The day is here and we can finally talk about Intel 12 Gen performance and how it stacks up against previous generations, as well as current leading chips from AMD. Spoiler alert: it does not suck! Mostly. More on that in a bit.
First let me do a quick high-level breakdown on what is new and different with Intel 12th gen. In our case we will be checking out the top of the line Intel i9 12900k.
Intel 12th Gen is now based on Intel 7 process, which is actually Intel's 10nm technology. It features a new hybrid design which consists of P and E cores. The traditional cores are now called Performance Cores, and the new type of cores are called Efficiency Cores, which are four times smaller on the die than P cores. There are 8 of each in the 12900k but if you look at the spec it mentions 24 threads. This is because the performance cores support multithreading while efficiency ones do not.
More cores is normally better as it allows for better multitasking, and with efficiency cores clocked lower they are perfect for background tasks, freeing up performance cores for the more critical utilisation. To further improve performance, Intel has introduced a new tool called Thread Director that helps operating systems schedule tasks to the correct cores. Unfortunately, this is only available on Windows 11 at the moment so we will have to see what performance difference we get.
Besides adding more cores and complicating whole core and thread count, the new P cores have reportedly 19% improvement over the last generation at the same frequency which is a pretty impressive jump.
There are a couple other major changes in this generation, like, for example, Intel now supports PCIE Gen 5 as well as DDR5 memory.
While we currently do not have any PCIE Gen 5 graphics cards, this move does seem like we may see them really soon. On the other hand, support for DDR5 is a great move by Intel as this starts the gradual migration into the world of faster RAM with higher bandwidth. Few things to note - the first memory modules that are scheduled to hit the market will have pretty average performance due to low speed and high latency but this will surely improve over time.
Lucky for everyone, these chips still support DDR4, but you will have to use a compatible board. At the moment it is unlikely that many boards will have dual memory support, and if they do, you will not be able to mix and match, you will simply be able to only use DDR4 or DDR5. Do watch out for compatibility - for example ASUS is using D4 in the product name to indicate DDR4 support on the motherboard.
Intel has also introduced new RAM overclocking profiles - XMP 3.0. They are working with vendors to validate it, and we should see many upcoming DDR5 modules support it in the near future.
While we are on the subject of overclocking - there is a new version of Intel extreme tuning utility, which has built-in auto mode to overclock by 100 MHz across all cores for those who just want to get a little bump, allowing to go more granular in manual mode. Let us know in the comments below, if you would like us to do a deep-dive in overlocking and what would be your use case.
With a quick overview out the way, let's jump into our benchmarking. In our test benches we used the following: for the last generation PC’s we have Intel i9 11900K and AMD Ryzen 5950X with Z590 and B550 motherboards respectively. Both use mid range Corsair RAM with 3600 MT speed and CL 18 timings. For the new Intel i9 12900k we have ASUS Maximus Z690 Hero motherboard and Geil DDR5 memory with 4800 MT speed running at 38 CL using ASUS RAM overclocking.
And yes - we are using separate platforms with different RAM generations which is likely going to slightly skew the results, but remember - just because RAM is 1200 MT faster, it does not automatically make it better as DDR5 Memory also has much higher timings and that contributes a lot to the performance.
In all examples we are running the same RX 6900 XT graphics card. For cooling we are using a 360mm liquid all-in-one cooler by ASUS. As far as general settings go - we set CPU cooler pump and fan speeds to 80% and GPU fan speeds to 75% to keep results consistent.
Few more things to add - we used the latest version of Windows 10, which I believe many enthusiasts would use for the time being, as Windows 11 has certainly had a rough start. Having said that, if Intel is to be believed - 12th generation chips should perform better on Windows 11 and improve over time as developers optimise their apps for the thread director.
The other thing we noticed is performance differences between BIOS revisions - this is pretty expected with the new platform. We have left both results in as an example.
Jumping straight into the benchmarks, starting with some games. We reduced the resolutions and graphics settings for each of the games in order to remove GPU bottlenecks and be able to really compare the CPUs. Ultimately, the goal is to show performance delta between these set-ups.
First game is Shadow of the Tomb Raider, and here we see a significant improvement from last gen at 19% on both average FPS and 1 percentiles. We have approximately 5% lead on average FPS and 15% lead on 1 percentiles over AMD.
Next is Doom Eternal, this game is pretty easy to run and the large FPS count makes this kind of testing even more pronounced. In this game we see that from 11900k there is about 30% improvement on average FPS, and 32% on 1 percentiles. When compared to AMD the new CPUs lead by 17% on average FPS and 10% on 1 percentiles.
Another light game is War Thunder - we threw this in basically for fun as framerate is just insane. In this game AMD takes a 6% lead on average FPS and only 2% on 1 percentiles.
In Horizon Zero Dawn there is only about 8% improvement on average FPS from the 11th Gen and 15% improvement on the 1 percentiles. The improvement here simply matches AMD’s performance. It is good, but probably won’t be called the best in the world just yet.
While analysing the data, we noticed a more unexpected pattern. The framerate is certainly competitive against AMD and in some cases better, but the performance per watt is actually even more impressive.
Here are the results of FPS per 10 watts on all of these chips across the same games - we see that in all titles the new Intel CPUs deliver more FPS per watt. In the heavier title - Horizon Zero Dawn - the difference between AMD and Intel is 17% and in Shadow of the Tomb Raider the difference is 51%. In the lighter Doom Eternal we see a difference of 26% and in War Thunder it is a whopping 68%.
So far, just from looking at gaming results I can certainly say that Intel is back, and there is a healthy competition happening here. This will certainly be good for gamers, but is the new architecture equally good in productivity?
This one is a bit of a mixed bag. Let's start with Cinebench R23. In this benchmark the new intel chips are ahead of AMD’s top of the line 5950X in both single core and also multicore results, and considering that Intel chip has only 24 threads against 32 from AMD, that is very impressive.
In Blender we also see Intel CPU trading blows between different tests with less than 15 seconds difference between them.
In Vray we see about 9% lead by AMD, this is where those extra threads are pushing it to victory. Even though Intel 12th Gen is not leading here, it is still miles ahead of its predecessor.
Next up we have 7zip benchmark and here AMD completely dominates on the decompression speed, but falls behind on the compression side, while Intel is a bit more balanced.
It is clear that the new chips from Intel have brought major improvements in productivity tasks. The hybrid architecture does not yet fully stack up against the large multithreaded beast from the competition. It is close but not quite there yet.
One thing that Intel has done is change its boost functionality - previously it would boost for a certain amount of time and then bring itself back down, but as we see in this blender run, it boosted for the whole duration.
Even with this beast of a cooler, it reached 90 degrees straight away and peaked at 95 towards the end. For reference, the room's ambient temperature was 28 degrees Celsius.
In terms of power consumption, this chip does not mess around. In this blender render it hit 224 Watt and stayed there. If we look at AMD, it is basically sipping power at 137 watts while delivering slightly better results.
To sum things up - has Intel delivered on its promise? In gaming, certainly! It is even doing pretty well in productivity. So for those who spend most of their time gaming and maybe doing some productivity tasks on the side, this is an incredible leap forward. With Windows 11 and thread director we can expect it to get even better.
I am glad to once again see innovation in the CPU market and we got AMD to thank for this with the pressure from Ryzen processors. I am really looking forward to seeing what team red brings to the table with their new chips in the upcoming release and will enjoy the view of both companies leapfrogging each other yet again.
____________________________________
This post contains affiliate links. An affiliate link means we may earn advertising/referral fees if you make a purchase through our link, without any extra cost to you. Thank you for your support!
コメント