What are you looking for in a Motherboard Review?

[TheoneandonlyMrK]

To be clear on this, you are for example copying data from one M.2 to another M.2, while another one is writing data to SATA AND you are using the forth M.2 to copy data to USB device?

I can see normal use case for copying between two drives, but not all at once.

Raid 0 on three nvme with another as a scratch drive on the south chip plus external via USB 3.2 nvme.

 

[ir_cow]

Raid 0 on three nvme with another as a scratch drive on the south chip plus external via USB 3.2 nvme.

Okay well that is a bit different lol.

Raid-0 a bit extreme. First off, no one who actual cares about their data is going to do that. Secondly ,who needs 18GB/s internal read/write speed on their desktop. A single Gen3 M2 is good enough for 8K RED footage, this is just overkill in my opinion.

Also you will be bandwidth limited by the CPU uplink.... But maybe that's what your getting at here?

 

[TheoneandonlyMrK]

Okay well that is a bit different lol.

Raid-0 a bit extreme. First off, no one who actual cares about their data is going to do that. Secondly ,who needs 18GB/s internal read/write speed on their desktop. A single Gen3 M2 is good enough for 8K RED footage, this is just overkill in my opinion.

Also you will be bandwidth limited by the CPU uplink.... But maybe that's what your getting at here?

Also data lane usage on sata can affect it as can usb 3.2 so valid.
Backed up games and files aren't essential.

 

[ir_cow]

Not to be rude, I'm just having a hard time understanding a use-case for that type of situation.

SATA, USB and M.2 drives on the PCH do all share lanes to the CPU. Now the question. How is it actually routed and what is the best way to test that.

I pulled this from the manual online. It shows how things are broken up.

 

[W1zzard]

How is it actually routed and what is the best way to test that.

As far as I understand there is no freedom and only one way to do storage. Of course you can choose to not have certain ports on your board, but it's kinda obvious that they are missing when they are missing

Maybe "which M.2 is connected to the CPU, which to chipset", that's mentioned in all manuals I think

 

[Assimilator]

Raid 0 on three nvme with another as a scratch drive on the south chip plus external via USB 3.2 nvme.

Sorry mate, you are an edge-case of an edge-case. Very few people have two NVMe drives, let alone 5; even fewer have them in RAID-0.

 

[TheoneandonlyMrK]

Sorry mate, you are an edge-case of an edge-case. Very few people have two NVMe drives, let alone 5; even fewer have them in RAID-0.

Are you the reviewer.
I appreciate your point but.

Your saying that because most use one nvme and a GPU.

That the three to four nvme ports , and any extra pciex space don't need testing.

Well don't ask what needs improving and speal the same shit out then.

They're on the board, why would you not use them?!

 

[dgianstefani]

Would be cool to do a cost estimate based on component prices, compare that to other mobos etc.

Buildzoid I think and probably others break down every important subunit of the boards, VRM chips, voltage regulators, sound codec chips etc.

I want to know how much I'm being upcharged for marketing.

 

[Assimilator]

They're on the board, why would you not use them?!

Because the reviewer has only a finite amount of time to do the review and thus must choose which tests to run and which to omit. Quite obviously, the ones that end up being included are the ones that the majority of the readers will find useful, while the ones omitted are the opposite. And your use-case is the opposite.

 

[ir_cow]

@TheoneandonlyMrK I don't want to shoot your idea down otherwise what is the point in asking people for feedback. Can you confirm what exactly testing would look like?

In my head this is what I'm thinking based on what said. Connect all M.2 drives on the PCH (This is 2-4 depending on the MB). Raid-0 them all and do a read test of a different drive connected via USB. This "should" show if any bandwidth restrictions exist for that USB port, like if the M.2 drives are using the bandwidth. The only flaw in this experiment would be if the Raid-0 is idle, it might not affect the USB transfer speed at all. I guess only one way to find out.

 

[cvaldes]

Would be cool to do a cost estimate based on component prices, compare that to other mobos etc.

Buildzoid I think and probably others break down every important subunit of the boards, VRM chips, voltage regulators, sound codec chips etc.

I want to know how much I'm being upcharged for marketing.

Estimating component costs isn't realistic. You'd need to know how each motherboard manufacturer negotiated supplier contracts for every single component with the expectation that volume purchases at wholesale result in lower cost per unit. You most certainly cannot use retail price for ____ capacitor or ____ microcontroller.

Do you think NZXT is paying the same prices for their mobo components as Asus? I don't. And neither shops on eBay or AliExpress.

And what does it matter if gross margin for one mobo manufacturer is 38.7% and another is 39.3%? (Yes, those are made up numbers.) Even if Asus handed ir_cow a BOM with COGS listed at the bottom, he still wouldn't know other costs specific to the board.

As I mentioned in my original response, no one can please everyone all the time. Reviewer ir_cow needs to prioritize what he can focus on and what he'll let drop.

He doesn't have time to write 200 page motherboard review with 100+ graphs and diagrams. And I don't want to read one anyhow. Who does? I bet some here would raise their hands and then complain that ___ wasn't covered as much as they had hoped.

Everyone needs to remember that for every minute ir_cow spends on working on Motherboard A's review, that's one less minute for Motherboards B, C and D. Do people here want exhaustive reviews of 8-10 boards per year or wider coverage over a wider number of products?

Look at it from a different angle. The four big motherboard manufacturers are AsRock, Asus, Gigabyte and MSI. If ir_cow reviews the top of the line boards (full ATX format) with premium Intel and AMD chipsets, that's eight reviews right there. Worse, when a new chipset releases, all four manufacturers release new flagship models in a relatively short span. Does he test them in the order received?

It would be like exhaustive reviews of sportscars with V12 motors. Great, that's $200K I don't have but the tech is cool. Would more people read reviews of Honda Accords and Toyota Camrys?

As we all know, website operators worship at the Temple of the Almighty Pageview.

Does W1zzard want ir_cow to focus on $1,000+ motherboards? $150 ones?

 

[TheoneandonlyMrK]

@TheoneandonlyMrK I don't want to shoot your idea down otherwise what is the point in asking people for feedback. Can you confirm what exactly testing would look like?

In my head this is what I'm thinking based on what said. Connect all M.2 drives on the PCH (This is 2-4 depending on the MB). Raid-0 them all and do a read test of a different drive connected via USB. This "should" show if any bandwidth restrictions exist for that USB port, like if the M.2 drives are using the bandwidth. The only flaw in this experiment would be if the Raid-0 is idle, it might not affect the USB transfer speed at all. I guess only one way to find out.

I would suggest on the modern board's with possibly four nvme ports, all filled and two filled with only all filled on raid 0, simply to absolutely max the io required through to usb too obviously,then with two separate drive's, you can do internal nvme transfers and external and people would have a great idea of typically used and worse case.

 

[ir_cow]

Everyone needs to remember that for every minute ir_cow spends on working on Motherboard A's review, that's one less minute for Motherboards B, C and D. Do people here want exhaustive reviews of 8-10 boards per year or wider coverage over a wider number of products?

Look at it from a different angle. The four big motherboard manufacturers are AsRock, Asus, Gigabyte and MSI. If ir_cow reviews the top of the line boards (full ATX format) with premium Intel and AMD chipsets, that's eight reviews right there. Worse, when a new chipset releases, all four manufacturers release new flagship models in a relatively short span. Does he test them in the order received?

Does W1zzard want ir_cow to focus on $1,000+ motherboards? $150 ones?

Without getting into the specifics of each vendor. Here is common way things happen.

Vendor sends a email asking if the reviewer wants to cover product X. If yes, product is sent. Products are reviewed in order received.

Things get shuffled around based on the conditions of getting a sample. "Publish within 5 weeks" or stuff like that. Or hey we sent you 3 things, can you do product B first. Generally though, its whatever is offered for sampling. I may have no choices, or 5 to pick from. Depends on the vendor. Some if you already established relations prior samples can be sent without contacting you first. Mystery boxes I call those. I get a Fedex tracking because its linked to my phone number. I can tell who its from, but not much else.

I've already talk to wiz about speedy reviews. He wants quailty. That means the review can be 2 paragraphs or 10 pages. Length does not matter, its the content within.

I agree with this generally. While I can't please everyone here, some requests can be accomplished without adding much time. I naturally write long reviews. Pros and cons to my writing style of coverage.

 

[cvaldes]

I've already talk to wiz about speedy reviews. He wants quailty. That means the review can be 2 paragraphs or 10 pages. Length does not matter, its the content within.

Yes yes, regardless of each review length, they could all be "good" as a standalone review.

However since you're reviewing motherboards, you basically have to cover the same things in each motherboard review from a practical standpoint. Consumers in the market for a motherboard naturally want to compare apples to apples. It's not like you are going to list all the ports on the I/O shield with speeds & protocols in a long Review A and say in a short Review B "huzzah, there are plenty of ports on this baby!"

It would be like a restaurant reviewer skipping the dessert course at Restaurant A but reviewing the wine list there and reviewing the dessert course at Restaurant B but ignoring the wine list. Or ignoring some aspects like dining room acoustics, access by public transit, etc. Or ignore vegetarian options at one place but not the other.

If you add new metrics to your motherboard reviews, wouldn't you be adding them to all future reviews when applicable?

If you start testing m.2 throughput, it's not terribly useful if you include the data for Motherboard A and leave it out on its direct competitor Motherboard B. Or leave out the unboxing photos. Or list of accessories. Or try overclocking Motherboard C but not do it on A, B, and D. Et cetera.

And by your own admittance, you are not writing 2 paragraph reviews anyhow.

It don't see anyone here asking you to write shorter reviews. In fact, every single request has you adding more data points, more testing, more detail, more time, more words. And I'm no different, I asked you to include more coverage on the software (UEFI/BIOS and motherboard companion software) like others have requested as well.

 

[ir_cow]

If you add new metrics to your motherboard reviews, wouldn't you be adding them to all future reviews when applicable?

Yep. This is why I am asking now, so when the new Ryzen stuff drops it can be fairly consistent onwards. Ask for stuff now or wait another cycle. That's pretty much it. Seems like a lot has been asked, which is great!

Now comes the balancing act. IF something had to be removed, what is the least important, informational or pointless general. This is a hard one because if it didn't make sense to have it, it would probably already be removed.

 

[freeagent]

I have no complaints with the reviews

 

[Voodoo Rufus]

I like the current reviews the way they are. I look at the naked board deep dives on the VRMs, their temperatures and board layouts, the stability during testing, and the analog audio performance.

 

[ir_cow]

I guess that was a stupid question to ask. Not much to remove, just things to add.

 

[Mussels]

This one might be hard to pull off or give funky results. I can batch script to test all drives at once with DiskSpd but to test all the usb ports, sata and nvme at once. That is a lot of drives and external enclosures I don't have. Who is accessing all this at once?

It's very very common for multiple NVME drives
As a simple test you could definitely have some smaller NVME drives, and just do a test from CPU->chipset and chipset-chipset and see if they slow noticeably

I agree that going balls to the wall testing *everything* at once is implausible - a user who wants that, wants threadripper with all its lane insanity. But testing two drives at once, or testing two chipset connected devices? sure.

Even the fastest USB port (with an NVME drive, ofc) to a chipset NVME slot, would be a great test that would be a common use case for a large amount of home users.

I've got an NVME drive that reads at 1.5GB a second, and not one system in my house breaks 800MB/s reads from it on a 10Gb USB device. On the 5Gb/s ports. it sits around 250-300MB/s and that just... doesn't add up.

Sorry mate, you are an edge-case of an edge-case. Very few people have two NVMe drives, let alone 5; even fewer have them in RAID-0.

Two NVME is fairly common, by now a lot of people have upgraded from the smaller earlier gen models or gen 3 to gen 4, and kept the first drive.

 

[MarsM4N]

6. VRM's and a discussion on why they matter.

Spoiler: Modern CPU wattages are insane.

Intels 10700 (non k) being a 65W CPU that uses 215W of power threw logic is out the window and it's never come back.
AMD's FX series started this with the FX 9590 and it's 220W TDP, causing major thermal throttling for unaware users on all but the best boards - this is not an intel only issue and could very easily go horrible on AM5 too
Even midrange CPU's are using enough power to thermal throttle boards at stock, and since TPU covers overclocking we'd need to know how they handle stock conditions and overclocked conditions. Even a simulated worst case (12 core ryzen with all core OC and higher than needed voltages lets you know how a 16 core would run, that an 8 core would never have issues, etc)
View attachment 259422View attachment 259423

This is actually the most important pice right now. After watching Hardware Unboxed's Review I was shocked how many boards from so called "premium manufacturers" where crippled by thermal throttling. So many low end boards can't handle Intel's boost power draw. And this is not only a matter of performance but also reliability (heat kills electronics). Granted, nobody runs their rigs without fans, but there will be a huge difference if you use a top down blower style CPU cooler or a water cooler where the VRM's don't get any direct air.

The power draw varies also very much from board to board, sometimes even without reflecting in performance.

I want to see more middle-/lower-end boards and fewer high-end ones reviewed. Halo products are technically interesting and the ones that the companies love to sample because they make them look good, but they are useless for the vast majority of consumers who can't afford such boards and thus won't be buying them.

The guy in the Hardware Unboxed's Review (where he reviewed 10 low/mid range boards) mentioned that he had to buy 6/10 boards with his own cash. High end boards always come free as a review sample. These companies who don't provide review samples of their low/mid end boards know they are semi trash, thus don't want to see reviews of them. That's the problem.

capacitors conected backwards and blowing mosfets, risk of fire, that's really bad to say the least

Hello, ASUS! Asus recalls 10,000 Z690 Hero motherboards because of a backward capacitor

I do have a few open ended questions to narrow down to specifics. You guys have to remember I am at constantly at war with time. I could spend months a a single review. All for a few people? It needs to be balanced.

1) In your opinion, is the motherboard power draw actually important? Are you looking for sleep, idle and load?
2) BIOS features - List out the things you want covered for AMD or Intel (both if they overlap).
3) Chipset temps - Is this important? If yes, why do you think so
4) budget overclocking - Would you suggest buying a budget MB (ASROCK B660 MB from video) with the purpose to put a 12700/12900 (non-k) in and overclock it? If so, is 100c on the MOSFETTs acceptable to you?
5) Would a Infrared camera picture be useful at all. It looks pretty, but the data isn't helpful. "Hot spots" are not in context. If a VRM heatsink is at 70c, but the MB is at 26c. That what does that tell you?

1) at our current time (global energy crisis) I think power draw is something a lot will look after
2) some BIOS flashback feature should get extra points; overclocking/underclocking results; just some screenshots of the BIOS menu (design)
3) if it's getting too hot it does matter for stability & reliability, so yes
4) if it's compatible & folks actually go for it, then sure it's a thing. And yes, 100c+ is not good for any component
5) don't know if it's useful, but it's a nice click bait. You'll see tech in a way you'll usually never see it (could add the hotspot temps/digits to the image for informative value)

 

[Mussels]

This is actually the most important pice right now. After watching Hardware Unboxed's Review I was shocked how many boards from so called "premium manufacturers" where crippled by thermal throttling. So many low end boards can't handle Intel's boost power draw. And this is not only a matter of performance but also reliability (heat kills electronics). Granted, nobody runs their rigs without fans, but there will be a huge difference if you use a top down blower style CPU cooler or a water cooler where the VRM's don't get any direct air.

The power draw varies also very much from board to board, sometimes even without reflecting in performance.



The guy in the Hardware Unboxed's Review (where he reviewed 10 low/mid range boards) mentioned that he had to buy 6/10 boards with his own cash. High end boards always come free as a review sample. These companies who don't provide review samples of their low/mid end boards know they are semi trash, thus don't want to see reviews of them. That's the problem.



Hello, ASUS! Asus recalls 10,000 Z690 Hero motherboards because of a backward capacitor



1) at our current time (global energy crisis) I think power draw is something a lot will look after
2) some BIOS flashback feature should get extra points; overclocking/underclocking results; just some screenshots of the BIOS menu (design)
3) if it's getting too hot it does matter for stability & reliability, so yes
4) if it's compatible & folks actually go for it, then sure it's a thing. And yes, 100c+ is not good for any component
5) don't know if it's useful, but it's a nice click bait. You'll see tech in a way you'll usually never see it (could add the hotspot temps/digits to the image for informative value)

Oh yes they do - they get these boards with high wattage CPUs, slap a budget garbage AIO on there and remove all cooling from the already bad VRMs

Heck that's worth testing - VRM's with and without airflow.
Too many CPU's dont have a stock cooler now, so an AIO and then with and without a fan would be enough (with the aforementioned stock and over-volted OC stress test)


100% agree on lower end boards needing testing too, but Self-Funding may be an issue.
I suppose donation drives and requests on the forums could work - it only takes one global user to be willing to pay for a cheaper board or send their hated budget board in for testing.

 

[cvaldes]

100% agree on lower end boards needing testing too, but Self-Funding may be an issue.
I suppose donation drives and requests on the forums could work - it only takes one global user to be willing to pay for a cheaper board or send their hated budget board in for testing.

If this sort of thing takes off, I'm gonna start reviewing restaurants and let GoFundMe pay for my dinner check.

Oh, and if you guys have any bottles of grower Champagne, feel free to send them to me. I'll be happy to review those for you too.

 

[Voodoo Rufus]

Oh yes they do - they get these boards with high wattage CPUs, slap a budget garbage AIO on there and remove all cooling from the already bad VRMs

Heck that's worth testing - VRM's with and without airflow.
Too many CPU's dont have a stock cooler now, so an AIO and then with and without a fan would be enough (with the aforementioned stock and over-volted OC stress test)


100% agree on lower end boards needing testing too, but Self-Funding may be an issue.
I suppose donation drives and requests on the forums could work - it only takes one global user to be willing to pay for a cheaper board or send their hated budget board in for testing.

All of this. Boards costing $400+ don't interest me much. Top end CPUs are fun to read about, but not really necessary for casual gaming. It's like Ferrari vs. Corvette. And the budget chips are the F150s, getting the job done.

 

[cvaldes]

I guess that was a stupid question to ask. Not much to remove, just things to add.

I know that the compensation terms between you and TPU are confidential however this discussion does bring up a question.

Let's say you decide to add three more metrics to your reviews. Including set up time, testing, equipment teardown, data analysis and writing, let's say two of these metrics take 30 minutes apiece and the third takes an hour. If you have ten boards queued up, that's twenty more hours of your time by my calculation. And what if some of these tests require additional equipment? Who pays for that?

Will TPU increase compensation for your additional effort?

One thing for certain, you aren't going to be adding more data points to your reviews without putting in more time and effort.

 

[geniekid]

I expect I'm in the extreme minority here, but I've often wondered about ECC support on consumer-ish motherboards. Particularly with AMD platforms, it feels like ECC support is sometimes present, though hardly ever supported officially.

I particularly enjoyed this article from Hardware Canucks - https://hardwarecanucks.com/cpu-motherboard/ecc-memory-amds-ryzen-deep-dive/