A few months ago, we investigated false assumptions about CPU cores and explained how overall processor performance is affected not only by the number of cores on a CPU, but other factors including cache levels and capacity. This was an interesting and unique look at Intel's 10th Generation Series in an article titled "How CPU Cores and Cache Affect Game Performance". Basically, we compared the Core i9-10900K, Core i7-10700K, and Core i5-10600K with the same 4.2 GHz frequency, memory, memory timings, ring bus frequency, and so on.
Then compare the three CPUs with only 6 cores / 12 threads enabled to see how much of a difference L3 cache capacity makes in terms of gaming performance. Then we compared this data with the 10700K and 10900K with 8 cores activated and finally with the 10900K with all 10 cores turned on.
Long story short, it turns out that in almost all games it is not the number of cores but rather the L3 cache capacity that is responsible for the improved performance of the higher-quality Intel components. Of course, the extra cores along the way will pull the high-end parts even further forward, but at least today's games are all about the L3 cache.
That research later turned into a quad core version where we included Core i3 models and a similar setting for AMD CPUs, looking at 10 years of AMD CPU advancement and back to Intel.
To summarize this content, we thought we should consider the new Intel Alder Lake CPUs 12th. While all of the other CPU architectures had one, two, or maybe three different configurations, the 12th Generation Core has three per CPU.
For example, the 10th generation CPUs had 20 MB of L3 cache on the Core i9 model, 16 MB on the i7 and 12 MB on the i5 models, i7 and 30 MB on the i9. But then we also had to consider which core configuration to test. Four P-cores, four E-cores or a mixture of both? The correct answer was all three configurations, of course, and that gave us a wealth of juicy data to go through.
To be clear, with four P-cores enabled we were using Hyper-Threading, so this is a 4-core / 8-thread configuration. Basically, SMT was enabled if it was supported for all test configurations. This means that the configuration with four e-cores was 4 cores with 4 threads because the e-cores do not support SMT. Then the mixed configuration with two P-cores with two E-cores was a 4-core / 6-thread configuration.
We used the MSI Z690 Tomahawk Wi-Fi DDR4 for testing because we wanted to use the same low latency DDR4-3200 CL14 memory that was used for testing all other CPU architectures that support DDR4. In our tests, the DDR5-6000 didn't prove to be faster for gaming, but most importantly, we wanted to keep the data as good as possible for this feature. Finally, all configurations were tested with the Radeon RX 6900 XT. Let's dive into the data.
There is a lot to talk about starting with Rainbow Six Siege, so be patient with me. Let's just take a look at the Core i9-12900K first, we see with four activated P-cores and locked at 4.2 GHz that this configuration was good for 510 fps, only 3% faster than AMD's Zen 3 architecture.
With two activated P-cores and two E-cores the performance then dropped by 15%, which is a pretty significant reduction, and then with only four activated E-cores another 12%, which is not that much and not nearly that I expected a decline. It's pretty shocking that four e-cores in this title were able to match the performance of the Core i9-11900K despite the 11th Gen architecture.
When comparing the various 12th generation processors, we see that the additional L3 cache increases the performance from 12600K to 12700K by 4% if only the P-cores are activated, or 7% only with the E-cores. From 12700K to 12900K we then see a further 5% increase in performance for the P-cores and a fairly significant 10% increase in performance for the E-cores.
If we compare all the data we have, we see that the CPUs are the 12th. It is also interesting that the 12900K with two P-cores and two E-cores was a lot slower than four Zen 3 cores. So this suggests that once games make heavy use of 16 cores … in about 10 years, a part like the 5950X will be much faster for gaming than the 12900K.
When we get to the Battlefield V results, we get some interesting insights. First, the E-cores suck a lot in this title, not only does the average frame rate cut in half compared to what we only see with the P-cores, but the 1% low performance is shattered.
We see a performance reduction of 22% on the 12900K when we switch from 4 P-cores to 2 P-cores and 2 E-cores. Then we'll see a further 31% reduction if we switch to e-cores only. Worse, that means the P-cores were 87% faster when you look at the average frame rate, and 170% faster when you look at the 1% low rate. So those efficient cores are devastatingly slow and far from efficient in this game.
We also see that the larger L3 cache capacity of the i7 and i9 models does not provide any additional performance or at least only very little additional performance when the E cores are activated. However, with only the P cores, the 12700K was 6% faster than the 12600K and then the 12900K was 7% faster than the 12700K.
When we compare this data to the rest of the CPU architectures we tested, there are some notable comparisons. Compared to Zen 3, Alder Lake is up to 12% faster when comparing the 12900K to the 5800X. However, the 12600K's smaller 20MB cache meant it was 2% slower, while the 25MB i7 was only 4% faster. So it's this larger 30MB L3 cache that puts the Core i9 firmly over the line.
However, if we were to force Intel to use the E-Cores for gaming, we would see that the mixed 2 P-Cores / 2 E-Cores configurations lag behind Zen 3. Then when you are exclusively using E-Cores, performance falls off a cliff and now we're nowhere near Skylake level of gaming performance, think more about Sandy Bridge.
Moving on to F1 2020, we see that the E-Cores are nowhere near as bad as they were in Battlefield V. We see a 65% increase in performance with the E-Cores when looking at the 12900K and a 43% increase with the 12600K increase. The 12600K seems stifled by its smaller 20MB L3 cache as the 12700K was 18% faster when comparing P-Core performance, while the 12900K was only 4% faster than the 12700K.
Compared to Zen 3, Alder Lake is slower when limited to 20MB L3 cache, then up to 10% faster with 25MB and 12% faster with 30MB. As far as the pure E-Core configuration is concerned, Alder Lake is comparable to Ivy Bridge in F1 2020 and far behind Skylake, for example the 7700K was 33% faster than the E-Core configuration of the 12900K.
The NPC heavy Hitman 2 test crushes the E cores. This is similar to what was found when testing with Battlefield V. The performance of all three 12th generation parts is similar, and that means we're seeing a 41% performance improvement with 2 P-cores and 2 E-cores compared to just using the E-cores. If we look at the low 1% performance, that's more like a 134% jump which is insane.
Then we see if we only use the P-cores, the average frame rate is improved by 27% compared to the mixed core configuration.
So if we compare the pure E-Core configurations with older CPU architectures, we see that the performance is nowhere near that of Skylake. The 1% low performance was as bad as AMD's Bulldozer, while the average frame rate was much closer to Ivy Bridge than Skylake.
Even in Horizon Zero Dawn, which isn't particularly CPU-intensive, only the E-Core configurations had problems, despite eating 4-cores / 4-threads, especially when they're slow. If we look at a low performance of 1%, we see a 104% increase from 4 E-cores to 2 E-cores plus 2 P-cores, while switching from the mixed core configuration to 4 P-cores just changes the performance another 14% increased. . We also see very little performance differences between the various L3 cache capacities in this game.
If we compare with the older CPU architectures, we find that Alder Lakes E-Cores are again not much better than AMD's FX series. The 1% low performance was almost identical and that means we're miles away from Skylake here.
Cyberpunk 2077 is another game where the E cores can't push 1% lows to 60 fps, not even close. As a result, we see a performance increase of 100% in the 12900K when comparing E-cores with the mixed core configuration and then only a further increase of 12% when only the P-cores are used. Interestingly, the mixed core configuration of the 12900K is quite good, while we see a significant drop in the 12700K and 12600K.
Compared to previous CPU architectures, we see that the E-cores are much slower than first-generation Ryzen and worlds slower than Skylake. We are looking at Sandy Bridge's level of performance here.
After all, we have Shadow of the Tomb Raider and here we only see very minor differences between the various Alder Lake CPUs, so that the cache capacity here has almost no tangible impact, at least for these core configurations. We had previously found with the 10th Gen series that the larger L3 cache is more useful when more cores are available.
Compared to older CPU architectures, the E-Cores have issues with gaming alone, with average frame rate performance comparable to Ivy Bridge and 1% lower performance comparable only to AMD's FX series. On the flip side, if only the P cores are used, Alder Lake is a beast who beats Zen 3 by 11% in this game.
That was opening eyes to say the least. These e-cores aren't good for gaming, and there's a good reason why we're going to get into in a moment. First, let's take a look at the 7-game average we collected.
In the 7 games tested, we see that the 12900K was only 3% faster than the 12700K with only active P-cores and 8% faster than the 12600K. These margins are entirely due to the difference in L3 cache capacity. The margins with two activated P-cores and two E-cores are similar and the same is true for only four E-cores.
What is interesting, of course, is the difference in performance between the various core configurations on the same CPU. Take the 12900K for example, we've seen a 44% increase in average frame rate when switching from 4 E-cores to a mix of P and E-cores, and an 81% increase with 1% lower power. Then the average frame rate was increased by a further 20% and the 1% low by 21% from the mixture of P and E cores to only P cores.
Obviously you would never run a 12th generation CPU with only the E cores, which would reduce performance by ~ 20%, but let's dig deeper into the analysis in our conclusion …
What we learned
Intel's 12th generation hybrid core design is really interesting and brings some obvious benefits to productivity workloads and will no doubt prove very useful in the mobile space. Now you're probably thinking, "Sure, I've seen the benchmarks, I understand that e-cores don't work well for gaming alone, but why?"
The answer is simple, and it's the same reason first generation Ryzen fell back on Intel when gaming when it matched the same core number. The core-to-core latency is very weak – we're talking about an average increase of 54%.
Typically, P-cores take 37ns to communicate with each other while E-cores take 57ns, and this cripples performance in games and for any other workload that relies heavily on core crosstalk.
The reason Intel has limited interconnection between the e-cores is to make them more efficient, both in terms of power consumption and die space required. For sequential workloads like the one we see in rendering, for example, where there is very little core-to-core communication, the e-cores work well and that's why Intel used SPECrate2017 to make its Skylake efficiency claim .
If you look at the overall picture, the hybrid design also makes sense on the desktop, at least for Intel. A part like the Core i9-12900K can claim to accommodate "16 total" cores with 24 threads, because technically that's what it packs, even if not all cores are the same.
On paper, the 12900K looks similar to the Ryzen 9 5950X, and when tested in applications that can take advantage of those core-heavy desktop pieces, the 12900K still looks great given the core-to-core E-Core weakness if communication isn't emphasized by these workloads, think of Blender as one such example.
When it comes to gaming, the 12900K still shines, because not a single game requires more than 8 Alder Lakes P-cores. Even if a game can distribute the load across 16 cores, that's not a problem. Even in the case of the 12600K, its cores are more than powerful enough to handle the load. If it weren't for that, the game would only be playable on a high-end CPU like the 12900K or 5950X, and that won't happen in this decade.
Of course, you'd never get a CPU of 12th or more, based on what we've seen here at least. But again, I don't expect the time to come within the realistic lifespan of this series.
Another reason e-cores sucks for gamers is because of the compatibility issue with DRM, and I ran into it on this benchmark test. I had previously tested all CPU architectures with Watch Dogs Legion and Assassin's Creed Valhalla, but both games failed in this test. Watch Dogs Legion only worked with E-Cores or only P-Cores, but the mix crashed the game, which is weird as the 12th Gen CPUs work just fine. Then Assassin's Creed Valhalla could not be loaded due to the DRM detection problem with the hybrid architecture of the 12th generation.
In short, the e-cores are a bug for games, and when they are needed they reduce frame rates. For gamers, the 12900K and 12700K are 8-core / 16-thread CPUs and nothing more. The E cores could help with background tasks, but honestly, they'd be better looked after on the desktop by two more P cores. There's no argument that gamers can make for the existence of E-cores, it's always much better to replace them with two additional P-cores.