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

[CMMChris]

Yup, same "fix" as back in the days pre 10.14.5. It just "works". The fans don't behave correctly. Slow spin down, slow reaction to temperature changes which can cause overheating. Not worth it given the real life performance differences with and without SMU firmware which are rather small.

[DrConflict]

This is a very interesting project. Have you thought about making a Github page for this project though? Thanks.

[Simonej]

Hi @CMMChris, apologies for the asking again, RX650 could be compatible replacing the device-id?
RX570 has the same device-id of RX580.

[WizeMan]

This is insane, thank you @CMMChris

 

RX580 Pulse 8GB

 

BEFORE:

OpenCL: 30823

Metal: 32984

 

AFTER:

OpenCL: 44887 (35%+ difference)

Metal: 45929 (32%+ difference)

 

[CMMChris]

New version 1.3 added to the first post.

[Pavo]

6 minutes ago, CMMChris said:

New version 1.3 added to the first post.

So... after asking if I could take this and add it to AGPMInjector and you telling me not too, you took AGPMInjector settings and added it to your kext injector. 

[CMMChris]

I don't get your point? You can use the device properties of RadeonBoost once optimum settings have been found. Likely this is the case now, so feel free to use them.

 

Would be great actually because people can then choose between a "one stop shop" kext that covers all relevant GPUs or generate one specifically for their set-up using your AGPM injector app.

 

Edit: Make sure to drop me a message once you implemented RadeonBoost, I will link your project then.

Edited April 20, 2020 by CMMChris

[Maf3r0]

Well, to me have no expressive impact:

 

Before:

 

After:

 

I've just put the kext at EFI/OC/Kext, updated config.plist and restarted.

 

SMBios iMac19,2.

 

Any tips?

 

Thanks!

[Labienus]

Great work Chris! This is really a big advancement and a spot of hope for those of us with affected cards who were hoping to get the most out of our systems! Any chance you can support can be expanded for the RX570? I've been having what looks like the same issue with that, as well as a bunch of other people over at AMD-OSD (which is how I found this post by the way, good news travels fast!). 

[neon-sunset]

It looks like there's no effect for RX580 when using v1.3 with MacPro7.1 smbios. If possible, could you share the v1.2 or consider keeping the previous versions of the kext?

(macOS Catalina 10.15.4, injected via OC 0.5.7)

Edited April 20, 2020 by neon-sunset

[CMMChris]

@Maf3r0 @IDON-NO What do you expect? Scores are up to the expected level in your cases. My kext only can improve scores if they are low. Example here:
https://forums.macrumors.com/threads/tired-of-low-geekbench-scores-use-radeonboost.2231366/post-28380276
https://forums.macrumors.com/threads/tired-of-low-geekbench-scores-use-radeonboost.2231366/page-2?post=28376589#post-28376589

 

@Labienus Not sure if it's necessary? How do the scores compare Windows vs. macOS?

Edited April 20, 2020 by CMMChris

[Maf3r0]

@CMMChris I just ask if there's a way to increase performance.

 

So, in my case has no difference b/c RX 580's performance at hackintosh is similar, or better, than in windows:

 

 

[Labienus]

1 hour ago, CMMChris said:

@Maf3r0 @IDON-NO What do you expect? Scores are up to the expected level in your cases. My kext only can improve scores if they are low. Example here:
https://forums.macrumors.com/threads/tired-of-low-geekbench-scores-use-radeonboost.2231366/post-28380276
https://forums.macrumors.com/threads/tired-of-low-geekbench-scores-use-radeonboost.2231366/page-2?post=28376589#post-28376589

 

@Labienus Not sure if it's necessary? How do the scores compare Windows vs. macOS?

 

 

@CMMChris, it's nuts that you are replying to all of the interest in this so quickly, thanks for being a real champ. 

 

I have have certainly noticed a big difference, and there are a bunch of people on Reddit and AMD-OSX forums as well. I just did a benchmark in Windows and here is now they compare: 

 

 

So if this issue affects the RX570 as well, which I think we might presume because we know of the issues with the 580, it does look like it could use the kext. 

 

Thanks again! Let me know if I can do other benchmarking or anything else to help. 

[CMMChris]

@IDON-NO What is it you don't understand? If you have no problem there is nothing to fix. So what do you expect?!

@Labienus I just checked, RX570 has the same device-id (67DF) as the Polaris cards already supported. So you can try RadeonBoost and check if it improves something.

Edited April 20, 2020 by CMMChris

[Labienus]

@CMMChris, I just gave it a shot, no luck unfortunately. Not sure what my issue is here but I guess the issue that I am having is unrelated. Thanks!

[bbmatias3]

i might be wrong but this issue mostly plagues amd users. Nvidia and amd gpus showing very low gpu test scores. not so much intel

[CMMChris]

Yeah there are a lot of issues with AMD hackintoshes. A lot of these machines also have issues with video encoding / decoding and DRM playback on AMD cards. I personally would never build or use an AMD hackintosh. At least not as a productive machine.

[maclinuxG4]

hello

 

it works well for my RX480... great job

 

 

[Paksman]

Hi, sorry if this is OT, i found a similar way from: https://github.com/acidanthera/WhateverGreen/blob/master/Manual/FAQ.Radeon.en.md. If you are using opencore with whatevergreen you can also inject gpu properties into config.plist, no need for kexts. It is working great and geekbench metal score skyrocketed from 39k to 66k+ on my rx 5700.

[Xanny]

Very good work @CMMChris the latest V1.3 bring a huge boost on metal and OpenCl scores.

@Paksman yes there are various way to obtain the same result, injecting with a kext, with Device Properties using Clover or Opencore, or thru a Dsdt, it's the same thing and bring the same result (using the same setup). The kext it's simpler in my opinion, you can remove it in a second in case of problems.

 

Edited April 21, 2020 by Xanny

[CMMChris]

@Paksman WEG doesn't automatically set the best settings for your card.

[Labienus]

@bbmatias3 ya, it's basically just people on AMD machines. For me my install went pretty smoothly and my MacOS is pretty stable, I haven't bothered getting iServices working though and obviously have this gfx issue. If someone already has AMD hardware and they are interested in the project, it's perfectly useable for most tasks, and can be stable if you aren't a total novice when you go to configure it. If I were to want to do any video editing with this device, or gaming (lol on a Mac), or to get ride of the other OS I have on here, I would definitely just spend the 200 bucks to swap-out to an intel chipset and compatible MoBo. 

[r!ppz]

On 4/19/2020 at 9:21 PM, CMMChris said:

Yup, same "fix" as back in the days pre 10.14.5. It just "works". The fans don't behave correctly. Slow spin down, slow reaction to temperature changes which can cause overheating. Not worth it given the real life performance differences with and without SMU firmware which are rather small. 

 

Are you sure that it is only buggy for Vega? On my 5700XT I notice that the fan is way louder under the same workload (gaming) with RadeonBoost enabled.

[CMMChris]

There is no fan bug on Navi cards. Undervolt it if you want it to be more silent.

[Tiem]

On 4/21/2020 at 4:21 AM, Paksman said:

Hi, sorry if this is OT, i found a similar way from: https://github.com/acidanthera/WhateverGreen/blob/master/Manual/FAQ.Radeon.en.md. If you are using opencore with whatevergreen you can also inject gpu properties into config.plist, no need for kexts. It is working great and geekbench metal score skyrocketed from 39k to 66k+ on my rx 5700.

 

This results in a KP on the latest Cat betas (10.15.5). Looks like 10.15.5 might break all this stuff. The Navi chip is really just all over the place.