Stressing it doesn’t make it turbo, on the contrary it can make it stop using the turbo.
Turbo just depends on the temperatures and getting a single core to turbo is a lot more likely then stressing the whole CPU to 100 because that produces more heat.
Folks, something to add which might be of use. I was reckless and went ahead and flash-updated the onboard Intel X550 firmware. This was a mistake. The intel updater happilly flashed and verified the update, but the result was that after a full power off, only one of the two ports is functional. The one that can be shared with the BMC works (although I disable that). The other gets no link. Lots of other people have been posting on Intel forums. This MB and others from ASRock Rack with the same Intel NIC onboard. Intel engineers have come back with a response after months of investigation;
"We believe the issue is that Debug Mode is enabled which results in the single port failure following the update.
Debug Mode is entered when the signal to X550 between the Flash’s SCLK pin is pulled up when the X550 powers on (like a boot strapping configuration).
This mode is for engineering and debug work only and is not supported in normal operation or basically in manufacturing.
We surmise that ASROCK is shipping the motherboards in this configuration.
The Debug Mode tells the Firmware onboard the X550 it is ok to make certain writes to the NVM it might (not) normally. This would explain the behavior we see evidence of, and it also correlates to the isolation to this one manufacturer.
It is highly recommended to request RMA to your boards and share the aforementioned information with the OEM, ASROCK."
I have an elaborate case and cooling setup, and really can’t live without this system for the months that an RMA will take to process. I could buy a spare board and swap that in(a PITA though), then have the RMA as a spare. That’s a lot of $ to drop on a spare. I’m only using one 10G and one 1G port (plus BMC port for OOB), so I’m just going to live with this. with all the other problems mentioned in this thread, the lesson is that unless you’re a huge org with specific custom mobo needs, avoid ASRock Rack. I’ve learned a lot from the problems; BIOS, BMC, video, network etc. So its not all a waste. my system now works fine and is performing well in all the ways i want it to. buying a HP Z server would have cost a LOT more. a Xeon based system would have cost more and performed less.
my almost-idle temps (AMD Ryzen 7 5800X). i have fans connected to the six motherboard headers.
The Intel X550 and AMD X570 chips get very hot. I’ve got Noctua 40mm fans directly attached and with some DIY ducting on each. My case has giant front and rear fans to give a lot of airflow. CPU has a Noctua NH-U12S which does a good job of keeping it cool. I’ve also added heatsinks to the ASPEED and SuperI/O chips.
under load the CPU goes up to 90°C and then gets throttled to keep the temp from going higher. i get spike of 4500 MHz or more, but under sustained load, the clock speed with all 16 threads loaded up is 4400-4450MHz after 5mins or so and then slowly sinks to 4360-4370MHz, which is still a boost over the standard max of 3800.
I’m using the fan curve settings in the BMC to link temps with fan speeds. I added a sensor cable to the TR1 header and stuck that onto the heatsink on an ASUS Hyper M.2 x16 Gen 4 Card that I’m using with four NVMe storage cards. I replaced the noisy inefficient fan on this card with a Noctua one attached externally. DIY ducting to reduce hot and cold air mixing.
Cooling shouldn’t be an issue here as the thermals remain in check. One thing I have noticed is that when the OS first boots, it will happily turbo up speeds for a bit (In single and all core loads, both were tested), but then stop after 30 seconds not turbo for the entire rest of the time the system is powered on. This was tested on windows server 2019 and hyper-v server, so I dont think it is an OS issue, and I still think the BIOS is largely to blame. That or I got unlucky.
Update: After doing further investigation, I have discovered that it is a BIOS power limit which seems to be capping the CPU
Thanks asrock.
Setting ignore_ppc does not make a difference here either.
Hi,
i have a 3700x on this board. cat bios_limit gives me 3600000 but my CPU goes over 4000 Mhz at Turbo. So i think this do not cause your problem.
In that case, I am completely stumped then. My only assumption is that the current BIOS has issues with turboing the RYZEN 5000 series.
The boost will engage for the first 90 seconds of the system being online, and then stay capped at 3.6 until the system is rebooted. The same happens in Hyper-V server, windows server 2016, windows server 2019, and various linux live CDs (partedmagic just to name one).
I can only test in Proxmox not in Windows. Did you have installed latest Beta Bios? (1.34) this have newest Agesa 1.2.0.2
UPDATE
It works just fine in proxmox. I believe that windows doesn’t have the proper drivers(?) for this yet, and the partedmagic ISO is from 2018 so it likely doesn’t have them either.
Praise OSS
OK, so it semms to be a OS problem, not a Board problem. Did you have installed latest chipset drivers in Windows?
Yes I had all the AMD drivers and stuff installed. The turbo would work until the system uptime hit 1:30, and then it would refuse to turbo till I rebooted it, so my best guess is a power management service on a delayed start, that just doesnt work nicely with this config.
I am running almost the exact same setup as you AffectedArc07 and I found simply running the AMD chipset driver installer when the server is Hyper-V server left out a bunch of drivers needed for the boost speeds to work.
I do the following when building out the Hyper-V OS:
Run the chipset installer this will install some but not all drivers but it creates C:\amd that has all the drivers
Next from command line (with admin) run the following command to install all the amd drivers:
pnputil /add-driver c:\amd*.inf /subdirs /install
Also be sure the power plan is set to high performance but I see you have that part already. Hope this helps. Screen shot is of my 5800x system showing boost speeds from windows admin center.
I finally upgraded an old fX8350 system and just bought all the following:
X570d4u
Ryzen 5600x
KSM32ES8/16ME x2 (32GB ECC)
Somehow my board came with bios 1.00 so I had to update to 1.1 through the BMC, no problem. Uptime so far 1 week with CPU running 100% 24/7 (re-encoding videos), no faults so far. OS is windows 10 AME 20H2 and I only needed to install the X570 chipset package and intel i210 proset drivers to get it fully going.
For those who don’t know, AME edition doesn’t receive any updates at all so this will be a good test to see if the system remains stable.
I set all fans to 100% with the BMC and so far the CPU hasn’t gone above 60c with a Noctua NH-D14, and the x570 chipset hasn’t gone above 50C. Ambient temperature here is 15-18c in the day, and the case is very well ventilated with 9x Akasa Viper fans.
With the stock cooler CPU was reaching 70c, and (alarmingly) 90c with PBO enabled, which at the moment seemingly isn’t worth it as the CPU is running at 4.25 - 4.3GHz all day without, and won’t go above 4.4Ghz with PBO enabled.
Total power consumption according to UPS is around 187w with 10x WD RED 3TB drives not really doing much, although will be a lot higher with PBO enabled.
Overall I’m really happy with this server and I’m pretty much just enjoying it at this point.
It appears Asrock screwed us all with RAM overclocking. The voltage setting in BIOS does not do anything, and it does not depend on BIOS version.
Because they used the cheapest available VRM controller for RAM, namely ISL 8121 IRZ which does not support control via I²C.
I managed to persuade it to give more millivolts to DRAM (need to change a resistor near PU25 marking), will test and post results later.
Good thread (and very long!). Been considering this m/board, and read a fair bit of this thread before I pulled the trigger…
I bought it B-Ware (open box but as-new) from a German ecommerce site. However after assembling the server this evening I immediately ran into problems.
The CPU is Ryzen 5950x which is on ARock’s CPU support list. RAM is 4x Kingston KSM32ED8/32ME (ditto). However the BIOS doesn’t post, the LED counter shows “21” and HDMI outputs no signal. So I figured I need to upgrade BIOS for it to support my Ryzen 5950x, and I managed to find the the IPMI web console (the management i/face used DHCP), HOWEVER the page just gives me “Login failed” with the default credentials (admin/admin).
So I’m guessing one of the following:
a) m/board faulty
b) RAM or CPU faulty which triggers weird (incorrect) behaviour in IPMI
c) the default IPMI credentials have been changed by the previous owner (who returned it) to something else
If c is the case, how does one reset IPMI credentials without having to boot…? (which I can’t)
I’ve sent a complaining email back to the vendor as well but I don’t expect to hear from them for a while.
Would appreciate other pointers or ideas in the meanwhile…!
PS just tried removing all RAM sticks, and get exactly the same outcome. So it’s not faulty RAM.
Yes almost certainly option C, as the first thing that happens when you login is that you are prompted to change credentials
That would be extremely irritating in that case. Is there an option to hard reset BMC/BIOS via physical access somehow? That also resets the IPMI settings, not just BIOS.
It should™ work to just take out the battery and pull the plug, wait for a minute or so and maybe press the power button without any actual power to get any residual charge out of the capacitors. That should reset the entire BMC to defaults.
I would recommend to return this board.
I suspect there was a reason it was returned by previous owner. And after resetting credentials you will be faced with a new trouble…
