MP All Models Tired of low Geekbench scores? Use RadeonBoost!

[leoaf79]

You need to add the follow in the config plist under Kernel -> add

            <dict>
                <key>BundlePath</key>
                <string>RadeonBoost.kext</string>
                <key>Comment</key>
                <string></string>
                <key>Enabled</key>
                <true/>
                <key>ExecutablePath</key>
                <string></string>
                <key>MaxKernel</key>
                <string></string>
                <key>MinKernel</key>
                <string></string>
                <key>PlistPath</key>
                <string>Contents/Info.plist</string>
            </dict>

Just put the kext in OC/Kexts won't do anything. We need to tell OC how to load this kext.

[code updated to remove Contents/MacOS/PropertyInjector]

After change config.plist, do I have to run Bless opencore.app again?
If you don't mind, I would like to know what this app (Bless opencore.app) do.

 

[h9826790]

After change config.plist, do I have to run Bless opencore.app again?

NO

If you don't mind, I would like to know what this app (Bless opencore.app) do.

I use the Automator approach, exactly because everyone can see what it does.

 

[Luca Maddalena]

It looks like you benchmarked in the wrong direction.

GB OpenCL / Metal, not GB CPU.

BruceX is more bottlenecking on encoding, not a Metal / OpenCL process. Especially true on the H264 exporting test, even with hardware encoding.

Luxmark is relavant, but we already know that's highly optimised in macOS, and this kext won't unlock anything for it.

Of course, this kext can be useless for your workflow. I won't argue that, you are the only one that know what you really need. But we also already proved that this kext definitely can unlock something. And few usages are proved can be benefited from it.

I obviously benched using GB OpenCL / Metal and not CPU.

 

[pierrox]

New version 1.3 added to the first post.

Cool, they're coming fast!
Do you publish a log of the version changes somewhere?
Edit: sorry I didn't realise you do, in the first post. 🤗
I have a RX580, anything specific you want me to check?

 

[leoaf79]

I use the Automator approach, exactly because everyone can see what it does.

Thanks for answer, I'm not familiar with automator.

 

[CMMChris]

Feel free to test more than just Geekbench with and without my kext (Luxmark, BasemarkGPU...). Maybe also games and real-life workloads if possible.

 

[startergo]

MacPro5,1
46140
Metal Score
Geekbench 5.1.0 Tryout for macOS x86 (64-bit)
Result Information

Upload Date	April 21 2020 02:06 AM
Views	1

System Information
�

System Information
Operating System	macOS 10.15.4 (Build 19E287)
Model	MacPro5,1
Motherboard	Apple Inc. Mac-7BA5B2D9E42DDD94

�

Processor Information
Name	Intel Xeon X5690
Topology	2 Processors, 12 Cores, 24 Threads
Identifier	GenuineIntel Family 6 Model 44 Stepping 2
Base Frequency	3.50 GHz
Package
Codename
L1 Instruction Cache	32.0 KB x 6
L1 Data Cache	32.0 KB x 6
L2 Cache	256 KB x 6
L3 Cache	12.0 MB x 1

�

Memory Information
Size	128.00 GB
Frequency	666 MHz
Type	DDR3

Metal Information
Device Name	AMD Radeon RX 480

Metal Performance

Metal Score	46140
Sobel	61566 15.9 Gpixels/sec
Canny	48263 3.02 Gpixels/sec
Stereo Matching	93732 132.6 Gpixels/sec
Histogram Equalization	31546 5.56 Gpixels/sec
Gaussian Blur	49302 2.71 Gpixels/sec
Depth of Field	113335 1.31 Gpixels/sec
Face Detection	9482 73.0 images/sec
Horizon Detection	42144 1.04 Gpixels/sec
Feature Matching	14845 307.1 Mpixels/sec
Particle Physics	172819 4603.2 FPS
SFFT	40090 552.2 Gflops

Compare
Set Baseline

A little bit less than version 2 results, but performance in HS is much smoother.
The highest GPU History spike is during "Histogram Equalization" test.

OpenCl test is much better:
MacPro5,1
41842
OpenCL Score
Geekbench 5.1.0 Tryout for macOS x86 (64-bit)
Result Information

Upload Date	April 21 2020 02:15 AM
Views	1

System Information
�

System Information
Operating System	macOS 10.15.4 (Build 19E287)
Model	MacPro5,1
Motherboard	Apple Inc. Mac-7BA5B2D9E42DDD94

�

Processor Information
Name	Intel Xeon X5690
Topology	2 Processors, 12 Cores, 24 Threads
Identifier	GenuineIntel Family 6 Model 44 Stepping 2
Base Frequency	3.50 GHz
Package
Codename
L1 Instruction Cache	32.0 KB x 6
L1 Data Cache	32.0 KB x 6
L2 Cache	256 KB x 6
L3 Cache	12.0 MB x 1

�

Memory Information
Size	128.00 GB
Frequency	666 MHz
Type	DDR3

�

OpenCL Information
Platform Vendor	Apple
Platform Name	Apple
Device Vendor	AMD
Device Name	AMD Radeon RX 480 Compute Engine
Compute Units	36
Maximum Frequency	1.27 GHz
Device Memory	8.00 GB

OpenCL Performance

OpenCL Score	41842
Sobel	54206 14.0 Gpixels/sec
Canny	31733 1.99 Gpixels/sec
Stereo Matching	129671 183.4 Gpixels/sec
Histogram Equalization	30983 5.47 Gpixels/sec
Gaussian Blur	46244 2.54 Gpixels/sec
Depth of Field	115053 1.33 Gpixels/sec
Face Detection	9330 71.8 images/sec
Horizon Detection	37808 931.9 Mpixels/sec
Feature Matching	12375 256.0 Mpixels/sec
Particle Physics	109123 2906.6 FPS
SFFT	39296 541.3 Gflops

And there is no spike during "Histogram Equalization" test. The highest GPU usage is during "Gaussian Blur" test.

RX-480 Video Card.

 

[startergo]

10.15.5 beta is even better:
MacPro5,1
42606
OpenCL Score
Geekbench 5.1.0 Tryout for macOS x86 (64-bit)
Result Information

Upload Date	April 21 2020 02:39 AM
Views	1

System Information

System Information
Operating System	macOS 10.15.5 (Build 19F53f)
Model	MacPro5,1
Motherboard	Apple Inc. Mac-7BA5B2D9E42DDD94

Processor Information
Name	Intel Xeon X5690
Topology	2 Processors, 12 Cores, 24 Threads
Identifier	GenuineIntel Family 6 Model 44 Stepping 2
Base Frequency	3.50 GHz
Package
Codename
L1 Instruction Cache	32.0 KB x 6
L1 Data Cache	32.0 KB x 6
L2 Cache	256 KB x 6
L3 Cache	12.0 MB x 1

Size	128.00 GB
Frequency	666 MHz
Type	DDR3
Memory Information

OpenCL Information
Platform Vendor	Apple
Platform Name	Apple
Device Vendor	AMD
Device Name	AMD Radeon RX 480 Compute Engine
Compute Units	36
Maximum Frequency	1.27 GHz
Device Memory	8.00 GB

OpenCL Performance

OpenCL Score	42606
Sobel	53469 13.8 Gpixels/sec
Canny	32169 2.01 Gpixels/sec
Stereo Matching	156525 221.4 Gpixels/sec
Histogram Equalization	30983 5.47 Gpixels/sec
Gaussian Blur	46506 2.56 Gpixels/sec
Depth of Field	115749 1.34 Gpixels/sec
Face Detection	9363 72.1 images/sec
Horizon Detection	36572 901.4 Mpixels/sec
Feature Matching	11414 236.1 Mpixels/sec
Particle Physics	117964 3142.1 FPS
SFFT	40570 558.8 Gflops

MacPro5,1
49636
Metal Score
Geekbench 5.1.0 Tryout for macOS x86 (64-bit)
Result Information

Upload Date	April 21 2020 02:41 AM
Views	1

System Information

System Information
Operating System	macOS 10.15.5 (Build 19F53f)
Model	MacPro5,1
Motherboard	Apple Inc. Mac-7BA5B2D9E42DDD94

Processor Information
Name	Intel Xeon X5690
Topology	2 Processors, 12 Cores, 24 Threads
Identifier	GenuineIntel Family 6 Model 44 Stepping 2
Base Frequency	3.50 GHz
Package
Codename
L1 Instruction Cache	32.0 KB x 6
L1 Data Cache	32.0 KB x 6
L2 Cache	256 KB x 6
L3 Cache	12.0 MB x 1

Size	128.00 GB
Frequency	666 MHz
Type	DDR3
Memory Information

Metal Information
Device Name	AMD Radeon RX 480

Metal Performance

Metal Score	49636
Sobel	63606 16.5 Gpixels/sec
Canny	47574 2.98 Gpixels/sec
Stereo Matching	140262 198.4 Gpixels/sec
Histogram Equalization	32025 5.65 Gpixels/sec
Gaussian Blur	49677 2.73 Gpixels/sec
Depth of Field	117989 1.37 Gpixels/sec
Face Detection	9627 74.1 images/sec
Horizon Detection	42367 1.04 Gpixels/sec
Feature Matching	13588 281.1 Mpixels/sec
Particle Physics	267382 7122.0 FPS
SFFT	38166 525.7 Gflops

 

[IanK MacPro]

Geekbench before
Metal 61859
OpenCL 57724

Geekbench before
Metal 91859
OpenCL 70619

OSX 10.14.6 Mojave, 80Gb ram

I've run various other benchmarks, Bruce X still around 8.5-9 seconds, I also output various graphic heavy Apple Motion projects and Final Cut Pro x projects before and after with no difference between either.

Luxmark also comes in the same at 105791

So not sure if there will be any real-world difference for most people other than possibly gaming?

 

[CMMChris]

Any GPU <-> Memory bandwidth demanding workload does benefit. If the workload ran into a bottleneck there it will perform better with loaded SMU firmware.

 

[IanK MacPro]

Any GPU <-> Memory bandwidth demanding workload does benefit. If the workload ran into a bottleneck there it will perform better with loaded SMU firmware.

That is true, its very software/project specific.

Would be interested to know if there is any speed difference in Final Cut Pro / Apple Motion etc between Mojave & Catalina, holding off for a while yet before I setup one of the machines as a test rig.

 

[leoaf79]

RadeonBoost 1.3 @ 5.1cMP - RX580 8GB (OpenCore):

• Mojave - everything worked well, but without performance improvement (Geekbench 5).
• Catalina - Geekbench 5 results drops to the half. If I deactivate SMBios spoofing, it backs to normal but Hwaccel doesn't work.

 

[pkdesign]

Non-OC data point:
* MP3,1 running Catalina, SIP disabled
* RX570 (same device-id as RX580, 0x67df)
* RadeonBoost linked into kernel (in /L/E)

The only difference RadeonBoost.kext seems to make is that the GPU fan hits 100% for about 5-10 seconds during mid-boot (remains slow/silent without the kext), which suggests that the kext is doing something. Various benchmarks were identical with/without the kext, and there is no noticeable difference in normal usage.
(EDIT: I do see the VBIOS Version and EFI Driver Version in System Information, so the kext is definitely being loaded.)

Do you think this is due to lack of explicit support for RX570, or something else? In every other context I've tested, the RX570 is functionally identical to the RX580, just slower.

I have the same question although I did not see any fan issues. I didn't see an answer. I have the RX570 card as well. This kext didn’t seem to make any difference. As the OP said the 570 is functional the same as the 580. These are all Ellesmere/Polaris cards.

Also, I see no mention of Clover. I have put this kext the the kext/other folder. Does it have to be placed in the L/E folder to work? It seems to have injected (as the OP mentioned) because I now see:
VBIOS Version: RadeonBoost
in System Info.

I'd love to see this work!

 

[inocraz]

I have put this kext the the kext/other folder. Does it have to be placed in the L/E folder to work? It seems to have injected (as the OP mentioned) because I now see:
VBIOS Version: RadeonBoost
in System Info.

Same thing. I injected alongside WG in OC. No noticeable change in Metal or OpenCL scores, hovering around 45K and 38K respectively. VBIOS version shows "RadeonBoost."

 

[CMMChris]

Score is fine, nothing that can be improved there.

 

[startergo]

Score is fine, nothing that can be improved there.

When I use 2 video cards with -wegtree the score drops in half and more. And then I can compare the same setup booting Vanilla without OC and any texts and the score in GB5 is as usual.
-wegtree places the cards at gfx0@0 and gfx1@0

 

[lirecsicnarf]

I saw that RX460 is NOT listed in the supported GPUs, but may I know why it isn't? Is it so much different with RX480? Or just because no test was done on these GPUs?

 

[efernandezjr]

Does this kext work for Clover users or is it Open Core only?

Clover 5112 -- MacOS 10.15.4 -- 9900k -- AsRock Z390 Phantom Gaming ITX/ac -- XFX Reference RX 5700xt

 

[tsialex]

Does this kext work for Clover users or is it Open Core only?

Clover 5112 -- MacOS 10.15.4 -- 9900k -- AsRock Z390 Phantom Gaming ITX/ac -- XFX Reference RX 5700xt

You are asking for the wrong crowd, Clover don't work with real Macs, here we only use native or via OpenCore.

 

[efernandezjr]

You are asking for the wrong crowd, Clover don't work with real Macs, here we only use native or via OpenCore.

Could you point a fellow Hackintosh user in the right direction?

 

[tsialex]

Could you point a fellow Hackintosh user in the right direction?

[KEXT] Tired of low Geekbench scores? Use RadeonBoost!

Hello everyone! Together with the help of @DSM2 I built a Kext aiming at removing limitations put in place by Apple for PC graphics cards that reduce performance in some benchmarks like Geekbench and some scenarios where memory bandwidth is crucial. Likely most of you already know the issue that ...

www.insanelymac.com

 

[pkdesign]

Score is fine, nothing that can be improved there.

So are you refering to the posts above for RX570 cards? There can be no improvemanent with these cards?

 

[dhl]

Awesome work, would love to try this with my Radeon VII on my x299 Catalina Hackintosh! But I'm not using OpenCore yet, still on Clover.

@CMMChris - can you give proper install directions for us Clover users please? Thanks!

 

[h9826790]

Awesome work, would love to try this with my Radeon VII on my x299 Catalina Hackintosh! But I'm not using OpenCore yet, still on Clover.

@CMMChris - can you give proper install directions for us Clover users please? Thanks!

If you don't know how to config Clover, you can do this

Tired of low Geekbench scores? Use RadeonBoost!

This kext does NOT require OpenCore, and it has nothing to do with OpenCore. OpenCore is just one of the method to load this kext, therefore, just need to do it once, than all installed macOS can benefit from it. No need to install that one by one for each single macOS installation. You can...

forums.macrumors.com

 

[pkdesign]

Awesome work, would love to try this with my Radeon VII on my x299 Catalina Hackintosh! But I'm not using OpenCore yet, still on Clover.

@CMMChris - can you give proper install directions for us Clover users please? Thanks!

From what I can gather from this site, which I believe may be the original posting (https://www.hackintosh-forum.de/for...hmark-scores-wie-am-echten-mac-unter-windows/) in German:

You simply place the kext in the kexts/Other folder in your Clover EFI