Intel's Core I7-14700K Benchmarked: More Cores, Higher Clocks | Hacker News Reader

Intel's Core I7-14700K Benchmarked: More Cores, Higher Clocks

tomshardware.com

27 pointsbelltaco45 comments

The laptop I got handed at work was surprisingly powerful. i7 13xxxk, i forget the exact model, but its spec'd at up to 5ghz.

Then i logged into MS Teams, Outlook in the browser...slow as ever. Probably just as 'fast' as if I opened them on my i5 3rd gen which is 10 years old at this point

Reminds me of a rant Jon Blow went on regarding the M1 [1], it doesn't matter how powerful the CPU is when the software running on it doesn't get any faster

[1] https://m.youtube.com/watch?v=unx3WCWZwUc&pp=ygULam9uIGJsb3c...

A big issue is that many processes are either pinned to a single core, or, dont utilize libraries that make it capable of using more processing power.

Games are stupidly notorious for this. All the GPU power and highest spec multi core CPU in the world wont help you when the bottleneck is the simple fact of the process not doing multiprocessing is in so many apps.

Its not just games either of course, other popular apps do it too.

modern game engines don’t really have that problem anymore.

agreed this video goes over the topic: https://www.youtube.com/watch?v=6dNuR16gJ2A&t=140s

How people with 10th gen feels snappier then the fastest of current gen for various reasons. I've noticed too that just a basic quad core mini pc can "feel" snappier to the eyes and normal usage then a gaming rig with top cpu gpu.

I have the same issue. Doubt its my processor.

Corp laptops have a bunch of things enabled, bitlocker in my experience slows my machine down considerably making it an IO bottleneck.

Meanwhile Android is probably running on 1/10 of the power or less and everything is perfectly snappy, for $200, including most browser stuff. I'm not sure why desktops are still laggy.

I really have never seen a 200$ android phone not borderline unuseable. I have a OnePlus 8 pro and it is very unsatisfying.. and that was a 1400$ phone when it came out

My LG G8 ThinkQ didn't have trouble till just recently when the battery started wearing. Do you run anything special or high performance?

There are some things I avoid, like switching away from Obsidian due to slow loading, but in general stuff is pretty snappy.

I paid $300 for my OnePlus 9 pro, and it's plenty fast.

Maybe you paid this through a plan but it's still a 1k+$ phone, and I bet it's still missing frames and having lags and latencies all over the place

I actually bought it unlocked on eBay. It may be "missing frames" as you say, but I can't tell if it is. It certainly doesn't have lags or latencies all over the place. It works great

Im not sure it counts, but refurbished flagships from Samsung might fit the bill.

Teams, like most current day “apps” is essena browser app in a web browser that pretends to be an app.

O wouldn’t be surprised if it turned out that teams is essentially single threaded, and that some remote code path is hogging the mainloop and/or doing some blocking operations.

No wonder it doesn’t get any faster.

Wow core i4900k and 14900ks editions rumored to have 6.0 ghz and 6.20 ghz clock speeds ! I thought we were had hit a 5 ghz ceiling with current tech. What a time to be alive !

The power draw is going to be insane though!

It feels like the Pentium 4 all over again, except no one is really objecting to skyrocketing power usage, and not just for CPUs... GPUs, RAM, even SSDs!

People only seem to care about power consumption when it’s AMD

Rocket Lake was criticized for it.

But Alder Lake/Ampere and everything after all feels insane and unopposed. Ryzen/Radeon 7000 is kinda turbo happy too, just not as bad.

125W at load isn't that much at all.

6ghz is crazy. We really are living in the future!

Edit: why does TFA say 12W for all three processors?

For the 13900K, its more like 300W over a 69W idle!

https://www.guru3d.com/articles-pages/intel-core-i9-13900k-r...

Yeah, it was a long day yesterday and I must've been exhausted - because my brain saw rows of "125W" in TFA's table, but looking at it today, that column is "PBP", and then there's another column with more meaningful values along the lines of what you've shared.

Now I'm wondering, what's PBP? Processor Base Power, is apparently an Intel-specific term introduced with 12th generation Alder Lake chips, and it's meant to convey the thermal energy which must be dissipated when said CPUs are running at their base frequency.

https://liliputing.com/pbp-vs-tdp-intel-changes-power-consum...

Thank you brucethemoose for helping clarify this for me!

Yeah, its complicated.

I believe the confusion is intentional on Intel's part. They want to suggest typical max power is 125W, when thats not the case at all, especially on modern motherboards that will hold boost clocks indefinitely.

https://www.nytimes.com/2004/05/17/business/technology-intel...

I remember the strain of Intel designing up to the "thermal wall" of an architecture before a refresh. Computer shops would use the same often passive heatsinks with fan shrouds that worked with previous models, and between the north bridge and CPU would dump enough heat into the case to either cause throttling or fry the graphics accelerator after a few weeks or months.

Or your computer would sound like a jet taking off every time it pulled a little load. Ridiculous wattages seem to be a historical sign that they're actively competing with AMD.

I know I've got a picture of a friend cooking cup noodles with egg by setting it on their heatsink while compiling. "Why waste it.JPG"

This generation of Intel CPUs is essentially more E-cores, higher clocks, and perhaps an updated DRAM controller which is guaranteed to work with higher-clocked RAM. I don't find any of this particularly compelling, and would probably wait for the next generation, which is supposed to represent a much more substantial upgrade. I would also like to see more processors offered with just P-cores - I would rather buy a $200 CPU with 8 P-cores than a $250 CPU with 8 P-cores and 12 E-cores (I understand that others may have different needs).

> I would also like to see more processors offered with just P-cores - I would rather buy a $200 CPU with 8 P-cores than a $250 CPU with 8 P-cores and 12 E-cores (I understand that others may have different needs).

On the contrary, I think there's a market for, say, 2 P cores and 24+ E cores.

They are extremely area efficient, and some workloads like compilation or chunked encoding just want more real cores. And this niche is (for the moment) an advantage Intel has over AMD.

Also, Intel already offers P core only SKUs.

I don’t see how P+E/bigLITTLE is giving Intel an edge when Ryzen 7th Gen is still whooping Intel CPUs in the power consumption department.

What are some cases where one would want 2 P cores and 24 E cores?

Compilation or video encoding, as said above.

But why would I want them to be E cores instead of P cores? Battery life?

Yes, I agree that there is a market for this. But I am not part of that market - I really want more beefy cores with lots of AVX units (I'm perfectly content if these beefy cores have low base clocks to conserve power).

I am curious to know what you are doing that is AVX heavy, but doesn't benefit from more cores?

The first thing that came to my mind is llama.cpp, but its actually core count bound for prompt ingestion and mostly RAM speed bound during token generation.

These intel cpus are power-hungry beasts.

Just doing a quick search on benchmark comparisions of the Ryzen 7040 series (a mobile cpu), and it's about 90% the performance of the intel flagship, and roasts it on GPU.

The CPU discussed has no iGPU.

The current 'flagship' (i9-13900k) discussed in comparison does.

And a whopping 253W at max.

Stop embarrassing yourself.

What's embarrassing is the performance/watt of these chips, compared to the competition. This is what happens when a product company gets MBA'd. Hopefully Gelsinger can turn the ship around.

I wonder how fast is enough? I'm using 9 year old Intel Core i7-4790K @ 4.00GHz.

I think this machine will last another 9 no problem?

And that is because we can but bigger workloads in the cloud at more economical prices for those rare times you meed it and most of the time we are thin clients to the web, or running simple UIs. Gaming excepted.

I've got a Core i7-3770K at 4ghz-ish.

As far as performance goes, yeah, totally fine for almost anything including gaming. Maybe not crushing big video editing projects but I don't do that. According to various benchmarks my other machine, a Core i5-13xxxx is "only" about 2x as fast single-threaded. Not a massive difference for CPUs released over 10 years apart.

The 3770K apparently won't be able to run Windows 11 because of lack of TPM support and Windows 10 will be discontinued in a few years. I guess the 4790 will be in the same boat?

Not sure I care. I guess it can still be a gaming machine as long as MS issues critical updates. I don't use these things for work anyway.

yep, 4790 is definitely SOL... re: Win 11 upgrade. I'm happy with Win 10. 95% of my computer usage is Chrome anyway. ;)

That's a shame, because those 3xxx and 4xxx CPUs are quite capable. Of course, they can still live on as Linux machines.

I have an unreasonable level of affection for that old box of mine.

It's depressing to know how many machines of that vintage will simply go to the scrap heap, especially once Windows 10 stops getting security updates.

"they can still live on as Linux machines. I have an unreasonable level of affection for that old box of mine."

So funny you mention that. That's exactly what happened to my older core-i5 machine. It's a media PC in the livingroom running on Ubuntu after I ported the Windows license over to my current machine ;-)

What is Intel's future? I find myself asking this lately. The desktop market is strong but niche, and their datacenter numbers are waning. Perhaps there's a future where an Intel/AMD consortium could take on the dominance of Nvidia in the GPU ML space. It really comes down to creating a viable alternative to CUDA, which maybe they could do together.

I think Intel will be fine. Their far future is probably based around fabbing mid-range chips for other customers, which should be pretty lucrative (with or without Samsung as a competitor). Maybe they'll rekindle their RISC-V research once their ARM partnership winds down, or double down on their GPU line. They have a lot of options right now.

> Perhaps there's a future where an Intel/AMD consortium could take on the dominance of Nvidia in the GPU ML space.

The thing that is cementing Nvidia's dominance is industry hostility. If you could ship a truly cross-platform GPGPU driver to every modern device, CUDA would be a harder sell. Instead, the best you can get are intermediary libraries like ONNX that do their best to cherry-pick and expose common and accelerated operations. CUDA eats it's lunch in featureset and optimization.

To fix this, you have to bring Microsoft and Apple to the table too. It is purely a matter of investment and cooperation, neither of which seems to exist right now.

Isn't Microsoft pretty invested in onnx?

They are. The problem isn't them, it's everyone else; their execution providers are all over the place, with many lacking fundamental features and others with insane compilation requirements.

And when you compare it to CUDA... they're in different leagues, literally. ONNX is solely looking at AI acceleration, not the field of GPGPU compute. If you want synchronized memory pages or generic GPU compute APIs, ONNX is not even part of the discussion.

It's a great effort, and one I'm definitely rooting for. I wouldn't expect it to dethrone Nvidia though, even if someone manages to beat their hardware.

I guess power-draw and clock speed is how we're going to be getting around Moore's law ending for a while... Doesn't feel like the right way forward. Would be cool to do more with less by switching off of x86 optimizations.

Reply on news.ycombinator.com