News Intel's Core i7-14700K Benchmarked: More Cores, Higher Clocks

[Admin]

Intel's Core i7-14700K packs 8P and 12E cores, beats predecessor in all tests.

Intel's Core i7-14700K Benchmarked: More Cores, Higher Clocks : Read more

 

[newtechldtech]

So nothing is new here , just rename i9 and call it a new Gen.

 

[TerryLaze]

Intel's Core i7-14700K packs 8P and 12E cores, beats predecessor in all tests.

Intel's Core i7-14700K Benchmarked: More Cores, Higher Clocks : Read more

While the reviewer does not compare the Core i7-14700K to the current flagship Core i9-13900K, it has the same configuration of cores (eight high-performance and 12 energy-efficient cores) and very close frequencies (the Core i9-13900K has 200 higher boost clock of 5.80 GHz). Therefore, performance of the new Core i7-14700K will be very close to that of the Core i9-13900K.

Intel® Core™ i9-13900K Processor (36M Cache, up to 5.80 GHz) - Product Specifications | Intel

Intel® Core™ i9-13900K Processor (36M Cache, up to 5.80 GHz) quick reference with specifications, features, and technologies.

www.intel.com

The 13900k has 8+16 cores for a total of 32 threads, not 8+12.

So nothing is new here , just rename i9 and call it a new Gen.

This is the wrong topic to make this comment, this topic is about the 14700 which will get 4 more e-cores increasing its multi performance by quite a bit, depending on if the price changes or not this will be a big bump for new builders.

 

[NightLight]

So nothing is new here , just rename i9 and call it a new Gen.

Tell me you're a fanboy without telling me you're a fanboy

 

[peterf28]

Will this work with B760 DDR4 mobo? Or will the motherboard be a bottleneck?

Will it be worth upgrading from a 13500? Not now, but maybe in a year or two.

 

[ManDaddio]

Will this work with B760 DDR4 mobo? Or will the motherboard be a bottleneck?

Will it be worth upgrading from a 13500? Not now, but maybe in a year or two.

That's quite a jump you're talking about. If these chips are still available and they fit in your board then sure it would be a fantastic upgrade. But in two years you probably would be interested in whatever is out then.

 

[ManDaddio]

Will this work with B760 DDR4 mobo? Or will the motherboard be a bottleneck?

Will it be worth upgrading from a 13500? Not now, but maybe in a year or two.

Here's an article I found which might help get some information you're looking for: https://wccftech.com/intel-core-i7-14700k-cpu-with-20-cores-spotted-6-3-ghz-msi-z690-motherboard/

 

[palladin9479]

Hmm I think they are going a little to nuts with the crappy "e-cores". Like having four of those around for web browsing / watching media or just generally putzing around the desktop is fine. Stuffing them in there to inflate MC benchmarks is kinda dumb. Like nobody, and I mean nobody, is going to actually be using 12 e-cores, heck 8 is pushing it.

 

[Giroro]

So the 14700k performs worse than the 13900k, has lower clocks, and consumes more power.

...So in other words, it's either just a really low bin 13900k reject, or Intel's "refresh" has changes which actually makes their architecture worse in every way. Or the benchmark is bad.

I'll just wait for a real benchmark from somebody more trustworthy.

 

[ilukey77]

yeah i was wondering myself if this 14700k would work in a B760 as i was looking at building a all intel build ( really just for fun nothing more ) ive got the ARC770 but was wondering and looking at a B760 ddr5 board with a 12th or 13th to start with the possibility of the 14700k at some point

 

[LabRat 891]

Please do a little side review:
i7-4770k vs. i7-14700k: What we've gained in 10 generations.

 

[TerryLaze]

Hmm I think they are going a little to nuts with the crappy "e-cores". Like having four of those around for web browsing / watching media or just generally putzing around the desktop is fine. Stuffing them in there to inflate MC benchmarks is kinda dumb. Like nobody, and I mean nobody, is going to actually be using 12 e-cores, heck 8 is pushing it.

The dumb part here are the reviewers, intel isn't doing this to inflate MC benches or to raise your power bill by providing even more cores for you to overclock.

The point of them is to be able to do a lot of work in the background while still having your normal 8 core CPU running at full speed for whatever you are doing in the foreground.
If you only need 4 e-cores you also don't need 8 p-cores.
Also if you use any program that uses many cores the e-cores are going to be used automatically, nobody has to do anything specific to use 8 or even 12 e-cores, they are being used by automatically.

View: https://www.youtube.com/watch?v=RDtJ5h0NjvQ

 

[newtechldtech]

Intel® Core™ i9-13900K Processor (36M Cache, up to 5.80 GHz) - Product Specifications | Intel

Intel® Core™ i9-13900K Processor (36M Cache, up to 5.80 GHz) quick reference with specifications, features, and technologies.

www.intel.com

The 13900k has 8+16 cores for a total of 32 threads, not 8+12.

This is the wrong topic to make this comment, this topic is about the 14700 which will get 4 more e-cores increasing its multi performance by quite a bit, depending on if the price changes or not this will be a big bump for new builders.

no it is not te wrong topic ... and the 14700 gets nothing it is actually the old i9 renamed i7

 

[in_the_loop]

no it is not te wrong topic ... and the 14700 gets nothing it is actually the old i9 renamed i7

It's not the old i9. The old i9 had 16 e-cores. The 14700 has only 12.

 

[TerryLaze]

no it is not te wrong topic ... and the 14700 gets nothing it is actually the old i9 renamed i7

The old i9, if you mean the 12900, has 8+8 the 13900 has 8+16 and the 14700 will have 8+12.
The 13700 was the old 12900, if core counts is all you go by.

 

[palladin9479]

The dumb part here are the reviewers, intel isn't doing this to inflate MC benches or to raise your power bill by providing even more cores for you to overclock.

The point of them is to be able to do a lot of work in the background while still having your normal 8 core CPU running at full speed for whatever you are doing in the foreground.
If you only need 4 e-cores you also don't need 8 p-cores.
Also if you use any program that uses many cores the e-cores are going to be used automatically, nobody has to do anything specific to use 8 or even 12 e-cores, they are being used by automatically.

View: https://www.youtube.com/watch?v=RDtJ5h0NjvQ

Again upping the e-core count to 12 is beyond silly, it's just there to cheaply inflate MC benchmarks. All the background tasks on your desktop will happily run on just 4 e-cores, the load is that light. The only remaining use cases are better solved by GPUs.

Paying for silicon that you will never use is... well I guess Apple doesn't have the market cornered on loyalists.

 

[TerryLaze]

Again upping the e-core count to 12 is beyond silly, it's just there to cheaply inflate MC benchmarks. All the background tasks on your desktop will happily run on just 4 e-cores, the load is that light. The only remaining use cases are better solved by GPUs.

Paying for silicon that you will never use is... well I guess Apple doesn't have the market cornered on loyalists.

Video was linked for a reason....
You can also look for other videos yourself that explain why intel has the e-cores but this one is short and to the point.

Also as I said before, if you are never going to use that much silicon then go for fewer cores in general, get an i5 or even i3.

 

[Elusive Ruse]

Underwhelming to say the least.

 

[palladin9479]

Video was linked for a reason....
You can also look for other videos yourself that explain why intel has the e-cores but this one is short and to the point.

Also as I said before, if you are never going to use that much silicon then go for fewer cores in general, get an i5 or even i3.

Appealing to authority of a video is kinda... yeah ok.

General purpose computing is largely limited by single threaded workloads, usually the main decision tree of a program. We spin auxiliary tasks up on other threads in the hope of moving as much work as possible away from whichever processing resource is running that decision tree. This is why we focus on things like IPC and clock speed, it directly translates into that main decision tree running faster. Everything else in a single program is a much smaller workload. MT is important when doing stuff like video editing since those extra processing resources can be used, but then GPU's are orders of magnitude much better at processing video then CPUs.

The last possible scenario is a gamer twitch streaming, which the 8/4 configuring is way more then enough to handle seemlessly.

 

[TerryLaze]

Appealing to authority of a video is kinda... yeah ok.

It a senior engineer of intel explaining why the e-cores are there, what higher authority do you want?

General purpose computing is largely limited by single threaded workloads, usually the main decision tree of a program. We spin auxiliary tasks up on other threads in the hope of moving as much work as possible away from whichever processing resource is running that decision tree. This is why we focus on things like IPC and clock speed, it directly translates into that main decision tree running faster. Everything else in a single program is a much smaller workload. MT is important when doing stuff like video editing since those extra processing resources can be used, but then GPU's are orders of magnitude much better at processing video then CPUs.

And if that is all you need from a PC you can go for a dual or quad core with htt, the main cores for your main workload and the htt threads for the extra/background work.

The last possible scenario is a gamer twitch streaming, which the 8/4 configuring is way more then enough to handle seemlessly.

It's not your decision to make what's enough for other people.
If you only need a 8/4 CPU that's fine but don't tell professionals that they have the luxury of waiting for one task to finish to start work on a different one.
Or the luxury of having to wait for one task to finish on only 4 ecores while they do something else.
This way they can have a long lasting task keep running in the background while they notice zero slow down for doing the rest of their work.
The more ecores they have the lower the time impact for the long lasting task.

Intel 13th Gen Desktop CPUs: Everything You Need to Know

Here's everything you need to know about the new Intel 13th-Gen Raptor Lake desktop processors, including the flagship Core i9-13900K.

beebom.com

 

[bit_user]

General purpose computing is largely limited by single threaded workloads, usually the main decision tree of a program. We spin auxiliary tasks up on other threads in the hope of moving as much work as possible away from whichever processing resource is running that decision tree. This is why we focus on things like IPC and clock speed, it directly translates into that main decision tree running faster. Everything else in a single program is a much smaller workload. MT is important when doing stuff like video editing

Web browsers are like bread and butter, for modern computing. Would you expect them to be mostly serial or to parallelize well?

I had an interesting experience, where the fan on my work laptop died (CPU: i5-1250P). I got no alert or warning, when it happened. I was just running Win 10 and noticed that everything got exceedingly slow. I checked Task Manager and saw that only 1 core was active and it was running at 0.39 GHz.

The first thing I did was a web search of the symptom, and someone suggested disabling the intelppm service, so I tried that and rebooted. The result was interesting, to say the least. The CPU now ran tasks on all 12 hardware threads (CPU is 4P + 8E), but Task Manager still showed the clock speed at 0.39 GHz. However, a web browser pointed at a Toms Hardware article went from completely unusable to being actually usable, but still not great.

This makes a lot of sense, if you think about it. Modern browsers run on both phones and PCs, and I'm sure the core code is all the same. Leaving aside questions about efficiency, phones went multi-core and scaled up core counts pretty aggressively. If you wanted to have good responsiveness on phones, you'd therefore do well to heavily multi-thread your browser. You can find some browser benchmarks which confirm they experience good multicore scaling.

Getting back to my story, it didn't take me long after that to figure out the real cause of my performance problems. Running the laptop's built-in diags confirmed the fan had stopped running and it was repaired under warranty.

 

[bit_user]

It a senior engineer of intel explaining why the e-cores are there, what higher authority do you want?

No, he's a Senior Technical Marketing Engineer. At a tradeshow, full of journalists. With his boss standing just off-camera. So, he's going to be like 1000000% on-message.

"Technical Marketing" is dealing mostly with OEMs, partner companies, and other industry players. The non-technical marketing folks are the ones dealing with consumers, branding, etc. A technical marketing person will be more comfortable getting into the nitty gritty of the product specs and details, but make no mistake: you're still talking to a marketing person, who will put the most positive spin on everything.

A higher authority would be one of the CPU architects, either interviewed anonymously or long enough after leaving Intel that they're no longer covered by NDA. That's when you get the real talk.

I think what he showed is mostly a gimmick and speaks poorly of Windows 11's thread scheduler, more than anything. A better behavior would be to simply de-prioritize background tasks, so that they can still benefit from the horsepower in the P-cores to finish sooner, while not impeding system responsiveness or bogging down higher-priority interactive tasks.

Let's put it this way: if I were doing CPU rendering, I'd be really annoyed if I had to keep my render window in the foreground, just to have it utilize the P-cores!

Using the raw data collected by Chips & Cheese, here are some performance projections I made, for how optimal clock management of P- vs. E-cores can boost performance at any power target, on a sufficiently well-threaded workload:

More details, here: https://forums.tomshardware.com/thr...14700k-cpus-benchmarked.3816188/post-23070245
​

More performance per Watt? That's the very definition of efficiency! So, the E-cores indeed do increase efficiency! To @palladin9479 's point, you need a workload with enough parallelism (or something like a laptop, with low power limits) to get the most benefit from them.

 

[palladin9479]

Web browsers are like bread and butter, for modern computing. Would you expect them to be mostly serial or to parallelize well?

Depends on the browser, anything chromium based will have each tab a different thread while firefox is a little different. Are you browsing multiple tabs simultaneously? Also browsers spend more time waiting for input then they do rendering a page, thus the entire user desktop experience can run on 4 E cores without issue. It's only when we load heavy programs like Games or Adobe Photoshop / Premier / etc.. that system resources start ramping up.

This isn't theoretical, it's based on metrics we pull from our company users Virtual Desktop Infrastructure (VDI) instances. Most endpoint devices are now virtualized with the users using WYSE terminals to login, this allows us to pull very detailed information on system utilization and we found that desktop programs use very little resources. Turns out the real resource eaters were the multiple security agents running on each desktop fighting with each other over who gets to scan something first.

So yeah, 8 P and 4 E cores is waay more the enough for all things desktop and gaming. The reason Intel bumped the numbers up is that E cores are super cheap silicon wise and they needed a way to distinguish the 14th gen from previous generations since they are still on the same milked out 10nm process. There won't be any real improvements until they shift to a new 5 or 7nm process node, until then expect silly stuff like this to try to justify people buying them.

I think what he showed is mostly a gimmick and speaks poorly of Windows 11's thread scheduler, more than anything. A better behavior would be to simply de-prioritize background tasks, so that they can still benefit from the horsepower in the P-cores to finish sooner, while not impeding system responsiveness or bogging down higher-priority interactive tasks.

Use Process Lasso, it's infinitely better then either Windows 10 or 11's thread scheduler. You can define profile rules that ensure anything with a specific file name runs on the cores you specify. At this very moment I have rules for anything vivaldi*, steam*, discord* or firefox* to be run on the E-cores. Doing this my normal resource utilization is 2~5%. The "heaviest" process is actually dwm.exe, which is the windows process for managing the desktop graphics and I let that run on whatever Windows wants it to run on.

More performance per Watt? That's the very definition of efficiency! So, the E-cores indeed do increase efficiency! To @palladin9479 's point, you need a workload with enough parallelism (or something like a laptop, with low power limits) to get the most benefit from them.

E-cores are basically what you would find on a tablet or laptop processor, they are super efficient but have a hard performance ceiling. We've discussed this before, about how x86 processors have all this extra silicon dedicated to branch prediction, instruction translation, cache and so forth. These resources are required to properly feed a powerful generic purpose processor lots of instructions at a high speed, but are extremely power expensive. If we strip most of that out and just do basic processing we can get x86 to act a whole lot like ARM, very power efficient with a much lower performance ceiling.

 

[TerryLaze]

Let's put it this way: if I were doing CPU rendering, I'd be really annoyed if I had to keep my render window in the foreground, just to have it utilize the P-cores!

You gonna be mad again for saying it but you again didn't even look at the picture or the video.
If you are doing CPU rendering but have to keep your normal workflow running alongside then even if you manage priorities, which nobody would do for every single time you change from one app to the other, you would lose either performance or respositivity.

This is about putting your CPU rendering in the background ON PURPOSE so it will only run on the e-cores so that you can continue doing your game development or video editing or whatever it is you are doing in the foreground, the CPU rendering will take a while longer but won't cut down on your workflow.

You choose your poison. Either hybrid or you have two systems one for working and a render box or your nerves are going to die trying to play thread director instead of focusing on your work.

More performance per Watt? That's the very definition of efficiency! So, the E-cores indeed do increase efficiency!

But only if there is limited space and/or a limit on acceptable cost.
You only get more performance per watt with the e-cores if you reduce the overall number of cores, going from a total of 16 to 12 or 10.
Do your math again comparing 4 e-cores to 6 and 8 p-cores...

You are saying it yourself, (power) efficiency is more performance per watt and not more performance per space, that is called space efficiency.

 

[bit_user]

Depends on the browser, anything chromium based will have each tab a different thread while firefox is a little different. Are you browsing multiple tabs simultaneously?

No, the only page I had open was Toms. Pretty much the first thing I did, when performance tanked, was to close the rest.

Also browsers spend more time waiting for input then they do rendering a page,

So many ads, videos, and scripts running in the background, though.

we found that desktop programs use very little resources.

That's what I thought, but then I had to edit an Excel worksheet with a couple thousand rows and was amazed at how it bogged down. No fancy formulas, even. Of course, it's only bogged when I'm doing something in it.

Same for MS Word. I haven't edited huge docs recently, but when you get into editing multi-hundred page technical documents, it can get slow at points.

Turns out the real resource eaters were the multiple security agents running on each desktop fighting with each other over who gets to scan something first.

For sure, I've noticed this on my work laptops. The difference in responsiveness and background CPU load between them and my personal Windows box is huge.

E-cores are basically what you would find on a tablet or laptop processor, they are super efficient but have a hard performance ceiling. We've discussed this before, about how x86 processors have all this extra silicon dedicated to branch prediction, instruction translation, cache and so forth. These resources are required to properly feed a powerful generic purpose processor lots of instructions at a high speed, but are extremely power expensive. If we strip most of that out and just do basic processing we can get x86 to act a whole lot like ARM, very power efficient with a much lower performance ceiling.

Yes, but... don't be too dismissive of Gracemont. Chips & Cheese did a great deep dive of it, and it's a lot closer to a P-core than you might expect.

​

IPC, for different instruction classes, is comparable or better than Skylake. Near Golden Cove, in a couple of cases:

Test	Gracemont IPC	Golden Cove IPC	Zen 3 IPC	Zen 2 IPC	Skylake IPC
Dependent Register to Register MOVs	4.56	5.62	5.72	4.55	1.65
Independent Register to Register MOVs	4.55	5.68	5.7	4.54	3.81
Zero Register using XOR	3.83	5.73	5.72	3.63	3.81
Zero Register using MOV r,0	3.34	5.64	3.81	3.64	3.82
Subtract Register from Itself	3.83	5.73	5.7	3.64	3.81
Dependent INC (increment by 1)	0.97	5.72	1.00	1.00	1.00
Dependent DEC (decrement by 1)	0.97	5.55	1.00	1.00	1.00
Dependent ADD, immediate value	1.00	5.61	1.00	1.00	1.00

Digging a bit further, this shows how the size of many of its structures compares. In some cases, it's deeper than even Zen 3!

Structure	Instruction needs entry if it...	Gracemont Capacity	Golden Cove Capacity	Zen 3 Capacity	Zen 2 Capacity
Reorder Buffer	exists	256	512	256	224
Integer Register File	writes to an integer register	214 (198+16)	280 (~248+32)	192 (173+32?)	180 (132+32?)
Flags Register File	sets flags (applies to many x86 ALU instructions)	214 (198+16)	280 (~248+32)	121	180 (132+32?)
FP/Vector Register File	writes to a fp/vector register	207 (191+16)	332 (300+32)	160 (139 measured+32?)	160 (143 measured + 32?)
256-bit FP/Vec Register File	writes a 256-bit fp/vector register	95+16?	332 (300+32)	160 (139 measured+32?)	160 (143 measured + 32?)
Load Queue	reads memory	80	192	116	116
Store Queue	writes to memory	50	114	64	48
Branch Order Buffer	affects control flow	116 Taken 126 Not Taken	128	48 Taken 117 Not Taken	32 Taken 128 Not Taken
MXCSR Register File	Changes FP unit configuration	1 (not renamed, always stalls)	8	9	9
Scheduler	Is waiting on an execution unit	221*	205	160	128
Scheduler + NSQs	Is waiting on an execution unit	299*	205	224	192

If we talk specifically about instruction scheduling, here's how they compare:

​

The article covers a lot more, so definitely check it out if you're interested:

Gracemont: Revenge of the Atom Cores

This article can be considered a part 2 to our Golden Cove article because today we are looking at the other core in Alder Lake, Gracemont.

chipsandcheese.com