Here's why Intel's Arrow Lake CPUs could be much better for gaming than Raptor Lake

[DragonSlayer101]

Something to look forward to: Intel's upcoming Arrow Lake processors could see a major bump in L2 cache compared to their predecessors. If the unofficial reports prove accurate, it would be a major win for PC gamers, as the added cache could improve performance across the board.

The news comes from prolific leaker Golden Pig Upgrade (via @9550pro), who claims that the 15th-gen Core Arrow Lake chips could have up to 3MB of L2 cache per Performance core. This would represent a significant upgrade over the current-generation Raptor Lake lineup, which offers 2MB of L2 cache per Performance core and 4MB of L2 cache per cluster of 4 Efficiency cores. In comparison, the older Alder Lake chips provide just 1.25MB of L2 cache per P-core.

Also read: How CPU Cores & Cache Impact Gaming Performance

This change suggests that the total amount of L2 cache in Arrow Lake processors could be markedly higher than that in existing Intel CPUs. For instance, the top-of-the-line Arrow Lake chip, potentially named Core i9-15900K, might boast as much as 56MB of L2 cache. This contrasts with the 32MB in the Core i9-13900K.

It is speculated that the flagship Arrow Lake chip will include 8 'Lion Cove' P-cores and 32 'Skymont' E-cores. Each E-core is expected to come with 1MB of L2 cache. Thus, while the 8 Performance cores could offer 24MB of L2 cache, the additional 32 E-cores might contribute at least 32MB more. This could render the Arrow Lake flagship notably faster than its predecessor in most applications.

ARL-S Big Core L2 3MB (per Core)https://t.co/31XsVl4x8Z

– HXL (@9550pro) August 12, 2023

The increased cache in Arrow Lake chips could lead to enhanced gaming performance. This is because having more built-in cache allows a CPU to store executable data closer to its cores, reducing the time needed to access it. The larger the cache, the more data the CPU can swiftly access, which in turn improves performance. While the L1 cache is the fastest due to its proximity to the cores, the L2 cache is the next in line, followed by the relatively slower L3 cache (as a side note, AMD's 3D V-Cache is essentially L3 cache layered over the CPU cores).

It's important to highlight that Intel has not yet released the full specifications for Arrow Lake. However, they have shared some information, including details about the manufacturing process nodes and a potential launch timeline. At their 2022 Investor Meeting, Intel CEO Pat Gelsinger revealed that Arrow Lake will utilize three process nodes: Intel 4, Intel 20A, and External N3 technology. He also confirmed that the lineup will launch in 2024, following the release of Meteor Lake chips in 2023.

Permalink to story.

https://www.techspot.com/news/99791-here-why-intel-arrow-lake-cpus-could-much.html

 

[i like foxes]

I expect they ll be cool but just one thing - Intel rolls out a new gen CPUs every year thus one cannot catch up .And the socket is often changed which is hard for upgrading in regard to the cost .

 

[Dimitriid]

I know this might be pretty unpopular, but I cannot bring myself to care at all about intel's IPC improvements and advancements because we know just how much power they take. Even on a desktop if AMD keeps getting you 90% of the way there in terms of raw IPC and single core or multi core performance at 50% of the power consumption you should not even consider intel.

This is even more endlessly frustrating in the laptop space where we know AMD just blows intel out of the water in terms of battery life is not even close how much more efficient they are but because they have decades long inertia it's still significantly more difficult to find a decent AMD laptop vs an intel one.

 

[Theinsanegamer]

I expect they ll be cool but just one thing - Intel rolls out a new gen CPUs every year thus one cannot catch up .And the socket is often changed which is hard for upgrading in regard to the cost .

Why are you upgrading constantly? CPUs have decade long lifespans, yes including for gaming. If you get a platform to "upgrade" with every CPU gen you are going to be wasting tons of cash no matter who you choose.

I know this might be pretty unpopular, but I cannot bring myself to care at all about intel's IPC improvements and advancements because we know just how much power they take. Even on a desktop if AMD keeps getting you 90% of the way there in terms of raw IPC and single core or multi core performance at 50% of the power consumption you should not even consider intel.

This is even more endlessly frustrating in the laptop space where we know AMD just blows intel out of the water in terms of battery life is not even close how much more efficient they are but because they have decades long inertia it's still significantly more difficult to find a decent AMD laptop vs an intel one.

If you run the chips in their TDP ranges, they are absolutely fine. Blame OEMs for pushing unlimited turbos to eek out 1% more performance.

 

[manowar51]

I know this might be pretty unpopular, but I cannot bring myself to care at all about intel's IPC improvements and advancements because we know just how much power they take. Even on a desktop if AMD keeps getting you 90% of the way there in terms of raw IPC and single core or multi core performance at 50% of the power consumption you should not even consider intel.

This is even more endlessly frustrating in the laptop space where we know AMD just blows intel out of the water in terms of battery life is not even close how much more efficient they are but because they have decades long inertia it's still significantly more difficult to find a decent AMD laptop vs an intel one.

I think I am the opposite, at least regarding the power consumption, I dont care if its Intel or AMD. If I buy a desktop CPU I would want it to pull as much power as I can get into it to give me the most performance. But I tend to buy laptops so for me, if im going to build something thats 10 times the size of a laptop, with a meaty expensive cooler on it and leave it permanently plugged in then it best make the most of that situation!

Also where I live on the US west coast energy is less than 10 cents a unit. You wouldn't notice the difference between AMD and Intel desktops here at those prices. But if I lived in Europe where energy costs are several times that of the USA then I'd probably think differently. I'd have to turn off my hot tub for a start. That thing uses 2000watts 24/7!

 

[pencea]

That sounds great and all regarding performance, but will power consumption and tempertures be improved as well compared to the existing 13900k?

 

[J Oelschl]

Only 8 PCores and 32 ECores! I dont want more ECores. I want more PCores. For transcoding you need all the power you can get! This is a non-starter for me. This would be great for a power laptop but on a desktop... who wants majority ECores? anyone?

 

[mbk34]

If I buy a desktop CPU I would want it to pull as much power as I can get into it to give me the most performance ... and leave it permanently plugged in then it best make the most of that situation! ... where I live on the US west coast energy is less than 10 cents a unit ... if I lived in Europe where energy costs are several times that of the USA then I'd probably think differently. I'd have to turn off my hot tub for a start. That thing uses 2000watts 24/7!

It's a shame that climate change effects everyone globally rather than just those countries that produce all the pollution.

 

[HardReset]

Problem with bigger L2 cache is fact that making bigger cache usually means it's also slower. Beauty of AMD V-cache is that while it has huge size, it does not slow down CPU caches. So Arrow lake may not be faster in games after all because caches most likely will be slower. That's also main reason why there aren't any "gaming CPUs" with huge L1 and L2 caches.

 

[waclark]

I know this might be pretty unpopular, but I cannot bring myself to care at all about intel's IPC improvements and advancements because we know just how much power they take. Even on a desktop if AMD keeps getting you 90% of the way there in terms of raw IPC and single core or multi core performance at 50% of the power consumption you should not even consider intel.

This is even more endlessly frustrating in the laptop space where we know AMD just blows intel out of the water in terms of battery life is not even close how much more efficient they are but because they have decades long inertia it's still significantly more difficult to find a decent AMD laptop vs an intel one.

The thing is, it's not going to deliver that level of performance at 50% less power. From what I've seen, the 13900K runs about 20% more power to get similar, and sometimes better performance in gaming. Then there's the question of idle power draw. I don't know if AMD has addressed this, but at one time the 79XX CPU were drawing higher power at idle than the Intel 13th gen. (Anand Tech)

In general the AMD CPUs are more efficient, but not 100% more efficient, and that efficiency comes at the cost of performance. I believe AMD could clock the 79xx CPUs higher but then they would be in the same power range as Intel and that doesn't seem to be where they want to go.

I don't know about mobile versions, so that may be a different conversation.

 

[Vanderlinde]

That sounds great and all regarding performance, but will power consumption and tempertures be improved as well compared to the existing 13900k?

Cache is stuff that needs to be turned on in all states or conditions of the CPU. So adding more cache will consume more (idle) power.

 

[hahahanoobs]

That sounds great and all regarding performance, but will power consumption and tempertures be improved as well compared to the existing 13900k?

Do we usually have data like that 1 year away from launch?

 

[hahahanoobs]

Cache is stuff that needs to be turned on in all states or conditions of the CPU. So adding more cache will consume more (idle) power.

Good thing process technology gets better every generation. L2 is faster and more efficient than L3 so we'll see. Clearly Intel can't maintain 250w CPUs forever and shouldn't be expected to.

 

[The Hardcard]

Only 8 PCores and 32 ECores! I dont want more ECores. I want more PCores. For transcoding you need all the power you can get! This is a non-starter for me. This would be great for a power laptop but on a desktop... who wants majority ECores? anyone?

For transcoding and other multithreaded tasks, the E cores give you more power. Intel‘s E stands for area efficiency - they actually don’t do a significant amount to boost power efficiency

Why they provide more power is that four of them in the same space as one P core yet those four can do double the work of the SMT P core .

 

[Mr Majestyk]

Please explain how they will do this cost effectively with SRAM scaling dead in the water and the fact you'll be trying to fit 6nm SRAM onto 3nm cpu. TSMC's latest SRAM for their 3nm for the first time ever has shown no improvements over previous gen. SRAM scaling is a result of physics and no one has come up with a way of breaking free from the constraints.

 

[Stiqy]

It's a shame that climate change effects everyone globally rather than just those countries that produce all the pollution.

Environmental and economic impact of running a CPU for a year is less than a few trips to McDonalds. If you are looking to make a lifestyle choice to help out the planet, there are other places to look.

 

[Thund3RChild]

Environmental and economic impact of running a CPU for a year is less than a few trips to McDonalds. If you are looking to make a lifestyle choice to help out the planet, there are other places to look.

I think mbk34 was referring to what manowar51 said about their hot tub running 24/7 that pulls 2000W constantly. Your calculation (before you edited your post) showed that, in fact, personal computing is negligible in the grand scheme of things. Having a hot tub pull 2000W 24/7 365 days a year? Now we're talking! That's 48kW/h a day, amounting to more than 17.520kW/h p.a., which is 17 times the annual electricity consumption of an average German and regarding pollution, we're at 17t CO2e when you consider a 1960s lignite plant that is said to emit 1kg CO2e per 1kW/h (which is worst case, but it gives you an idea).

 

[neeyik]

Please explain how they will do this cost effectively with SRAM scaling dead in the water and the fact you'll be trying to fit 6nm SRAM onto 3nm cpu.

The L2 cache doesn't take up a huge amount of space in Meteor Lake P and E-cores, so there's no reason to assume that it will be an issue in Arrow Lake:

[Source]

The CPU tile in Arrow Lake will be made on the same node as that used for the same tile in Meteor Lake. This suggests, barring any structural rearranging, AL-S P-cores will have a larger die area than ML ones, but nothing massively problematic.

 

[VcoreLeone]

Big l2 means less heat problems than big, stacked l3. was gonna pick 5800x3d earlier this year, but guys told me it's pretty much clc only for this to avoid seeing 90 degrees in the summertime.

 

[J Oelschl]

For transcoding and other multithreaded tasks, the E cores give you more power. Intel‘s E stands for area efficiency - they actually don’t do a significant amount to boost power efficiency

Why they provide more power is that four of them in the same space as one P core yet those four can do double the work of the SMT P core .

I used an unfortunate ambiguous generic word "power". I meant "processing power". Every description I have read describes the E cores as ultra power efficient and best used for background tasks. Again, for laptops where battery life is important they may be great. However "power efficiency" is not really a concern for me in a desktop system (but I acknowledge that it may be for some where the cost of a electricity is very expensive). My fear is that these heavy Ecore focus is being driven by the green agenda... for businesses that have fleets of desktops....

"When you’re looking at specifications of the processors, you’ll now see P-Cores and E-Cores:

1. P-Cores: designed for heavy-duty work and are your standard, high-performance CPU cores.
2. 
   
3. E-Cores: Ultra-efficient and are used for background tasks that run constantly but don't require much processing power.
4. 
   
5. In simpler terms, P-Cores are the power cores while E-Cores are the efficient ones in the system."

Regarding your comment where the Ecores "provide more power in the same space as 1 Pcore with SMT"..... well perhaps IFF a given task is multithreaded... but if not then a Pcore is going to win for those tasks that are single threaded and benefit by higher IPC and higher clock speed.

However a recent article I read may change the calculus a bit. The high ECore count could be an evolution where they are fundamentally changing how the processors are utilized. I could buy into that argument... but I would have to see various benchmarks between old vs. new before I am convinced.

Future Intel CPUs May Dump Hyper-Threading For Partitioned Thread Scheduling

An Intel patent describes an interesting technique that could spell the end for Hyper-Threading.

hothardware.com