An examination full of new knowledge
Over time, we have tested ten motherboards with Intel B660 and Z690 chipsets in great detail. From more than 5000 different measurements, we can thus confidently deny some speculations that are usually spread on the Internet from the ignorance of the authors of articles or comments (in discussions). But this is natural. The less substantial the reviews are, the more fertile ground they create for various confusions.
An examination full of new knowledge
According to our research, nobody is really doing in-depth motherboard tests. The reason for this is quite simple – to test a motherboard from all perspectives is an extreme amount of work that readers (or in the context of videos, viewers) won’t appreciate – “they are not interested”. Motherboard test views are some of the lowest among components. In proportion to the time spent, from the editors’ point of view, this would be time wasted that could be spent more productively.
The lower popularity of motherboard tests (not only in our country, but worldwide and historically) is certainly due to several reasons. One, and probably the main one, is that it is not a topic that would divide people into several camps, where some people are fans of one or another brand and regardless of the test results they stand by it, they feel the need to shout their opinions to the world and in short they create a certain buzz. This happens typically with processors and graphics cards.
Motherboards do not arouse much “passion”, that is, unless flames from burning voltage regulators appear in preview photos. But even this is artificially created interest, usually based on ill-conceived methodologies, where processors that are too powerful are used to test cheap motherboards. And yet those “bad” motherboards don’t expect to run with 300W Core i9s, but are perfectly designed for, say, 65W Core i5s.
Motherboard tests are also quite difficult to understand, as long as they are done correctly. And the many obscurities that surround them don’t help either. We do try to set them straight with detailed analyses, but let’s say the “generalities” in motherboard subtests don’t stand out that much and deserve attention in a separate article. We group the individual findings and generally valid conclusions thematically, by chapter – one topic, one chapter. The biggest “issues” are discussed in these chapters.
But then there are also a few things that we didn’t target from the beginning (of the motherboard tests) and only started to show up over time. For example, the different quality of PWM control on motherboards. We only picked this up in the Gigabyte Z690 Gaming X DDR4 test, where too high speeds were being achieved in basic tests of the connectors on the system fans with the Noctua NF-A12x25 PWM. Higher than on other motherboards.
After testing the 10 more motherboards, we will surely come across other things that remain “under wraps” for the time being. With that we hope that even these, unattractive tests, are of value to you. We’ll certainly stick with them and we share the sentiment that sometimes you have to do things that don’t pay off, as long as they make sense. Now, on to the important stuff (on the next pages of the article).
Over time, we have tested ten motherboards with Intel B660 and Z690 chipsets in great detail. From more than 5000 different measurements, we can thus confidently deny some speculations that are usually spread on the Internet from the ignorance of the authors of articles or comments (in discussions). But this is natural. The less substantial the reviews are, the more fertile ground they create for various confusions.
Possible differences in CPU performance
As you know, we ran all motherboards with two processors in all performance tests. In addition to the Core i9-12900K, the weaker Core i5-12400 model. Why? For the hypothesis that a more expensive motherboard can boost the performance of a cheaper processor. And that’s misrepresentative in CPU tests with builds built on expensive boards. Supposedly… However, tests have clearly shown this hypothesis to be false.
Sure, it may happen that a cheaper processor will perform better on a more expensive board, but it’s definitely not possible to generalize. In fact, the exact opposite can also be true, namely that the higher performance of the cheaper processor will be achieved on the cheaper board. This is always determined by how which board approaches which processor. If we are talking about the smallest differences in performance, then it is definitely not enough to remove the power limits or on the contrary to enable them (and limit Tau after a short time limit, typically to the TDP/PL1 level). Boards approach the settings of different processors individually within their management, for example also with respect to whether the application uses (then most boards have a negative offsets preset for the multiplier and the all-core boost tends to be 100–200 MHz slower) or does not use AVX instructions.
For example, the x264 video encoding test in HandBrake perfectly refutes the claim that a weaker processor performs better on a more expensive board. On the contrary, on the most expensive board tested, the Asus ROG Z690 Maximus Hero, the Core i5-12400 processor is the weakest, but with the Core i9-12900K, this board already achieves the highest performance.
This is because the BIOS optimizations were geared more towards the Ci9-12900K, which we were counting on using, rather than the locked Core i5, which only we installed in the board. Surely, among hundreds of our tests, you would find such results that will be a good “proof” that in an expensive motherboard a weaker processor delivers higher performance. However, these are always the exceptions, and in the end you’ll find that higher performance than with a 700-euro motherboard can be lower even with the Core i9-12900K than on one of the cheapest Intel Z690 chipset motherboards – the Gigabyte Z690 Gamimg X DDR4.
On average, however, the board used doesn’t matter in terms of CPU performance, unless it has an outright undersized VRM. Whether the motherboard has a B660 or Z690 chipset doesn’t play a role for CPU performance either. The days when there were multiple chipset manufacturers (including Nvidia) where the chipset had an impact on performance are long gone. After moving the north bridge to the processor, the south bridge doesn’t really have an impact.
Over time, we have tested ten motherboards with Intel B660 and Z690 chipsets in great detail. From more than 5000 different measurements, we can thus confidently deny some speculations that are usually spread on the Internet from the ignorance of the authors of articles or comments (in discussions). But this is natural. The less substantial the reviews are, the more fertile ground they create for various confusions.
Power draw under control…
“Processor power draw from a 24-pin”, this is sometimes the argument of testers, who derive processor power draw by system power draw with a wattmeter on the wall socket. Aside from the large distortions that arise at the level of varying chipset power draw, the controllers used, or the effect of ARGB LED lighting on overall power draw, this too can be described as a myth.
In addition to the power draw on the CPU power cables in the motherboard tests, we also analyze the power draw on the 24-pin connector. We have every single wire under control. This means that we know what current ever flows through the 12-volt, 5-volt and 3.3-volt wires.
There are various speculations as to which voltage from the 24-pin powers what. And to some extent it can be different, depending on the specific design of a particular motherboard. These are always small things like some manufacturer doesn’t trust the weak 3.3V branch (with very low current load) and therefore even for devices where the VRM works with 3.3V (typically SSDs), this voltage is produced (changed) from stronger branches, often also from the 12-volt one.
The main load on the 5-volt branch is memory. Some materials state that the memory controller of AMD processors could be powered through the same delivery (through the memory phase). How it really is you will see in motherboard tests, we won’t comment on it now. In the case of the Intel platform, however, there is no 24-pin 5 V (or 12 V) power supply to the CPU. From an electrical design point of view, especially in the case of a monolithic chip, this would also be a very awkward solution.
The fact that the 12-volt 24-pin wire is guaranteed not to power the processor is evidenced by the very low draw on it in a pure processor load. With one exception (Asus TUF Gaming B660 Plus WiFi D4 with obviously more aggressive graphics card management in that mode), the 24-pin 12-volt wires draw 3.7–4.7 W at maximum load. This is used by the graphics card in idle mode, which we have also verified with measurements directly on the PCI Express slot.
The power draw on the 12 V 24-pin increases only under gaming load, because the internal power supply of the graphics card is provided along this route. The difference in regulating the internal power supply (i.e. through the slot) of the graphics card is minimal across boards.
We consider evaluating CPU power draw by deriving it from system power draw at the socket to be irrelevant for the reasons stated above, and the fact that “everybody does it” doesn’t change that.
Over time, we have tested ten motherboards with Intel B660 and Z690 chipsets in great detail. From more than 5000 different measurements, we can thus confidently deny some speculations that are usually spread on the Internet from the ignorance of the authors of articles or comments (in discussions). But this is natural. The less substantial the reviews are, the more fertile ground they create for various confusions.
… more efficient VRM does not equal lower power input
In the context of measuring CPU power draw, there is one more thing you need to know. Namely, how big the difference in total power draw can be for the same computational (or gaming) power, which already includes VRM inefficiencies.
For this reason, we use sufficiently oversized boards so that the VRM efficiency does not cause significant differences between processors of different performance. However, VRM efficiency is not the only thing that determines what power draw you measure.
Boards with typically “strong” VRMs overdrive the power supply a bit in order to achieve the highest possible stability. Conversely, some boards with “weaker” VRM (with lower current carrying capacity) push it in the opposite direction. That is, the power supply is optimized to be as low as possible, so as to heat up the voltage regulators as little as possible for self-preservation.
A prime example is the Gigabyte B660 Aorus Master DDR4, with which the lowest power draw is achieved with the Core i9-12900K, although this motherboard does not have the most efficient VRM. It is much more efficient with the Asus ROG Z690 Maximus Hero with higher measured power draw.
In both cases, the difference in computational power is significantly less than the difference in input power. Hence, in processor tests it makes no sense to fixate on very small differences in power draw, these can vary from board to board and there is no way to fight with limiting this bias. But it is important to know about it, to be aware of it.
Over time, we have tested ten motherboards with Intel B660 and Z690 chipsets in great detail. From more than 5000 different measurements, we can thus confidently deny some speculations that are usually spread on the Internet from the ignorance of the authors of articles or comments (in discussions). But this is natural. The less substantial the reviews are, the more fertile ground they create for various confusions.
Motherboards are also network, slots and…
In addition to the impact of the motherboard on CPU and graphics card performance, we also measure the bandwidth of Ethernet, M.2 SSD slots and USB ports. Here it is nice to note that in the vast majority of cases there are no big performance differences across the boards and the performance is usually at the limit of the bus bandwidth (or the test device).
Ethernet adapters tests
We use a Sonnet Solo 10G 10-gigabit card to test network adapters within a local area network, and nine of the ten boards with 2.5 Gb adapters achieve speeds of 286–295 MB/s in both directions. These are very close to the limits of 2.5-gigabit connectivity.
We only saw significantly slower speeds with the MSI MAG B660M Mortar WiFi board with slow uploads. What exactly is behind this is hard to say. The chip used is a Realtek 8125BG similar to other boards that achieved expected speeds. When we encounter something similar again, we will subject it to a more rigorous analysis.
SSD slot tests
M.2 SSD speeds are virtually identical across the board, and the difference between the fastest and slowest SSD is 2 %. Significantly slower speeds like 6700 MB/s (sequential read) and 4950 MB/s (sequential write) with the Samsung 980 Pro SSD are only achieved in slots that either only support PCI Express 3.0 or are only connected from the chipset by two lanes, or a combination of both. This is the situation, for example, on the ASRock B660 Steel Legend (in the second M.2 slot on the SSD).
Over time, we have tested ten motherboards with Intel B660 and Z690 chipsets in great detail. From more than 5000 different measurements, we can thus confidently deny some speculations that are usually spread on the Internet from the ignorance of the authors of articles or comments (in discussions). But this is natural. The less substantial the reviews are, the more fertile ground they create for various confusions.
USB port tests
The picture is similar for the USB connectors, of which we are testing the external ones (on the rear panel). You don’t have to worry about ASRock giving you slower ports than, say, Gigabyte. Speeds across standards are always the same, or with differences of less than 1 %.
On rare occasions, however, you may find that you get significantly faster USB ports than you thought you had. For example, the MSI Pro Z690-A DDR4 is presented in the specifications with fewer 5 Gb USB 3.2 Gen 1 ports than it actually has. According to the specs, there should only be two to four USB 2.0 ports. But that ratio is reversed, four are faster ports with 5 Gb speeds, although two of them deceive with the body and mask the black color, typical for USB 2.0 ports.
This is probably due to cheaper production (with that faster configuration), as they don’t mass produce two different designs, but only one, shared with the more expensive boards. But at the same time, MSI probably needs to increase the attractiveness of the higher-end, higher-margin boards to give more incentive to buy them. And that’s because of the larger number of faster ports, for example, which this board actually has (MSI Pro Z690-A DDR4).
Over time, we have tested ten motherboards with Intel B660 and Z690 chipsets in great detail. From more than 5000 different measurements, we can thus confidently deny some speculations that are usually spread on the Internet from the ignorance of the authors of articles or comments (in discussions). But this is natural. The less substantial the reviews are, the more fertile ground they create for various confusions.
SSD coolers are good enough, but they waste material
Motherboard are also usually equipped with SSD coolers. Quite a lot has changed since MSI’s thin plate that kicked off the M.2 SSD cooler era, so to speak.
Instead of a thin shell, which can also act as a thermal insulator in certain circumstances, it is usually already solid aluminium profiles. These always work well from a cooling perspective and you don’t have to worry about even the most powerful SSDs overheating. And certainly not in a gaming PC, where SSDs can do without a heatsink, but that’s another story, which we’ll also get to sooner or later.
In any case, SSD coolers on motherboards always knock a few degrees Celsius off the controller and SSD memory temperatures compared to if they weren’t there at all.
However, when we take it from the point of view of design efficiency in terms of cooling power/weight ratio, it is already significantly weaker compared to standalone coolers. There is some hint of a more articulated surface, but the “brute force” – the amount of material used – prevails. This often outperforms third-party coolers, such as those from BeQuiet!, many times over, for the same cooling performance. We recently compared a cooler from them (MC1) to motherboard solutions and it was in the top half in performance despite its below average weight.
For the most efficient cooling of SSDs, motherboard manufacturers should also work on other things than heatsink designs. Namely, that the connector (M.2) and the top of the standoff post for screwing the other side of the SSD should be in perfect alignment. This doesn’t always happen and even when you mount the SSD correctly, with adequate force, there is often some flexing, bending. To which side, it depends on the characteristics of the standoff post and, of course, the influence of the heatsink itself, which may not be perfectly flush with the SSD across the entire surface. But we won’t elaborate on that now, there will be room for it later, in the thematic test, which is full of measurements.
English translation and edit by Jozef Dudáš