Motherboard reviews: What matters the most to you?

[dragontamer5788]

Theoretical values are so far off from reality, that real testing needs to be done

Electronic models and the way people teach electronics is surprisingly crap. (But apparently is good enough that enough computers happen to work?)

Slight changes to testing methodologies will lead to different results in practice. It's the nature of the field.

 

[agent_x007]

I would like to see which boards have better memory layout (usually support higher memory speeds), but this can be retrofitting nightmare on new platforms (adding higher speced memory to website is easy).
So, maybe instead, listing how many layers board has could be done (sinco not all list it) ?

Secondary functions tests :
Is Dual Bios working (manual swap/auto swap) ?
Does board "recover" from bad memory settings/CPU settings, or does it need a clear CMOS (if so, what it can recover from) ?
Is removing battery required to force a BIOS reset ?
Do we lose saved profiles after CPU swap ?
Is BIOS Flashback updating both BIOSes or only one ?
(on AMD) How many options are hidden in AMD Overclock menu vs. "regular" one ?
How long memory training takes and are there options in EFI BIOS to tweak it ?
CSM/TPM Toggles (on/off only, or are there more options, if so list) ?
RGB software test (system load vs. usefullness) ?

^Some random thoughts I got.

EDIT :
Forgot about the most import part...
Can your finger unlock latch of all PCI-e x16 (mechanical) slots when CPU tower cooler is mounted in vertical (ie. "[]" ) position with standard 25mm thick fan attached in front, and there is full size GPU (2 slot minimum, with a backplate), installed in each available PCI-e x16 slot (mechanical) ?
"Can't do it/card or cooler damage required" = Fail
^Could be also measured in number of cuts on your hand, lower = better /s

 

[Mussels]

1) Can we get 1 low end CPU SKU tested as well during the review, even if only abbreviated. Like yeah test the i9 or Ryzen 9 hard, but also see what changes in the BIOS and how well the board handles with the Pentium and the APU.

I want this myself, but it's tricky because then it just becomes a CPU review.

The closest i've been able to rationalise is testing with lowered power limits to match the cheaper CPUs, on board that have power/heat issues - if a board can't handle a 350W CPU, can it handle 250? 125? 65?

Electronic models and the way people teach electronics is surprisingly crap. (But apparently is good enough that enough computers happen to work?)

Slight changes to testing methodologies will lead to different results in practice. It's the nature of the field.

It's more like marketing at each step shows off the best possible outcomes.
VRM's advertise their amperage, but then those values might be 50 amps for 10 milliseconds at 25c at 15 volts - and then none of those criteria are met when used in a PC, so the end up much lower.

That's why the focus has been pushed to more real world testing (stressing the shit out of them, measuring everything possible - and 'whats possible to measure' has been expanded a few times)

I would like to see which boards have better memory layout (usually support higher memory speeds), but this can be retrofitting nightmare on new platforms (adding higher speced memory to website is easy).
So, maybe instead, listing how many layers board has could be done (sinco not all list it) ?

Secondary functions tests :
1*Is Dual Bios working (manual swap/auto swap) ?
2Does board "recover" from bad memory settings/CPU settings, or does it need a clear CMOS (if so, what it can recover from) ?
3Is removing battery required to force a BIOS reset ?
4Do we lose saved profiles after CPU swap ?
5Is BIOS Flashback updating both BIOSes or only one ?
6(on AMD) How many options are hidden in AMD Overclock menu vs. "regular" one ?
7How long memory training takes and are there options in EFI BIOS to tweak it ?
8CSM/TPM Toggles (on/off only, or are there more options, if so list) ?
9RGB software test (system load vs. usefullness) ?

^Some random thoughts I got.

EDIT :
Forgot about the most import part...
Can your finger unlock latch of all PCI-e x16 (mechanical) slots when CPU tower cooler is mounted in vertical (ie. "[]" ) position with standard 25mm thick fan attached in front, and there is full size GPU (2 slot minimum, with a backplate), installed in each available PCI-e x16 slot (mechanical) ?
"Can't do it/card or cooler damage required" = Fail
^Could be also measured in number of cuts on your hand, lower = better /s

Almost all of your questions have well known answers.

I added numbers to them
1. Since that has dozens of ways to trigger, that's impossible to test. It may trigger for me artificially doing so, but your system may just crash because you have RAM settings stable enough to POST, that crashes a second too late

2. Again with impossibilities - I am testing some RAM settings out, and you can alter things by 50Mhz and that'll change from POSTing but unstable to not POSTing to needs a CMOS clear... and that varies by the CPU and ram used, not by the motherboard.

3. That's never required on any board, that's what the CMOS clear jumpers are for.

4. You damned well should, never ever reuse a BIOS profile after a BIOS update or CPU change. EVER.

5. That's aorus only, and both always update.

6. The AMD menus are universal and always the same.

7. That's impossible to answer - every CPU is different, every RAM kit is different and people keep blaming 'training' with 'I cant overclock RAM'. I've found the slow 'training' only happens with unstable settings, the moment stable settings are used you get *one* slower POST only (approx 22 seconds, only with high RAM speeds) and then every following boot is faster.

8. ? Every board has on/off/auto as the only options for this. The moment things like rebar are used, off is the only option.

9. I'm definitely checking for BIOS RGB settings to avoid software bloat, too few boards have even the ability to turn RGB off, let alone control it.

 

[Shrek]

BIOS support seems important to me; how long did the vendor support previous boards.

 

[rusty caterpillar]

1.BIOS and drivers.
2.Good onboard sound codecs.
Not codecs like 4080 and 4082 with hissings and pops not fixable even trough firmware. Also bad implementation methods trough USB that cause noise and other issues.
3.Positioning of M2 drives slots as far as possible from the hot video cards and main PCIe.
4. No PCBs written all over with the brand or some stupid phrases like "dare to game" > "GL, cause this board has no proper memory support" "dare to pay us? >get some BSOD party" increasing the overall cost for some nonsense gibberish.
5. Better screening from EMF of sound chips and network chips.
6. Mainboard manufacturers should stop making us pay for unwanted features - Wi Fi and RGB. In some cases "gold" capacitators which are not there actually. Just the higher cost is there.
7. Reintroduce back on every motherboard tier the error codes checker.
8. No malware driver installers, malware and miners OC suites and FAN/ RGB controllers. We had over a decade your malware, is time to stop.
9. When you disable something BIOS - in windows that device should not be visible- creating more vulnerabilities. No thanks, Asrock.
10. Nobody should void their warranty when using a Beta BIOS from the manufacturer website. " We want to test on your machine but, we also want you to pay in case...our lazy codes bricks your mainboard"

 

[P4-630]

1.BIOS and drivers.
2.Good onboard sound codecs.
Not codecs like 4080 and 4082 with hissings and pops not fixable even trough firmware. Also bad implementation methods trough USB that cause noise and other issues.
3.Positioning of M2 drives slots as far as possible from the hot video cards and main PCIe.
4. No PCBs written all over with the brand or some stupid phrases like "dare to game" > "GL, cause this board has no proper memory support" "dare to pay us? >get some BSOD party" increasing the overall cost for some nonsense gibberish.
5. Better screening from EMF of sound chips and network chips.
6. Mainboard manufacturers should stop making us pay for unwanted features - Wi Fi and RGB. In some cases "gold" capacitators which are not there actually. Just the higher cost is there.
7. Reintroduce back on every motherboard tier the error codes checker.
8. No malware driver installers, malware and miners OC suites and FAN/ RGB controllers. We had over a decade your malware, is time to stop.
9. When you disable something BIOS - in windows that device should not be visible- creating more vulnerabilities. No thanks, Asrock.

It's not about what you prefer on a motherboard.

It's about reviewing motherboards, what you'd like to see/read in a motherboard review.

 

[dragontamer5788]

It's more like marketing at each step shows off the best possible outcomes.
VRM's advertise their amperage, but then those values might be 50 amps for 10 milliseconds at 25c at 15 volts - and then none of those criteria are met when used in a PC, so the end up much lower.

Yes, but also not necessarily.

I've just finished studying capacitors and have noticed three values for "ESR": Impedance, ESR, and Dissipation Factor. All three have different tests associated with them but ultimately give you the "parasitic resistance" associated with a real capacitor (instead of an ideal mathematical capacitor, which as ESR of 0). But depending on how you do the test, this ESR value changes dramatically, by as much as 300% in a recent capacitor I used in a recent project.

Electronics as a whole are just a bunch of little lies that we tell other EEs that fit simple models... models that no one bothered to prove correct in the first place. Its not malicious, its just really complex, so these simple models have significant warts in their predictive behavior. In most cases, you'll only get similar results with the tests if you exactly follow the other guy's test setup. This naturally applies to your tests as well, its likely going to be difficult for others to replicate your tests even in the best of circumstances.

--------

Where things start to go sideways is that the motherboard is a real component in this whole setup. A relatively thick 35um (standard PCB height of copper) 1mm wide trace (most traces are on the order of 0.15mm on a modern PCB) with 6-inches of length will have 0.0769 Ohms of resistance, sounds small... except at 50 Amps that's a 3.845V drop. Of course, real world engineers know that even a 1mm wide trace is insufficient for 50Amps, but there's a lot of subtle effects here that are certainly affecting your measurements.

Another aspect: each via / drill-hit has a limited amount of current they support as well, those holes have limited amounts of copper in them after all, so there's complications with pulling power up-and-down vertically through a board. Especially because of the high currents that modern computers pull these days.

And that's not even getting into trace-capacitance and trace-inductance. I'm just covering the simpler resistive-like effects of modern PCBs. But at 10-millisecond timescales, trace-inductance and via-inductance (aka: the holes/drill locations) will absolutely play a role.

---------

But yeah, the goal here isn't to "understand" the motherboard per se, but instead to test them. The reader is likely a video gamer who ultimately just cares about FPS, and vague concepts like "reliability" that are basically untestable. Frames-per-second should (in theory) be modeled as power-loss in the VRMs and Motherboard (akin to the ESR tests I started this post with the real-capacitor vs ideal-capacitor). The less power that gets to the CPU, the more the CPU will throttle back and slow down.

After all: all the power that goes into heating up the motherboard and VRMs is power that's lost on the way to the CPU.

The good news to all this complications means that you too can lean into the truth of EE, and recognize that you're creating a model (whether you explicitly have acknowledged it or not), and you can choose models that are easier to test. Other EEs will accept your model for what it is, your best guess given the level of effort you want to put into these tests. Don't feel like you need to "find the truth" with your tests, but instead find something that's a good balance of "truthiness" and "easy enough to repeat". I don't think the Masters-degree or PH.D EEs at universities have figured this crap out yet, so lean into it and find a model that works.

-------

Where does this all go? What is the easiest thing to do? Well... maybe its as simple as taking a temperature-measurement of the motherboard and VRMs after 10-minutes of real world usage. That's certainly one valid way to measure the power-loss in the motherboard/VRMs itself, as long as you can keep the temperature effects somewhat consistent. Not saying you should "really" do this, but just kinda telling you this to help the creative juices flow. Of course, the easiest thing could just be reporting FPS numbers in various video games, which would be the "direct" result people would be clicking on these reviews for.

 

[rusty caterpillar]

It's not about what you prefer on a motherboard.

It's about reviewing motherboards, what you'd like to see/read in a motherboard review.

Arrh damn! my bad. Too much wine

 

[InVasMani]

I don't feel you can please everyone with MB reviews, but you can try to do a good job of showcasing typical usage and expectations of the board's design appeal either in terms of overall features especially sticking points in the design that stand out like WIFI7 or 10G or memory support or external clock generator and general OC support and VRM robustness. Like look at the board design and try to think what would most people buying it be most interested in and cover those area's.

 

[Mussels]

BIOS support seems important to me; how long did the vendor support previous boards.

That's relevant in a way but not something that's viable to predict the future.

AM4 for example, AMD themselves forced boardmakers to support older boards out of nowhere (5800x3D support on 300 series chipsets) - but they could have gone 3 years with no updates prior to that, and that would vary per BOARD let alone per series.

I don't feel you can please everyone with MB reviews, but you can try to do a good job of showcasing typical usage and expectations of the board's design appeal either in terms of overall features especially sticking points in the design that stand out like WIFI7 or 10G or memory support or external clock generator and general OC support and VRM robustness. Like look at the board design and try to think what would most people buying it be most interested in and cover those area's.

I got sick with a cold, then covid, then covid and the flu. I could review plenty of tissue brands quite well, but the motherboard side fell behind.

A really distilled summary would be:

How any first-time user would feel, using this board.
Then, a smaller section on how experienced users would feel using it.


Anything beginner friendly, is also experienced user friendly. The reverse is not true.

As an example i'm planning to physically measure and monitor Vdroop and SoC/SA voltage if at all possible. The average user will just skip that, they don't care.
We can't make them care without videos of CPU's exploding.

What do both care about? XMP/EXPO support and how the system behaves on a failed boot.

What do i want people to care about? Avoiding the pitfalls of using data from other types of reviews.

An RTX 4090 review requires the worlds fastest CPU at it's highest clock speeds with no power limits, or it becomes a limiting factor and you aren't able to fairly test the GPU.

That CPU performance is nothing like a home user would experience, since their CPU will be limited by the PL1/PL2 or PBO settings in their BIOS, and their XMP/EXPO settings.

If i had to distill the argument down as simple as possible, it's that people are seeing the ST performance of 125W to a single 14900K P core, then see it has 20 cores and think "Ah yes, my performance is this times 20"
When after 60 seconds or X cores, that first core slows down massively.


These are CPU architecture problems, but how the motherboard deals with them controls everything about their performance so it's my job to figure this out - despite in theory, being something that should be covered in the CPU reviews (which they cant, as they'd also need multiple motherboards)

Electronic models and the way people teach electronics is surprisingly crap. (But apparently is good enough that enough computers happen to work?)

Slight changes to testing methodologies will lead to different results in practice. It's the nature of the field.

Because everything has wild ranges of tolerance, and marketing.


A device could have a 20v maximum and a 10A maximum - marketing could call that 200W peak.

The unit itself may only handle 10V and 5A for more than a single second (50W sustained) but still be "true" so marketing it that way is legal.
When each component does this, the total values at the end of the chain are massively blown away from reality.


At least now with known wattage limits like PL1/PL2, testing can be focused on those values to save me effort and time

 

[InVasMani]

Sounds like you've been thru quite a bit of hell in a relatively short period of time. Covering the full scope of MB options would be a huge endeavor and peoples still would have a hard time wrapping their heads around all over it in key places without a lot of careful reading and without it being very concise and well written at the same time.

My experience with Asus Z790-H STRIX WIFI has been postive thus far and has lived up to my expectations well enough. It feels nice quality and well built overall. The ASROCK Nova looks quite nice as well though, but either are good and I didn't really need like 6 M.2's on board. The WIFI 7 would have been great, but you always swap in a newer adapter for that.

 

[Wirko]

What do both care about? XMP/EXPO support and how the system behaves on a failed boot.

As you mentioned that... Do you have any plans to test the systems with four RAM sticks too, to see if even at relatively low speeds of 4400 or so, the motherboard can be a limiting factor?

 

[Mussels]

As you mentioned that... Do you have any plans to test the systems with four RAM sticks too, to see if even at relatively low speeds of 4400 or so, the motherboard can be a limiting factor?

That's all in the CPU.
I've already ran into that myself with my personal AM4 motherboards, i can swap the half a dozen AM4 CPUs i have around and memory support changes with them.


My results with a 12900K and 14900K would be entirely different, and they'd be useless to users who bought any other CPU. It's like on AM4 where we took ages to find the 'average' supported speeds (3200 for almost everyone, 3600 with SoC voltage tweaks) - while the rest was basically blind luck as to how your motherboard and CPU chose to behave as a combination. Change either one and results would be whacky.

GA-AX370 Gaming 5 with 5800x, 3133MT/s maximum. Higher had cold boot issues. Same board with a 2700x ran 3200 just fine.
That 5800x then runs 3800MT/s stable on every other board i've got - just an easy example of how it's always combinations that matter.

Boards are often tuned for certain types of RAM too, early AM4 was tuned in for single rank samsung B-die since thats all they had to test the boards with at high speeds in the early days - so as time went on BIOS updates changed timings to work with newer RAM as it came out. Four sticks of B-die may have been easy at 3600, but impossible on hynix.

 

[kanamari]

Just recently, I figured out that motherboard layout is quite important in reviews. With many third party heatsink for ssd I think it is quite important to know the clearance of ssd heatsink (many online market reviews suggest that they cannot fit the heatsink to motherboard because it is located too close to GPU, RAM, or CPU heatsink

 

[OkieDan]

4. You damned well should, never ever reuse a BIOS profile after a BIOS update or CPU change. EVER

It's a shame we shouldn't expect MBD Mfgs to write a profile loader intelligent enough to store what CPU/RAM was used in a profile and to only load settings from a profile that are compatible with the current hardware. They probably never will if we just accept what they give us instead of expecting better from them. Maybe there's an Mfg already doing a good job at this and should get kudos in a review so other Mfgs take notice and do the same.

 

[ir_cow]

It's a shame we shouldn't expect MBD Mfgs to write a profile loader intelligent enough to store what CPU/RAM was used in a profile and to only load settings from a profile that are compatible with the current hardware. They probably never will if we just accept what they give us instead of expecting better from them. Maybe there's an Mfg already doing a good job at this and should get kudos in a review so other Mfgs take notice and do the same.

There is so much that goes on behind the scenes. I much rather my settings be whipped out per update instead of killing my CPU.

 

[Space Lynx]

@Mussels hope you are doing well

I just bought my dream mobo, so thought I'd share why I bought it:

the only Mobo in existence that has 20k caps to my knowledge, I know this doesn't really matter, but I think it's a nice touch, and one of the reasons I have always like the Steel Legend line is they came with 12k caps, when other brands usually only do 5k caps.

8 layer PCB which I have heard rumors makes EXPO work better, not sure if it true or not, but none the less nice to have... @ir_cow any thoughts on this?
beautiful white/gray theme matches my white ram nicely (I think I will call my rig Rubrum (Latin for red) Paladin) white Mobo/white ram, silver heatsink, all Paladin theme, then all AMD for red and my case is red/black for contrast/AMD as well. so yeah Rubrum Paladin it is.
gen 5 graphics slot (I may upgrade to rtx 6090 in 2-3 years so it may indeed come in handy)
gen 5 storage (don't really care but nice to have)

optical port for my Modi DAC, which is nice cause lot of AM5 even high end ones, don't have optical strangely enough
wifi 6e, I won't use and it will stay shut off most of the time, but nice to have if I ever move house someday and lose access to ethernet
good amps and power delivery - 14 phase and 80 amps
great vrm cooling, this baby will be able to handle 11800x3d, which is when I plan to retire the 7800x3d

price was only 179 for all these features, and it beats out higher end mobo's in various areas, which is kind of strange
It ended up only being a little more extra out of pocket over my original choice of Mobo which was the ASRock b650m HDV board, and for the prices paid, I felt that was a fair price for all the bonus/icing on cake things you get.

 

[Shrek]

20k hours at what temperature?

 

[Space Lynx]

20k hours at what temperature?

I just took a flashlight and looked up close on the Mobo, the cap says:

20k
560j
d23

I have no idea what that means, but AsRock has always advertised 12k caps as being a benefit over 5k caps on other steel legend boards. I have no idea to tell for sure though to be honest. I would doubt they would lie, but you are right it doesn't state temps anywhere.

 

[Shrek]

20k is probabaly not hours

 

[Space Lynx]

20k is probabaly not hours

from their official website: this picture says d19, mine says d23, no idea what the difference is though lol

source:

ASRock B650 Steel Legend WiFi

Supports AMD Socket AM5 Ryzen™ 8000 and 7000 Series Processors; 14+2+1 Power Phase, 80A Dr.MOS for VCore; 4 x DDR5 DIMMs, Supports Dual Channel, up to 7200+ (OC); 1 PCIe 5.0 x16, 1 PCIe 3.0 x16, 1 PCIe 4.0 x1; Graphics Output Options: 1 HDMI, 1 DisplayPort; Realtek ALC4082 7.1 CH HD Audio Codec...

www.asrock.com

 

[Shrek]

Ah, so 20K is indeed hours, good to know. My guess is at 105°C

 

[Space Lynx]

Ah, so 20K is indeed hours, good to know. My guess is at 105°C

and since they rarely if ever get that hot, realistically its probably more like 40k hours? which is why this is more of a nice detail/icing on the cake thing and not actually a big deal in the world of mobo's? a 5k caps probably lasts 15k because they never reach 105 Celsius, except on badly designed boards or bad airflow cases, etc

is my thinking on this right or no? regardless, I still think its cool to have 20k caps, just a nice detail

 

[Shrek]

No, way way more

For every 10°C cooler, the life approximately doubles.

 

[Space Lynx]

No, way way more

For every 10°C cooler, the life approximately doubles.

ok so this is really icing on the cake, makes sense why no other mobo advertises it. still cool to have as a nice attention to detail thing, but really doesn't matter.

I wonder how hot they actually get during an average gaming session, its probably not much at all really