Core i9-12900K performance in Intel charts

When it comes to gaming performance, let’s start with it. One of the charts that deals with it was made public shortly before the end of the NDA. It compares top models of generations Rocket Lake (Core i9), Zen 3 (Ryzen 9) and new Alder Lake (Core i9). Before we look at it alone Alder Lake, allow a small detour. We see a test on the slide in nine games, in which he achieves in seven of them Zen 3 higher game performance than Rocket Lake:

However, if we look at the chart from the beginning of this year, from the announcement Rocket Lake at CES 2021, presented by Intel Rocket Lake as a processor that is clearly faster in games than Zen 3:

CES 2021: Game comparison of Rocket Lake (blue) and Zen 3 (orange) in Intel’s January slides

At the March issue, he had Rocket Lake achieve even better game results, the lead was to increase from an average of 4.4% to 7.5%:

Game comparison of Rocket Lake (blue) and Zen 3 (orange) in Intel’s March slides

Now, after about half a year, it is already Rocket Lake in games slower than Zen 3. Officially according to Intel.

You can think of the situation whatever you want. What light this binds to the credibility and at the same time the information consistency of Intel’s official materials is clear.

Let’s go back to Alder Lake. According to the official slide, he outperforms higher in games on average Zen 3 the 12.3%.

When compared to Rocket Lake the lead should be on average even 20.7% (1080p on GeForce RTX 3090).

However, the presentation also lists a slide with a higher number of tested games, but the representation is only graphical, without the stated percentage difference (the numbers are not even listed at the end of the presentation, where the note on the slide refers). Graphically, it can be deduced that the game lead is in this larger average Alder Lake over Rocket Lake 13.6% (less than the selection in the previous slide). Only 9 out of 31 tested games achieve a lead of over 20%.

Intel continues to report a 19% increase in IPC when comparing desktop processors Rocket Lake with large cores Alder Lake (Golden Cove).

We discussed the comparison of Core i9-11900K and Core i9-12900K at different TDPs in yesterday’s article. This slide would be correct if it did not look like a comparison of specific models, but a comparison of architectures. Manipulating the Core i9-12900K specifications ceases to be the results of the Core i9-12900K, so the claim of a quarter of the energy requirements at the same power is essentially misleading. By lowering the power limit to 65 watts, there was a de facto underclocking, shifting the processor to another part of the power-performance curve. Such an intervention would significantly change (improve) the power / performance ratio for the Core i9-11900K, but the behavior of the Core i9-11900K and Core i9-12900K, which are specific products with specific specifications, says nothing. Nor does it say anything about the hypothetical Core i9-12900 without-K, which could appear next year, as it will have different clock frequencies and 65 watts will not be its energy limit, but the base value (TDP / PL1 / base).

In “Content Creation”, ie content creation, according to Intel tests, the Core i9-12900K is 22-100% faster than the Core i9-11900K. On average by 45%. Given the 19% increase in IPC for large cores and the addition of 8 small cores, such a difference is possible.

In productivity (language purists may prefer terms such as fertility or yield), the intergenerational difference appears to be significantly lower, at only 15-19% (16.3% on average). It seems as if only large cores are involved here (the power corresponds more or less to an increase in IPC minus a 100 MHz boost), or the latency of small cores reduced the power to this level.

The last slide presents intergenerational increases in IPC. In the first column we see Comet Lake, ie. core Skylake, in the second Rocket Lake (modified Ice Lake), in the third large cores Alder Lake (Golden Cove) and in the fourth small cores Alder Lake (Atoms Gracemont).

First of all, it will surprise you Rocket Lake versus Comet Lake does not show Intel’s 18% increase, but only 12%. The difference between big cores Alder Lake a Rocket Lake is only 14% (1.28 / 1.12), not the above average of 19%. It is possible that Intel deliberately chose a test here to get the smallest cores possible (Gracemont), which puts this graph practically on a par with the cores Skylake. However, it is still necessary to keep in mind that this is a comparison of IPC (ie at the same clock), with the kernel Skylake in processors Comet Lake boost up to 5.3 GHz while cores Gracemont reach 3.9 GHz at best. They will then reach about 74% of the performance of the cores Skylake present in generation processors Comet Lake.

But even such a comparison has certain flies. While the core Gracemont is fully loaded by one thread, in the case of the kernel Skylake one core is not fully utilized and more power can be obtained from it with a parallel load of two threads. Thus, it cannot be said that four cores, for example Gracemont at up to 3.9 GHz will achieve the same performance as, for example, the Core i7-6785R, which carries four cores Skylake at up to 3.9 GHz. That would be true to load four threads. At five, six, seven, eight (and basically more) it will be Gracemont due to lack of HT / SMT support to lose.


Source: Diit.cz by diit.cz.

*The article has been translated based on the content of Diit.cz by diit.cz. If there is any problem regarding the content, copyright, please leave a report below the article. We will try to process as quickly as possible to protect the rights of the author. Thank you very much!

*We just want readers to access information more quickly and easily with other multilingual content, instead of information only available in a certain language.

*We always respect the copyright of the content of the author and always include the original link of the source article.If the author disagrees, just leave the report below the article, the article will be edited or deleted at the request of the author. Thanks very much! Best regards!