core i9-14900K	100.00%	100.00%
Core i9-13900KS	96.68%	94.91%
Core i9-13900K	100.00%	94.75%

On single-threaded performance, it managed a 3.4% win vs. the i9-13900KS, which exactly corresponds to the difference you'd expect between 5.8 GHz and 6.0 GHz (presumably, that i9-13900K was overclocked). On multi-threaded, it only manages a bit over 5%, which sounds like what I'd expect from the rumored bump in PL2.

However, where does the author get 10% ??? Seems very sloppy, to me. We need answers!

 

[_Shatta_AD_]

So this is essentially a 13900KS repackaged with probably a better IHS and soldering material to help boost longer and Intel gonna market this as a whole new generation? Intel back to its 14nm ways…

 

[bit_user]

I do not like to be CPU refreshed.

I actually do like refreshes, if it gives us the original product as it was meant to be. I just don't like when they're passed off as an entirely new product, and especially not if they cost more.

I will gladly wait 6-8 months longer with hopefully a more real bang for the buck!

Maybe Intel is hedging their bets on Arrow Lake-S. How would you feel about waiting 13 months, in case they hit some problems and have to delay it that long? As mediocre as the gains from the refresh are, they're better than continuing to sell the Gen 13 product for a full 2 years.

 

[bit_user]

So, we're looking at a potential very marginal 3-4% increase in performance per clock,

I don't think so. For single-threaded, there's no improvement in IPC. The increase we see (compared to the KS) is exactly what's predicted by the clockspeed increase. IPC is everything not accounted for by clockspeed!

For multi-threaded, you'd have to look at the actual clockspeeds to determine any IPC increase. Given they're exactly the same cores and exactly the same cache sizes, I doubt simply improving memory speeds can amount to much multi-threaded IPC increase, but maybe. I chalk up the ~5% multi-core improvement mostly to boosting higher.

For multithreaded, the discrepancy we need to explain is only 1.5%, though. There's not room for a 3-4% IPC increase, anywhere in this data.

 

[M42]

Wrong!


OMG? Using what kind of math?

Here's the relative speedup between the i9-14900K and the others:

​

​

If you use the average CPU-Z 13900K score, which is roughly 890ish, then the performance increase would be close to 10%. I think we should wait and see what reviewers find when it is launched. It may not be 10% as Intel doesn't have much incentive to improve single-thread performance since it already puts AMD's fastest single-thread performance to shame.

978/890 => 1.09887.

 

[ottonis]

If these results are anywhere near of the final production samples, then Intel's 14th gen may not provide any substantial increase in IPC - meaning that Alder Lake was the last major microarchitectural update in quite a long time, only to be followed by the infamous meteor lake architecture, which is due in fall next year.

In other words, there will be quite a window of opportunity for AMD to fill that gap with Zen 5 in 2024 H1.

 

[bit_user]

If you use the average CPU-Z 13900K score, which is roughly 890ish, then the performance increase would be close to 10%.

Okay, weird of the author not to quote that, however.

I think we should wait and see what reviewers find when it is launched.

Oh, for sure. I'm a big fan of waiting for controlled benchmarks on final product, rather than whatever the heck this leak is.

My beef is just that the headline is wildly inconsistent with the data presented in the article. That's unacceptable.

It may not be 10%

It can't be. We know the cores and caches are the same - only the clockspeed and PL2 changed. That means zero single-threaded IPC change. Single-threaded performance should be up to 3.4% faster, and that's only from clock speed.

If these results are anywhere near of the final production samples, then Intel's 14th gen may not provide any substantial increase in IPC - meaning that Alder Lake was the last major microarchitectural update in quite a long time,

But Raptor Lake should've achieved some IPC improvements via larger caches & maybe other small tweaks.

Also, Ice Lake introduced the Sunny Cove microarchitecture, but was only seen on laptops & servers (lack of a desktop version was thought to be due to that 10 nm node's poor frequency scaling).

 

[TerryLaze]

It can't be. We know the cores and caches are the same - only the clockspeed and PL2 changed. That means zero single-threaded IPC change. Single-threaded performance should be up to 3.4% faster, and that's only from clock speed.

It's called Instructions per cycle and not Instructions per X cycles.
The amount of cycles you can muster up, especially with designs that are so complex, is a very big deal.
Cycles is half of the equation of IPC.

Instructions per cycle - Wikipedia

en.wikipedia.org

The calculation of IPC is done through running a set piece of code, calculating the number of machine-level instructions required to complete it, then using high-performance timers to calculate the number of clock cycles required to complete it on the actual hardware. The final result comes from dividing the number of instructions by the number of CPU clock cycles.

It's fine if reviewers test different CPUs at the same clocks to see differences but it's not the same thing.

Also where did you see that PL2 changed?!

 

[bit_user]

Yes, I think we're on the same page on IPC vs. clock speed.

Also where did you see that PL2 changed?!

Some speculation or leak, somewhere - not any great authority.

Are you saying that it definitely didn't? That could be newsworthy, in itself.

 

[TerryLaze]

Yes, I think we're on the same page on IPC vs. clock speed.


Some speculation or leak, somewhere - not any great authority.

Are you saying that it definitely didn't? That could be newsworthy, in itself.

13900K to KS didn't change PL2, that was better binning, now the binning got so much better in general that they can release the KS as a new gen normal K.
If they increase PL2 it's going to be only because the process node got so much better that it can handle much more power without issue.

Intel product specifications

Intel® product specifications, features and compatibility quick reference guide and code name decoder. Compare products including processors, desktop boards, server products and networking products.

ark.intel.com

 

[AgentBirdnest]

Wrong!
[...]
However, where does the author get 10% ??? Seems very sloppy, to me. We need answers!

Good gog... and the original article and title still remains a day after you brought this up. -__- GN needs to do an expose about this site. : P

I really need to stop reading TH. I love the forums, and I think Paul and Jarred do outstanding articles and reviews. But the daily barrage of junk just hurts my brain. Well done, TH - you broke me.

 

[bit_user]

Good gog... and the original article and title still remains a day after you brought this up. -__- GN needs to do an expose about this site. : P

It's a holiday weekend, in the US. So, I can understand if the site doesn't have the usual amount of minders.

BTW, I know some staff members sometimes read the comments, but often not. I wonder if there's a proper way to report factual errors or inconsistencies, in the articles.

I really need to stop reading TH. I love the forums, and I think Paul and Jarred do outstanding articles and reviews. But the daily barrage of junk just hurts my brain. Well done, TH - you broke me.

I feel you. I think it's probably better to just browse the news archives a couple times per week, and only click the articles that really seem important.

Anyway, it was nice chatting with you. If you take leave of us, we'll miss you in the forums.

 

[AgentBirdnest]

Anyway, it was nice chatting with you. If you take leave of us, we'll miss you in the forums.

Oh, I'll stick around on the forums. I love chatting with the lovely and helpful folks here. : )
I just might skip the articles - or do what you said and only check out the important stuff instead of clicking every leak.

 

[JamesJones44]

Performance of Intel's Core i9 in CPU-Z Built-In Benchmark​

Row 0 - Cell 0	Single Thread	Multi Thread
Core i9-14900K	978	18,118.5
Core i9-13900KS	945.5	17,197
Core i9-13900K	978	17,167.5

Maybe I missed it in the article, I'm crazy or the chart is wrong... but according to the chart in the article the 14900K has the same single thread performance as the 13900K and somehow the 13900KS regressed in single thread performance.

I'm also not sure where 10% is coming from, at least based on these numbers ((978 - 945.5) = 32.5 / 945.5 = 0.034 or 3.4%)

3% seems to line up with the slide we saw from one of the MB vendors (ASRock I think?)

 

[bit_user]

3% seems to line up with the slide we saw from one of the MB vendors (ASRock I think?)

You mean this?

Intel Raptor Lake Refresh, Arrow Lake CPU Performance Projections Leaked

Arrow Lake's up to 21% faster than Raptor Lake with an over 2X faster iGPU.

www.tomshardware.com

Yes, 3% is just about the most improvement it shows on single-threaded benchmarks, in that article. However, the SPECrate2017_int (1-copy) does show a range of 3% to 4%.

However, 4% is also the highest improvement shown on any of the multi-threaded benchmarks included. That puts a ~5% improvement on CPU-Z's benchmark in the realm of the plausible, but pretty much rules out 10% as a speedup on even multi-threaded benchmarks.

 

[Lucky_SLS]

14700 is the only highlight with the extra E cores I guess.

They could have tried 3D cache with their 14th gen and priced them accordingly, missed opportunity by intel...

 

[Axell]

Am I missing something obvious here?
I see the 13900K and 14900K tied and the 13900KS beaten by both.

Screenshot

Captured with Lightshot

prnt.sc

Performance of Intel's Core i9 in CPU-Z Built-In Benchmark​

Row 0 - Cell 0	Single Thread	Multi Thread
Core i9-14900K	978	18,118.5
Core i9-13900KS	945.5	17,197
Core i9-13900K	978	17,167.5