Intel's Lion Cove P-Core and Gaming Workloads

https://chipsandcheese.com/p/intels-lion-cove-p-core-and-gaming

by zdw • 1 year ago

260 123 1 year ago

samrus• 1 year ago

122 points and no comments? Is this being botted or something?

Could be. Usually it means the subject is too advanced for the average HN user yet something that they are interested in.

adrian_b2• 1 year ago

Such articles are very interesting for many people, because nowadays all CPU vendors are under-documenting their products.

Most people do not have enough time or knowledge (or money to buy CPU samples that may prove to be not useful) to run extensive sets of benchmarks to discover how the CPUs really work, so they appreciate when others do this and publish their results.

Besides learning useful details about the strengths and weaknesses of the latest Intel big core, which may help in the optimization of a program or in assessing the suitability of an Intel CPU for a certain application, there is not much to comment about it.

HelloNurse• 1 year ago

It is an interesting but particularly non-actionable analysis: only a handful of engineers at Intel are in a position to design an improved Lion Cove, while the main takeaway for the game programmers who care about game workload performance is that nothing performs too badly and therefore general performance improvement techniques like accessing less memory are a good fit for this processor.

whatever10• 1 year ago

I mean what is there to comment. Intel botched another product release. It is just a sad state of affairs.

Nursie• 1 year ago

How so?

Not that I disbelieve, I just wasn't especially picking that up from the article.

FirmwareBurner• 1 year ago

122 points and no comments?

Better no comments than having to trod through the typical FUD or off topic rants that tend to plague Intel and Microsoft topics.

baq• 1 year ago

exactly. I'm very happy to notice there are no 'x86 bad arm good' comments here as of now. a welcome change.

also - or maybe, first and foremost - it's just a very good article.

xxs2• 1 year ago

it's a very good article.

wuming2• 1 year ago

Substack.

onli7• 1 year ago

To see what that means in practice, in my multi generational meta benchmark the 285K lands currently only on rank 12, behind the top Intel processors from the last two generations (i7-13700K and 14700K plus the respective i9) and several AMD processors. https://www.pc-kombo.com/us/benchmark/games/cpu. The 3D cache just helps a lot in games, but the loss against the own predecessor must hurt even more.

vladvasiliu3• 1 year ago

I don't fully follow this, so what has been gained with the new models?

I seem to remember you'd need dedicated industrial cooling for the 14700k. Does the new model at least pump much less power?

onli7• 1 year ago

So, for one in other software the new processors do better. The 285K beats the i9-14900KS by a bit in my app benchmark collection (which is less extensive, but still). And second yes, according to https://www.computerbase.de/artikel/prozessoren/intel-core-u... for example they are less extreme in their energy usage and more efficient in general, albeit not more efficient than the AMD processors.

But it is a valid question.

KronisLV• 1 year ago

I don't fully follow this, so what has been gained with the new models?

There were power efficiency gains, as well as nice productivity improvements for some workloads: https://www.youtube.com/watch?v=vjPXOurg0nU

For gaming, those CPUs were a sidegrade at best. To be honest, it wouldn't have been a big issue, especially for folks upgrading from way older hardware, if only their pricing wasn't so out of line with the value that they provide (look at their GPUs, at least there the MSRP makes the hardware good value).

Aurornis2• 1 year ago

I seem to remember you'd need dedicated industrial cooling for the 14700k.

Those CPUs run hot, but it got exaggerated a lot online. It’s not hard to handle their heat with a good air cooler (even some of the $50 units like the Peerless Assassin) or a run of the mill closed loop water cooler.

There are a lot of old notions in the gaming community that you need to keep CPUs under arbitrary temperature thresholds or that any throttling is bad. Modern CPUs and GPUs run themselves deep into the performance curves and slight throttling is barely noticeable.

blacklion• 1 year ago

Is your benchmark trustworthy?

I see strange discrepancy: my "old" i7-12700K vs i7-13700K:

Games: 170 vs 270 Software: 271 vs 1875 (!!!)

I can believe into 170 vs 270 (though, these two CPUs are not so different!) but 7x difference in software!? Is it believable?

onli7• 1 year ago

I see how that can be misleading. It's globally not a percentage based score.

The benchmark is creating a global order based on which one is faster, including indirect comparisons. But that only takes care of the position, not the rating. That one is based on the percentage between direct neighbor iff there is a direct comparison in the benchmark data, otherwise it's a small .1 increment. So many small percentage increases that don't necessarily match the direct comparion between parts that are not neighbors. Sometimes that works great, sometimes not.

Here the example looks a bit pathological, that difference is further off than I expected when introducing the calculation recently. For the 12700K and 13700K, the direct comparison sees the 12700K at 85% of the 13700k:

https://www.pc-kombo.com/us/benchmark/games/cpu/compare?ids%...

For apps it's 79%:

https://www.pc-kombo.com/us/benchmark/apps/cpu/compare?ids[]...

So yeah, sorry, that part is misleading right now. I'll check the calculation.

But the ingested individual benchmark numbers and the ranking, so the position, is very solid I'd say. With the caveat that ranking position can change with more data.

onli7• 1 year ago

I found a bug in the score calculation that inflated the score, this case is closer now.

jfindley• 1 year ago

If you want people to take your benchmark seriously, you'd need to provide a very great deal more information on how those numbers are generated. "It's complicated, just trust me" isn't a good enough explanation.

If you want people to listen, you need to have a link where you explain what hardware you're using, what settings you're using, what apps/games you're running, what metrics you're using and how you compute your Magical Number.

My already high level of sceptism is compounded by some scarcely-believable results, such as that according to your testing the i9-14900K and i9-13900K have essentially identical performance. Other, more reputable and established sources do not agree with you (to put it mildly).

onli7• 1 year ago

Hey, I do try to make the site as transparent as possible - but admit that the site does not make it obvious. For such a doubt, go into the comparison of the two (https://www.pc-kombo.com/us/benchmark/games/cpu/compare?ids%...) where all benchmarks used that the two processors share are listed. The benchmark bars are clickable and go to the source.

It does get really complicated to address something like that when all comparisons are indirect. Thankfully, that's not the case here.

The 13900K and 14900K in games really have been that close, see https://www.computerbase.de/artikel/prozessoren/intel-core-i... for an example, where the two have a 2% FPS difference.

mort962• 1 year ago

E-Cores are turned off in the BIOS, because setting affinity to P-Cores caused massive stuttering in Call of Duty.

I understand doing this for the purpose of specifically analyzing the P-core microarchitecture in isolation. However this does make the test less interesting for potential customers. I don't think many people would disable E-cores in BIOS if they bought this CPU, so for the purpose of deciding which CPU to buy, it would be more interesting to see results which factor in the potential software/scheduling issues which come from the E-core/P-core split.

This isn't a criticism, just an observation. Real-world gaming results for these CPUs would be worse than what these results show.

pjmlp6• 1 year ago

I think many haven't yet grasped the future is heterogeneous computing, especially, when many desktops are actually laptops nowadays.

Software working poorly in such setup means no effort was made to actually make it perform well in first place.

Games requiring desktop cases looking like a rainbow aquarium with top everything will become a niche, in today's mobile computing world, and with diminishing sales and attention spans, maybe that isn't the way to keep studios going.

jama2112• 1 year ago

Not “no effort to make sure it performs well in the first place”, that isn’t fair. Lots of effort probably went into it performing well, just this case isn’t handled yet and to be fair this only impacts some people currently and there is a chance to update it.

This just reads like “if they haven’t handled this specific case they’ve made no effort at all across the board” which seems extreme.

raverbashing• 1 year ago

Which is funny because the pushback the PS3 got from developers was too much. Maybe it was "too heterogeneous"

But I guess the future was already set.

flohofwoe• 1 year ago

Software working poorly in such setup means no effort was made to actually make it perform well in first place.

How do you optimize for a system architecture that doesn't exist yet?

(e.g. CoD could probably be fixed quite easily, but you can't do that before the hardware where that problem manifests even exists - I'd say it's much more likely that the problem is in the Windows scheduler though)

Games requiring desktop cases looking like a rainbow aquarium with top everything will become a niche

PC gaming has been declared dead ever since the late 90s ;)

gruturo• 1 year ago

There's a giant pile of software - decades worth of it, literally - which was already written and released, much of it now unmaintained and/or closed source, where the effort you cite is not a possibility.

By all means anything released in 2025 doesn't have that excuse - but I can't fault the authors of 15+ years old programs for not having a crystal ball and accounting for something which didn't exist at the time. Intel releasing something which behaves so poorly in that scenario is.... not really 100% fine in my eye. Maybe a big warning sticker on the box (performs poorly with pre-2024 software) would be justified. Thankfully workarounds exist.

P.S. At least I would have expected them to work more closely with OS vendors and ensure their schedulers mitigate the problem, but nope, doesn't look like they did.

dandellion• 1 year ago

People have been making the same argument for more than a decade now, meanwhile PC gaming has only kept growing. If that was the case, why hasn't it happened yet?

StillBored4• 1 year ago

Is it?

No one has come close to solving the problem of optimizing software for multiple heterogeneous CPU's with differing micro-architectures when the scheduler is 'randomly' placing threads. There are various methods in linux/windows/etc which allow runtime selection of optimized paths, but they are all oriented around runtime selections (ex pick a foo variation based on X and keep it) made once, rather than on cpu/context switch.

This means that one either uses a generic target and takes the ~1.5-2.5x perf hit, or one optimizes for a single core type and ignores the perf on the other cores. This is probably at least partially why the gaming IPC numbers are so poor. The games are optimized for a much older device.

So a 1.5-2.5x perf difference is these days could be 5-10 years of performance uplifts being left on the table. Which is why AMD seems to be showing everyone a better path by simply using process optimizations and execution time optimization choices with the same zen cores. This gives them both area, and power optimized cores without having to do a lot of heavyweight scheduler/OS and application optimization. Especially for OS's like Linux which don't have strong foreground/background task signalling from a UI feeding the scheduler.

In other words heterogeneous SMP cores using differing micro-architectures will likely be replaced with more homogeneous looking cores that have better fine grained power and process optimizations. Ex, cores with even wider power/performance ranges/dynamism will win out over the nearly impossible task of adding yet another set of variables to schedulers which are already NP-hard, and adding even finer grained optimized path selection logic which will further damage branch prediction + cache locality the two problems already limiting CPU performance.

smolder4• 1 year ago

You can still buy and build (I do) powerful gaming-capable systems that do not emit rainbow-puke or have transparent panels and other gimmicky stuff like that.

amy214• 1 year ago

desktops are actually laptops nowadays.

With concepts such as "chromebook" (semi-merging with android phones), apple convergence between iPhone and iComputer chipset and code, Samsung DeX (run samsung phone as a desktop), in my opinion the endpoint is:

- Your phone is your desktop is your laptop. It may dock with a keyboard/monitor. This will happen when software kit is integrated between phone and computer (Microsoft has tried this in the past - too early), and the formfactor is sorted. Also a key gating item is phone CPUS matching laptop performance - Also some form of local AI will be present everywhere - Also, sadly, the way things our headed, such a system will be hardcore locked down if it inherits more from the phone ecosystem than the OS ecosystem

giingyui• 1 year ago

It’s a problem of compatibility of those games, not an issue with the processor. The kind of thing a game or windows update solves.

mcraiha• 1 year ago

Old games won't get updates. That is why there are multiple separate tools that try to force the situation. e.g. Process Lasso

xxs2• 1 year ago

Real-world gaming results for these CPUs would be worse than what these results show.

That's mostly an application and/or OS issue, not really a CPU one.

onli7• 1 year ago

Tell that to AMDs bulldozer. There is something to be said for considering theoretical performance, but one can't ignore how hardware works in practice, now.

Havoc2• 1 year ago

Yeah intel either needs to ensure much better day 1 support for E / P core split or drop it entirely.

People see them as an active negative right now rather than how intel pitches them

vladvasiliu3• 1 year ago

Is this a problem with Intel or with the OS scheduler? Haven't they had this kind of CPUs for a few years now, with this being touted as a reason to move from Win10 to 11?

masklinn• 1 year ago

Scheduling is the OS’s decision not the CPU’s.

Although Intel’s configurations are unhelpful to say the least: it’s hard to make good scheduling decisions when Intel sells a 2P, 8E, 2LE as a 12 cores to the user.

zamadatix• 1 year ago

The recommended actions are built around the assumption optimal gaming performance is the only market segment to please. Of course, efficiency cores were never about pleasing that segment anyways - gamers want fewer huge cores, not lots of big cores or tons of efficient cores. That Windows gaming is slightly imperfect due to scheduling difficulties will not stop AMD, Intel, ARM, etc from continuing to use different sized cores on non-uniform interconnects if it means larger changes in perf/watt and perf/cost for that price.

In practice this is more of a problem for gaming news site headlines than users anyways. For all but the biggest enthusiast the uplift from going to a new system is usually so much they'd not notice it could be a couple percentage points better or that one game is still a bit hitchy. It's us ultra-nerds that care about the 1% lows between different top end CPUs, and we tend to be quickly convinced the problem is Microsoft's to solve for us when the scheduling problem hits ARM, AMD, and Intel based Windows devices equally.

smileybarry• 1 year ago

It’s also a relatively “old” Call of Duty entry (2020-2021), in a series that famously ditched its older yearly entries quickly. The newer ones probably work fine with E-cores.

Aurornis2• 1 year ago

I don't think many people would disable E-cores in BIOS if they bought this CPU, so for the purpose of deciding which CPU to buy, it would be more interesting to see results which factor in the potential software/scheduling issues which come from the E-core/P-core split.

Every mainstream review includes the E-cores. It’s not interesting to continue to repeat that testing.

The purpose of this article is an architectural exploration. Disabling the E-cores is a way to reveal that.

If you’re looking for simple gaming benchmarks to decide which CPU to buy, this isn’t the site for it. There are countless outlets for generic gaming reviews. This site examines architectural details.

barkingcat• 1 year ago

Pretty sure the mainstream gaming advice is to turn off E-Cores so the vast majority of actual gamers would be doing just that.

MangoToupe• 1 year ago

I wonder if there's also something to do with how the window scheduler (or allocator, or driver interaction, or some system component—I just guessed the scheduler because it seems like the most likely thing to be different enough to drive this sort of phenomenon) approaches the task. I notice remarkably smoother gameplay with the same games under wine—even, confusingly, games developed in house at Microsoft like Flight Simulator—with the same hardware.

MobiusHorizons• 1 year ago

Yep, I don't think chipsandcheese is trying to write a review of the CPU for practical purposes. Their content is more on technical microachitechture minutiae. In pretty much all of their articles they work to isolate performance characteristics of subsystems to show the differences that can be attributed to just the changes in that system gen-on-gen or between manufacturers.

bdavbdav• 1 year ago

Frustrated with this in the Linux world too - trying to get proxmox to play nicely with P/E is a nightmare. The OS scheduler side of things is a PITA

moffkalast4• 1 year ago

Say, is there any talk about Intel working on an AMD Strix Halo competitor, i.e. quad channel LPDDR5X in the consumer section?

jeffbee4• 1 year ago

I am still waiting on evidence that memory architecture helps anyone.

moffkalast4• 1 year ago

I'm pretty sure memory is still the main bottleneck in the vast majority of operations, and why the X3D is such a monster with so much L3 cache. There's almost nothing that doesn't benefit from having twice the memory bandwidth.

Rohansi• 1 year ago

Apple's M-series of chips make use of more than two memory channels. I don't think there is evidence that it is actually beneficial for anything but marketing though.

fortmeier• 1 year ago

I would like to understand more of the article, which book should I read?

yvdriess3• 1 year ago

https://archive.org/details/computerarchitectureaquantitativ...

throwaway31131• 1 year ago

That's more of a graduate level book. If you would like a book by the same author, that is lighter in content but much more approachable, I recommend this version: https://www.cs.sfu.ca/~ashriram/Courses/CS295/assets/books/H...

And don't forget to look at the online appendices on the MK website. They are also all very good.

celeritascelery• 1 year ago

Another thing you could do if you want to understand more of the details is open this article with an LLM and ask it questions. Have it explain unfamiliar terms and comparisons to you

fschutze2• 1 year ago

Fantastic article -as always. Regarding the top-down analysis: I was a bit surprised to see that in ~1/5 of the cases the pipeline stalls b/c the pipeline is Frontend Bound. Can that be? Similarly, why is Frontend Bandwidth a subgroup of Frontend Bound? Shouldn't one micro-op be enough?

yvdriess3• 1 year ago

Take front end bound with a grain of salt. Frequently I find a backend backpressure reason for it, e.g. long-tail memory loads needed for a conditional branch or atomic. There are limitations to sampling methods and top down analysis, consider it a start point to understanding the potential bottlenecks, not the final word.

fschutze2• 1 year ago

Interesting. You realize this by identifying the offending assembly instructions and then see that one operands comes from memory?

Avlin67• 1 year ago

I suggests using a Xeon w7 2595X so that you have 26P cores and 0E cores

ece• 1 year ago

Another article with some original slides from Intel. https://www.hwcooling.net/en/intels-new-p-core-lion-cove-is-...

Sohcahtoa82• 1 year ago

Seeing that the Lion Cove L3 cache takes ~83 cycles while the previous generation was only ~68 cycles explains why Lion Cove is an utter dud of a CPU for gaming and why it loses to the Raptor Cove in so many gaming benchmarks.

Meanwhile, the Zen 5 is only 47 cycles, and if you get the X3D variant, you get a TON of that L3 cache, which just turbo-charges games.

How did Intel allow such a major performance regression in their L3 cache?

Intel's Lion Cove P-Core and Gaming Workloads

Related Stories