G'day Buildzoid, 🤔about the VRM 50A vs 70A could it be different revisions? When you look at the Screen Printing of the A SUS Spec Page The DIMM Layout is Top Right of Socket after WI-FI vs Your's the DIMM Layout is Bottom Right of Socket next to PROART with R1.03 after WI-FI Plus while both the Pics you showed have 2x 90A VDDMISC in the PCCG listing for "PROART X670E-Creator WI-FI" (Shop down here in Australia) when you scroll down to the Pic with no VRM Cooler it only has 1 90A VDDMISC
Buildzoid, if you search community on the AMD site, you can find an entry by properMBsearch in "General Discussions" titled "X670 motherboard with many PCIe-slots" with CAD drawing and block diagram.
"it doesn't make sense to build a storage server around a high end am5 board" - well you see that's where thunderbolt matters, ingest stations / media servers love that stuff. production companies carry around these huge OWC DAS boxes to shoots and guess what, it's all thunderbolt.
7:34 Professional audio engineers wouldn't use the integrated audio chip that comes with a motherboard. They all use external audio interfaces nowadays that provide much better input latency. That being said, the signal-to-noise ratio has certainly improved over the years, even on entry-level motherboards.
I can't find an introduction date for the C787 but the datasheet is 5/23/2024. The chips on the board are made is in 2022 week 16. The regular 86992 datahseet is from 2019. So I'm gonna stick with these being the 50A chips because as far as I can tell the 80A variant didn't exist when this board was made.
@@ActuallyHardcoreOverclocking I never checked the datasheet publishing dates as I generally find them somewhat unreliable, Monolithic Power arent the quickest in publishing datasheets alongside new IC releases either. For example I recieved some RISC-V dev boards last year with MP1612 & MP1657 step down converters that still do not have any datasheets published/publicly available, yet these converters are used in Xbox consoles released years before these boards were fabbed - which is a real shame as those ICs would be ideal step down converters for a pcb design I have in the works - but no datasheet to work from. :) I have owned a ProART X670E since it was released, it is a great board and I have a 7900X mounted in it with 2x32GB for work, I got it primarily due to the onboard Marvell/Aquantia 10GbE which performs awesome, no overheating that I have noticed (I havent exactly been looking, but I would notice degraded network perf), the rest of the X670E lineup across all the vendors is pretty sucky overall to be honest, they are all full of compromised configuration choices [its a result of what happens when PR takes priority over the engineering, sadly too prevalant these days], or just overpriced halo showpieces / tech demonstrations. :)
@@ActuallyHardcoreOverclocking The datasheet is a newer version, which changed 70A into 80A one month ago😅. And before that C787 was only labelled as 70A parts. I've also seen some Z690 boards with these parts, which I think are the same age as those x670s
i wonder if the abundance of MLCCs on the output is the secret behind the low number of polymer caps. if the main factor dominating performance in this application is output capacitor ESR and not value, it might be beneficial to reduce the net output capacitance in exchange for a better frequency response. conventional wisdom is "more capacitance more better", and when you add more output caps in parallel, you are also reducing the total ESR, so confirmation bias could lead engineers to assume it's the increased total capacitance causing the performance improvement (or just, newer voltage regulators having better control loops, allowing some of the work previously done passively by capacitors to be done actively by the controller. the output capacitors are acting like a low pass filter between the extremely high-frequency switched power demand of the CPU and the relatively slow response of the regulator; as faster regulators come out, your cutoff frequency can increase). in which case they could be capitalizing on this realization, since polymer caps are pretty efficient at turning dollars into farads, but pretty poor at turning dollars into 1/ohms
Always nice to see you go over a part I have. My personal reasons for grabbing this board was the 10 Gbit LAN port, secondary 2.5 Gbit LAN and dual 8x PCIe 5.0 slots. The M.2 layout does have its quirks but isn't bad. I do wish it had a POST code display instead of those simple LED lights. Those 40 Gbit USB-C ports are indeed Thunderbolt compatible. I've attached an external 4x PCIe 3.0 chassis to the board just to see if it'll work. Yeah, it does. I'm also of the impression that you don't need to use the DP input for these USB-C ports to drive external displays either: they'll use the simple integrated GPU to drive them by default which is fine for desktop work but not for gaming. Both LAN ports support AVB audio over the network. Windows doesn't have drivers for it but they do exist on the Linux side. AMD does require some CPU attached M.2 slots per X670E spec, at least at launch. I think this as since change as they've launched the Epyc 4004 line up that uses the same AM5 socket. Having three 8x PCIe 5.0 slots on that low end Epyc would have some server value though it'd leave all the M.2 slots hanging off of the chipset and at 4.0 speeds.
Great video as always! :D Full-featured USB4 and Thunderbolt are effectively interchangeable, so its just down to it not being Intel certified which I doubt they will ever issue for AMD boards. And oh the CMOS battery... It was a nightmare to get to it on my TRX40 Aorus Xtreme, you have to take off the entire board backplate to be able to take off the plastic bits that cover the battery... And yeah, this is one of the sanest/best AM5 boards in terms of IO configuration, though for actual workstationy use you would probably want the extra IO available on the W680 Intel version tbh, same for storage boxes and such since they will idle a fair bit lower.
And for capacitors for every 10c lower temperature, their insurance dabbles so if it runs at 100 5k is 10k if its 90c its 20k and they usually raun it 70 or lower so they last 80k plus hours
Chelsio is also a solid option for optical networking. Mellanox has worked brilliantly for me in Windows and macOS, but not in Linux or FreeBSD. Aquantia AQC107 has also been solid on macOS, but not other OSes.
You should definitely go for an Electrical Engineer degree. It will enrich your life, make the see things from a different angle, you can even make a side channel or live sessions with EE stuff. If you keep doing the same things over and over without broadening your knowledge could keep you from growing.
Have you considered that the VRM MOSFET is MP86992-C787 which is 16V 80A rated? It's common to use a power controller rated for higher capacity than what is advertised - to obviously cover any OC-related increased change in power usage. Otherwise, I enjoyed your video, thank you.
You should be a concultant for Asus, gigabite, msi etc. Before they send their stuff for mass production, they should send the motherboards and GPS to you and ask "we are about to produce this sht for x price....can we?"
@ActuallyHardcoreOverclocking On page 4 of the MP86992A datasheet, it lists the high-side current limit as 75A for up to 8 cycles (8 periods of what ever frequency the VRM is running at?). This might just be a case of marketing choosing the highest number they could justify. No idea if the cycle count gets adjusted for the current, i.e. it would not make much sense to have only 8 cycles at 75A and an infinite number of cycles at 74.5A, but maybe it is expected that it would go through a thermal shutdown at that point anyway. With that in mind, it is possible that the ASUS engineers were testing this power stage and found that it could do 70A continuously as long as it had appropriate heatsinking, given the 75A hard limit, then wrote that number down in their internal test documents, and marketing just ran with it.
I own this board & it's intel counterpart with the same USB4/Thunderbolt chipset onboard. Can confirm the X670E USB4 works just as intended, identical performance with the Intel Thunderbolt hardware devices. If memory serves, the Thunderbolt branding is exclusive trademark to Intel platform chipsets -- hence why for AMD platforms they get Intel's blessing so long as AMD calls it by a different name.
I don't think it is exclusive, at least the AsRock X570 ITX/TB3 uses Thunderbolt branding. Maybe Asus just didn't want to certify it with Intel to get the branding.
Apple uses Thunderbolt branding as well. If my understanding is correct, anything with Thunderbolt capability can be called Thunderbolt, no matter the implementation.
@@Sebyllis7350k Intel developed Thunderbolt together with Apple, so they probably are a special case in that regard. Until Thunderbolt 4 Intel required strict certification to use Thunderbolt branding. They eased the requirements when they opened up the specification to USB, but not fully.
i would almost say that asus has a very distinctive take on vrm design versus the other manufacturers. this and the gene are the best examples ive seen. its like theyre the only company thats bothered to follow what the rest of the electronics market has learned over the past decade or two the other manufacturers are like theyre frozen in 2006. zero innovation thats not literally forced on them by their suppliers
basically what theyre doing is like if actually remove power supply ripple effectively at the input .... having a whole pile of capacitance on the output can actually reintroduce its own ripple into the circuit. and it will make regulation worse not better
problem is that if you try to implement this style of vrm and eff it up... now you got no output filtering. typically you wanna a lot of simulation and then a lot of testing to define ur goldilocks zone, and then your gonna need to be able to account at the controller for whatever future chips and agesa updates might do to your boards already out in the wild and i mean i can go on and on about this , once you have only stage of filtering to speak of, now you end up a far greater risk of feedback loop developing between the switching vrm and the psu that can cause insane swings in the vrm and those single little protection components to break the feedback are really precisely chosen, to filter out very specific frequencies at which what are called "freak waves" can develop honestly the guy who designed this knows what hes doing and knows that he knows what hes doing and everyone on the design team takes his word as like mandate of heaven
the thunderbolt ports on this board are really picky about being hotplugged, the manual even says not to do it (contrary to the USB spec) also i couldn't get 20Gbps peer-to-peer networking working with this thing at all.
Wait really? Shit I missed that in the manual. I’ve bought this board at launch and hotplug the whole time. Countless times in the last two years. In fact every time I start my pc, I have to hot plug it to get the thunderbolt dock to work properly. But sometimes I do it from the dock side.
@@cassiohui if I need to restart my pc, then I need to, reboot twice, hot plug twice or three times, on the second or third hotplug it will recognize. This is why I’ve had to hotplug so many times it’s the only way to get it to work whenever my pc shuts down or restarts. I probably rekt something because I always hotplug between pcs, I was chalking it up to sketchy amd implementation. Feelsbadman. I definitely will try not to do it anymore but it seems to be the only way to get keyboard and mouse to recognize.
Asus ROG STRIX X670E-E GAMING WIFI is a better board without the 10G which you can add in a slot if required. Also dual x8 and has the digital audio and cheaper. Great review btw keep up the good work.
The way it's listed as 70Amps on a 50Amp power phase seems to me like that's laziness on Asus's end. After the repeated behavioral patterns, it'll be a matter of time before these boards become a QA issue and Asus won't be able to figure out why the components fail quicker in them... Or maybe they will and will shirk the blame on the consumers
I mean I have an Asus X670E-A which I’m betting is the same power phase and it’s labeled as the same 70A on my boards webpage to… I mean by law, they have a duty to get the numbers they advertise correct, otherwise it’s false advertising… in any event if I don’t have 70 amps I’m not happy…
@@fanofentropy2280 The issue with that approach, is simply the software support... or more accurately, the lack of it. I still use Audigy and X-fi sound cards entirely because of support for Legacy EAX code for games made in that era. I could be wrong (and for once I am hoping I am!), but last I looked into it, I could not find a single external unit that supports 5.1 or 7.1 analog, _and_ supports EAX code, either natively with a Creative branded offering or something that supports at least Creative's ALchemy software. Realtek famously refuses to support it despite claiming their audio chips could even emulate it internally for all their boasting.
On the X670E you get two 4 lane for M.2 or one 4 lane M.2 and USB4. With the next generation 70 class board the USB4 is required instead of optional so as I've been saying the entire time - AM5 is still just a 20 lane platform with the occasional rare better board that offers the additional 4.
Hello, I am from Russia and using this MB for almost half of the year. Bought it for 500 dollars. Yes, it has insane PCB, and lol, yeah ASUS tell everyone: "this is workstation MB for creators, not for extreame OC", ahahaha. It has tb4 2 ports, 15w PD, they even include display port to display port from your pci-e gpu to display port IN of the MB, so you can incapsulate video output from it to usb4/tb4 port. Also usb 3.2 gen2x2 header to front panel has additional x6 pci-e power pin. So you will have 60 or 65 PD for your devices from this port on your system case. Only one problem is found while using this MB - WOL works for some reason only for 10g port it seems can't do WOL PC for 2.5g port 😢 And no Spdif out, but do even anyone care about it? There is pci-e 4.0 x2 for your sound card, usb4/tb4 ports, you can plug anything anythere. Please make a video for that MB about OC 7800x3d with ram. Something about 6000-6400mhz and curve optimizer. There is TONS of settings for that, it is too hard even for me. I've achiced only -15CO and 6000 32-40-40-28 CR1 2167 fclk Upd: just updated to 2204 bios, they have added even more things for OC, omg...
I have this board and I use the 2.5 Gbps NIC all the time, including for WoL which works ok… I don’t use the 10 Gbps Aquantia due to performance problems in Windows and the interface randomly disappearing, which I think is due to thermal issues or buggy firmware - I ended up buying an X550 PCIe NIC :)
@@ChristopherWoods Wow, lol, is it actually exact board? X670E proart? Then it is really strange. My 10gb works perfectly, but no wol for 2.5g, it is the only problem. Also there is no thermal problems.
@@Misterixmax yeah :) the machine is running Windows 11. I don't remember doing anything special, it worked immediately for me. I can send WoL packets to the PC, and a couple of other PCs, using a Raspberry Pi or an Android phone app. Are you using a specific driver for the 2.5 Gbps or do you have machines on different network segments? I found that WoL packets don't route properly between VLANs on my Ubiquiti UDM, which was annoying.
@ChristopherWoods My PC is connected directly to my keenetic router, keenetic (via mobile app) can do wol perfectly inside own subnet. I am using win 10 (latest updates) and latest drivers from the Asus site for MB, including network drivers. Idk why I am having a problem with wol specifically for 2.5gb, but not for 10gb. In drivers settings for both adapters wol is on, wol from pci-e device in uefi is also toggled on.
15:36 I think it was msi or asrock has a whole bunch of post code back for the upcoming AM5 boards, maybe someone has seen the crys for post costs instead of LEDs back.
I picked up the Asus Prime X670E Pro WiFi (Microcenter open box for $260 vs the no discount available $450 ProArt) instead of this board due to the only real differences being, Back I/O (not worth the extra money to me): no 10gb Ethernet (no secondary RJ45 port at all), no 40gb USB-C (10gb USB-C instead), and it has a 5 Audio connectors stead of 3. Internal layout: the second M.2 PCIe is 3.0 instead of 5.0, and the CMOS battery location is below the Primary PCIe x16 slot, (potentially inconvenient) but there is an actual CLR_CMOS button (which stopped working on my board) and the second PCIe x16 is also just an x4 slot electrically, with the middle PCIe slot x4 being a true x4.
Just connecting Hyperdrive USB4 40Gbps enclosure into that 40Gbps port. The read speed is just 2000mb/s (same as 20Gbps port), and write speeed is only 1200mb/s (worst than 20Gbps port which is 2000mb/s)
I wonder if any motherboard multiplexes 4.0 or 3.0 PCI into 5.0 to have 2x or 4x the lanes at full bandwidth. 16 lanes would become 64 3.0 lanes then (basically).
i think the post code on ws boards used to be a feature cause lots of workstations were booting from pcie network or storage cards that required csm mode enabled to be bootable. and thats just no longer the case.
Just curious: is there any chance they are actually pushing the 50A parts to 70A? Maybe the 50A rating is conservative, perhaps at 80 deg C ambient temperature, or something?
So the thing is that the nominal current ratings is already impractical for real world applications as at 50A output an 86992 will produce over 7W heat(an in my exp any over 3-4W per component is basically uncoolable). Though I've gone through a couple of datasheets for higher nominal current rating MPS parts and they basically have the same performance at 50A output. So from what I can see MPS doesn't make any parts with significantly better efficiency just parts with higher current limit settings.
i had a midget land in the 2nd x16 slot on a B550-F and fry itself/short board, no post, took ages to figure that shit out. the boards when you review look sweet without the heatsinks. might nude my board lmao
ASUS X570 Crosshair Impact is damn awful in regards to the CMOS battery, it has to be completely disassembled to get behind the IO shield. I am definitely looking out for battery placement for my next board :D
Why do you want to have/pay another power supply when one good already exists inside your pc? also paying the enclosure, knobs, tax premium brand, when none of it gives better audio lol The max you need is a 200$ Pci-e sound cards with Dolby. Jealous much? Which if you look at component choice dunks on thousand$ externals. and for less money....
Well, the reason for that, is that asus has a duty to get what they advertise correct, because otherwise it’s false advertising…. And this isn’t the only board that uses the same power phases and advertises 70A also…
What is even the point of M.2 slots. Would be simpler with just regular PCI-E slots for SSD drives. And why are motherboard manufacturers fiddling around with VRM design. Surely the best architecture is settled long ago. Everything is just about marketing bs.
I had 2700x, 5800x and back in the day various Athlon. I am not against AMD, but I went Intel 13th when released because the boards often have 4x m2 sockets at low price. My board was 180€ and I dont understand why this board is 400€. OK it has 10G LAN, but I dont see any other advantages.
I don't know, Thunderbolt? Dual PCIe x8 off of the CPU? VRM? Loads of USB ports? Also, Intel boards only have one NVMe m.2 slots off of the CPU. AMD boards can have two.
even if it doesn't meaningfully improve performance, the bulk capacitor placement could help with EMI compliance (or just the marketing benefit of engineers going "nice" when they see it). i wonder if that's what those MLCCs in the middle of nowhere might be fore; it's not meaningfully reducing impedance to the power stages, but it could be reducing local voltage fluctuations in those planes. i'm not an engineer, just a hobbyist, but the impression i get is that EMC testing is so expensive that designers will just go completely over the top because the a few cents extra on the BOM is more palatable than the threat of failing and having to pay for another round of testing on another revision. (not to mention getting embarrassed at the fancy EMI lab!). So i wouldn't be surprised if those parts are entirely vibes-based.
ok i'm thinking about it more and some of the placement seems too weird for human superstition. i wonder if they have simulation software that's (semi-)automatically adding MLCCs at the noisiest spots, and that's what's causing some of the weird arbitrary looking positions and values.
Thanks for this very in-depth & informative video. I use this exact board for my machine and love it. Running 4 x 48-gb ram kit ( 192 gb ) hasn't been a problem.
I’m running 2x32 GB Corsair sticks on the QVL, and I have crazy POST times due to the memory training unless I enable the related UEFI options to reduce the boot time. Crazy difference when coming from 9th gen i9 which boots in
many people have reported problems with gene, and Asus has discontinued the board apparently due to some quality control issues. have you encountered any of the problems?
