netkas.org forum

General Category => OSX and Graphics Cards => Topic started by: netkas on June 06, 2017, 08:33:55 AM



Title: NEW amd drivers in 10.13
Post by: netkas on June 06, 2017, 08:33:55 AM
Seems to be support for Fiji, polaris10 without patching


Title: Re: NEW amd drivers in 10.13
Post by: Ciro82 on June 06, 2017, 12:43:24 PM
Seems to be support for Fiji, polaris10 without patching
Tonight i'll test 10.13 with my RX580, we will see ;)


Title: Re: NEW amd drivers in 10.13
Post by: Pval on June 06, 2017, 02:34:32 PM
From egpu.io : supported AMD GPUs in macOS 10.13 High Sierra:

### AMDLegacySupport.kext
* pci device: 6704 - Cayman PRO GL [FirePro V7900]
* pci device: 6718 - Cayman XT [Radeon HD 6970]
* pci device: 6719 - Cayman PRO [Radeon HD 6950]
* pci device: 6720 - Blackcomb [Radeon HD 6970M/6990M]
* pci device: 6722 - unknown device
* pci device: 6738 - Barts XT [Radeon HD 6870]
* pci device: 6739 - Barts PRO [Radeon HD 6850]
* pci device: 6740 - Whistler [Radeon HD 6730M/6770M/7690M XT]
* pci device: 6741 - Whistler [Radeon HD 6630M/6650M/6750M/7670M/7690M]
* pci device: 6745 - unknown device
* pci device: 6750 - Onega [Radeon HD 6650A/7650A]
* pci device: 6758 - Turks XT [Radeon HD 6670/7670]
* pci device: 6759 - Turks PRO [Radeon HD 6570/7570/8550]
* pci device: 6760 - Seymour [Radeon HD 6400M/7400M Series]
* pci device: 6761 - Seymour LP [Radeon HD 6430M]
* pci device: 6768 - Caicos
* pci device: 6770 - Caicos [Radeon HD 6450A/7450A]
* pci device: 6779 - Caicos [Radeon HD 6450/7450/8450 / R5 230 OEM]
* pci device: 6840 - Thames [Radeon HD 7500M/7600M Series]
* pci device: 6841 - Thames [Radeon HD 7550M/7570M/7650M]
* pci device: 6898 - Cypress XT [Radeon HD 5870]
* pci device: 6899 - Cypress PRO [Radeon HD 5850]
* pci device: 68a0 - Broadway XT [Mobility Radeon HD 5870]
* pci device: 68a1 - Broadway PRO [Mobility Radeon HD 5850]
* pci device: 68b0 - unknown device
* pci device: 68b1 - unknown device
* pci device: 68b8 - Juniper XT [Radeon HD 5770]
* pci device: 68c0 - Madison [Mobility Radeon HD 5730 / 6570M]
* pci device: 68c1 - Madison [Mobility Radeon HD 5650/5750 / 6530M/6550M]
* pci device: 68d8 - Redwood XT [Radeon HD 5670/5690/5730]
* pci device: 68d9 - Redwood PRO [Radeon HD 5550/5570/5630/6510/6610/7570]
* pci device: 68e0 - Park [Mobility Radeon HD 5430/5450/5470]
* pci device: 68e1 - Park [Mobility Radeon HD 5430]
* pci device: 9440 - RV770 [Radeon HD 4870]
* pci device: 944a - RV770/M98L [Mobility Radeon HD 4850]
* pci device: 9488 - RV730/M96-XT [Mobility Radeon HD 4670]
* pci device: 9490 - RV730 XT [Radeon HD 4670]
* pci device: 94c8 - RV610/M74 [Mobility Radeon HD 2400 XT]
* pci device: 9501 - RV670 [Radeon HD 3870]
* pci device: 9540 - RV710 [Radeon HD 4550]
* pci device: 9553 - RV710/M92 [Mobility Radeon HD 4530/4570/545v]
* pci device: 9583 - RV630/M76 [Mobility Radeon HD 2600 XT/2700]
* pci device: 9588 - RV630 XT [Radeon HD 2600 XT]
 
### AMD7000Controller.kext
* pci device: 2200 - unknown device
* pci device: 2600 - unknown device
* pci device: 6780 - Tahiti XT GL [FirePro W9000]
* pci device: 6790 - unknown device
* pci device: 6798 - Tahiti XT [Radeon HD 7970/8970 OEM / R9 280X]
* pci device: 679a - Tahiti PRO [Radeon HD 7950/8950 OEM / R9 280]
* pci device: 679e - Tahiti LE [Radeon HD 7870 XT]
* pci device: 6800 - Wimbledon XT [Radeon HD 7970M]
* pci device: 6801 - Neptune XT [Radeon HD 8970M]
* pci device: 6806 - Neptune
* pci device: 6808 - Pitcairn XT GL [FirePro W7000]
* pci device: 6810 - Curacao XT / Trinidad XT [Radeon R7 370 / R9 270X/370X]
* pci device: 6818 - Pitcairn XT [Radeon HD 7870 GHz Edition]
* pci device: 6819 - Pitcairn PRO [Radeon HD 7850 / R7 265 / R9 270 1024SP]
* pci device: 6820 - Venus XTX [Radeon HD 8890M / R9 M275X/M375X]
* pci device: 6821 - Venus XT [Radeon HD 8870M / R9 M270X/M370X]
* pci device: 6823 - Venus PRO [Radeon HD 8850M / R9 M265X]
* pci device: 6825 - Heathrow XT [Radeon HD 7870M]
* pci device: 6827 - Heathrow PRO [Radeon HD 7850M/8850M]
* pci device: 682b - Venus LE [Radeon HD 8830M]
* pci device: 682d - Chelsea XT GL [FirePro M4000]
* pci device: 682f - Chelsea LP [Radeon HD 7730M]
* pci device: 6835 - Cape Verde PRX [Radeon R9 255 OEM]
* pci device: 6839 - unknown device
* pci device: 683b - unknown device
* pci device: 683d - Cape Verde XT [Radeon HD 7770/8760 / R7 250X]
* pci device: 683f - Cape Verde PRO [Radeon HD 7750/8740 / R7 250E]
 
### AMD9520Controller.kext
* pci device: 67c0 - Ellesmere [Polaris10]
* pci device: 67df - Ellesmere [Radeon RX 470/480]
* pci device: 67e0 - Baffin [Polaris11]
* pci device: 67ef - Baffin [Radeon RX 460]
* pci device: 67ff - Baffin [Polaris11]
 
### AMD2400Controller.kext
* pci device: 94c8 - RV610/M74 [Mobility Radeon HD 2400 XT]
 
### AMD6000Controller.kext
* pci device: 6704 - Cayman PRO GL [FirePro V7900]
* pci device: 6718 - Cayman XT [Radeon HD 6970]
* pci device: 6719 - Cayman PRO [Radeon HD 6950]
* pci device: 6720 - Blackcomb [Radeon HD 6970M/6990M]
* pci device: 6722 - unknown device
* pci device: 6738 - Barts XT [Radeon HD 6870]
* pci device: 6739 - Barts PRO [Radeon HD 6850]
* pci device: 6740 - Whistler [Radeon HD 6730M/6770M/7690M XT]
* pci device: 6741 - Whistler [Radeon HD 6630M/6650M/6750M/7670M/7690M]
* pci device: 6745 - unknown device
* pci device: 6750 - Onega [Radeon HD 6650A/7650A]
* pci device: 6758 - Turks XT [Radeon HD 6670/7670]
* pci device: 6759 - Turks PRO [Radeon HD 6570/7570/8550]
* pci device: 6760 - Seymour [Radeon HD 6400M/7400M Series]
* pci device: 6761 - Seymour LP [Radeon HD 6430M]
* pci device: 6768 - Caicos
* pci device: 6770 - Caicos [Radeon HD 6450A/7450A]
* pci device: 6779 - Caicos [Radeon HD 6450/7450/8450 / R5 230 OEM]
* pci device: 6840 - Thames [Radeon HD 7500M/7600M Series]
* pci device: 6841 - Thames [Radeon HD 7550M/7570M/7650M]
 
### AMD5000Controller.kext
* pci device: 6898 - Cypress XT [Radeon HD 5870]
* pci device: 6899 - Cypress PRO [Radeon HD 5850]
* pci device: 68a0 - Broadway XT [Mobility Radeon HD 5870]
* pci device: 68a1 - Broadway PRO [Mobility Radeon HD 5850]
* pci device: 68b0 - unknown device
* pci device: 68b1 - unknown device
* pci device: 68b8 - Juniper XT [Radeon HD 5770]
* pci device: 68c0 - Madison [Mobility Radeon HD 5730 / 6570M]
* pci device: 68c1 - Madison [Mobility Radeon HD 5650/5750 / 6530M/6550M]
* pci device: 68d8 - Redwood XT [Radeon HD 5670/5690/5730]
* pci device: 68d9 - Redwood PRO [Radeon HD 5550/5570/5630/6510/6610/7570]
* pci device: 68e0 - Park [Mobility Radeon HD 5430/5450/5470]
* pci device: 68e1 - Park [Mobility Radeon HD 5430]
 
### AMD9300Controller.kext
* pci device: 7300 - Fiji [Radeon R9 FURY / NANO Series]
 
### AMD8000Controller.kext
* pci device: 4500 - unknown device
* pci device: 4600 - unknown device
* pci device: 6640 - Saturn XT [FirePro M6100]
* pci device: 6641 - Saturn PRO [Radeon HD 8930M]
* pci device: 6646 - Bonaire XT [Radeon R9 M280X]
* pci device: 6647 - Bonaire PRO [Radeon R9 M270X]
* pci device: 6650 - Bonaire
* pci device: 6651 - Bonaire
* pci device: 665c - Bonaire XT [Radeon HD 7790/8770 / R7 360 / R9 260/360 OEM]
* pci device: 665d - Bonaire [Radeon R7 200 Series]
* pci device: 67b0 - Hawaii XT / Grenada XT [Radeon R9 290X/390X]
 
### AMD9000Controller.kext
* pci device: 6920 - unknown device
* pci device: 6921 - Amethyst XT [Radeon R9 M295X]
* pci device: 6930 - unknown device
* pci device: 6938 - Tonga XT / Amethyst XT [Radeon R9 380X / R9 M295X]
* pci device: 6939 - Tonga PRO [Radeon R9 285/380]
 
### AMD4800Controller.kext
* pci device: 9440 - RV770 [Radeon HD 4870]
* pci device: 944a - RV770/M98L [Mobility Radeon HD 4850]
 
### AMD2600Controller.kext
* pci device: 9583 - RV630/M76 [Mobility Radeon HD 2600 XT/2700]
* pci device: 9588 - RV630 XT [Radeon HD 2600 XT]
 
### AMD9510Controller.kext
* pci device: 67ef - Baffin [Radeon RX 460]
 
### AMD4600Controller.kext
* pci device: 9488 - RV730/M96-XT [Mobility Radeon HD 4670]
* pci device: 9490 - RV730 XT [Radeon HD 4670]
 
### AMD9515Controller.kext
* pci device: 67ef - Baffin [Radeon RX 460]
 
### AMD9500Controller.kext
* pci device: 67c0 - Ellesmere [Polaris10]
* pci device: 67c0 - Ellesmere [Polaris10]
* pci device: 67df - Ellesmere [Radeon RX 470/480]
* pci device: 67df - Ellesmere [Radeon RX 470/480]
* pci device: 67e0 - Baffin [Polaris11]
* pci device: 67e0 - Baffin [Polaris11]
* pci device: 67ef - Baffin [Radeon RX 460]
* pci device: 67ef - Baffin [Radeon RX 460]
* pci device: 67ff - Baffin [Polaris11]
* pci device: 67ff - Baffin [Polaris11]
 
### AMD3800Controller.kext
* pci device: 9501 - RV670 [Radeon HD 3870]
 
### AMDRadeonX4100.kext
* pci device: 67e0 - Baffin [Polaris11]
* pci device: 67ef - Baffin [Radeon RX 460]
* pci device: 67ff - Baffin [Polaris11]
 
### AMDRadeonX3000.kext
* pci device: 6720 - Blackcomb [Radeon HD 6970M/6990M]
* pci device: 6722 - unknown device
* pci device: 6738 - Barts XT [Radeon HD 6870]
* pci device: 6739 - Barts PRO [Radeon HD 6850]
* pci device: 6740 - Whistler [Radeon HD 6730M/6770M/7690M XT]
* pci device: 6741 - Whistler [Radeon HD 6630M/6650M/6750M/7670M/7690M]
* pci device: 6745 - unknown device
* pci device: 6750 - Onega [Radeon HD 6650A/7650A]
* pci device: 6758 - Turks XT [Radeon HD 6670/7670]
* pci device: 6759 - Turks PRO [Radeon HD 6570/7570/8550]
* pci device: 6760 - Seymour [Radeon HD 6400M/7400M Series]
* pci device: 6761 - Seymour LP [Radeon HD 6430M]
* pci device: 6768 - Caicos
* pci device: 6770 - Caicos [Radeon HD 6450A/7450A]
* pci device: 6779 - Caicos [Radeon HD 6450/7450/8450 / R5 230 OEM]
* pci device: 6840 - Thames [Radeon HD 7500M/7600M Series]
* pci device: 6841 - Thames [Radeon HD 7550M/7570M/7650M]
* pci device: 6898 - Cypress XT [Radeon HD 5870]
* pci device: 6899 - Cypress PRO [Radeon HD 5850]
* pci device: 68a0 - Broadway XT [Mobility Radeon HD 5870]
* pci device: 68a1 - Broadway PRO [Mobility Radeon HD 5850]
* pci device: 68b0 - unknown device
* pci device: 68b1 - unknown device
* pci device: 68b8 - Juniper XT [Radeon HD 5770]
* pci device: 68c0 - Madison [Mobility Radeon HD 5730 / 6570M]
* pci device: 68c1 - Madison [Mobility Radeon HD 5650/5750 / 6530M/6550M]
* pci device: 68d8 - Redwood XT [Radeon HD 5670/5690/5730]
* pci device: 68d9 - Redwood PRO [Radeon HD 5550/5570/5630/6510/6610/7570]
* pci device: 68e0 - Park [Mobility Radeon HD 5430/5450/5470]
 
### AMDRadeonX4500HWServices.kext
* pci device: 7300 - Fiji [Radeon R9 FURY / NANO Series]
 
### AMDRadeonX4150.kext
* pci device: 67e0 - Baffin [Polaris11]
* pci device: 67ef - Baffin [Radeon RX 460]
* pci device: 67ff - Baffin [Polaris11]
 
### AMDRadeonX4000.kext
* pci device: 6640 - Saturn XT [FirePro M6100]
* pci device: 6641 - Saturn PRO [Radeon HD 8930M]
* pci device: 6646 - Bonaire XT [Radeon R9 M280X]
* pci device: 6647 - Bonaire PRO [Radeon R9 M270X]
* pci device: 6650 - Bonaire
* pci device: 6651 - Bonaire
* pci device: 665c - Bonaire XT [Radeon HD 7790/8770 / R7 360 / R9 260/360 OEM]
* pci device: 665d - Bonaire [Radeon R7 200 Series]
* pci device: 6780 - Tahiti XT GL [FirePro W9000]
* pci device: 6790 - unknown device
* pci device: 6798 - Tahiti XT [Radeon HD 7970/8970 OEM / R9 280X]
* pci device: 679a - Tahiti PRO [Radeon HD 7950/8950 OEM / R9 280]
* pci device: 679e - Tahiti LE [Radeon HD 7870 XT]
* pci device: 67b0 - Hawaii XT / Grenada XT [Radeon R9 290X/390X]
* pci device: 67c0 - Ellesmere [Polaris10]
* pci device: 67df - Ellesmere [Radeon RX 470/480]
* pci device: 6800 - Wimbledon XT [Radeon HD 7970M]
* pci device: 6801 - Neptune XT [Radeon HD 8970M]
* pci device: 6806 - Neptune
* pci device: 6808 - Pitcairn XT GL [FirePro W7000]
* pci device: 6810 - Curacao XT / Trinidad XT [Radeon R7 370 / R9 270X/370X]
* pci device: 6818 - Pitcairn XT [Radeon HD 7870 GHz Edition]
* pci device: 6819 - Pitcairn PRO [Radeon HD 7850 / R7 265 / R9 270 1024SP]
* pci device: 6820 - Venus XTX [Radeon HD 8890M / R9 M275X/M375X]
* pci device: 6821 - Venus XT [Radeon HD 8870M / R9 M270X/M370X]
* pci device: 6823 - Venus PRO [Radeon HD 8850M / R9 M265X]
* pci device: 6825 - Heathrow XT [Radeon HD 7870M]
* pci device: 6827 - Heathrow PRO [Radeon HD 7850M/8850M]
* pci device: 682b - Venus LE [Radeon HD 8830M]
* pci device: 682d - Chelsea XT GL [FirePro M4000]
* pci device: 682f - Chelsea LP [Radeon HD 7730M]
* pci device: 6835 - Cape Verde PRX [Radeon R9 255 OEM]
* pci device: 6839 - unknown device
* pci device: 683b - unknown device
* pci device: 683d - Cape Verde XT [Radeon HD 7770/8760 / R7 250X]
* pci device: 683f - Cape Verde PRO [Radeon HD 7750/8740 / R7 250E]
* pci device: 6920 - unknown device
* pci device: 6921 - Amethyst XT [Radeon R9 M295X]
* pci device: 6930 - unknown device
* pci device: 6938 - Tonga XT / Amethyst XT [Radeon R9 380X / R9 M295X]
* pci device: 6939 - Tonga PRO [Radeon R9 285/380]
 
### AMDRadeonX4200.kext
* pci device: 67c0 - Ellesmere [Polaris10]
* pci device: 67df - Ellesmere [Radeon RX 470/480]
* pci device: 67ef - Baffin [Radeon RX 460]
* pci device: 67ff - Baffin [Polaris11]
 
### AMDRadeonX4250.kext
* pci device: 67df - Ellesmere [Radeon RX 470/480]


Title: Re: NEW amd drivers in 10.13
Post by: surfin on June 09, 2017, 10:27:14 PM
the 10.13 was quite easy to install over 10.12. nvme is about 10% faster on my 960 evo, the new disk format I suspect.

No acceleration on my Nano though.
The AMDRadeonX4500HWServices.kext isn't loading.
The AMD9300Controller.kext is.

Any ideas?



Title: Re: NEW amd drivers in 10.13
Post by: mysticalos on June 10, 2017, 04:51:22 AM
I'm hoping official enclosure support without patching means they'll take more care to support more GPUs out of box for said enclosures.

hoping it also forces them to update the STOCK nvidia drivers too to support something newer than kepler.


Title: Re: NEW amd drivers in 10.13
Post by: pigsyn on June 12, 2017, 02:46:10 PM
the 10.13 was quite easy to install over 10.12. nvme is about 10% faster on my 960 evo, the new disk format I suspect.

No acceleration on my Nano though.
The AMDRadeonX4500HWServices.kext isn't loading.
The AMD9300Controller.kext is.

Any ideas?


Maybe try to spoof 0X73001002 into AMDRadeonX4000 ?
Strange thing is Fiji is present in AMDRadeonX4000GLDriver, has a dedicated controller (AMD9300Controller.kext), but no reference in Graphics accelerator kext and plist. (AMDRadeonX4000.kext),


Is it a volcanic island card ? There is also references in 4100 and 4200 kexts.

It's like a section for AMDRadeonX4000_AMDVIGraphicsAccelerator in AMDRadeonX4000.kext is missing

Looks like a complete mess tbh
   


Title: Re: NEW amd drivers in 10.13
Post by: surfin on June 16, 2017, 01:57:33 AM
the 10.13 was quite easy to install over 10.12. nvme is about 10% faster on my 960 evo, the new disk format I suspect.

No acceleration on my Nano though.
The AMDRadeonX4500HWServices.kext isn't loading.
The AMD9300Controller.kext is.

Any ideas?


Maybe try to spoof 0X73001002 into AMDRadeonX4000 ?
Strange thing is Fiji is present in AMDRadeonX4000GLDriver, has a dedicated controller (AMD9300Controller.kext), but no reference in Graphics accelerator kext and plist. (AMDRadeonX4000.kext),


Is it a volcanic island card ? There is also references in 4100 and 4200 kexts.

It's like a section for AMDRadeonX4000_AMDVIGraphicsAccelerator in AMDRadeonX4000.kext is missing

Looks like a complete mess tbh
   

I have no idea what I'm doing and it even looks like a mess to me! haha.
In Sierra I could modify the X4000 kext and get acceleration.
In high Sierra any modification of the kext breaks it! Evidently Apple has decided to beef up kext protection and signing. Or hopefully that's present only in the beta and when it goes public it'll be able to be modified... Or they might fix R9 nano / fury - 0x73001002 kexts...

The takeout over this is that we might have to find another way to modify kexts that live in S/L/E


Title: Re: NEW amd drivers in 10.13
Post by: Squishy Tia on June 18, 2017, 12:50:30 PM
You do have SIP disabled, right? You're not going to be modifying any kexts while that's active. Typically when you install a new OS, SIP gets re-enabled.


Title: Re: NEW amd drivers in 10.13
Post by: surfin on June 19, 2017, 05:18:50 AM
You do have SIP disabled, right? You're not going to be modifying any kexts while that's active. Typically when you install a new OS, SIP gets re-enabled.

SIP is disabled.

I'm not too fussed. It hasn't even made it to public beta yet and Sierra is working fine.
I've got it running well on my backup X99 (12 core) xeon with an HD7970, and I'm liking that it doesn't need anywhere near as many patches or kexts to work.
edit: Running even better with the R9 380x, both oob without ATI injection or fake ids or anything.


Title: Re: NEW amd drivers in 10.13
Post by: mysticalos on June 19, 2017, 09:43:28 PM
I'm keeping a real eye on AMD drivers in 10.13 since I may be switching over to AMD next upgrade. With apple alienating nvidia and nvidia pushing out buggier and bugger web drivers, the inferior hardware may be worth it if the software support is better.


Title: Re: NEW amd drivers in 10.13
Post by: surfin on June 20, 2017, 04:00:34 AM
I'm keeping a real eye on AMD drivers in 10.13 since I may be switching over to AMD next upgrade. With apple alienating nvidia and nvidia pushing out buggier and bugger web drivers, the inferior hardware may be worth it if the software support is better.

The new vega cards should be good. Native support on account of next years mac pro will have them.
The X299 motherboards should also work well for the same reason. I quite fancy 18 cores for my design work :)
I've postponed this years upgrading to wait for them.


Title: Re: NEW amd drivers in 10.13
Post by: tomtomgps on June 22, 2017, 01:42:00 PM
I'm testing 10.13 on my mac pro 5.1. The R9 380X does seem to work better. At least the animations are a lot smoother. SInce the RX 580 is natively supported by high sierra doe sthat mean that we can get full 5,0 GT/sec bandwith with the 5.1 mac pro ?


Title: Re: NEW amd drivers in 10.13
Post by: nekton1 on June 25, 2017, 12:43:10 AM
PCIe Gen5 is coming soon and supports 32 GT/s, so what are the odds the 2018/19 Mac Pro will support it?  That's a lot of compute power with the right card 18 months out.


Title: Re: NEW amd drivers in 10.13
Post by: Squishy Tia on June 26, 2017, 07:52:56 AM
PCIe Gen5 is coming soon and supports 32 GT/s, so what are the odds the 2018/19 Mac Pro will support it?  That's a lot of compute power with the right card 18 months out.

2018 model? Not happening. If it isn't already out or due out shortly, there isn't enough time for Apple to R&D + order customized parts for their new Mac Pros. 2019? Unless Apple hits a grand slam out of the park with the new MPs, 2019 ain't gonna happen.


Title: Re: NEW amd drivers in 10.13
Post by: brockbackups on June 29, 2017, 10:11:36 PM
Interesting, 10.13 beta 2 update 1 is now showing PCI Information when using non flashed gfx cards
(http://i.imgur.com/uQL4AFw.png)


Title: Re: NEW amd drivers in 10.13
Post by: Spacedust on June 30, 2017, 03:34:47 PM
Interesting, 10.13 beta 2 update 1 is now showing PCI Information when using non flashed gfx cards
(http://i.imgur.com/uQL4AFw.png)

At least it's not displaying the name correctly ;)


Title: Re: NEW amd drivers in 10.13
Post by: brockbackups on June 30, 2017, 06:06:08 PM
Anyone know of the resistor that needs to be taken off of a RX 480 to get 5.0GT Link Speed?


Title: Re: NEW amd drivers in 10.13
Post by: tomtomgps on July 02, 2017, 10:23:44 PM
The reason we know which resistor  to remove on the 7950,  7970 or 280X is because they are basically  the same cards and there is a 7950 Mac edition. By comparing the 7950 Mac edition  to the 7950 Pc edition someone found out which resistor to remove, simply because it was missing on the 7950 Mac Edition card.  TBH, the 7950 was designed that way but from what I understand nothing guarantees that the RX 480 has a resistor  that can be removed without damaging the card and that would provide 5.0GT/sec on a MAC, in fact it's actually very unlikely. Since  AMD APPLE authentic cards were more expensive than their PC counterparts,  AMD probably used that  resistor   as a cheap way to differentiate  a PC Card from an APPLE card. From what I understand they used the resistor trick twice, once for the HD4850 and once for the HD7950. I do not know if AMD used that resistor trick for the HD5870 too, but from what I have read simply flashing a 5870 does not allow 5GT/sec on a Mac Pro.

For some reason NVIDIA cards on boot camp need to have a Mac EFI [EDIT: no need for a Mac EFI but it needs a firmware mod] in order to provide 5.0GT/sec , but they do not require any hardware modification.  In reality there very likely is a way to remove the 2.5GT/sec limitation on the RX 480 by altering its bios/ Mac Pro firmware or Mac OS X driver.

Has anyone tried to boot the Mac Pro with linux and the open source  MESA drivers to see what type of connection speed between the GPU and the Mac Pro is possible ? If the connection is 5GT under linux then we would know that the limitation is purely a software one.


Title: Re: NEW amd drivers in 10.13
Post by: Fl0r!an on July 03, 2017, 01:52:52 PM
From my understanding it's a hard strap (AMD) or soft strap (Nvidia) to configure the PCI slots properly at boot time. It's possible to do the same from an EFI shell: http://forum.netkas.org/index.php?topic=13541.0

Rominator has found the necessary modification on lots of cards which never had an official "Mac Edition".


Title: Re: NEW amd drivers in 10.13
Post by: nekton1 on July 31, 2017, 11:30:31 PM
Despite 10.13 beta 2 and later seeming to have support for the RX 460 (Baffin) in a host of AMD kexts, I cannot get my cMP 5,1 to boot with a Sapphire RX 460 in place (and nor could Rominator).
Before the Safe mode bug appeared in beta 4, it was possible to boot to an unaccelerated desktop in Safe mode but even that option is gone now.  Adding an old HD2500 XT as a helper card does not work and the cMP just shuts down.
When I try to boot with the RX 460, it gets to a whitish screen filled with snow of mixed RBY pixels which then cycles to a black screen with a white rectangle at top left and then sometimes to scrolling script down the left edge for a width of about 10 pixels and then the cycle repeats.
I have looked through the kexts for the correct board ID and it seems to be present and correct in the kexts.
What could be stopping this RX 460 in High Sierra? (It works fine in Sierra.)
Has Apple got a bug in the AMDX40xx.kexts or is this is a deliberate block on this card?
Any pointers for a fix other than buying an RX580 or Vega?
Thanks.


Title: Re: NEW amd drivers in 10.13
Post by: Ciro82 on August 04, 2017, 09:57:10 AM
Sierra 10.12.6 - Lilu.kext + Whatevergreen.kext

(http://i.imgur.com/JYVkvfE.png)


Title: Re: NEW amd drivers in 10.13
Post by: Rominator on August 10, 2017, 01:36:06 AM
Ciro82, is that a Hack?

If so, not surprising. The 2.5 lock is on the real Macs.

I was able to find this resistor on other AMD cards by using 2 tools

1. Voltmeter
2. Brain

I have no doubt that it can be found on Fury and Nova and whatever other AMD cards have come out.


Title: Re: NEW amd drivers in 10.13
Post by: SMIKX on August 10, 2017, 04:55:27 AM
Rominator. That caught my attention too. Under heavy load would this link speed be dangerous to eg a 5,1's PCIe bus ?


Title: Re: NEW amd drivers in 10.13
Post by: Squishy Tia on August 11, 2017, 01:50:27 PM
Rominator. That caught my attention too. Under heavy load would this link speed be dangerous to eg a 5,1's PCIe bus ?

Link speed? No. The link speed ideally, is determined by PCIe generation. 2.5 GT/sec for PCIe 1.x, 5.0 GT/sec for PCIe 2.x, ~8 GT/sec for PCIe 3.x. That doesn't factor in bitrate for each of those transactions, just the link speed itself. So even though a GPU is capable of 8 GT/sec, it will report only 5 GT/sec in any Mac Pro from 2008 onwards as none of the cMPs utilize PCIe 3.0.

Now, power draw is another can of worms altogether. You need to stay within the Mac's capabilities unless you add in external power. That generally means that your TBP ceiliing will be approximately 225 watts for safe operation (this assumes you are using both an SATA to PCIe 8-pin connector and the motherboard's two six pin connectors, likely bridged into a 2x 6-pin to 1x 8-pin adapter. If your GPU is rated for more than 225 watts, it is potentially unsafe for use in a Mac Pro***.

*** Under load. Idle and mild loads usually won't trigger a shutdown into protection mode. However, as Rominator found out, it's a case by case basis as some Pascal GPUs worked great regardless of load, and some didn't even make it to load at all.


Title: Re: NEW amd drivers in 10.13
Post by: Pval on August 11, 2017, 10:53:19 PM
The 6-pin connectors are 75watt each, the PCI bus another 75watt, so that's the 225watt you are talking about. A 5870 has a TDP of 188, 228 for the 2GB version. The 7950 has a TDP of 200watt. All supported in a standard Mac Pro.

Adding a SATA to 8-pin would raise your limit to 375watt, more then adequate for modern GPUs.

If you are concerned about pushing your PSU; my Mac Pro with a R9 290X (OC to 1100MHz), a W3690, 24GB, 2x PCIe M.2 SSDs, will draw just over 400watt max, measured at the wall, while playing games at 3440x1440@75Hz on max settings (Pixlas power mod). Idle it will sit around 90-140watt.

The PSU in a Mac Pro is rated at 980watt, the R9 290X has a TDP of 250watt (no OC, 1000MHz clock), the W3690 has a TDP of 130watt.
Yes, I removed all HDD and the DVD, never use the DVD anyway. For storage I use a NAS, or you could go eSATA.


Title: Re: NEW amd drivers in 10.13
Post by: Squishy Tia on August 12, 2017, 10:37:14 AM
Some GPUs don't even make use of the PCIe bus power except for booting up. Granted AMDs aren't among them (they're power hungry), but don't count on that PCIe bus power limit being able to be counted toward the cap - it isn't always in use. The 225w level is the only truly "safe" power ceiling for the Mac Pros. Getting above that is a YMMV and will undoubtedly be case by case depending on the GPU in use. You can get 300w with the two mobo connectors + SATA to 8-pin connector, but for some reason the Mac Pros are super finnicky with regard to power under load. Just ask Rominator about those quirks. He's seen power draw that defied logic many times. Usually the problematic suspect is the motherboard PCIe power lines since mobo traces just weren't meant to transfer that kind of power. Apple chose unwisely back then. Entrepreneuring users can of course find ways around that. :)

Still, if it's a first gen Polaris GPU, I'd stay the hell away from those even though it looks like Apple's AMD drivers contain AMD's PCIe bus draw limit "fix" in them.


Title: Re: NEW amd drivers in 10.13
Post by: h9826790 on August 12, 2017, 08:26:06 PM
Some GPUs don't even make use of the PCIe bus power except for booting up. Granted AMDs aren't among them (they're power hungry), but don't count on that PCIe bus power limit being able to be counted toward the cap - it isn't always in use. The 225w level is the only truly "safe" power ceiling for the Mac Pros. Getting above that is a YMMV and will undoubtedly be case by case depending on the GPU in use. You can get 300w with the two mobo connectors + SATA to 8-pin connector, but for some reason the Mac Pros are super finnicky with regard to power under load. Just ask Rominator about those quirks. He's seen power draw that defied logic many times. Usually the problematic suspect is the motherboard PCIe power lines since mobo traces just weren't meant to transfer that kind of power. Apple chose unwisely back then. Entrepreneuring users can of course find ways around that. :)

Still, if it's a first gen Polaris GPU, I'd stay the hell away from those even though it looks like Apple's AMD drivers contain AMD's PCIe bus draw limit "fix" in them.

May I know which graphic card that has 6pins or even 8pins connectors but doesn't use PCIe slot power?

I know lots of card won't pull 75W from slot, may be just 25-35W even under stress, but really don't know a card that only pull power from slot during boot (but pulling zero under Furmark).

That 225W limit is definitely assuming we pull 75 from slot, 75 from 6pin A, and 75 from 6pin B.

Also, if we talking about 100% technically safe config, then we should not use SATA to 8pin, not even dual SATA to 8pin. A 8pin is rated up to 150W, a SATA port only rated up to ~55W, even dual SATA port's limit is nowhere near 150W.

In general, we don't really need to care the card's TDP, but only the config of the card. If it has a 6pin, then connect it to the mini 6pin. If it has a 8pin, connect it to 2x mini 6pin. If it has 6+8pin, then connect the 6pin to 2x SATA, and the 8pin to 2x mini 6pin.

If the card's TDP is 225, but only come with dual 6pin, then it is safe to run this card in a cMP with just the dual mini 6pin.

If the card's TDP is 375W, then connect it to 2x mini 6pin + 2x SATA is also safe. I can't see how a card that rated > 225W TDP is unsafe for cMP especially you mentioned about the SATA power.

Anyway, the mini 6pin on the cMP can practically deliver up to ~120W. 225W even zero power from slot is still OK for the cMP. And I really don't know any card that has 225W TDP, come with dual 6pin, but drawing zero power from slot, and natively draw way more than 75W from the 6pins.

But I totally agree that we should avoid the gen 1 RX480. I don't believe that MacOS's driver has the power draw fix. And I really don't want the card draw too much and kill the slot (or even the logic board).


Title: Re: NEW amd drivers in 10.13
Post by: Squishy Tia on August 13, 2017, 07:05:33 PM
I can't remember which one it was, but Rominator ran into a Pascale nVidia GPU that essentially used virtually no power off of the PCIe bus whatsoever. It all really depends on how the card is designed. And the newer and more power hungry cards can cause problems if any other cards are in the machine because that 75w PCIe bus power is for all slots combined. It would actually make more sense to design a GPU to sip power from the PCIe bus (primarily for boot functionality) and draw the remaining power needed from the PCIe power connections directly from the PSU. PC mobo manufacturers get away with being able to SLI power hungry cards by using a PEG connector (SATA power connector directly on the mobo itself) to give the PCIe bus extra juice for stability. Mac Pros don't have that luxury, so if your GPU sucks down 75w of power, good luck powering that RAID card too if it doesn't have external power connections.


Title: Re: NEW amd drivers in 10.13
Post by: h9826790 on August 14, 2017, 10:12:26 PM
I can't remember which one it was, but Rominator ran into a Pascale nVidia GPU that essentially used virtually no power off of the PCIe bus whatsoever. It all really depends on how the card is designed. And the newer and more power hungry cards can cause problems if any other cards are in the machine because that 75w PCIe bus power is for all slots combined. It would actually make more sense to design a GPU to sip power from the PCIe bus (primarily for boot functionality) and draw the remaining power needed from the PCIe power connections directly from the PSU. PC mobo manufacturers get away with being able to SLI power hungry cards by using a PEG connector (SATA power connector directly on the mobo itself) to give the PCIe bus extra juice for stability. Mac Pros don't have that luxury, so if your GPU sucks down 75w of power, good luck powering that RAID card too if it doesn't have external power connections.

It's 75W per slot, 300W in total for all 4 slots.


Title: Re: NEW amd drivers in 10.13
Post by: Rominator on September 28, 2017, 09:03:37 PM
Plugged a Vega 56 in, saw the 8.0 GT/s link speed in PCI section. Must be cosmetic, just like when Nvidia cards say "2.5" but are really at 5.0.

Vega seemed OK until I threw Uningine at it. At Extreme HD it had issues that didn't show up in Medium settings.


Title: Re: NEW amd drivers in 10.13
Post by: Spacedust on October 02, 2017, 11:06:16 PM
Plugged a Vega 56 in, saw the 8.0 GT/s link speed in PCI section. Must be cosmetic, just like when Nvidia cards say "2.5" but are really at 5.0.

Vega seemed OK until I threw Uningine at it. At Extreme HD it had issues that didn't show up in Medium settings.

Another candicate for SATA bay power connector?


Title: Re: NEW amd drivers in 10.13
Post by: mysticalos on October 07, 2017, 03:13:33 AM
BTW, I've confirmed with MANY users now on macs with AMD hardware, the low performance bug after sleep bug is back in full force. if you use sleep. after sleep AMD gpu performs terribly in game and bencharms until you power cycle to get card out of low power state.

this bug first appeared in 10.7 on only AMD hardware and it wasn't fixed until 10.9. it came back for some users in 10.10 or 10.11 as well forget which and was again fixed.

it's back AGAIN in 10.13.

I'd be curious if rominator can do some testing with his vast array of GPUs.

This time around at least it's affecting metal, not sure if it's affecting opengl or not. in the previous version of bug it affected opengl but at the time there was no metal.. but YES it affects metal in 10.13.


Title: Re: NEW amd drivers in 10.13
Post by: Spacedust on October 08, 2017, 12:51:08 PM
Can you check PCI-Express speed after resuming from sleep?


Title: Re: NEW amd drivers in 10.13
Post by: mysticalos on October 08, 2017, 09:24:16 PM
I cannot, but I'm hoping some others can/will

I've also reached out to people who contacted apple who confirmed the bug so it might get fixed, for the third time, soonish

I have just seen the bug tanking performance on opengl/metal games on multiple forums from multiple users all using AMD hardware (real macs).

I don't know if it'd affect a hackintosh or not, probably? most hackintosh are on nvidia though.


Title: Re: NEW amd drivers in 10.13
Post by: tomtomgps on December 30, 2017, 04:39:51 PM
Looks like some people were able to enable HEVC decoding on the GPU  https://www.tonymacx86.com/threads/radeon-rx-4xx-5xx-standalone-system-amdradeonx4250-kext-gva-support-h264-does-not-support-hevc-hw.240353/