For almost 10 years, all the top “consumer” processors in the Intel range were released under the Core i7 brand. It cannot be said that everything was so clear and understandable – since both dual-core laptop models and ten-core HEDTs fell into this family, but everything was predictable up to the segment: the best possible. Desktop models completely “bronzed” on one formula – four cores / eight computation threads. Process technology and microarchitecture have changed – so that performance has grown one and a half times, and power consumption (despite this) has decreased. The built-in GPU (and from the second to the eighth generation of Core i7 it was an indispensable element in the design of desktop processors) has completely improved radically. But fundamentally nothing has changed.

Then Intel decided to change the label for top models – Core i9 appeared. First, these came out in the HEDT family under LGA2066, then the practice was extended to all other models. The main confusion, accordingly, has shifted to this family. But one more problem has been added – how to separate the Core i7 and i9. In the ninth generation of desktop Cores, for example, both were eight-core – it’s just that the Core i9 supported Hyper-Threading, while the Core i7 did not. What were similar to the Core i3 and i5 of the same generation – but differed in the number of cores: four, six or eight. And Core i9 – also eight, but already 16 computation threads. Moreover, such Core i7s turned out to be the only ones in history that use “single-threaded” cores. Since in the tenth generation, on the contrary, Hyper-Threading became the standard – only Celeron did not support this technology. And the whole system became simple and logical – since the Core i9 received 10 cores. The Core i7 has eight left, the Core i5 has six, and the Core i3 has four. But “two-threading” was added, which, all other things being equal, increased performance and made it easy for the new LGA1200 platform to compete with the previous L:GA1151: most of all . The new Core i3 had the performance of the old Core i5, the new Core i5, in turn, competed with the old Core i7, the same overtook the former Core i9 – and the new generation Core i9 simply occupied a new performance bar.

Had this happened a couple of years earlier, everyone would have been happy. However, by 2020, AMD has risen from the ashes – so this was still not enough to compete with Ryzen. The development of new microarchitectures was tied to the development of new technical processes – and problems continued with them for the third year. Mostly ended by this point, but all 10nm production lines were occupied by Tiger Lake laptops. Moreover, more laptop processors are needed for a long time, and their marginality is higher – so this is a more important direction. Therefore, Rocket Lake appeared as a temporary solution in the desktop segment – with a microarchitecture similar to Ice Lake (the tenth generation of Core laptops based on a 10-nanometer process technology), but produced on old 14-nanometer capacities. The latter led to the fact that 10 cores did not fit into a reasonable area crystal, so the company produced only eight-core crystals – from which both Core i7 and Core i9 were obtained. And six-core Core i5 – too: just a couple of cores turned off.

The “reasonableness” of Rocket Lake’s release can be measured in different ways. These processors really solved some problems – although they did not linger too long on the market. But the perniciousness of selling the same (in fact) device under two different brands and at different prices is unlikely to be challenged by anyone. We wrote about all this, as well as the history of the Core i7 family in detail – when we tested the Core i7-11700K. Now this page of history has been turned over – processors from the Alder Lake family and the new LGA1700 platform have entered the market.

Upside down – but not completely. What do we mean? It becomes clear if you look at the technical characteristics of older desktop models. Almost all new Core i5s, with the exception of the i5-12600KF and i5-12600K, are “regular” processors with six P-cores and no E-cores. The latter (in the amount of four pieces, i.e., one cluster) appear just in the mentioned pair – in addition to the six P-nuclei. As a result, the positioning is different – most processors will not have any compatibility problems with old software, including when running under Windows 10. With the exception of the “hybrid” older pair – but it is also aimed at enthusiasts who are sometimes even ready to endure any then deprivation for the sake of technical novelty. And how can you get eight P-cores? Buy Core i7 or Core i9. Yes – in the twelfth family, as well as in the eleventh, both are eight-core as a first approximation. But they are no longer the same, since the total number of cores is 12 and 16, respectively. However, the whole difference is only due to one cluster of E-cores in Core i7 and two in Core i9. The L3 cache capacity and clock speeds are also different (of all processor cores – and the GPU too) – but does this justify the almost one and a half times difference in price …

Here you need to think carefully. If, for example, you need only eight P-cores and there are fears that the E-cores will only interfere (and this happens to them under Windows 10 – as we have already seen), then it makes sense to save money. Many people would certainly prefer a “regular” processor with eight P-cores – but Intel does not offer this. You can only disable it – but it’s better to do it with one cluster than with two. In any case, it won’t be such a pity for the paid money 🙂 And the main purpose of both processors is the same: users who need the maximum, are ready to change their habits for this and generally pay at least. Do the performance differences justify the extra charge, at least in part? And this is exactly what needs to be checked.

Of course, it makes sense to do this exclusively under Windows 11 – as we have recently established, its use is preferable for hybrid processors. Once again, this hypothesis can be tested, but not necessary – it is a priori clear that under the “ten” Core i7-12700K will differ little from the i9-12900K in the “right” applications and from the i5-12600K in the “wrong” applications. With the E-cores disabled, it will stably turn out to be very close to the first – noticeably different from the second. But first of all, we want to evaluate the usefulness of increasing the number of E-cores, which means we need to provide processors with the “right” conditions for comparison.

Along the way, it was decided to slightly expand the list of systems tested under Windows 11. Not too much – to avoid unnecessary work. However, it is useful to add the results of the tenth generation Core to the results already obtained – after all, architecturally Intel desktop processors did not change from 2015 to 2020, so they can be extended to the corresponding models of the eighth or ninth generation. It is clear that you should not expect any benefits from the new Windows for old (relatively) systems, but why not? Yes, and the three triples of Core are more interesting than two in terms of retrospective.

Test participants

Moreover, the oldest trinity, as already mentioned, is beautiful ideologically – each family differs from the neighboring one by exactly a couple of cores and 4 MiB L3. The maximum frequencies differ only slightly, so everything was simple and clear with the choice – the more “serious” the tasks, the more attractive the older lines begin to look. Just because it has long been customary to solve such problems using multi-threaded code – so more cores are always better. On the other hand, this is not always the largest number of threads needed, but single-threaded performance is almost the same – which leads, for example, to the fact that for games, regardless of the video card, Core i5 is enough with a margin. Great opportunity to save money. And the microarchitecture licked to a shine since 2015 – why the “dozens” are still quite popular.

With one caveat – in this form, the LGA1200 strongly resembles the LGA1151, and both versions. Unless there were no ten-core processors at that time – but six and eight were quite enough (the first incarnation of LGA1151, however, officially supports only quad-core processors – but with a strong desire and accuracy, you can “screw” more to it). But Rocket Lake is the first new microarchitecture in many years. And new peripherals – the same PCIe Gen4, for example, both for video cards and SSDs: in the form of an additional dedicated connector (AMD had such a scheme earlier – but the fact itself is more important here). Here, the processors, as has also been said more than once, let us down – firstly, the maximum number of cores was again reduced to eight (ten “did not fit”), and secondly, there were so many in both Core i7 and Core i9. A reason to save money – and a disappointment for those who needed the maximum.

However, certain reasons for disappointment have remained even now – after all, there are still only six or eight “large” P-cores. But this is a leap through the microarchitecture and through the process technology relative to Rocket Lake, firstly. Secondly, “small” E-cores are architecturally comparable to Skylake and its peers. However, there will be no full compliance with performance due to differences in the environment and clock frequencies, but in general they will definitely not get worse (with appropriate software support – do not forget about this important condition). But whether it will get so much better to justify the price difference between Core i7 and Core i9 at this step is just the most important question at the moment. Moreover, an additional cluster of E-cores and 20% L3 are just the main differences. As for clock frequencies and power consumption limits, they, as usual, can be adjusted in their own way. For both models – and in both directions. But this is already a creative process – so for a start we will limit ourselves to regular modes.

Let’s stop there for now. It would be possible to add the results of several AMD Ryzen, but only the Ryzen 9 5900X could be new (and potentially the most needed today) – and, unfortunately, it is temporarily not at hand. With others, everyone has already been compared and repeatedly, so doing it again is not too necessary. The main question is Core over the past two years. Moreover, as already mentioned, the tenth generation does not fundamentally differ from the sixth or ninth, but last year was a time of great leap for Intel. Which I want to evaluate – and already on the top three processors. And at the same time see how the situation in the troikas themselves has changed.

We were not able to completely equalize other test conditions: DDR5 was “gotten” only in the form of two modules of 16 GB each, and the rest of the processors were tested with 16 GB of memory in total. However, we specially optimized the tests for the latter value and with some reservations for 8 GB, so this cannot affect the results (but in the new method, of course, we will already switch to higher values). The AMD Radeon Vega 56 video card and SATA SSD are the same for everyone, they also do not affect. The memory clock speed is the maximum according to the processor specification. Intel Multi-Core Enhance and AMD Precision Boost Overdrive technologies are disabled – for the latter, this is typical by default, but many motherboards strive to turn on the former quietly. Here, along with the memory frequency, they can affect performance, and their use is made more specific by the requirements for the board and chipset, but in normal mode there are no problems.

Test Methodology

The testing methodology is described in detail in a separate article. The only change we made for some of the materials was the use of Windows 11. Accordingly, we did not add the results of all tests to the general table – they are available in a separate one (as usual, in Microsoft Excel format) . Directly in the articles, we use the processed results: normalized with respect to the reference system (Intel Core i5-9600K with 16 GB of memory, AMD Radeon Vega 56 video card and SATA SSD) and grouped by computer application areas. Accordingly, on all diagrams related to applications, dimensionless points, so here “more is better” everywhere. And starting from this year, we are finally transferring game tests to an optional status (the reasons for which are discussed in detail in the description of the test methodology), so that only specialized materials will be available for them.

iXBT Application Benchmark 2020

Many did not like the previous stage of modernization because the Core i9-11900K often lagged behind the i9-10900K – and these programs are just a typical illustration of this. Now there is no ground for such claims – the Core i5-12600K is already faster than all earlier Intel desktop developments. In general, everyone. The Core i7-12700K in the new line is closer to the Core i9-12900K than to the Core i5-12600K – however, this was also observed in the tenth generation, to a slightly lesser extent. The shame of the eleventh generation , in any case, is not even close.

Similar picture. However, it is easy to see that a linear increase in the number of homogeneous cores from six to ten gave a much larger range of results than the current complex scheme. In which, too, it is the increase in the number of P-nuclei that is most noticeable. That is, from the point of view of the race for performance at any cost, it would be better not to deal with E-cores at all – but try to “shove” 10 P-cores into the Core i9 again. With regard to desktop processors – so. But, we repeat, the main share of deliveries has long been provided by laptop models, and they have their own mechanics. Actually, hybridity is exactly what she needs. But it also works well on desktop models. And it is clearly seen that the current Core i9 is not exactly a Core i7 on steroids. Although on them too. Recall that the limits in long-term mode for the latter are much softer – however, as a result, some users are frightened by the level of energy consumption, of course.

As we have repeatedly seen, in these programs the “quality” of cores is no less important than their number. As a result, the eleventh generation behaved better than the tenth. Although the quantity should not be neglected either – as we already remember, in this group of programs, disabling E-cores significantly reduced the performance of Alder Lake hybrid processors . So we’ll see how the younger “non-hybrid” models of this family will behave in such conditions in the near future. But with today’s set of test subjects (when the tops of Core i5, i7 and i9 are taken), everything is quite clear – even the Core i5-12600K radically outperforms all Intel processors for LGA1200 (which means for previous platforms), not to mention faster counterparts.

Once again, we are convinced that this group of programs for older processor models is primarily a task “for architecture” – everyone has more than enough cores, so single-threaded performance comes first. Which depends on the clock frequency, among other things – but the frequencies do not differ too much in this group either. Only one thing is striking here – how much Intel managed to improve the microarchitecture. And it’s understandable why the company delayed this moment until the development of new technical processes: Rocket Lake is also significantly better than Comet Lake, but against the background of the fact that Alder Lake processors were brought into the world, the effect turns out to be very blurry. And it remains only to regret that the company succeeded in all this only now – initially, the development of 10-nanometer standards was planned for 2017, so something similar to Intel 7 and Golden Cove (Alder Lake P-core microarchitecture) with “normal” we could have seen the development of events three or four years ago. And all the competition in the processor market would be completely different. But the history of the subjunctive mood does not tolerate.

The reverse situation is when the number of cores and computation threads is still critical. But every cloud has a silver lining – the E-cores are given the opportunity to show everything they are capable of. Which is not so little – as we remember from experiments with disabling E-cores, it was in this program that this led to a very serious decrease in performance. Accordingly, the Core i9-12900K has a chance to noticeably break away from the i7-12700K, and the i5-12600K reaches the level of the tenth-generation Core i9 (on P-cores alone, it lagged behind those of the Core i7).

As we already wrote, archivers have preferences similar to FineReader – but there are nuances. First, performance scales much worse in terms of the number of computation threads. Secondly, the influence of the memory system is noticeable. In the full sense of the word – including caches. As a result, you won’t get much out of the architecture of processor cores alone – although there’s nothing to seriously criticize the new models here either. And the difference in L3 capacity allows the Core i9-12900K to quite convincingly overtake the Core i7-12700K – the absurdity of the previous line is certainly not observed.

And again everything “plays” – clock frequencies, L3, and E-cores are far from useless here. Although P-cores again turn out to be the main ones here – on only six of them, the Core i5-12600K “left” to the level of the Ryzen 7 5800X and overtook (in this group) all the “old” Intel processors. The figure eight, of course, is even faster a priori – but then secondary (but also important) factors begin to work.

The main question that worried us today is whether the performance of the Core i7-12700K and Core i9-12900K is significantly different? Yes, it’s significant. Is it worth the price increase? And now everyone will have to decide on their own. In the end, the effectiveness of “investment” in processors has always declined as performance increases – why the love of some buyers for the abstract combination of “price / performance” looks a little … stupid. To prevent this from happening, it is necessary, at a minimum, to set boundary conditions. Although these are already more serious questions than simple testing of several processors – and more abstract ones too. Returning to the topic – Intel had more opportunities to smash the twelfth generation Core i7 and i9 compared to the eleventh, and it took full advantage of all of them. It turned out even better than in the “reference” tenth.

Energy consumption and energy efficiency

But there is one serious nuance – the energy consumption setting was also included in the number of methods used. However, it was still impossible to do without it – there are more different blocks in the Core i9, their operating clock frequencies are also higher – so it should consume more. And the company practically does not limit his appetite – unlike others. Where and outstanding voracity. Really outstanding – all other processors are much more modest. As for the twelfth generation Core i5 and i7, they are strictly more economical than the eleventh, and even against the background of the older models of the tenth, they look acceptable.

Considering the increase in performance, it is difficult to criticize the Core i9-12900K either – it is inferior to its counterparts in terms of energy efficiency, but still bypasses all models of previous generations. As for the absolute level of electricity consumption, which is completely converted into heat, the target audience of such processors learned to deal with such problems a long time ago. And even often enjoys the very process of struggle, and not just the results. Therefore, many (though not all) buyers will be generally satisfied with the current state of affairs.

With intercompany competition, not everything is so smooth, since even the Core i5 and i7 are inferior to many AMD Ryzen models in terms of energy efficiency – but in absolute power consumption, on the contrary, they surpass almost everyone. However, Intel was also good at selling the LGA1200, where everything was worse in this regard, and in terms of performance, and even with functionality at the start. And with “perspective” – too. From all points of view, the LGA1700 looks much better than its predecessor. But AMD AM4 is already at the end of its life , and it also has its drawbacks, the correction of which has been postponed until AM5 (and whether all of them, time will tell). However, we have already digressed to global issues that we did not plan to touch on today. The main thing I would like to note is that Intel has done a very serious job on the new platform and processors. Could it have been better? Maybe. But what turned out in reality is much better than the company’s proposals of the last three or four years.

Total

We needed to close the gestalt with the starting trio of Alder Lake desktop models, but this is not the end of the introduction to the new Intel processor family. With the models that appeared at the beginning of the year, everything is much simpler: they are not hybrid, so platform compatibility with Windows 10, and generally different old habits , will come in handy here. In general, there is no brilliance of novelty – but there will be two, four or six powerful cores with the Golden Cove architecture. But there won’t be eight cores: the new Core i7 and Core i9 are made on a different chip (it is also traditionally used in the “overclocker” Core i5), which includes a pair of E-core clusters (Gracemont architecture). And the main difference between Core i7 and Core i9 in the new family is just the number of included clusters: one or two. Accordingly, if you are interested in a “pure eight-core”, it looks more logical to buy a Core i7 – it is much cheaper, and solves the task. Disabling one cluster of E-cores is more logical than two, and clock speeds and power consumption limits can also be manually moved. You can’t do anything with the L3 capacity, and E-cores in the right software environment only increase performance, so you still can’t talk about the sameness of these two processors, while in the previous generation Core i7 and Core i9 were not physically different, so there was no practical sense in the Core i9. Now there is a sense, but whether it justifies the difference in price is, as already mentioned, a purely subjective question. You will have to find the answer to it yourself – and it is not necessary to limit the choice only to processors of this family, since other products on the market have their own advantages.