We have already tested two processors of the Intel Alder Lake line – the top-end Core i9-12900K and the youngest of the older Core i5-12600K. For the first acquaintance with the new family, the “standard” testing methodology was used: all programs were running Windows 10, which is fully supported by the new processors – but there are nuances. As we already wrote, Intel, of course, conducted almost all internal tests under the control of this operating system (and there were no others during the development of Alder Lake), so, if possible, it took into account all the features of its operation. But some features cannot be changed without the participation of Microsoft.
Hybrid architecture for the x86 world is new. So there is nothing surprising in the fact that most of the software products for it until recently have not been sharpened in any way. And we are talking about both application and system software. In particular, the task scheduler of the vast majority of desktop operating systems has no idea that the processor can have cores of “different quality”. Similarly, at one time such a banal thing as correct support for SMT (in Intel processors is implemented through Hyper-Threading technology, the first versions of which appeared 20 years ago), the software also had to be specially trained. And even earlier there were problems with multiprocessor configurations. Well, as problems – just all sorts of Windows 9x, installed on computers with a pair of processors, “saw” only the first one. Further, multiprocessor support was constantly improved, but not by itself, but as a result of the work of programmers in close cooperation with processor manufacturers.
Hybridity is just a new stage of multiprocessing. And the stage, apparently, is necessary, since it is not easy and expensive to “fill” a processor with many large cores, for this you still have to make a lot of unpleasant compromises, and any (even a large) array of small cores is not capable of providing a high “single-threaded” performance – which was, is, and will be the main one in interactive programs (for a personal computer, in practice, this is, in general, much more important than peak performance in multi-threaded software, since it directly determines the comfort of work, and not some quantitative trifles) . And the only way to solve this problem is to use different kernels. Productive ones will “chew” heavy code streams and / or provide a quick response to user actions, while efficient ones will “rake up” background loads and sometimes help productive ones by additionally serving heavy threads (at a slower speed, but it’s still better than nothing).
However, for the full implementation of this concept, just the correct distribution of the load among the threads is required. Which, as we already wrote, are not even two, but three types – Hyper-Threading support by productive cores has not disappeared, but such computational threads are not fully independent. The main problem with Windows 10 and earlier versions of Windows is that threading is done statically, while the nature of the load can change dynamically. This does not give rise to fatal problems , as we have already seen, but sometimes the performance turns out to be lower than it could be, since the code flows that hit the E-cores remain there. As a result, in some applications, we observed the full load of these clusters with idle P-cores, which saves energy, but beats the speed. We turn off the E-cores (which we also did) – the processor becomes simple and understandable for the system, so the speed of work often grows radically.
We can conclude that for the “budget” (in quotation marks – since most Core i5 models also fall here) desktop processors of the new generation that are now appearing on sale, the use of Windows 10 will not lead to any problems at all: they have no E-cores at all , but there are only radically improved (compared to earlier developments) P-kernels. This will be especially useful for Core i3 and Pentium buyers: there are still a lot of applications for quad- and even dual-core processors, but these processors themselves have hardly changed since 2017 (the only real exception is the Ryzen 3 5300G, which has remained exclusively in OEM segment). New models with new performance and functionality, and even at a low price, are a good reason to upgrade. Moreover, there are no software “troubles” in this case.
But what should buyers of older processor models do, where they have to pay extra just for E-cores? And okay, even desktops – but in mobile processors the balance is generally shifted towards the latter. You can, of course, turn them off – and reduce the problem to an already solved one. But this is not our method. I would like all resources to really work. And the way it should be.
“The right way” is promised in Windows 11. Which, of course, does not cause much enthusiasm among the broad masses of workers : as already mentioned, the only way to make the aforementioned masses fall in love with some version of Windows is to release the next one. So this time the situation repeated itself: almost everyone (with the exception of the most persistent fans of ancient artifacts, such as Windows 7) somehow immediately forgot about the claims against Windows 10 and switched to searching for flaws in Windows 11. find the same. And not only imaginary shortcomings, but also quite real ones. Some of them may be fixed over time, and some will remain unremovable features that you need to get used to – but the workers are not set up to wait or get used to: they need everything at once. On the other hand, it is clear that all these questions of choosing a system actually affect only a very narrow (albeit noisy) layer – the majority calmly purchase ready-made computers and use them under the control of the OS that will be installed by the manufacturer. And “will be” there in the near future, as a rule, exclusively Windows 11. And not only in the latest systems with Alder Lake, but also in computers with the eleventh generation Core, and in those using AMD processors.
Tasks and testing methodology
What does it change? We decided to just take it and see. Why not right away? There was no certainty right away that the programs we used would correctly respond to a change in the operating system, and a complete update of the test methodology is not a quick matter. Along the way, there was time to try out the “old” one in a modified form – simply by taking and updating the OS (we know that the default settings for updating and for a clean installation are slightly different – but in this case it certainly does not interfere; rather, on the contrary). It turned out that nothing terrible was happening: almost all programs perceived Windows 11 as just another build of Windows 10 – which, by and large, it is. The only exception was the SolidWorks package, in which activation flies tightly (and not only the “native”, but even the pirated activator does not save) – this is a known problem in most versions of the program. Officially, Dassault Systemes for SolidWorks 2021 and below does not support Windows 11 – and is not going to: the new version 2022 will work with the new OS, which itself is not yet officially available. Accordingly, if this program is the main working tool, this is where the choice of Windows version ends. However, tests can be carried out in trial mode – which we took advantage of.
The testing methodology is described in detail in a separate article, and the results of all tests are available in a separate table in Microsoft Excel format , since testing today is special. 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 scores, 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.
Thus, we have the full results of a pair of 12th generation Cores in the same programs running Windows 10 and Windows 11. Comparing them is already useful, and if not only them, even more useful. Indeed – how does the performance of “obsolete” computers change and does it change? Yes, and new items in a vacuum are also not as interesting as when compared with their predecessors and main competitors.
Test participants
Intel Core i5-11600K | Intel Core i5-12600K | Intel Core i9-11900K | Intel Core i9-12900K | |
---|---|---|---|---|
Kernel name | rock lake | Alder Lake | rock lake | Alder Lake |
Production technology | 14 nm | Intel 7 | 14 nm | Intel 7 |
Core frequency, GHz | 3.9/4.9 | 2.8/3.6(E) – 3.7/4.9(P) | 3.5/5.3 | 2.4/3.9(E) – 3.2/5.2(P) |
Number of cores/threads | 6/12 | 10/16 | 8/16 | 16/24 |
L1 cache (total), I/D, KB | 192/288 | 256/128(E) – 192/288(P) | 256/384 | 512/256(E) – 256/384(P) |
L2 cache, KB | 6×512 | 1×2048(E) – 6×1280(P) | 8×512 | 2×2048(E) – 8×1280(P) |
L3 cache, MiB | 12 | 20 | sixteen | thirty |
RAM | 2×DDR4-3200 | 2×DDR4-3200 / 2×DDR5-4800 | 2×DDR4-3200 | 2×DDR4-3200 / 2×DDR5-4800 |
TDP, W | 125 | 125 / 150 | 125 | 125 / 241 |
Number of PCIe lanes | 20 (Gen4) | 16 (Gen5) + 4 (Gen4) | 20 (Gen4) | 16 (Gen5) + 4 (Gen4) |
Integrated GPU | UHD Graphics 750 | UHD Graphics 770 | UHD Graphics 750 | UHD Graphics 770 |
The main four of the test subjects look like this – since the new Intel models actually replace the old ones, and the recommended retail prices in each pair are as close as possible. Of course, we are not talking about the parity of real retail prices – and the environment (especially DDR5 memory) costs differently. Therefore, we will not be attached to prices – they tend to change.
AMD Ryzen 5 5600X | AMD Ryzen 7 5800X | AMD Ryzen 9 5950X | |
---|---|---|---|
Kernel name | Vermeer | Vermeer | Vermeer |
Production technology | 7/12 nm | 7/12 nm | 7/12 nm |
Core frequency, GHz | 3.7/4.6 | 3.8/4.7 | 3.4/4.9 |
Number of cores/threads | 6/12 | 8/16 | 16/32 |
L1 cache (total), I/D, KB | 192/192 | 256/256 | 512/512 |
L2 cache, KB | 6×512 | 8×512 | 16×512 |
L3 cache, MiB | 32 | 32 | 64 |
RAM | 2×DDR4-3200 | 2×DDR4-3200 | 2×DDR4-3200 |
TDP, W | 65 | 105 | 105 |
Number of PCIe lanes | 20 (Gen4) | 20 (Gen4) | 20 (Gen4) |
Integrated GPU | No | No | No |
Accordingly, we were also free to choose AMD processors for comparison, but we decided not to be free. We took three models – the first two are technically similar to the eleventh generation Core i5 and i9, and the third is already an ideological analogue of the Core i9. In fact, AMD’s fastest “desktop” processor is up against Intel’s fastest desktop processor. It would be possible to expand the list of subjects – but this will not give additional useful information, but it will take additional time: under Windows 11, we had to test everyone in general. Unlike most materials from the main line – where you have to tinker with only one or two main characters, and the results of the rest have already been obtained earlier. Fortunately, all the models we have today are “fast” – but still, none of them fit into a day.
To reduce the amount of work, we did not change anything in the hardware configuration. Accordingly, the LGA1700 has some head start in the amount of RAM – it is 32 GB here, and not 16 GB like the others. However, as already mentioned, we optimized the tests specifically for the second value (in fact, the results only slightly change even when switching to 8 GB of memory), but the new build version of the OS may make its own adjustments. But today, just such features are of no less interest to us than the comparison of different processors with each other in itself.
iXBT Application Benchmark 2020
Windows 10 | Windows 11 | |
---|---|---|
AMD Ryzen 5 5600X | 151.1 | 152.3 |
AMD Ryzen 7 5800X | 189.7 | 192.2 |
AMD Ryzen 9 5950X | 265.2 | 265.6 |
Intel Core i5-11600K | 145.8 | 144.9 |
Intel Core i5-12600K | 90.7 | 202.8 |
Intel Core i9-11900K | 176.1 | 175.2 |
Intel Core i9-12900K | 158.1 | 266.6 |
As we remember, two out of three programs worked “incorrectly”, and questions arose about the performance of the third. Switching to Windows 11 fixed all the problems – and even compared to disabling the E-cores, it added performance to both processors. As a result, there is nothing corny and there is nothing to compare – in normal operation, the Core i5-12600K easily overtakes the “obsolete” eight-core processors (this clearly includes not only previous Intel developments, but all AMD processors too), and the Core i9-12900K is simply equal to top-end Ryzen 9. It would seem that its success is less, but … You can find some kind of head start in the number of cores in the 12600K – after all, it has 10 of them. And in the second case, the cores seem to be equally divided – but in Ryzen they are all productive , and in Core half are just effective . However, everything is like in a joke about the funeral of a preference – and two was enough .
It is clear that this whole effect is more correct resource planning, since the performance of other processors simply did not change. More precisely, Ryzen started to work a little faster, but Core – a little slower. But both are somewhere on the verge of measurement error – so you can ignore it. Unlike Alder Lake, where mastering a new version of Windows is not something optional, but, perhaps, still mandatory.
Windows 10 | Windows 11 | |
---|---|---|
AMD Ryzen 5 5600X | 153.7 | 154.7 |
AMD Ryzen 7 5800X | 195.8 | 199.9 |
AMD Ryzen 9 5950X | 301.5 | 302.8 |
Intel Core i5-11600K | 149.9 | 147.8 |
Intel Core i5-12600K | 127.7 | 218.4 |
Intel Core i9-11900K | 195.7 | 193.6 |
Intel Core i9-12900K | 244.6 | 309.4 |
Well, all the trends remained in place. The performance of Ryzen from the transition to Windows 11 increases slightly, the old Core is reduced, the new Core is nothing to even compare. Disabling E-cores is still only a temporary method of solving problems. Yes, and not effective – if any, must be used. There is a lot of benefit from it – but with the “correct” use. Alas, Windows 10 is contraindicated in this case. And with Windows 11, it can be the same as with SolidWorks: there are no more silent films, but there is no sound movie yet .
Windows 10 | Windows 11 | |
---|---|---|
AMD Ryzen 5 5600X | 137.5 | 138.1 |
AMD Ryzen 7 5800X | 161.3 | 163.1 |
AMD Ryzen 9 5950X | 201.8 | 204.2 |
Intel Core i5-11600K | 142.5 | 141.8 |
Intel Core i5-12600K | 187.2 | 191.7 |
Intel Core i9-11900K | 167.2 | 166.2 |
Intel Core i9-12900K | 224.1 | 234.9 |
In this group, disabling the E-cores and previously only reduced performance – from which it could be concluded that the load is distributed “correctly”. However, it turned out that even in this case there was room for improvement. And this to a large extent also applies to Ryzen 9 – which added comparable to the Core i5-12600K. Here are the “previous” Core, on the contrary, once again slowed down a bit. But, we repeat, it is better never to consider fluctuations within 5% seriously. And, if you do this, it’s easier to consider both versions of Windows equivalent for these programs. Including for older Alder Lake models. Good news too, if you think about it.
Windows 10 | Windows 11 | |
---|---|---|
AMD Ryzen 5 5600X | 140.3 | 141.8 |
AMD Ryzen 7 5800X | 146.1 | 148.1 |
AMD Ryzen 9 5950X | 153.5 | 156.9 |
Intel Core i5-11600K | 131.5 | 131.7 |
Intel Core i5-12600K | 188.9 | 195.4 |
Intel Core i9-11900K | 140.5 | 139.9 |
Intel Core i9-12900K | 211.6 | 223.5 |
As well as the fact that imaging programs behave similarly. In general, for a creative designer , the systems are equivalent. As a first approximation, of course, there are a lot of various programs, so the list of “bad” ones we have discovered does not end there. And here Photoshop turned out to be like that, we recall – the difference in the test results is only about 10%, but it is there.
Windows 10 | Windows 11 | |
---|---|---|
AMD Ryzen 5 5600X | 186.3 | 184.4 |
AMD Ryzen 7 5800X | 245.9 | 245.5 |
AMD Ryzen 9 5950X | 436.3 | 437.3 |
Intel Core i5-11600K | 171.6 | 172.6 |
Intel Core i5-12600K | 248.3 | 245.2 |
Intel Core i9-11900K | 216.7 | 216.1 |
Intel Core i9-12900K | 388.9 | 389.3 |
One of two cases where even a “fix” for the scheduler does not allow the Core i9-12900K to keep up with the Ryzen 9 5950X. But nothing surprising – both processors are 16-core, but one simultaneously executes 24 code threads, and the other – 32. The code itself is simple, mostly integer and perfectly parallelized, so this also matters. On the other hand, we would not be surprised if this happens more often – after all, in Ryzen all the cores are “productive”. But it turned out that this is no longer the case 😉
Windows 10 | Windows 11 | |
---|---|---|
AMD Ryzen 5 5600X | 158.6 | 157.2 |
AMD Ryzen 7 5800X | 182.2 | 181.9 |
AMD Ryzen 9 5950X | 248.2 | 247.9 |
Intel Core i5-11600K | 162.8 | 161.9 |
Intel Core i5-12600K | 174.8 | 174.8 |
Intel Core i9-11900K | 187.6 | 187.6 |
Intel Core i9-12900K | 228.8 | 228.4 |
And here is the second – where the same reasons worked. It is possible that not only them – there is really a lot of cache memory in Ryzen 9, and this “loves” both FineReader and archivers. In general, there is so much in common between these two scenarios that it would be time to combine them into one.
Windows 10 | Windows 11 | |
---|---|---|
AMD Ryzen 5 5600X | 146.4 | 144.6 |
AMD Ryzen 7 5800X | 179.8 | 176.4 |
AMD Ryzen 9 5950X | 227.9 | 224.3 |
Intel Core i5-11600K | 136.5 | 136.1 |
Intel Core i5-12600K | 197.4 | 202.8 |
Intel Core i9-11900K | 169.0 | 168.7 |
Intel Core i9-12900K | 265.3 | 271.8 |
Another case where everything seemed right anyway – but there was something to improve. The increase is small – but the rest under Windows 11 even slowed down.
Windows 10 | Windows 11 | |
---|---|---|
AMD Ryzen 5 5600X | 152.7 | 152.7 |
AMD Ryzen 7 5800X | 183.7 | 184.6 |
AMD Ryzen 9 5950X | 250.0 | 250.9 |
Intel Core i5-11600K | 148.1 | 147.5 |
Intel Core i5-12600K | 166.2 | 203.4 |
Intel Core i9-11900K | 177.6 | 176.8 |
Intel Core i9-12900K | 238.0 | 270.0 |
Previously, the Core i9-12900K was slightly behind the Ryzen 9 5950X in the overall standings, but now it is ahead – but it should be understood that both are just a statistical result. There are scenarios in which the top AMD is still ahead – for objective reasons. It’s also more goofy – older Alder Lake models need the right environment for all their goodies. Which, as you can see, are many. And performance is one of them.
Energy consumption and energy efficiency
Max power | Minimum power | Average power | |
---|---|---|---|
AMD Ryzen 5 5600X (Win10) | 120.1 | 63.3 | 109.5 |
AMD Ryzen 5 5600X (Win11) | 103.7 | 46.5 | 93.4 |
AMD Ryzen 7 5800X (Win10) | 182.1 | 65.6 | 149.9 |
AMD Ryzen 7 5800X (Win11) | 166.6 | 50.2 | 131.6 |
AMD Ryzen 9 5950X (Win10) | 186.0 | 76.3 | 166.8 |
AMD Ryzen 9 5950X (Win11) | 182.2 | 67.0 | 153.5 |
Intel Core i5-11600K (Win10) | 208.3 | 67.5 | 161.8 |
Intel Core i5-11600K (Win11) | 205.1 | 60.3 | 161.9 |
Intel Core i5-12600K (Win10) | 170.0 | 55.7 | 122.6 |
Intel Core i5-12600K (Win11) | 181.1 | 59.8 | 145.9 |
Intel Core i9-11900K (Win10) | 233.9 | 68.4 | 198.2 |
Intel Core i9-11900K (Win11) | 219.5 | 65.1 | 183.2 |
Intel Core i9-12900K (Win10) | 310.9 | 62.9 | 186.7 |
Intel Core i9-12900K (Win11) | 303.0 | 63.5 | 230.3 |
But power consumption is not one of the strengths of the new Intel platform. And the fact that the new version of Windows “learned” to better load Alder Lake processors with work (more precisely, the scheduler stopped “mistakes” with the cores) only increases it. Interestingly, for other subjects – just a decrease. Especially noticeable on AM4 is almost without exception 15-20 W on average, and this is not so little. And not at the expense of performance. For the eleventh generation Core, not everything is so beautiful, but the “hot temper” of the i9-11900K also managed to be somewhat curbed. That is, it cannot be said that the changes are useful only for the latest platforms – in fact, Windows 10 and Windows 11 generally behave somewhat differently. And, if the performance of processors (with the exception of heterogeneous ones) turns out to be comparable, the new task scheduler managed to use some reserves of energy savings. Perhaps not only to him.
Windows 10 | Windows 11 | |
---|---|---|
AMD Ryzen 5 5600X | 1.39 | 1.64 |
AMD Ryzen 7 5800X | 1.21 | 1.40 |
AMD Ryzen 9 5950X | 1.50 | 1.63 |
Intel Core i5-11600K | 0.92 | 0.91 |
Intel Core i5-12600K | 1.36 | 1.36 |
Intel Core i9-11900K | 0.90 | 0.97 |
Intel Core i9-12900K | 1.27 | 1.17 |
However, it is easy to see that for the Core i5-12600K, the increase in (average) power consumption was offset by a change in (average) speed. For the Core i9-12900K, there is no such correspondence. On the other hand, compared to Rocket Lake, the new processors are much faster and noticeably more efficient. The competition with Ryzen is… ambiguous. The company succeeded in “pulling” the Core i9 to the level of Ryzen 9 in terms of performance, but losing in terms of energy efficiency. The lower models are in a better position, so they don’t have to be boosted so hard – with a natural result. But again, Ryzen are more energy efficient. This is especially noticeable under Windows 11 – which, apparently, is preferred for the latest AMD processors under AM4 (despite the horror stories of the first months after its appearance). So Intel still has something to work on – but AMD is not going to sit idly by either. Where this will eventually lead, time will tell. In the meantime, one thing can be said: Intel’s stagnation is over, new architectures and technical processes are much better than the old ones, but at the moment it has not been possible to return unconditional leadership (as it happened in the days of Core 2 Duo). You will still have to choose between the products of the two companies – they have different strengths and weaknesses.
Total
Most computer buyers do not choose an operating system – they just use the one that is installed. But sometimes you have to make a choice, including for some purposes you can update the OS on an existing PC – the benefit of switching to Windows 11 from Windows 10 is free (it is updated in the same way as 7 and 8.1 to 10 at one time). Sometimes the reverse need arises if some software or hardware requires exactly the “old” system. For example, we were faced with the fact that SolidWorks does not officially work under Windows 11 yet, although tests can still be carried out. And this is hardly the only such program – especially when it comes to professional applications or a variety of custom software in large companies. In principle, Microsoft did its best to ensure that the new version of Windows behaves just like a new build of the old one, but everything happened in the Windows 10 line with different builds. And even earlier, sometimes you had to “linger” on a certain Service Pack – because something turned out to be incompatible with newer ones.
In general, situations where you have to choose not even a family of operating systems, but a specific version of Windows have met, are encountered and will continue to occur. But this choice itself usually does not cause problems. Especially if we talk about self-assembled home computers, where many enthusiasts use officially unsupported versions of operating systems (and it doesn’t matter – for business or fun). As long as it works properly.
But “normal” for hybrid processors, as tests have shown, is suitable only for Windows 11. We emphasize that this is true for hybrid processors, and not for those desktop models of the Alder Lake family that have only P-cores, and not for the expected budget laptop processors with E-cores only. But in the near future, hybrid models will certainly appear in AMD’s assortment, and they are unlikely to behave in a fundamentally different way. Problems are caused by themselves “different” kernels in one device. Yes, many applications work fine on such processors and under Windows 10. We will not be surprised if some compatibility roughness is found under Windows 11. However, in our test suite, all the problems we found have disappeared. In addition, if necessary, the hybrid processor can be turned into a conventional one – at least as a temporary compromise solution.
The main conclusion is simple: it cannot be unequivocally stated that only Windows 11 is suitable for the new platform. Moreover, for many buyers of the LGA1700 desktop platform, nothing has changed in this regard: up to the Core i5-12600, all processors are “regular”, and hybrid models start only with Core i5-12600KF. But for the latter, the new version of Windows is preferred. It is in it that everything works as intended by Intel. In Windows 10 – when how. In general, this is a less serious problem than the need to abandon Windows 9x after the appearance of multithreading in x86 personal computers (it does not matter which method: SMP, SMT or CMP – even Windows 98 could not fully use any of the above). Rather, this is a problem with the compatibility level of Windows 2000 and processors with Hyper-Threading – and many have not noticed problems with this compatibility at all. But still, buyers of older desktop or most laptop processors of the Alder Lake family will have to take these features into account.