Yeah but I’m hoping when the refresh comes out the price will drop a bit. I think it ridiculous what they are charging for this now.
2 Likes
@normanu It depends. If you run any type of software disk redundancy (aka software RAID/ZFS/UnRAID) I would recommend unbuffered ECC. If you are not, I would save your money and use standard unbuffered RAM.
Yea… At that price, I can consider an Intel server mobo and CPU and still be cheaper than AMD. What the heck?
Unless that is a misprint for the X470D4U2-2T?
Thats x399 MB ranges
1 Like
But that is x470 chipset on that mobo, no TR4 socket, only 4 memory slots, and a lot less PCI lanes.
This was supposed to be a budget server motherboard. At 450 USD, that’s hardly budget/entry level motherboard. Is this trade war issues or just supply and demand, or gouging?
I’ve already tried these “advanced techniques”, too.
3700X came today, system performance is still messed up - but differently (CPU_PROCHOT no longer appears).
Clock speeds under load etc. look normal, however usable oerformance is only a fraction of what it should be.
Going to RMA the board 
Yeah I’m rmaing my board too. Hopefully they have enough lol
On a final note: Of course, 5 min after my previous posting, CPU_PROCHOT started to show up in the IPMI log again 
That’s weird. I thought for sure going Zen 2 would’ve fixed that. It has to be a board issue. Let me know how the RMA goes because I have never done that with ASRock even though I’ve used three boards from them.
Good luck!
Can now confirm that the issues I reported are definitely coming from my X470D4U unit.
Got an ASRock X570 Taichi* and a second 3700X today, only swapped the motherboard, added a Radeon Pro WX 8200 and everythng is running as fast as expected.
The first 3700X that was initially used in the X470D4U even seems to be the better unit silicon qualitiy-wise, boosting 4,400 GHz reliably, the other one “only” 4,375 GHz. But who knows what future UEFI updates may bring (CPU heat sink on the ASRock X570 Taichi is a Noctua NH-U12A).
Got the 4 DR three year-old Samsung 2400 ECC UDIMM sticks running at 2933 [email protected] V (everything else Auto), which is a nice performance increase, higher fequencies need more manual tuning.
Could Samsung and Micron/Crucial please (re-)release 32 GB ECC UDIMMs with current DRAM chips?!
*Goddamn that M.2/chipset heatsink cover is a piece of work, I get the need for the little fan etc. but the stupid fixed M.2 covers and the chipset heat sink are actually not separable (unsrewed every srew I could find but the little chipset fan cooler cannot be separated). I intended to use the M.2 PCIe lanes with adapter cables for other devices which doesn’t work since the M.2 cover is completely blocking access unless you remove the entire thing exposing the chipset die.
Will have to look for an aftermarket solution.
Did you RMA it? ASRock support emailed me just now that they have a BIOS to test out.
I’m reading allot about motherboards having issues with the new Ryzen series because of bad VRM power supplies on the motherboards.
As in not being able to run at full speed etc.
Anyone have a idea what the VRM’s are like for this board?
Will it run the 3900x without a issue?
Not yet - that would be great if it really was just a firmware bug. Could you supply that new BIOS file since on the ASRock Rack website there’s still only the 3.04 from May 29th, 2019 listed.
The X470D4U handles CPUs up to a TDP of 105 W meaning every Ryzen 3000 up to the 3950X is supported. Regarding the reports of not running at full speed (that has nothing to do with “our” CPU_PROCHOT bug here), there seems to be a lot of confusion caused by AMD’s stupid marketing terms and buggy early BIOS versions.
Just today, Gamers Nexus released a quite well condensed video clearing these things up.
Additionally the X470D4U’s VRM heatsinks are real heatsinks that increase the surface area and not just a mass of aluminium like at most consumer boards so there should be no problem at all if the case has proper air flow.
Overclocking by increasing CPU frequency doesn’t really work with 7 nm Ryzen 3000 in a practical way anyhow - by increasing DDR4 frequency up to 3800 MHz with “tight” timings should lead to a real performance gain.
Could verify that with my four old Samsung 2400 ECC UDIMM sticks that work up to 3200 in the X570 Taichi.
Finally, the 7 nm CPUs like low temperatures very much, they draw less current for the same frequency/workload at lower temperatures reducing load on the VRMs.
PM sent, the US support has an unreleased BIOS update. I’ve sent who to contact to get it.
I sent ASRock Support my board and they cannot recreate the problem so I’m sending my CPU later to see. I’m also curious if potentially 16GB DIMMs are a factor as he only used 4 GB DIMMs for testing.
1 Like
Thanks for these details!
Will also be able to test an independently purchased second X470D40 soon, should be educational.
Some updates!
As mentioned I could get a second X470D4U for further testing.
- Firmware issue that was present on both motherboard units and was fixed with pre-release BIOS version L3.09A:
Fixed: No motherboard temperature sensor readings -> no fan control, fan speed set to default BIOS value not reacting to increasing CPU Tdie.
- Time Bug in BMC (?)
On both motherboard units the BMC date and time synchronization seems to be faulty - this may (?) cause weird issues.
The X470D4U motherboards and the separate system for BMC access get their date and time data from an NTP server in the local network; this works fine with various Windows, Linux and macOS systems.
The BMC however gets the time correctly but sets a wrong time zone (in my example I’m in GMT+0, the BMC for some reason gets GMT+2 (?) from the local NTP server, this is why the time stamps in the IPMI log jump around a bit).
- The CPU_PROCHOT issue is related to the power supply
The European support gave the advice to swap the used Seasonic G-450 for a different PSU. I tried this with a Seasonic Focus Plus Platinum 550 W and the CPU_PROCHOT issue is no longer triggered.
This is weird since the G-450 works fine in various other systems that are more power-hungry (ASRock X570 Taichi, 3700X and Vega 56; ASRock X99 WS, Intel Xeon E5-1650v4, 4 x 32 GB RDIMM).
(Edit: Correction: CPU_PROCHOT now also gets triggered when using Linux.)
I’ve attached various files for documentation from the second X470D4U unit I could test (the same hardware components were used as with the first unit), its IPMI log is complete, from the very first Power On with BIOS 3.04 and the G-450 where CPU_PROCHOT gets triggered and after the L3.09A update that didn’t fix CPU_PROCHOT and finally the PSU swap where the motherboard’s performance is normal.
I’m not competent on the field of PSUs and what could be wrong here (or if it is a power management firmware bug that gets only triggered with certain PSUs), but the G-450’s voltages seem fine and as mentioned above various other systems also work without any issue with the very same used G-450.
X470D4U_Second_Unit_2019-07-20_Seasonic_G-450_CPU_PROCHOT Other_PSU_CPU_OK_Complete_SELLog.txt (8.0 KB)
1 Like
All because you were not giving the board enough juice? Wow. I was using an Athena Power 500W PSU that was Bronze rated so perhaps I just hit the mark?
It’s interesting to me that something as simple as a lower wattage from the rails can cause this issue though.
It’s highly unlikely that this has something to do with the PSU’s maximum output wattage, possibly a PSU defect that isn’t a clear case that is triggered in full load/very light load conditions.
I just have no idea how to properly diagnose that since it’s a single rail PSU and the 5 V/12V voltages meassured at a Molex connector are well within spec (used a Fluke 117 multimeter).
1 Like
It was fine for him on normal AM4 boards. So its not about wattage. Unless the GPU is creating the issue maybe… not sure where it gets its power from probably the 24pin, or does it have its own 8 pin?
But Id think that probably more about voltages than wattages. But I dont own the board nor do I have any idea how to test that.
Could be related to output ripple.
1 Like