trx40 [Discussion] - TRX40 Bare Metal - Vanilla Patches

Driftwood · September 7, 2020

11 hours ago, iGPU said:

EDIT:

I cleaned up the compilations errors and added several GPRW calls that were missing (same as with MSI). Let me know how it boots.

Thanks @iGPU couldn't see a great deal of difference in BS on benchmark tests. Slight improvement on Apple Compressor rendering!

Big Sur BM Tests - NO XMP(Above 4G Enabled, Macpro7,1)

Edited September 7, 2020 by Driftwood

Driftwood · September 7, 2020

Catalina Tests with Asrock DSDT.aml

Geekbench 3 64-bit test

Edited September 7, 2020 by Driftwood

Driftwood · September 7, 2020

Im definitely getting mediocre results when looking at other TRX40 / Asrock board owners (rocket88). Im going to have to fine tune things...

fabiosun · September 7, 2020

Hi @DriftwoodI think you have some task consuming cpu cycle, it seems also your cpu benchmark is not so stable

optimal results for 3970x is about 17500 on CB20

maybe you can check with monitor activity

It is weird that radeonVII as a benchmark about 100 FPS (only IGPU has a more proper value for this card after patching his DSDT)

Maybe a poor support in BS also

fabiosun · September 7, 2020

a bit off topic here

Would like to advice this pretty app I have using during its beta test period.

Useful for number conversion, mounting EFI, installer creation and much more..

Take a look:

a mini guide here:

Pavo · September 7, 2020

10 hours ago, iGPU said:

I'd already removed this when reducing the patch list down to 13 total patches. The GPU performance on C15 increased from a very low 67 to 97 FPS without Fix PAT.

When I added the modified DSDT file, the GPU further increased to 140 FPS.

Did you try the modified DSDT file? It should work since we use the same mobo. I derived it with Above 4G enabled, and did not check if this setting affects the DSDT file.

I did not because I do not used custom DSDTs, if you can provide the exact modifications you did I can replicate them with a SSDT. I also created a ACPI delete entry to remove the ACPI error that we were getting see below in the pic. This removes the ShakTooth table that is trying to overwrite the WMI1 device that already exist in the DSDT and causes the ACPI error.

Edited September 7, 2020 by Pavo

Ploddles · September 7, 2020

I finally had a chance to test the MMIOWhitelisting, disabling the last 3/4 entries but enabling all the others.

There was no difference between disabling the last 3 or 4 and the system booted fine. Shutdown and restart still work as previous. Sleep still does not work.

I also calculate my Slide value, which worked out as 80, but again, this doesn't seem to affect anything.

I cannot boot at all with Above 4G enabled and the debug version of OC does not show any MMIO values in the report.

Attached is my DSDT file to add to the collection - Gigabyte AORUS Xtreme.

DSDT.aml.zip

iGPU · September 7, 2020

54 minutes ago, Ploddles said:

I finally had a chance to test the MMIOWhitelisting, disabling the last 3/4 entries but enabling all the others.

There was no difference between disabling the last 3 or 4 and the system booted fine. Shutdown and restart still work as previous. Sleep still does not work.

I also calculate my Slide value, which worked out as 80, but again, this doesn't seem to affect anything.

I cannot boot at all with Above 4G enabled and the debug version of OC does not show any MMIO values in the report.

Attached is my DSDT file to add to the collection - Gigabyte AORUS Xtreme.

DSDT.aml.zip 14.25 kB · 0 downloads

I don't see anything to 'fix' on your DSDT file. Both of the GB DSDT file compile without errors and have the proper GPRW injections.

I not certain that you need to use a Slide value. I don't believe any one is using; I know I am not. It is not necessary to use Above 4G; if not working ignore it.

You do need to have a proper MMIO section, which is specific for your mobo. Show us the MmioWhitelist values and we can help you select the correct ones.

Edited September 7, 2020 by iGPU

Ploddles · September 7, 2020

Thanks for taking a look. I have removed the Slide value now I have tested it and found it didn't change anything.

I followed your method on page 7 to obtain and calculate the Whitelist, it is only the last 3 or 4 values that are different from everybody else's and following the recommendations earlier I have enabled them all except the last 4.

Next thing to try is to remove all the USB devices, except the keyboard and mouse, to see if they affect Sleep. When I put it to sleep, the keyboard, mouse and monitors turn off and then the MB clicks and the BIOS output LEDs start to count up. The mouse, keyboard and monitor do not turn back on and the BIOS codes stop at 30. On a normal start/restart it counts up to AA. There is nothing in the manual about what '30' indicates and a google search wasn't fruitful.

Driftwood · September 7, 2020

@iGPU After trying the DSDT, and looking at comparable builds I am at a losss to explain your 149 fps? There must be something else you are doing or can you get down to the nitty gritty of the DSDT to find out the cause of this graphics performance in CB15? I have to O/c to 4Ghz just to get over 100fps!

I really need to understand how. You are not o/c GPU's etc..?. I know u have a decent water block setup but 50% greater than most of us is really quite stunning.

Edited September 7, 2020 by Driftwood

fabiosun · September 7, 2020

9 minutes ago, Ploddles said:

Thanks for taking a look. I have removed the Slide value now I have tested it and found it didn't change anything.

I followed your method on page 7 to obtain and calculate the Whitelist, it is only the last 3 or 4 values that are different from everybody else's and following the recommendations earlier I have enabled them all except the last 4.

Next thing to try is to remove all the USB devices, except the keyboard and mouse, to see if they affect Sleep. When I put it to sleep, the keyboard, mouse and monitors turn off and then the MB clicks and the BIOS output LEDs start to count up. The mouse, keyboard and monitor do not turn back on and the BIOS codes stop at 30. On a normal start/restart it counts up to AA. There is nothing in the manual about what '30' indicates and a google search wasn't fruitful.

code to 30 after a rewake from sleep could be fine

With Asus I have had always code 30

also in windows

Bad news is with gigabyte I never had a working sleep/wake cycle in my experience with x299 board

iGPU · September 7, 2020

31 minutes ago, Driftwood said:

@iGPU After trying the DSDT, and looking at comparable builds I am at a losss to explain your 149 fps? There must be something else you are doing or can you get down to the nitty gritty of the DSDT to find out the cause of this graphics performance in CB15? I have to O/c to 4Ghz just to get over 100fps!

I really need to understand how. You are not o/c GPU's etc..?. I know u have a decent water block setup but 50% greater than most of us is really quite stunning.

I've always added DevProper which some here don't like so I don't promote. But I've used before and after the custom DSDT file. I'll attach below as a plist file. To use, open in editor and copy/paste into your DevProp OC section. It is for dual Radeon VII in slots 1 & 3, which is what I recall you're using. These do inject some speed properties. I think they'll be okay with air-cooled. If it throttles or crashes during a stress test, let me know and I'll provide a less-energetic version. There are two sections for each GPU: one is for the display, the other for audio. The audio portion is simply descriptive.

If anyone wishes to use for 1 Radeon VII, only use the slot-1.

As far as OC, I have not. XMP is off. And none of these should much influence GPU performance. Until we trimmed the Kernel Patch list and I added the custom DSDT, my FPS were in the 60s. Now MarkLux has always been good and your values are the same as mine. So I have the feeling that your set up is just fine. (In my opinion, the tests are somewhat strange in and of themselves.)

DevProp-Dual-RadeonVII.plist.zip

Ploddles · September 7, 2020

Just out of interest, here is my score for a single lowly RX580. Before disabling the PAT patch it was 100.73, so that is quite a big performance boost.

iGPU · September 7, 2020

On 9/7/2020 at 11:57 AM, Ploddles said:

Thanks for taking a look. I have removed the Slide value now I have tested it and found it didn't change anything.

I followed your method on page 7 to obtain and calculate the Whitelist, it is only the last 3 or 4 values that are different from everybody else's and following the recommendations earlier I have enabled them all except the last 4.

Next thing to try is to remove all the USB devices, except the keyboard and mouse, to see if they affect Sleep. When I put it to sleep, the keyboard, mouse and monitors turn off and then the MB clicks and the BIOS output LEDs start to count up. The mouse, keyboard and monitor do not turn back on and the BIOS codes stop at 30. On a normal start/restart it counts up to AA. There is nothing in the manual about what '30' indicates and a google search wasn't fruitful.

CREATING MMIOWHITELIST

What I originally posted is unnecessarily cumbersome. There is no longer any need to test and re-boot after each modification. Instead, of re-doing the earlier post, I'll write an updated methodology here on how to make an MmioWhitelist (and link the original post to this one).

Here are the instructions in a nutshell.

First gather your debug text file, which is normally saved on the booted EFI partition. (If you're not certain how to create the debug text file, see the earlier MmioWhitelist post here.) Once you've gotten this debug text file, locate the area of interest within this file, by searching for "MMIO" as shown in the Spoiler below (an excerpt is taken from a file with a name such as "opencore-2020-08-31-150945.txt").

Excerpt from "opencore-2020-08-31-150945.txt":

Spoiler

Copy and paste this section, placing in a new text file for processing and future reference:

Spoiler






09:203 00:002 OCABC: MMIO devirt 0xCB100000 (0x81 pages, 0x8000000000000001) skip 1
09:206 00:002 OCABC: MMIO devirt 0xD7180000 (0x81 pages, 0x8000000000000001) skip 1
09:211 00:005 OCABC: MMIO devirt 0xE3180000 (0x81 pages, 0x8000000000000001) skip 1
09:214 00:002 OCABC: MMIO devirt 0xE3300000 (0x100 pages, 0x8000000000000001) skip 1
09:217 00:002 OCABC: MMIO devirt 0xEF100000 (0x181 pages, 0x8000000000000001) skip 1
09:222 00:005 OCABC: MMIO devirt 0xFEA00000 (0x100 pages, 0x8000000000000001) skip 1
09:225 00:002 OCABC: MMIO devirt 0xFEC00000 (0x1 pages, 0x8000000000000001) skip 1
09:228 00:002 OCABC: MMIO devirt 0xFEC10000 (0x1 pages, 0x8000000000000001) skip 1
09:233 00:005 OCABC: MMIO devirt 0xFED00000 (0x1 pages, 0x8000000000000001) skip 1
09:236 00:003 OCABC: MMIO devirt 0xFED40000 (0x5 pages, 0x8000000000000001) skip 1
09:239 00:002 OCABC: MMIO devirt 0xFED80000 (0x10 pages, 0x8000000000000001) skip 1
09:245 00:005 OCABC: MMIO devirt 0xFEDC2000 (0xE pages, 0x8000000000000001) skip 1
09:247 00:002 OCABC: MMIO devirt 0xFEDD4000 (0x2 pages, 0x8000000000000001) skip 1
09:250 00:002 OCABC: MMIO devirt 0xFEE00000 (0x100 pages, 0x8000000000000001) skip 0
09:256 00:005 OCABC: MMIO devirt 0xFF000000 (0x1000 pages, 0x8000000000000001) skip 1
09:259 00:002 OCABC: MMIO devirt 0x4040000000 (0x10400 pages, 0x8000000000000001) skip 1
09:262 00:002 OCABC: MMIO devirt 0x6F70000000 (0x10400 pages, 0x8000000000000001) skip 0
09:267 00:005 OCABC: MMIO devirt 0x9EA0000000 (0x10400 pages, 0x8000000000000001) skip 0
09:270 00:002 OCABC: MMIO devirt 0xCDD0000000 (0x10400 pages, 0x8000000000000001) skip 0

From this, copy the hex values, creating the following list, which may be in the same, new text file. (I've color-coded for purposes of description; you do not need to do this step).

Spoiler

0xCB100000
0xD7180000
0xE3180000
0xE3300000

0xEF100000
0xFEA00000
0xFEC00000
0xFEC10000
0xFED00000
0xFED40000
0xFED80000
0xFEDC2000
0xFEDD4000
0xFEE00000
0xFF000000

0x4040000000

0x6F70000000

0x9EA0000000

0xCDD0000000

These are for my MSI Creator mobo with Above 4G enabled. The red values will be used; the green will not be used (don't even bother entering them into the MmioWhitelist section). That is, the bottom 4 entries do not seem to be necessary for MmioWhitelist.

*****

Clarification (5/11/21):

Vit9696, the main person behind OpenCore, recommended not using the last 4 entries. However, the most up-to-date method of creating the MMIOWhitelist suggests disabling all entries in the above Spoiler that say "skip 0" and enabling all entries that say "skip 1". This is not discussed below (it is shown in the final code example), but is probably the best method of choosing whether or not to enable or disable a given entry. The color codes have been accordingly updated.

*****

Next, run calculations for each red hex value, creating another set of data, shown in white (the decimal conversion of the adjacent hex values). The bottom 11 hex values are in common with all of our mobos. If you have Above 4G enabled, they'll be on the bottom; if not, there will be only 2 values above 0xEF10000 and 2 other values below 0xEF10000, namely 0xFA180000 and 0xFA300000 (shown in orange in the 4G Disabled section).

So there is a bit of a pattern... the differences between our mobos seem to lay in the top 4 values if Above 4G is enabled, or, the top 2 values, if Above 4G is disabled, (shown in yellow below for the MSI Creator mobo):

Spoiler

4G Enabled:

0xCB100000 3406823424
0xD7180000.   3608674304
0xE3180000      3810000896
0xE3300000      3811573760

0xEF100000 4010803200
0xFEA00000    4271898624
0xFEC00000 4273995776
0xFEC10000 4274061312
0xFED00000    4275044352
0xFED40000   4275306496
0xFED80000     4275568640
0xFEDC2000   4275838976
0xFEDD4000      4275912704
0xFEE00000   4276092928
0xFF000000 4278190080

0x4040000000 275951648768

4G DISabled:

0xB2100000 2987393024

0xB3180000 3004694528

0xEF100000 4010803200

0xFA180000 4195876864

0xFA300000      4197449728
0xFEA00000    4271898624
0xFEC00000 4273995776
0xFEC10000 4274061312
0xFED00000    4275044352
0xFED40000   4275306496
0xFED80000     4275568640
0xFEDC2000   4275838976
0xFEDD4000      4275912704
0xFEE00000   4276092928
0xFF000000 4278190080

0x4040000000 275951648768

These values will, in turn, be entered into the MmioWhitelist section in OC (again, these are my values; yours will be somewhat different) for each fo the 15 values.

Note that all of these values will be enabled (set to "Yes"). Also note that these values are what were labelled above as "skip 1" ("skip 0" means that entry was set to "No" in MmioWhitelist).

Since we've removed the bottom 4 entries, which would otherwise have been disabled (set to "No"), all values in our new MmioWhitelist will be enabled (again, set to "Yes").

Spoiler

And the code will look like this (Above 4G enabled):

Spoiler




		<key>MmioWhitelist</key>
		<array>
			<dict>
				<key>Address</key>
				<integer>3406823424</integer>
				<key>Comment</key>
				<string>MMIO devirt 0xCB100000</string>
				<key>Enabled</key>
				<true/>
			</dict>
			<dict>
				<key>Address</key>
				<integer>3608674304</integer>
				<key>Comment</key>
				<string>MMIO devirt 0xD7180000</string>
				<key>Enabled</key>
				<true/>
			</dict>
			<dict>
				<key>Address</key>
				<integer>3810000896</integer>
				<key>Comment</key>
				<string>MMIO devirt 0xE3180000</string>
				<key>Enabled</key>
				<true/>
			</dict>
			<dict>
				<key>Address</key>
				<integer>3811573760</integer>
				<key>Comment</key>
				<string>MMIO devirt 0xE3300000</string>
				<key>Enabled</key>
				<true/>
			</dict>
			<dict>
				<key>Address</key>
				<integer>4010803200</integer>
				<key>Comment</key>
				<string>MMIO devirt 0xEF100000</string>
				<key>Enabled</key>
				<true/>
			</dict>
			<dict>
				<key>Address</key>
				<integer>4271898624</integer>
				<key>Comment</key>
				<string>MMIO devirt 0xFEA00000</string>
				<key>Enabled</key>
				<true/>
			</dict>
			<dict>
				<key>Address</key>
				<integer>4273995776</integer>
				<key>Comment</key>
				<string>MMIO devirt 0xFEC00000</string>
				<key>Enabled</key>
				<true/>
			</dict>
			<dict>
				<key>Address</key>
				<integer>4274061312</integer>
				<key>Comment</key>
				<string>MMIO devirt 0xFEC10000</string>
				<key>Enabled</key>
				<true/>
			</dict>
			<dict>
				<key>Address</key>
				<integer>4275044352</integer>
				<key>Comment</key>
				<string>MMIO devirt 0xFED00000</string>
				<key>Enabled</key>
				<true/>
			</dict>
			<dict>
				<key>Address</key>
				<integer>4275306496</integer>
				<key>Comment</key>
				<string>MMIO devirt 0xFED40000</string>
				<key>Enabled</key>
				<true/>
			</dict>
			<dict>
				<key>Address</key>
				<integer>4275568640</integer>
				<key>Comment</key>
				<string>MMIO devirt 0xFED80000</string>
				<key>Enabled</key>
				<true/>
			</dict>
			<dict>
				<key>Address</key>
				<integer>4275838976</integer>
				<key>Comment</key>
				<string>MMIO devirt 0xFEDC2000</string>
				<key>Enabled</key>
				<true/>
			</dict>
			<dict>
				<key>Address</key>
				<integer>4275912704</integer>
				<key>Comment</key>
				<string>MMIO devirt 0xFEDD4000</string>
				<key>Enabled</key>
				<true/>
			</dict>
			<dict>
				<key>Address</key>
				<integer>4276092928</integer>
				<key>Comment</key>
				<string>MMIO devirt 0xFEE00000</string>
				<key>Disabled</key>
				<true/>
			</dict>
			<dict>
				<key>Address</key>
				<integer>4278190080</integer>
				<key>Comment</key>
				<string>MMIO devirt 0xFF000000</string>
				<key>Enabled</key>
				<true/>
			</dict>
          <dict>
				<key>Address</key>
				<integer>275951648768</integer>
				<key>Comment</key>
				<string>MMIO devirt 0x4040000000</string>
				<key>Enabled</key>
				<true/>
			</dict>
		</array>

Edited May 11, 2021 by iGPU
"Clarification (5/11/21)" entry.

Driftwood · September 7, 2020

@iGPU Gave it a whirl in Catalina: No real performance increase (slight decrease). Im getting 98fps avg on BS without DevProp, but will try it over on BS shortly.

Here's OC debug text DevProp rip which I assume success means the DevProps were loaded ok.

Edited September 7, 2020 by Driftwood

iGPU · September 7, 2020

23 minutes ago, Driftwood said:

@iGPU Gave it a whirl in Catalina: No real performance increase (slight decrease). Im getting 98fps avg on BS without DevProp, but will try it over on BS shortly.

Here's OC debug text DevProp rip which I assume success means the DevProps were loaded ok.

Ploddles post above shows overall C15 scores about like mine and even with a single RX580 (but the RX580 won't compare as well if tested in LuxMark or using in DaVinci).

Ploddles · September 7, 2020

Just a slight difference 🙂

Driftwood · September 7, 2020

47 minutes ago, iGPU said:

Ploddles post above shows overall C15 scores about like mine and even with a single RX580 (but the RX580 won't compare as well if tested in LuxMark or using in DaVinci).

No its all good. Methinks I don't need any Device Properties I guess. Thanks so much for letting me try. Of course on Air, I don't wanna damage these babies! But I think you gave me a rather safe power table. I get the drift, we're all good in Metal and that's what matters.

meina222 · September 8, 2020

11 hours ago, Pavo said:

I did not because I do not used custom DSDTs, if you can provide the exact modifications you did I can replicate them with a SSDT. I also created a ACPI delete entry to remove the ACPI error that we were getting see below in the pic. This removes the ShakTooth table that is trying to overwrite the WMI1 device that already exist in the DSDT and causes the ACPI error.

@Pavo, I also get this ACPI error. Does the above apply to all boards or just MSI Creator? Thanks.

iGPU · September 8, 2020

2 hours ago, meina222 said:

@Pavo, I also get this ACPI error. Does the above apply to all boards or just MSI Creator? Thanks.

It should work on all of our mobos. Code is below to copy/paste (I added to comments to show how it is derived; data are basically ascii to hex conversions).

<key>Delete</key>
		<array>
			<dict>
				<key>All</key>
				<false/>
				<key>Comment</key>
				<string>Delete SHAKTOOH (hex=oemtab); hex SSDT=tablesig; 4271=0010AF (from "log show --last boot | grep ACPI")</string>
				<key>Enabled</key>
				<true/>
				<key>OemTableId</key>
				<data>
				U0hBS1RPT0g=
				</data>
				<key>TableLength</key>
				<integer>4271</integer>
				<key>TableSignature</key>
				<data>
				U1NEVA==
				</data>
			</dict>
		</array>

ckrueger99 · September 8, 2020

On 9/6/2020 at 2:49 PM, Driftwood said:

HEVC Encode Test on AMD Threadripper.

Adobe Media Encoder Metal Works on AMD Hackintosh - Apple's Compressor doesn't seem to be able to use Metal with AMD Hackintosh.

In 5 minutes, Adobe ME Beta had finished transcoding a 10 minute long 12-bit 4444 UHD video file to a HEVC 420 10-bit mp4, in the same time Apple Compressor had barely started - clearly its not making use of Metal and is Software/CPU only.

Handbrake is quicker than Compressor using CPU software encoding doing same file 20% quicker than Compressor.

Compressor would have taken over an hour to complete so I stopped it!

UPDATE: After much reading it appears the following:-

HEVC is hardware-accelerated only in single-pass, 8-bit mode. This is reasonably fast, but not as fast as compressing media into H.264.

10-bit HEVC, which is required for HDR media, is compressed in software – and, that takes forever!

If you are going to do any form of HEVC encoding, use ffmpeg or Handbrake GUI because they implemented tried and test, fast and
efficient encoders developed over many years, AND they fully implement the use of the hardware where it can.

@DriftwoodConsidering a professional media editor, whose primary work flow is in FCPX and, to a lessor extent, Adobe After Effects but not Premier, would the lack of HEVC hardware acceleration kill the decision to go with a TR hackintosh? Or would the same condition apply to an Intel Xeon 3175X machine? (Nearly as fast but much more expensive and using more power, but maybe without the HEVC compression problems?) Or does Apple Compressor depend on the T2, so we'd have the same problem with any brand of CPU , even a real Mac that doesn't have a T2?

Edited September 8, 2020 by ckrueger99

fabiosun · September 8, 2020

1) Without patch

2) proxmox

3)with patch

dtek · September 8, 2020

Hey guys,

I got my first computer build up and running and would love to join you guys with some testings. I got the installation screen up with EFI-v061-8e2755f-BareMetal-USB-installer-Public but stuck at [PCI configuration begin]. Any idea what the problem is?

Edited September 8, 2020 by dtek

Ploddles · September 8, 2020

10 minutes ago, dtek said:

Hey guys,

I got my first computer build up and running and would love to join you guys with some testings. I got the installation screen up with EFI-v061-8e2755f-BareMetal-USB-installer-Public but stuck at [PCI configuration begin]. Any idea what the problem is?

Try disabling Above 4G Encoding in the BIOS.

Pavo · September 8, 2020

55 minutes ago, fabiosun said:

1) Without patch

2) proxmox

3)with patch

Thought you said on discord yesterday that removing the Fix PAT effected Bare Metal in HS heavily?

Edited September 8, 2020 by Pavo

Sign In

trx40 [Discussion] - TRX40 Bare Metal - Vanilla Patches

Recommended Posts

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Create an account or sign in to comment

Create an account

Sign in

Recently Browsing 0 members

Important Information