Monterey on Lenovo P340tiny / M90Q - The even better Mini/Tiny Hackintosh.

[rafale77]

Background:

 

After using "a perfect MacMini Hackintosh" for the past 2 years going from the HP Powerdesk/Elitedesk Mini G4 to G5 inspired by @deeveedee, I have decided to build a very similar system leveraging some the knowledge accumulated over the years. It has been a great machine but there were some limitations/quirks I was looking to overcome:

1. PCIe Expansion capability using cheaper non proprietary cards

2. CPU power limitation as I discovered recently that my 65W core i9 was underperforming due to its PL2 limit.

3. The elusive CPU MSRx02/CFGLock forcing an additional patch on the HP.

4. The one thing which doesn't function on the hack is DRM support. (I am still working on that one)

5. The reactivity of the BIOS command of the CPU fan in response to temperature hikes was lacking causing more thermal throttling than necessary.

 

For sake of learning about it and in the hopes of addressing the DRM issues, I settled on the one AMD GPU I could get my hands on which would fit the box: The Radeon Pro WX4100. It is single slot half height. These machines are normally sold with nVidia P620/P1000/T600/T1000 which can't be hacked... but are even smaller.

 

The Lenovo is a much more upgradable unit allowing for a greater realization of higher end CPUs potential. The Dell micro appear to be very similar to the HP Mini in terms of power supplies and for most configurations are limiting the CPU’s performance by decreasing its PL2.

Details on the power limitations on the CPU for those who prefer data in a table format comparing the HP Mini to the Lenovo Tiny for Q370/Q470/Q570 chipset units:

Model and configuration	35W Core i3/i5 PL2 (65W PL2)	35W Core  i7/i9 (90W PL2)	65W Core i3/i5 (135W PL2)	65W Core i7/i9 (235W PL2)	Power/dGPU Upgradable
HP 35W G4/5/6/8 65W PSU. PL2=50W	Limited. dGPU very limited	Very Limited	N/A	N/A	No. Limited by motherboard
HP 65W G4/5/6/8 90W PSU. PL2=65W	Ok *CPU downgrade	Limited	Very Limited	Very Very Limited	No. Limited by motherboard
HP 35W G4/5/6/8 with dGPU 150W PSU. PL2=90W	Ok	Ok	Limited	Very Limited	No. Nothing to upgrade to.
Lenovo M920Q/M920X with 90W PSU. PL2=65W	Ok	Limited	N/A	N/A	Yes
Lenovo M920X/ M90Q G1/G2. 135W PSU. PL2= 90W	Ok even with dGPU	Ok even with dGPU	Ok. Limited with  dGPU	Very Limited. no dGPU	Yes
Lenovo P330/P340/P350 170/230W PSU. PL2=135W	Ok with dGPU	Ok with dGPU	Ok with dGPU	Limited.	Yes

 

Side benefit: the NVMe SSDs are tool-less. SSD and DRAM are on the bottom side of the unit which frees up space to the top side.

Dimension wise the units are very similar and are part of the mini/micro/tiny 1L class. The Lenovo is 1mm thinner, about the same width but deeper by 5mm. The rear external wifi antenna design is definitely nicer on the HP.

My System Specs:

• Lenovo P340 (Core i9 10900 / UHD 630 / Q470 / 64GB DDR4 3200)
• AMD Radeon Pro WX4100 (In order to fit it, I had to remove the front wifi antenna panel and tape the antenna to the back of the GPU)
• 1 x DP port, 1x HDMi port for the iGPU
• 1 x USB type-C 3.2 Gen1, 2x USB 3.2 Gen1, 3x USB 3.2 Gen2
• 2x NVMe SSD (Hynix P31 2TB)
• 1x SATA6 (space occupied by the dGPU)
• 1x intel i219 Gigabit RJ45
• Realtek ALC235
• Apple BCM94360CS2 BCM943602CS with an M.2 adapter (upgraded to BT 4.1 and 3rd antenna with shielding stuck into the front panel with shielding.)
• 230W Power Adapter (CPU runs up to 135W in burst, dGPU takes up to 50W so the lower 170W would be a limiter as the system could need up to around 200W)

 

A few more details:

Audio ALC235 layout is 11 due to having only one headphone jack upfront.

I am initially providing 2 EFIs (dGPU and iGPU) but will only be maintaining the dGPU one eventually. The only difference will be in the WEG framebuffer where the dGPU version is using a headless Comet Lake Framebuffer (9BC80003), the iGPU version uses the Coffee Lake desktop (3E9B0007) with port patching. 

 

Working:

Everything as it was on my HP Mini:

• Continuity, watch unlock, sidecar
• Ethernet
• HDMI and DP Audio
• Boot Chime from the onboard speaker
• iGPU video decoding
• dGPU processing all the videos.
• Without dGPU, both DP and HDMi work with the framebuffer patching in the enclosed EFI.
• All USB ports mapped
• Sleep/Wakeup
• iServices

 

Some comparisons with the HP below: While the HP was limiting the PL2 to 65W, assuming that it had a 90W Power brick, the Lenovo uses 135/170/230W power supplies depending on the model. With 230W version and I can see from the CPU power go up above 135W under stress. The performance beats out more powerful 95W/125W TDP CPUs like the 10700K or my much older and bigger 9900K machine.

 

 

What is not working:

As on the HP Mini, DRM remains a challenge as the shikigva patch doesn't appear to work on Monterey with the IMac20,1 or iMac20,2 SMBIOS.

One solution using the AMD GPU could be to move to an iMacPro SMBIOS and disable the iGPU altogether but there are other potential issues with Sidecar and additional need to patch the CPU power management with CPUFriend which I will stay away from for now.

The workaround for me has been to permanently force Apple streaming sources to use the dGPU by executing:

defaults write com.apple.AppleGVA gvaForceAMDKE -boolean yes

and then use a chromium based browser like Brave for other source, thereby relying on software decoding.

The summary below shows the problem: Can't have the cake and eat it too on this one. The P350/M90Q Gen2 using a Rocket Lake Gen11 CPU would force the use of the iMP/MP SMBIOS...

 

Update on DRM: I have come to an acceptable solution by using unfairgva=5 in boot-arg which allows for Sidecar and Apple DRM without having to input a Terminal command. See this separate thread dedicated to the subject.

 

 

BIOS configuration: （version M2WKT54A M2WKT55A）

Set the Video output to Auto

Disable all the secure boot features

Enable all the Power Management States

Use the CFGLock.efi tool to unlock the MSRx02 bit and reboot before attempting a MacOS install/boot MacOS.

 

I have attached the iGPU version of the EFI. Moving to the dGPU requires simply changing the framebuffer patching of the dGPU to 9BC80003 which removes all the ports.

There are certainly better ways to patch the ACPI so any input is welcome...

 

After Playing around with the BIOS logo, I was able to replace it with an Apple one so now it really boots like a Mac.

The tool is embedded in the UEFI version of the BIOS update packages.
1. I created a USB boot drive for firmware updates of my P340.
2. Created a jpg logo file which needs to be relatively small, <20kb in my case, and put it on the USB drive.
3. Upon booting the firmware update USB drive, I rejected the firmware upgrade to access the UEFI shell.
4. The following commands then need to be executed:
Compress.efi logo.jpg
ChgLogo.efi compress.bin
If the file is small enough post compression, the logo will be loaded into the BIOS.

 

iGPU 0.8.0 Rev2.zip iGPU 0.8.2.zip

 

 

 

Edited August 4, 2022 by rafale77

[miliuco]

@rafale77

Good summary!

I think the best way to have DRM is to set SMBIOS MacPro or iMacPro with dGPU as main card and iGPU disabled in BIOS.

[rafale77]

On 4/21/2022 at 12:18 PM, miliuco said:

@rafale77

Good summary!

I think the best way to have DRM is to set SMBIOS MacPro or iMacPro with dGPU as main card and iGPU disabled in BIOS.

Thank you! Yes it seems to be the nuclear option to fix the DRM problem for good but I am a bit worried about losing the features like Sidecar as well.

In the meantime I have applied this in terminal:

 

defaults write com.apple.AppleGVA gvaForceAMDKE -boolean yes

It is not recommended in WEG docs to run on a permanent basis but it seems to work for now with no negative side effects for AppleTV. Amazon and Netflix on Safari don't work but I can use an alternate browser for these.

 

Also this machine seems to be even easier to hack than the HP as I am streamlining my ACPI patches... It appears to work fine. I have disabled HPET and rtcClock and further removed unnecessary patches.

 

I am attaching my updated dGPU version of the EFI.

 

 

Change Log :

 

OC 0.8.0

 

Rev1: Initial Rev

 

Rev2: Updated ACPI using configurations from the HP predecessor, mostly cosmetic and related to Power Management:

• SSDT-XSPI
• SSDT-DMAC
• SSDT-PPMC
• SSDT-PMCR
• SSDT-SBUS-MCHC

 

Rev3:

• Fixed DRM for AppleTV/iTunes using "unfairgva=5" boot-arg without sacrificing SideCar. See discussion here
• Increased boot chime volume
• Added SSDT-FWHD to support intel 82802 Firmware Hub (cosmetic)
• Removed Slot info from device properties to make it behave like a real mac
• Disabled picker and enabled Poll Apple Hotkeys in UEFI also to make it behave like a real mac
• Added framebuffer-unifiedmem to increase iGPU shared memory and removed unnecessary framebuffer-force-enable

 

Rev4:

• Overall outcome of further testing with SideCar and AppleTV DRM and comparison to real iMac ioreg output
• Removed SSDT-PPMC
• Added SSDT-ALS0-PNLF (dummy ambient light sensor and LCD backlight control), SSDT-P340tiny (Rename of SSD and airport card), SSDT-RM-PLIM (removes power Lenovo power limit settings, use at your own risks!) all disabled by default. All are cosmetic except for SSDT-RM-PLIM courtesy of @theroadw which potentially can improve CPU performance
• Added Voltageshift.kext but this file will likely not work as you will either need to recompile your own kext to work with your SIP settings or use a generic one with SIP disabled. This enabled changing power limits and undervolting of the CPU. The combination of undervolting and power limits mods yielded a net ~20% improvement in Cinebench multicore scores.
• Cleanup of iGPU properties to only have what is needed: rpc control and apple guc firmware load for performance and framebuffer-patch-enable to switch on WEG patches. Everything else was removed because they showed no difference on ioreg. -> This made Sidecar reliable and smooth running.
• Cleanup AMD dGPU properties to inject only missing properties to the AMD generic framebuffer. -> Improved Metal and OpenGL performance while also lowering power consumption and temperature.
• Further testing with Sidecar contradicted my conclusion from rev3. I have decided to run with SideCar fully functional and having to run a terminal command if DRM is needed so unfairgva and FeatureUnlock.kext were removed.

 

Rev5:

• Self-compiled lilu current 1.6.1 version from source with no modifications. Not sure why it is ~3x smaller than the compiled one on github.
• Self-compiled WEG version 1.5.9 with the only modification being the unfairgva bitmask 4 injecting macmini8,1 board-id instead of iMacPro1,1.
• Added unfairgva to AMD device properties and enable-metal to the iGPU properties. This combination allows for smooth sidecar running on AMD when GVA is switched to dGPU.

 

OC 0.8.1

 

Updated to OC 0.8.1.

Updated AppleALC.kext with self built one (again much smaller than the pre-built one on github)

 

 

OC 0.8.2

 

Rev1:

Updated OC to 0.8.2

Updated all WEG, NVMEFix, AppleALC, lilu to latest (self built from source)

Removed unnecessary kernel patches: Third Party Drive and ExtendBTFeatureFlags

Modified GPU injections to match iMac behavior (removed enable-metal for iGPU, and removed unfairGVA injection of fake board-id) as further testing indicated no significant benefits.

 

Rev2:

Removed EFICheckedDisabled.kext which is not needed on this rig.

Updated version comments for kexts in config.plist

Recompiled AppleALC.kext following @5T33Z0 guide which shrunk from >4MB -> 2.3MB (self compiled) -> 103KB

Recompiled my own fork of VoltageShift.kext from XCode UI which shrunk it from 233KB->78KB

Posted an updated iGPU version corresponding to dGPU 0.8.2 rev2 in the first post.

 

OC 0.8.3

 

Rev1:

Updated OC to 0.8.3 (opencore.efi and driver files)

Updated kexts to self compiled (with high optimization) from unmodified acidanthera source.

Removed unnecessary patches for dGPU in WEG.

Combined ACPI patches into one single optimized SSDT.

Removed SSDT-USBX as properties are already injected by USBMap.kext.

 

Rev2:

config.plist -- Removed unnecessary rtc-blacklist from add/remove section of NVRAM as well as some of the deletes since they never get changed through nvram configs

Removed USB port mapping kext as this device only declares 14 USB ports and doesn't exceed the MacOS limit. Added ACPI patch to enable the BT USB port which is disabled in the lenovo firmware.

Added back USBX patch but modified parameters to match real iMac20,2.

 

 

 

dGPU 0.8.3.zipdGPU 0.8.3 rev2.zip

 

Edited August 7, 2022 by rafale77

[rafale77]

I was browsing the GB5 scores and this little machine as configured with the iGPU is very competitive with the M1 MacMini:

 

My P340 Vs. MacMini

Single-Core: 1,350 Vs 1,700

Multi-Core:  9,600 Vs 7,200

Compute Metal: 22,600 Vs. 21,400

Compute OpenCL: 22,000 Vs 19,800 

 

Going with a P350 or M90Q Gen2 with a cheaper 11700 CPU would beat beat M1 in the Single-Core.

 

 

[rafale77]

DRM is definitely a pain on these machines... I finally ran into a problem with the command:

defaults write com.apple.AppleGVA gvaForceAMDKE -boolean yes

It forces sidecar to try to decode using the dGPU and makes it very flaky. It works and connects but breaks very easily and flickers occasionally. I initially thought I had an RF antenna issue but disabling the command above fixes the problem... I guess I will be using it only when I need itunes/AppleTV then...

 

I am practically done tinkering with OC so I disabled the picker and all debug/logging. The machine now almost boots like a Mac: Apple Logo(from BIOS)-Boot Chine-Apple Logo with Progress bar. Boot time is 5s longer than with the HP 800 G5 Mini. 26s Vs 21s from chime to login screen. It appears to be due to loading of the AMD drivers...

 

 

Edited April 26, 2022 by rafale77

[rafale77]

Posted 0.8.0 Rev3 which likely will be the final one until 0.8.1. Made some progress on making it behave like a real mac at boot and getting Apple DRM to work on the GPU while maintaining the iGPU for Sidecar. I discovered that on a real my MBp that both also cannot run simultaneously so I was on a goose chase. I think I am now as close as I can be from being fully functional with the exception of the notorious Amazon/Netflix DRM on Safari which can be overcome through using SW decode on a different browser.

Edited April 27, 2022 by rafale77

[rafale77]

A couple of updates:

I have just verified that the DisableRTCChecksum is not needed on these machines as it reboots and wakes from sleeps very smoothly so I took it out. I also cleaned up the OC tool folder and will post it in the next update.

I also had some fun testing the Monterey Universal Control (beta) feature which works amazingly well in combination with SideCar.

 

Adding a fun fact (well not so fun only a few hours ago when I was trying to figure this out): I suddenly noticed that my CPU power limit dropped to 15W so machine became awfully slow. I could not figure out how to recover, rebooting, cold booting, resetting BIOS, tinkering with OC as I was wondering if I messed something up... As it turns out, the motherboard on these Lenovo tinys adjust the CPU power limits according to the power brick they detect is powering them. The power bricks do not use simple dipolar connectors. On another forum, previous gen of tinys' users have reported performance improvements swapping out power supplies even with 35W TDP CPUs. Turns out for me that a power flicker caused by laser printer likely messed something up in the power supply and the motherboard went into low power mode... The only way to recover was to unplug the power brick for 5s and plug it back... The CPU PL2 went back to 135W. Note that the 95W i7 and i9 CPUs have an intel default PL2 of 224W!

Edited April 27, 2022 by rafale77

[rafale77]

More of a hardware selection discussion. I added a comparison table of different tiny/mini models and how they limit the CPU performance from a power perspective which has led me to pick this machine for my mini hack. Of course how it impacts real life depends on what the machine is used for. My OCD tendencies had me get really bothered by how much the multi core performance of my machine was handicapped/undermined by the power limitations of my previous hack.

[rafale77]

Upgrade:

 

Uneventful Upgrade to Monterey 12.4

 

Note: For universal control to work, the iOS device also needs to be updated to 15.5. I ran into the problem.

 

 

Performance:

Most of my current activities on the hack since I have come back have been around trying to "overclock" it by undervolting it and then seeking to increase the Power Limits in the BIOS.

These findings and discussions are in this other thread.

 

The gist of it is that this lenovo machine has hidden BIOS options which can be uncovered using the ControlMsrE2 tool. This tool oddly cannot write these bits, on this Lenovo machine even the CFGLock bit can't be changed by ControlMsrE2 but CFGLock.efi works. I had to use the RU.efi tool to edit the addresses I found from ControlMsrE2.

The process is as follows:

Launch ControlMsrE2.efi from the opencore openshell tool with the interactive argument.

Change the search string. 

Write down the found offset ControlMsrE2 outputs.

Launch  RU.efi and follow the process in the link to find the CPUSetup section

Go to the address offset previously written down and edit

Save the change by inputing CTRL-W.

 

I have disabled the OC Lock that way and enabled the Overclocking Features flag (which so far doesn't appear to do anything)

Then using VoltageShift, I have been able to find out that by default, with the 230W power brick, both PL1 and PL2 appear to be already set at 150W and I was able to undervolt the CPU.

These PL values reported do not match with my observations of the CPU volt under load or the performance of the machine which appear to be PL1=65W and PL2=120W

One oddity I discovered is that the i9 10900 performance actually degrades due to lower power limits once the OCLock flag is disabled.

I switched to an i9 10850K which does not exhibit this behavior and was able to improve the performance of this tiny... The 10850K however is powered with the same power limits as the 10900 running it essentially at 65W TDP CPU.

The performance is on par with a real iMac20,2 but with 65W actual TDP instead of 125W. It is interesting that the scores seem to have less variability with undervolting than before as I am seeing 3-8% improvement in benchmarks.

 

 

Comparing 10900 to 10850K undervolted (left) and 10850K before and after undervolting (right).

 

 

 

Edited May 17, 2022 by rafale77

[rafale77]

A new discovery related to SideCar and FeatureUnlock:

 

I tested removing FeatureUnlock.kext as its documentation indicated that I shouldn't need it given the SMBIOS at play not being covered but it appears that my use of unfairgva=5 makes Sidecar to run video encoding on the dGPU which doesn't work very well. As soon as I put it back, Sidecar is smooth again as it appears to have switched back to run on the iGPU... Just in case one day I would question why I have it in there.

Edited May 17, 2022 by rafale77

[rafale77]

Just wanted to show how the Radeon Pro fits in such a tiny chassis. Note that the fan position for this GPU is ideally located under the case vent opening.

Upon further optimizations of both the GPU which I will share in the next EFI and further undervolting to -130mV on Vcore, -30mV on iGPU and -120mV on VCache, temperatures dropped lower.

 

CPU Package Volt/CPU Min Temp/CPU Max Temp/CPU ave. Freq/VCore/dGPU Freq/dGPU Temp/dGPU Freq/dGPU fan speed/CPU fan speed/SSD1 Temp/SSD2 Temp

Ambient is 20C

[rafale77]

Uploaded rev4 of 0.8.0. I was hoping to wait until 0.8.1 to be released but I have made enough changes to warrant an update.

Changes are driven by ioreg comparison and matching with a real iMac20,1 to match ACPI device naming, performance improvements and addition of dummy devices and deletion on unnecessary/redundant device properties injections.

The whole iGPU/dGPU combination working to get SideCar/FaceTime work with DRM is quite an adventure... Hopefully a solution will be found soon to be able to dynamically force switching the AppleGVA from one GPU to the other dynamically or as a retry.

[rafale77]

Uploaded rev5. Still not perfect on the Sidecar/DRM side but this setup allows for a much smoother Sidecar experience when run on AMD (eliminated the freeze and drop issues)

 

When AppleGVA runs on intel iGPU:

• Sidecar is perfect
• No DRM

 

When AppleGVA runs on the AMD dGPU:

• Sidecar runs with very few graphical artifacts
• Apple TV DRM works.

 

I have attached below a little script I wrote to switch gva between dGPU and iGPU so that one doesn't need to go into terminal and submit a command.

gvaswitch

[rafale77]

Posted update with OC 0.8.1 release and associated kext updates.

[rafale77]

Decided to tinker with a native apple BCM943602CS card (circa 2015) which previously gave me troubles with watch unlock on the HP Mini G5. Somehow it seems to work fine now on the lenovo. The card fits perfectly but I had to add a third antenna which is now located in the front panel. It's an upgrade from the previous 2 channel BCM94360CS2 (circa 2013). ioreg also shows on lesser level of USB hub for the BT side of the card.

 

[rafale77]

So it turns out that either I was lucky with the bcm94360CS2 with the stock lenovo setup or I did not run enough long term/intensive testing, or the 3x3 bcm943602CS is much more sensitive/picky with antennas than the 2x2 bcm94360CS2...

 

I got curious about antennas and started using my 2 separated SSID setup between 2.4 and 5G wifi, the wifi scan tool and the combination of watch unlock and sidecar stress testing to find out that the lenovo stock antennas work great for 5Ghz but are very poor for 2.4GHz. I am now suspecting that Sidecar is heavily relying on the 2.4GHz radio (watch unlock heavily does since it communicates mostly over BT).

 

Watch unlock for example was working 100% to unlock apps (The unlock which requires double clicing the button on the watch) but was hit and miss when unlocking the computer itself (auto-unlock).

I discovered that I was communicating at full speed on 5GHz and was getting poor signal levels from my 2.4GHz SSID to the point of frequent connection failures.

 

So I gave up on pursuing a tiny antenna for looks which behaves about the same as the lenovo stock antenna and installed 3 third party antennas, 2 internal ones rated at 2dBi which appear to be stronger on both 2.4 and 5GHz from the wifi scan and a basic 3dBi one I had laying around and got much better reliability and signal strength on 2.4GHz...

 

Apple must have mastered quite an art with their aerials...

[rafale77]

So after much experimenting, I have finally found a working setup for my three antenna to get my bcm94602CS working:

 

connector J1 is connected to the external antenna in the back of the unit with a cheap 3dBi standard antenna.

connector J2 is connected to a Dell 0DJ3VC antenna from an XPS 7760 AIO desktop. I lodged it in the front top left of the unit, taped to the front bezel.

connector J3 is connected to one of the tiny antenna which as I found it is good for 5GHz but not so great for 2.4GHz based on the RSSI of my 2.4GHz network. Using two of these antenna, made SideCar unreliable. This antenna is also lodged in the bezel, in the front top right where it fits perfectly between clipping tabs of the bezel.

 

Very happy with the final result as I have now SideCar has become rock solid.

 

 

[Knuxyl]

I am trying to get my SFF P340 working with Opencore. I'm using 0.7.8 for compatibility with the latest dortania guide.

I have a Xeon W-1290T, nvidia T1000 with a patched ACPI to disable it, 2 nvmes, and intel wifi.

Intel wifi is working on my laptop and I've had this working before.

The problem I'm facing is Opencore keeps restarting PC right after boot before boot picker. The last line before the neverending NUL lines is

03:913 00:007 OCABC: Recovering trashed GetMemoryMap pointer

 

What are these aml files

SSDT-DMAC (I'm guessing memory addresses?)

SSDT-PMCR

SSDT-XSPI

 

and I don't believe our cpus need the SSDT-PPMC, but if I'm wrong please let me know. I would just throw your config on USB and test but too much stuff is commented out, i don't know if that's a mac thing, where file starts with ._

 

And the Tower, SFF, and Tiny P340s all share the exact same hardware

I was actually looking at getting a WX 4100 or 3200 in my SFF P340 for mac compatibility, but I'd have to use it in a x1 slot because I still want my nvidia because it's a lot better. Does the WX 4100 work good with mac?

Edited June 26, 2022 by Knuxyl

[rafale77]

@Knuxyl,

 

The three SSDTs you are asking about are mostly cosmetic and did not prevent me from booting. They are there to reassign the ACPI names to match how apple does it and I have them enabled on my setup.

You actually have a very different setup from mine just from the fact that you are using Xeon processor. This means that if you are disabling the nVidia T1000, you have to enable the iGPU which in your case probably has a very different ID from what MacOS would be expecting. I don't know what graphic ports are available on the SFF as they may not match what I have on my tiny. Take a look at my iGPU version of the EFI and in particular at the iGPU framebuffer injection.

 

PPMC is not needed. It is not even there on the iMac20,1/20,2.

 

The WX4100 works very well out of the box on macOS...

 

 

[Knuxyl]

.

Edited June 27, 2022 by Knuxyl

[Knuxyl]

Well I was wrong, our hardware is different. I guess it's just the sff and tower the same. You have q470 chipset and mine has w480 chipset. I got it working but I have no idea how.

 

Hows your temps in that thing? I know Pro AMD typically runs hot, Ive had a pro w5700 and w6600. 

[rafale77]

@Knuxyl, scroll up a few posts and you will see my temperatures at idle. I have never managed to get my GPU hot enough to throttle.

[rafale77]

Posted updated OC version to 0.8.2 with my most current configuration in post #3.

Edited July 5, 2022 by rafale77

[rafale77]

To give a look at what my hack setup look like on my desk... the tiny's footprint is barely larger than my trackball and about half of my mechanical keyboard...

I am also now using a tiny rear antenna.

 

[rafale77]

Nervously updated the BIOS on my unit to:

 

Quote

CHANGES for M2WKT55A
  - Enhancement to address security vulnerability.
  - Update CPU Microcode.

 

And was surprised to see that all the MSR settings stuck. NVRAM was wiped in the process though. My nervosity came from the CPU microcode update as I was worried my non officially supported 125W CPU would be removed.