Discussion about how the motherboard affects overall system performance

B

bfeist

n00b
Joined
Oct 4, 2020
Messages
3
The way motherboards tend to be reviewed and discussed is around the max specs--number of PCIe lanes, etc., or by compatibility with different RAM or hardware.

I'm about to spec out a new high-end workstation to replace my 5-year-old system and I find myself wanting to see a discussion somewhere about the small irritations that is part of the PC experience, and how good or bad different motherboard manufacturers are at dealing with them. For whatever reason, I've always purchased ASUS motherboards (the last 3 PCs at least) and there are always issues that I never blame ASUS for because historically, PC users come to expect some level of device gremlins that they have to be able to work out. My tolerance for this has eroded down to zero. It's 2020, the PC industry should have this crap worked out by now.

Two examples of the kinds of small irritations I'm talking about:
  • The onboard Intel LAN port on my "asus x99 pro usb 3.1" (the name just rolls off the tongue, doesn't it?) sporadically switches from 1Gbps to 100Mbps. Unplugging the cable and plugging it back in several times makes it switch back to 1Gbps. I know, I know, but I've switched the cable to Cat 7--no change, I've changed which port it's on, and then even switched the actual switch on the other end--no change. It seems to be related to interference from the USB-3 ports next to the ethernet port because which USB ports I use changes how often the rate switch problem happens...or maybe I've just become superstitious. Tried adding a TP-Link 1Gbps PCIe card so I can stop using the onboard altogether, but reasons I never figured out, the card doesn't even appear in Windows. It's like the motherboard refuses to recognize it or something even though there are 3500 reviews on amazon saying that it's fine on Windows 10. C'est la vie.
  • USB Bluetooth connector doesn't stay sane after reboot. After almost every reboot I have to uninstall the USB bluetooth device in device manager, physically remove the dongle and reinsert it to make it active again. Come on! Where is this kind of thing ever caught in a review. Sure, this issue could be anything and isn't necessarily the motherboard (registry problem?). Nobody is ever to blame about interoperability because in the PC world nobody owns the overall system. But, the motherboard is the closest thing we have to the integration point of everything. They are responsible.

There are many windows-level irritating issues like switching a monitor to another input causes all of the windows across multiple monitors to rearrange randomly. This is clearly a MS issue but I bet they site their inability to know what's happening at the hardware level as the reason they can't handle it better.

So here I am, about to do this all again in the hope of creating a fast systems that also feels fast. I'd like to see a benchmark for how quickly a system responds to the user rather than only seeing CPU-bound benchmarks, and I'd like to see a benchmark for pains in the ass the didn't happen with a given system.

Ok, end rant. Anyone have any thoughts on this?
 
My thoughts on this can be summed up in a word: "No." That is, what your asking for isn't really reasonable, nor is it really doable. I don't think you understand the processes used by reviewers and why it is the way that it is. I've been reviewing motherboards for 15 years. The fact is that we can only share our experienced based on what we see while testing the hardware in question. If I never encounter a problem, then I'll tell you that. If I did encounter a problem with a given piece of hardware, I'll share all the details of that experience. I've had trouble with NICs and other integrated features and I cover those issues whenever they occur. The fact is, problems with these boards are fairly unusual. But, I do cover them when they occur. Whether that's problems setting up a drive array, or installing the OS. I cover that whenever I see issues. Recently, I mentioned an issue with an i225v NIC in a review. I resolved it simply by updating the UEFI BIOS to the latest version. Ordinarily, I do that first, but forgot to update it before I started installing the OS and setting everything up.

As for system responsiveness, there are too many variables that come into play to really provide useful data on that. Wireless mice or keyboards, running things through USB hubs, monitor latency, etc. all factor in. I use the same monitor, keyboard and mouse for all my reviews. As a result, the testing is as level and as fair as it can be. That is, every X570 board I review feels the same as the rest. All Z490's feel the same as the others do. Sometimes there are some differences I can perceive between an X570 and a Z490 based system, but those too are often difficult to quantify. About all I can say there is that the Intel's often feel smoother, but I only notice this if I have both an Intel and an AMD system side by side with the same GPU, keyboard, monitor, and mouse. I've tried to correlate these differences by checking things like DPC latency, but that's all over the map and I've never seen DPC latency have an impact on anything unless it was atrociously high for some reason.

Things like your Bluetooth issue aren't caught in reviews because we aren't testing or using such things in reviews. It's not about Bluetooth dongles. Were I reviewing those, that would be different. But any testing I did would only be applicable to whatever specific make and model Bluetooth device I tested. It would be meaningless for anything else. Like your example with multiple monitors, there are just too many variables for a reviewer to try and replicate. Honestly, issues with that may be Microsoft's fault to some degree, but other factors like GPU model, driver revisions and specific monitors, cabling choices etc. all factor in there. We can't replicate each use case scenario. We simply don't have either the time or the necessary hardware to do that. My reviews take 25-30 hours to complete as it is.

Lastly, the title of your post needs addressing. Since memory and PCIe controllers became integrated into the CPU, the motherboard is virtually meaningless in terms of how they effect performance. Most of the time, the data shows a 1-3% variance in the test data between system configurations. That's basically the accepted margin for error in the tests themselves. It's meaningless unless you see an anomaly. Even then, those anomalies usually have an explanation. A CPU isn't boosting correctly, or the VRM's start throttling due to insufficient cooling. Essentially, I can explain those variances that step outside the line pretty easily in most cases. When all factors are equal, the motherboards generally don't mean anything when it comes to performance. All the factors that impact performance were moved off the motherboard a long time ago.
 
Last edited:
I’m curious why you thought “cat 7” cable would solve the problem with you network link downgrading too 100M, especially since a true CAT7 cable requires a special connector that’s not compatible with the ports on any consumer motherboard (hence the quotes). What you had was a hacked together S/FTP cable with 8p8c ends that likely wasn’t properly grounded and maybe causing more problems that you thought it would solve.

As for the PCIe card problem, are you sure the PCIe slot you used wasn’t disabled? A lot of motherboards disable slots if certain combinations of GPUs, SATA drives, or m.2 slots are also in use.
 
Honestly a lot of those issues are Windows related.

I have two workstations currently:

9900K
Gigabyte Z390 Designaire
5700XT (for use in MacOS)
RTX 2080TI (for gaming in windows)
32GB Ram
2x 2TB NVME

And for number crunching:

AMD 3990x
Gigabyte Aorus Extreme
128GB ram
4x RTX 2080TI
3x 2TB NVME

The 9900k is running opencore Hackintosh and it just works including updates right from the Mac app store. This build is pretty much the worst case for compatibility - heavy use of Thunderbolt, two different brand graphics cards, a hefty overclock, dual booting Windows 10 and MacOS. In MacOS everything is absolutely perfect; in Windows 10 I get a lot of the same bizarre intermittent issues that you are talking about. In MacOS Thunderbolt 3 is working perfectly and I have over 20 peripherals connected to it including 10GBe. Sleep works, time machine works, 10gbe works with zero drivers, TB3 works, updates work, everything just fucking works with no hassle and no bloat.

The 3990x build is running PopOS and again - it just works. 10GBe, all 64 cores able to be maxed out at 100% use with zero tweaks, etc. Don't have to deal with Windows weird thread scheduling, etc.

If you want a system that feels fast, move away from Windows unless you are locked to it due to software compatibility.
 
As I've been saying for years, motherboards rarely have any impact on overall system performance for many of the reasons stated above. Taking that thought a bit further, I can objectively tell you when a VRM is better, or a motherboard has better physical build quality. However, even that's usually meaningless outside of some fringe cases where those factors may come into play. By that I mean, LN2 overclocking. That's really the only time the extra BIOS settings, higher power VRM's, and build quality really and truly come into play. For example: The ASUS Maximus XI Formula has a four phase VRM. The Maximus XI APEX has a 14 phase VRM. It's not 12+2, or any of that. It's 14 phases. No doublers. Yet, both can overclock a 9900K to 5.0GHz and run that way 24/7. I can tell you the APEX is a better board in the sense that it has a better VRM, but the VRMs on the Formula are "good enough" for most people and most use cases.

For that matter, the Hero and the Formula have the same exact VRM design. The difference between these two boards is their feature sets. Performance is functionally identical.
 
The only reason I buy high end boards is that they tend to have better construction (bit more tolerant of ham fisting - IO shields and the like, and better shielding for electronic noise/etc), tend to get updates better over the life of the system, and (at least in my experience), last longer, so I can pass them down from use case to use case. And since they use more name-brand parts for things like TB / Ethernet controllers, you often (not always, but often) have better drivers/support there too. An extra couple of hundred bucks amortized over 10+ years of use in various configurations is very very little in the grand scheme of things.
 
Succinctly, on the desktop, it does not matter. /thread
 
Thank-you everyone for your thoughtful replies.

Dan_D said:
My thoughts on this can be summed up in a word: "No." ...
Click to expand...

Dan, I appreciate and agree with everything that you wrote. I chose to make this a motherboard discussion because it's the integration point of a "PC" and as you mentioned, when nobody is responsible for higher order smooth interaction, we have to start somewhere. I've used Apple a lot as well (mostly at work) any they definitely own the integrated operation of everything together. That said, they lack transparency and when things go wrong they go very wrong. (I also dislike MacOS, so don't mistake this as an Apple fan moment). It would be nice to get something like this seamless integration of components into the PC world somehow.

@ryan_975 I bought this. It may not actually be "cat 7" but it works just as well as my older Cat5e cables at 1Gbps. The point I was making in my original post was that I've troubleshooted the situation to no avail. Thanks for the PCIe reco. I inadvertently ruled your suggestion out when I removed my working PCIe 10GBe card and replaced it with the not-working TP-Link card as a test. Still a mystery.

@paradoxical I'm with you and agree that my examples are all very likely windows issues. My system is 5 years old and I'm not gentle with it. I use a lot of different software with lots of different peripherals. I'm sure my installation is bloated all to hell at this point. The prospect of "just reinstall Windows" is so time consuming and frustrating that I never get around to it.

Back to the motherboard discussion, when I do reinstall windows, it comes with a process of going and hunting for all of the 3rd party integrated components on the motherboard. Suddenly I'm not an Asus customer, I'm a Realtek audio and Intel NIC customer in my own little hell getting every little piece of the motherboard online. Intel installs some "turbo boost max 3.0" kind of tray icon as the only indication of some chipset choice made by Asus. No explanation or ownership of this choice and I'm left unplugging and plugging in the ethernet cable for 5 years. I totally get that this is just how things are in the PC industry at this point. Has anyone seen any glimmers of hope that it might be changing? Car manufacturers outsource the inventions of all of the subsystems in their cars to OEMs, but you don't see me having to suddenly call Garmin when my Audi satnav is broken.

Thanks again everyone.
 
As an Amazon Associate, HardForum may earn from qualifying purchases.
That’s because Audi owns the integration- much like Apple does (it’s just a custom intel board with intel cpu, and other random OEM parts)- they design for precisely ones set of hardware (Mac Pro aside), and that’s it. How many different vendors exist for network cards? Peripherals? Etc?
Asus doesn’t own that integration; by building your own, you do.
If you want that level of completeness on the PC side, you’re looking at major OEMs (to a point) or custom shops like Digital Storm. I guarantee you theyll help troubleshoot things if you buy a top end system from them.
Build your own brings flexibility. Flexibility brings compatibility issues potentially.

That being said, windows has gotten better. I built my new Threadripper workstation a month ago. Windows booted, found the network card, I connected to wireless, and up popped the armory crate for asus. Boom, drivers handled.
 
I agree 10 has come a long way for personal pc customization since it's inception. Recently upgraded my old HTPC to an ssd and win 10. Only the onboard azalia audio driver was missing from the device manager. Inconsequential as it is since HDMI audio from the rx 580 will serve audio. The fx6300 still gets it done. :) One of my all time favorite systems as it has ALWAYS just worked.
 
You must log in or register to reply here.
Back
Top