- Joined
- Apr 24, 2020
- Messages
- 2,875 (1.55/day)
With regards to @Mussels , congrats! But aren't motherboards basically glorified PCBs? That means your tests are basically PCB-testing, at least from a first-principles perspective. IMO, that's:
1. Power Delivery -- VRMs are obviously the #1 factor with regards to cost. But even the PCB-itself forms a capacitor/inductor between components (aka: the parasitic elements). I've seen some features that make me feel better as a EE / knowledgeable PCB guy (albeit college-level training, no professional experience), like split loops, parallel power delivery and whatnot. I don't know how to test for that however, but these things matter as much to power-delivery as VRMs. Capacitors, VRMs, and the PCB-board parasitics form the power-delivery network, a master's degree level engineering problem in its own right.
Nominally, most people shortcut the process and basically tab up how costly the VRMs are, under the overall (likely fair) assumption that if Motherboard#1 spends $15 on VRMs, it probably has a better power-network than the Motherboard#2 that only spent $10 on VRMs. But IMO, its best to keep the whole picture in mind if at all possible.
2. Electromagnetic-interference (aka: EMI) -- Audio is likely the only component that is going to care about EMI-issues. However, there were also a number of USB-interference issues vs Bluetooth and/or WiFi that cropped up in the past. All PCB-traces could potentially be a radio-antenna if poorly designed (or perhaps: a PCB is an antenna fundamentally whose EEs have specifically designed it to not act like an antenna). And of course, wireless components (WiFi/Bluetooth) are being added to motherboards in practice, so plenty of interference opportunity since a "real antenna" is on board. A "poorly designed" Arduino could likely emit wideband noise, and in theory you could wave it around the audio-bits and see if any of the audio-lines get interference.
3. Heat -- PCBs are also crude heatsinks that absorb the heat from misc. components. Not enough for the CPU, but lesser-chips like RAM or the Chipset will sink heat into the PCB. All components today will automatically detect overheat conditions and self-shutoff, but this means that heat-design is still paramount for performance. Any chips that are entering thermal shutdown are chips that are slower, less performant, etc. etc. But I have big questions of how to even test the chipset in practice, so maybe its a punt on this one.
4. Features -- The PCB has to offer what the engineer wants it to offer. Correct USB ports, WiFi, Ethernet, Audio, etc. etc.
5. Software -- The BIOS is key to usability in practice. I don't know what the big list of features that's useful for everyone... or even if going through each item line-by-line is useful to anyone. BIOSes are used by both beginners (where ease-of-use is biggest), to experts (who will want full control over obscure features, like NUMA-configuration items or PCIe port-bifurcation).
------------
Nominally, the "big connection" will be between CPU and RAM. This is a high-speed connection that requires hundreds-of-pins, with the bulk of them delay-matched and impedance-controlled. I presume that the quality of this connection will relate to how well RAM can overclock, but I'm not sure how to test it in practice. Like all digital components, the connection "works until it doesn't work".
----------
The ultimate test to push everything to the limits would be an overclock test, but there's too many variables at play here. I don't even know how to normalize an overclock such that X motherboard vs Y motherboard would be better (and manufacturing variance could just fundamentally limit you). I recognize this is very far off on the wishlist of things to do. But I figured "reaching for the stars" is useful for the discussion, even if they're outside of anyone's scope. At least for me, I don't expect that CPU-speeds would vary too much between motherboards (aside from #3: heat issues).
PCBs (and therefore motherboards) are surprisingly complex beasts. Best of luck to you as you generate a methodology!
1. Power Delivery -- VRMs are obviously the #1 factor with regards to cost. But even the PCB-itself forms a capacitor/inductor between components (aka: the parasitic elements). I've seen some features that make me feel better as a EE / knowledgeable PCB guy (albeit college-level training, no professional experience), like split loops, parallel power delivery and whatnot. I don't know how to test for that however, but these things matter as much to power-delivery as VRMs. Capacitors, VRMs, and the PCB-board parasitics form the power-delivery network, a master's degree level engineering problem in its own right.
Nominally, most people shortcut the process and basically tab up how costly the VRMs are, under the overall (likely fair) assumption that if Motherboard#1 spends $15 on VRMs, it probably has a better power-network than the Motherboard#2 that only spent $10 on VRMs. But IMO, its best to keep the whole picture in mind if at all possible.
2. Electromagnetic-interference (aka: EMI) -- Audio is likely the only component that is going to care about EMI-issues. However, there were also a number of USB-interference issues vs Bluetooth and/or WiFi that cropped up in the past. All PCB-traces could potentially be a radio-antenna if poorly designed (or perhaps: a PCB is an antenna fundamentally whose EEs have specifically designed it to not act like an antenna). And of course, wireless components (WiFi/Bluetooth) are being added to motherboards in practice, so plenty of interference opportunity since a "real antenna" is on board. A "poorly designed" Arduino could likely emit wideband noise, and in theory you could wave it around the audio-bits and see if any of the audio-lines get interference.
3. Heat -- PCBs are also crude heatsinks that absorb the heat from misc. components. Not enough for the CPU, but lesser-chips like RAM or the Chipset will sink heat into the PCB. All components today will automatically detect overheat conditions and self-shutoff, but this means that heat-design is still paramount for performance. Any chips that are entering thermal shutdown are chips that are slower, less performant, etc. etc. But I have big questions of how to even test the chipset in practice, so maybe its a punt on this one.
4. Features -- The PCB has to offer what the engineer wants it to offer. Correct USB ports, WiFi, Ethernet, Audio, etc. etc.
5. Software -- The BIOS is key to usability in practice. I don't know what the big list of features that's useful for everyone... or even if going through each item line-by-line is useful to anyone. BIOSes are used by both beginners (where ease-of-use is biggest), to experts (who will want full control over obscure features, like NUMA-configuration items or PCIe port-bifurcation).
------------
Nominally, the "big connection" will be between CPU and RAM. This is a high-speed connection that requires hundreds-of-pins, with the bulk of them delay-matched and impedance-controlled. I presume that the quality of this connection will relate to how well RAM can overclock, but I'm not sure how to test it in practice. Like all digital components, the connection "works until it doesn't work".
----------
The ultimate test to push everything to the limits would be an overclock test, but there's too many variables at play here. I don't even know how to normalize an overclock such that X motherboard vs Y motherboard would be better (and manufacturing variance could just fundamentally limit you). I recognize this is very far off on the wishlist of things to do. But I figured "reaching for the stars" is useful for the discussion, even if they're outside of anyone's scope. At least for me, I don't expect that CPU-speeds would vary too much between motherboards (aside from #3: heat issues).
PCBs (and therefore motherboards) are surprisingly complex beasts. Best of luck to you as you generate a methodology!
Last edited:
