Radeon VII compatible with Mac?  

@Kelvin Tan

Nice setup, quite happy you could get a taste on how these AMD card should run.
To unlock further the card power limit, you can check if soft modding the registry power table works like in Vega56/64.

About the fan issue, i really hope it is not a pcb revision issue and rather be fixed by a simple software update.
I already pointed out if someone could possibly check the VegaII DSDT file output under windows, then check the ACPI features.
Repeat the same process with Vega56/64, compare both DSST file.
If the issue come from a missing ACPI features, one could patch and load a custom DSDT file at boot with the right ACPI table and maybe fix the fan issue.

Great contribution, @kelvin_tan

@wimpzilla did that with the AMD memory tweaker but I’m not a pc gamer (only pc games played are mtg arena and everspace, everything else is on the Switch). so didn’t really push it. I mainly use the RVII for provis software such as autocad mechanical, solidworks, etc. In fact it’s back down to undervolted settings [email protected] Can’t wait for KVM linux to be ported to Mac, that way I don’t need to bootcamp just to view schematics.

you all are in luck! While I was ripping the card apart, I was thinking given my luck at these sort of things, the fans issue will be resolved in the next 5 days. If it really happens, I expect a like at least 😉

@oliverb thanks.

@kelvin_tan dude that looks very nice, did you do the washer mod? i think you could get your VII with better temperatures, i did that with mine and it definitely helped. Also, I received the 3d printed for my core X and it looks nice (than you again), however i think my fan sucks (in the bad way) because it doesn't seem to improve at all (it is an airflow fan, so maybe a static pressure one would be better).

I also 3d printed the mac mini stand that you shared a while ago, it is helping. (I'm still calibrating my 3d printer lol)

I can't wait for your finds for the sleep issue.

If this fan issue with the VII is serious then I may have to settle for a Liquid Cooled 64. It's a really great looking card and covers my needs.

Another update on the Radeon VII fan issue...  

I reset my NVRAM after @ningauble77 posted yesterday.  It now seems like the sleep fan issue is resolved in macOS 10.14.5 beta 5.  I've tried putting the MacBook to sleep in both open and clamshell mode - both with PowerNap on and off - several times.

@juan_c_yunis your core X looks great! Check the temps under load (mine fell about 7C), and also see if your fans are both facing either outwards or inwards - if both of them are inverse direction this setup does nothing and in fact makes it worse as the air from one is just immediately expelled by the other. The core X by default side fan is intake IIRC.

The waterblock comes with new mounting screws and it seems like the plate is making full contact, so no need the washer mods. Compared temps online with others on WC and the temps seems right, considering I'm using just a single 120mm radiator. And really not chasing world records for temps at idle it's at room temperature and at peak loads the delta (ambient is 28C) chip temp is just 17C, and junction is just about 41C. Can't complain about those and don't see a need to tweak further.

@kushnyc and @ningauble77 I resetted my SMC and pram when I tried before the WC but to no avail. Also it doesn't explain why even for windows machines such as my X1 tablet gen 3 it faces the same issue, unless you are saying that b5 fundamentally changes the way normal tb3 protocol operates only on Mac machines, or it writes something to the enclosure firmware or set a parameter? I don't recall a firmware update in b5 for the Mac...

Else it could be because your Macs aren't really sleeping, just the display turns off. Maybe run sudo pmset -g assertions and see if anything is keeping it awake?

Check sleep using pmset -g log|grep -e " Sleep  " -e " Wake  "

2019-05-11 00:41:49 +0800 Sleep                Entering Sleep state due to 'Software Sleep pid=99':TCPKeepAlive=active Using AC (Charge:100%) 115 secs 
2019-05-11 00:44:29 +0800 Sleep                Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using AC (Charge:100%) 7213 secs 
2019-05-11 02:45:27 +0800 Sleep                Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using AC (Charge:100%) 7213 secs 
2019-05-11 04:46:25 +0800 Sleep                Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using AC (Charge:100%) 7212 secs 
2019-05-11 06:47:22 +0800 Sleep                Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using AC (Charge:100%) 7213 secs 
2019-05-11 08:48:20 +0800 Sleep                Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using AC (Charge:100%) 7212 secs 
2019-05-11 10:49:17 +0800 Sleep                Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using AC (Charge:100%) 7212 secs 
2019-05-11 12:50:14 +0800 Sleep                Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using AC (Charge:100%) 5615 secs 
2019-05-11 14:23:55 +0800 Wake                 DarkWake to FullWake from Standby [CDNVA] due to UserActivity Assertion: Using AC (Charge:100%) 759 secs 
2019-05-11 14:36:34 +0800 Sleep                Entering Sleep state due to 'Idle Sleep':TCPKeepAlive=active Using AC (Charge:100%) 4737 secs 
2019-05-11 15:56:16 +0800 Sleep                Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using AC (Charge:100%) 136 secs 
2019-05-11 15:59:17 +0800 Sleep                Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using AC (Charge:100%) 90 secs   
2019-05-11 16:01:32 +0800 Sleep                Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using AC (Charge:100%) 154 secs 

Also, Geekbench in MacOS under WC seems to have improved another 8k, not sure if it's just a fluctuation, as I can't undervolt nor overclock in MacOS.

Heaven (Extreme settings) for reference under WC in MacOS 10.14.5b5

Also another Metal benchmark made by @psonice here! which I haven't ran in a long time - previously my Vega 56 was hitting about 3200~

Now with b5 and RVII WC, 5000~

So glad I can finally be on the latest MacOS and have reasonable performance (everspace extreme [email protected]) without having to resort to OS hacks, stick with High Sierra, or have to shutdown before ejecting gpu. If they do the rumored 16" MBP right, it's going to be a great combo!

@kelvin_tan tried both commands and i'm getting this:

pmset -g log|grep -e " Sleep " -e " Wake "
2019-05-11 04:51:28 -0500 Sleep Entering Sleep state due to 'Software Sleep pid=98':TCPKeepAlive=active Using AC (Charge:0%) 29 secs
2019-05-11 04:51:57 -0500 Wake Wake from Normal Sleep [CDNVA] due to EC.Bluetooth/UserActivity Assertion: Using AC (Charge:0%)
2019-05-11 05:00:04 -0500 Sleep Entering Sleep state due to 'Software Sleep pid=98':TCPKeepAlive=active Using AC (Charge:0%) 17 secs
2019-05-11 05:00:21 -0500 Wake Wake from Normal Sleep [CDNVA] due to EC.Bluetooth/UserActivity Assertion: Using AC (Charge:0%) 271 secs
2019-05-11 05:04:52 -0500 Sleep Entering Sleep state due to 'Software Sleep pid=98':TCPKeepAlive=active Using AC (Charge:0%) 17 secs
2019-05-11 05:05:09 -0500 Wake Wake from Normal Sleep [CDNVA] due to EC.Bluetooth/UserActivity Assertion: Using AC (Charge:0%) 523 secs
2019-05-11 05:13:52 -0500 Sleep Entering Sleep state due to 'Software Sleep pid=98':TCPKeepAlive=active Using AC (Charge:0%) 52 secs
2019-05-11 05:15:29 -0500 Sleep Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using AC (Charge:0%) 20 secs
2019-05-11 05:15:50 -0500 Wake DarkWake to FullWake from Normal Sleep [CDNVA] due to UserActivity Assertion: Using AC (Charge:0%)

pmset -g assertions
2019-05-11 05:16:02 -0500
Assertion status system-wide:
BackgroundTask 0
ApplePushServiceTask 0
UserIsActive 1
PreventUserIdleDisplaySleep 0
PreventSystemSleep 0
ExternalMedia 0
PreventUserIdleSystemSleep 0
NetworkClientActive 0
InternalPreventSleep 1
Listed by owning process:
pid 57(powerd): [0x00000414000d84af] 00:00:12 InternalPreventSleep named: "PM configd - Wait for Device enumeration"
Assertion applied on Battery power also
pid 57(powerd): [0x00000414000d84b1] 00:00:12 InternalPreventSleep named: "com.apple.powermanagement.acwakelinger"
Timeout will fire in 33 secs Action=TimeoutActionRelease
pid 104(hidd): [0x0000001d00098186] 00:00:00 UserIsActive named: "com.apple.iohideventsystem.queue.tickle.4294971722.3"
Timeout will fire in 600 secs Action=TimeoutActionRelease
Kernel Assertions: 0x12c=USB,BT-HID,THNDR,MAGICWAKE
id=500 level=255 0x20=THNDR mod=5/11/19, 5:16 AM description=NHI0 owner=AppleThunderboltHAL
id=517 level=255 0x100=MAGICWAKE mod=12/31/69, 7:00 PM description=en1 owner=en1
id=518 level=255 0x8=BT-HID mod=12/31/69, 7:00 PM description=com.apple.driver.IOBluetoothHIDDriver owner=AppleHSBluetoothDevice
id=531 level=255 0x4=USB mod=12/31/69, 7:00 PM description=com.apple.usb.externaldevice.14600000 owner=AN-VC500 Camera
id=534 level=255 0x8=BT-HID mod=12/31/69, 7:00 PM description=com.apple.driver.IOBluetoothHIDDriver owner=IOBluetoothHIDDriver
Idle sleep preventers: IODisplayWrangler

2019-05-05 16:25:43 -0400 Sleep Entering Sleep state due to 'Software Sleep pid=102':TCPKeepAlive=active Using AC (Charge:100%) 67 secs
2019-05-05 16:27:35 -0400 Sleep Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using AC (Charge:100%) 2330 secs
2019-05-05 17:07:35 -0400 Sleep Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using AC (Charge:100%) 8750 secs
2019-05-05 19:33:25 -0400 Wake Wake from Deep Idle [CDNVA] due to EC.LidOpen/UserActivity Assertion: Using AC (Charge:100%)
2019-05-06 11:00:47 -0400 Sleep Entering Sleep state due to 'Clamshell Sleep':TCPKeepAlive=active Using Batt (Charge:100%) 229 secs
2019-05-06 11:04:46 -0400 Sleep Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using Batt (Charge:100%) 3201 secs
2019-05-06 11:58:31 -0400 Sleep Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using Batt (Charge:100%) 450 secs
2019-05-06 12:06:06 -0400 Wake DarkWake to FullWake from Deep Idle [CDNVA] due to Notification: Using AC (Charge:100%)
2019-05-06 13:47:36 -0400 Sleep Entering DarkWake state due to 'Clamshell Sleep':TCPKeepAlive=active Using AC (Charge:99%) 17 secs
2019-05-06 13:47:53 -0400 Wake DarkWake to FullWake from Invalid [CDNVA] due to HID Activity: Using AC (Charge:99%) 4830 secs
2019-05-06 15:08:23 -0400 Sleep Entering Sleep state due to 'Clamshell Sleep':TCPKeepAlive=active Using Batt (Charge:100%) 231 secs
2019-05-06 15:12:14 -0400 Wake Wake from Deep Idle [CDNVA] due to EC.LidOpen/Lid Open: Using BATT (Charge:100%)
2019-05-06 15:19:45 -0400 Sleep Entering Sleep state due to 'Clamshell Sleep':TCPKeepAlive=active Using Batt (Charge:100%) 307 secs
2019-05-06 15:25:02 -0400 Sleep Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using Batt (Charge:100%) 1491 secs
2019-05-06 15:50:02 -0400 Sleep Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using Batt (Charge:100%) 1979 secs
2019-05-06 16:23:11 -0400 Sleep Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using Batt (Charge:100%) 1311 secs
2019-05-06 16:45:12 -0400 Sleep Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using Batt (Charge:100%) 1793 secs
2019-05-06 17:15:05 -0400 Wake Wake from Deep Idle [CDNVA] due to EC.LidOpen/Lid Open: Using BATT (Charge:99%)
2019-05-06 18:55:28 -0400 Sleep Entering Sleep state due to 'Clamshell Sleep':TCPKeepAlive=active Using Batt (Charge:68%) 23 secs
2019-05-06 18:56:03 -0400 Sleep Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using Batt (Charge:68%) 3918 secs
2019-05-06 20:01:25 -0400 Wake DarkWake to FullWake from Deep Idle [CDNVA] due to HID Activity: Using AC (Charge:67%)
2019-05-07 00:49:33 -0400 Sleep Entering Sleep state due to 'Software Sleep pid=102':TCPKeepAlive=active Using AC (Charge:100%) 30 secs
2019-05-07 00:50:03 -0400 Wake Wake from Deep Idle [CDNVA] due to EC.Bluetooth/HID Activity: Using AC (Charge:100%)
2019-05-07 09:03:33 -0400 Sleep Entering Sleep state due to 'Software Sleep pid=102':TCPKeepAlive=active Using AC (Charge:100%) 34 secs
2019-05-07 09:04:51 -0400 Sleep Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using AC (Charge:100%) 1940 secs
2019-05-07 09:37:56 -0400 Sleep Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using AC (Charge:100%) 2529 secs
2019-05-07 10:20:25 -0400 Wake DarkWake to FullWake from Deep Idle [CDNVAPB] due to Notification: Using AC (Charge:100%) 9 secs
2019-05-07 10:20:34 -0400 Sleep Entering DarkWake state due to 'Maintenance Sleep':TCPKeepAlive=active Using AC (Charge:100%)
2019-05-07 10:21:38 -0400 Sleep Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using AC (Charge:100%) 33 secs
2019-05-07 10:22:56 -0400 Sleep Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using AC (Charge:100%) 1849 secs
2019-05-07 10:53:45 -0400 Wake Wake from Deep Idle [CDNVA] due to EC.PowerButton/UserActivity Assertion: Using AC (Charge:100%)
2019-05-07 19:46:38 -0400 Sleep Entering Sleep state due to 'Clamshell Sleep':TCPKeepAlive=active Using AC (Charge:100%) 21 secs
2019-05-07 19:47:44 -0400 Sleep Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using AC (Charge:100%) 2354 secs
2019-05-07 20:27:55 -0400 Sleep Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using AC (Charge:100%) 3313 secs
2019-05-07 21:24:16 -0400 Sleep Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using AC (Charge:100%) 3403 secs
2019-05-07 22:20:59 -0400 Wake Wake from Deep Idle [CDNVA] due to EC.LidOpen/UserActivity Assertion: Using AC (Charge:100%)
2019-05-08 02:45:16 -0400 Sleep Entering Sleep state due to 'Clamshell Sleep':TCPKeepAlive=active Using AC (Charge:100%) 692 secs
2019-05-08 12:22:39 -0400 Sleep Entering Sleep state due to 'Idle Sleep':TCPKeepAlive=active Using Batt (Charge:100%) 236 secs
2019-05-08 12:26:42 -0400 Sleep Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using Batt (Charge:100%) 562 secs
2019-05-08 12:36:17 -0400 Sleep Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using Batt (Charge:100%) 1926 secs
2019-05-08 13:08:23 -0400 Wake Wake from Deep Idle [CDNVA] due to EC.LidOpen/Lid Open: Using BATT (Charge:100%) 2084 secs
2019-05-08 13:43:07 -0400 Sleep Entering Sleep state due to 'Idle Sleep':TCPKeepAlive=active Using Batt (Charge:95%) 151 secs
2019-05-08 13:45:38 -0400 Wake Wake from Deep Idle [CDNVA] due to EC.KeyboardTouchpad/User: Using BATT (Charge:95%) 1368 secs
2019-05-08 14:08:26 -0400 Sleep Entering Sleep state due to 'Clamshell Sleep':TCPKeepAlive=active Using Batt (Charge:91%) 592 secs
2019-05-08 14:18:23 -0400 Sleep Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using Batt (Charge:91%) 2576 secs
2019-05-08 15:01:29 -0400 Sleep Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using Batt (Charge:91%) 7133 secs
2019-05-08 17:00:22 -0400 Sleep Entering Sleep state due to 'Idle Sleep':TCPKeepAlive=active Using Batt (Charge:90%) 7215 secs
2019-05-08 19:00:44 -0400 Sleep Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using Batt (Charge:90%) 7213 secs
2019-05-08 21:01:03 -0400 Sleep Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using Batt (Charge:90%) 7212 secs
2019-05-08 23:01:21 -0400 Sleep Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using Batt (Charge:90%) 666 secs
2019-05-08 23:12:27 -0400 Wake Wake from Standby [CDNVA] due to EC.LidOpen/UserActivity Assertion: Using BATT (Charge:89%) 1474 secs
2019-05-08 23:37:01 -0400 Sleep Entering Sleep state due to 'Clamshell Sleep':TCPKeepAlive=active Using Batt (Charge:53%) 23 secs
2019-05-08 23:37:24 -0400 Wake Wake from Deep Idle [CDNVA] due to EC.LidOpen/UserActivity Assertion: Using BATT (Charge:56%) 580 secs
2019-05-08 23:47:04 -0400 Sleep Entering Sleep state due to 'Clamshell Sleep':TCPKeepAlive=active Using Batt (Charge:54%) 119 secs
2019-05-08 23:49:03 -0400 Wake Wake from Deep Idle [CDNVA] due to EC.LidOpen/Lid Open: Using BATT (Charge:54%) 249 secs
2019-05-08 23:53:12 -0400 Sleep Entering Sleep state due to 'Clamshell Sleep':TCPKeepAlive=active Using Batt (Charge:53%) 932 secs
2019-05-09 00:09:23 -0400 Sleep Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using Batt (Charge:50%) 4829 secs
2019-05-09 01:29:52 -0400 Wake Wake from Deep Idle [CDNVA] due to EC.USBC/Notification: Using AC (Charge:48%) 34 secs
2019-05-09 01:30:26 -0400 Sleep Entering DarkWake state due to 'Software Sleep pid=102':TCPKeepAlive=active Using AC (Charge:48%)
2019-05-09 01:30:41 -0400 Sleep Entering Sleep state due to 'Software Sleep pid=102':TCPKeepAlive=active Using AC (Charge:48%) 67 secs
2019-05-09 01:32:33 -0400 Sleep Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using AC (Charge:50%) 2282 secs
2019-05-09 02:11:44 -0400 Sleep Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using AC (Charge:82%) 1901 secs
2019-05-09 02:44:56 -0400 Sleep Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using AC (Charge:97%) 1850 secs
2019-05-09 03:16:31 -0400 Sleep Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using AC (Charge:100%) 2079 secs
2019-05-09 03:52:15 -0400 Sleep Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using AC (Charge:100%) 17428 secs
2019-05-09 08:42:43 -0400 Wake Wake from Deep Idle [CDNVA] due to EC.Thunderbolt/HID Activity: Using AC (Charge:100%) 126 secs
2019-05-09 08:44:49 -0400 Sleep Entering Sleep state due to 'Clamshell Sleep':TCPKeepAlive=active Using Batt (Charge:100%) 388 secs
2019-05-09 08:51:27 -0400 Sleep Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using Batt (Charge:100%) 5245 secs
2019-05-09 10:18:52 -0400 Wake Wake from Deep Idle [CDNVA] due to EC.LidOpen/Lid Open: Using BATT (Charge:100%)
2019-05-09 11:33:11 -0400 Sleep Entering Sleep state due to 'Clamshell Sleep':TCPKeepAlive=active Using Batt (Charge:59%) 44 secs
2019-05-09 11:34:15 -0400 Sleep Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using Batt (Charge:59%) 5456 secs
2019-05-09 13:05:19 -0400 Sleep Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using Batt (Charge:58%) 1308 secs
2019-05-09 13:27:15 -0400 Wake DarkWake to FullWake from Deep Idle [CDNVA] due to Notification: Using AC (Charge:57%) 4 secs
2019-05-09 13:27:19 -0400 Sleep Entering DarkWake state due to 'Maintenance Sleep':TCPKeepAlive=active Using AC (Charge:57%) 14 secs
2019-05-09 13:27:33 -0400 Wake DarkWake to FullWake from Deep Idle [CDNVA] due to HID Activity: Using AC (Charge:57%)
2019-05-09 15:20:12 -0400 Sleep Entering Sleep state due to 'Idle Sleep':TCPKeepAlive=active Using AC (Charge:100%) 398 secs
2019-05-09 15:27:35 -0400 Sleep Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using AC (Charge:100%) 1254 secs
2019-05-09 15:48:32 -0400 Wake DarkWake to FullWake from Deep Idle [CDNVA] due to HID Activity: Using AC (Charge:100%)
2019-05-09 16:49:25 -0400 Sleep Entering Sleep state due to 'Idle Sleep':TCPKeepAlive=active Using AC (Charge:100%) 381 secs
2019-05-09 16:56:31 -0400 Sleep Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using AC (Charge:100%) 432 secs
2019-05-09 17:04:28 -0400 Sleep Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using AC (Charge:100%) 262 secs
2019-05-09 17:09:15 -0400 Wake DarkWake to FullWake from Deep Idle [CDNVA] due to HID Activity: Using AC (Charge:100%)
2019-05-09 17:38:32 -0400 Sleep Entering DarkWake state due to 'Maintenance Sleep':TCPKeepAlive=active Using AC (Charge:100%)
2019-05-09 17:38:47 -0400 Sleep Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using AC (Charge:100%) 334 secs
2019-05-09 17:44:21 -0400 Wake Wake from Deep Idle [CDNVA] due to EC.PowerButton/UserActivity Assertion: Using AC (Charge:100%) 396 secs
2019-05-09 17:50:57 -0400 Sleep Entering Sleep state due to 'Software Sleep pid=102':TCPKeepAlive=active Using AC (Charge:100%) 60 secs
2019-05-09 17:52:17 -0400 Wake DarkWake to FullWake from Deep Idle [CDNVAP] due to Notification: Using AC (Charge:100%) 10 secs
2019-05-09 17:52:27 -0400 Sleep Entering DarkWake state due to 'Maintenance Sleep':TCPKeepAlive=active Using AC (Charge:100%) 9 secs
2019-05-09 17:52:36 -0400 Wake DarkWake to FullWake from Deep Idle [CDNVA] due to HID Activity: Using AC (Charge:100%)
2019-05-10 14:17:45 -0400 Sleep Entering DarkWake state due to 'Clamshell Sleep':TCPKeepAlive=active Using AC (Charge:100%) 7 secs
2019-05-10 14:17:52 -0400 Wake DarkWake to FullWake from Invalid [CDNVA] due to HID Activity: Using AC (Charge:100%) 1615 secs
2019-05-10 14:44:47 -0400 Sleep Entering Sleep state due to 'Idle Sleep':TCPKeepAlive=active Using AC (Charge:100%) 277 secs
2019-05-10 14:50:09 -0400 Sleep Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using AC (Charge:100%) 258 secs
2019-05-10 14:55:12 -0400 Sleep Entering Sleep state due to 'Maintenance Sleep':TCPKeepAlive=active Using AC (Charge:100%) 507 secs
2019-05-10 15:03:39 -0400 Wake Wake from Deep Idle [CDNVA] due to EC.KeyboardTouchpad/UserActivity Assertion: Using AC (Charge:100%)
2019-05-11 10:02:16 -0400 Sleep Entering Sleep state due to 'Software Sleep pid=102':TCPKeepAlive=active Using AC (Charge:100%) 28 secs
2019-05-11 10:02:52 -0400 Sleep Entering Sleep state due to 'Dark Wake Thermal Emergency':TCPKeepAlive=active Using AC (Charge:100%) 45 secs
2019-05-11 10:03:37 -0400 Wake Wake from Deep Idle [CDNVA] due to EC.PowerButton/UserActivity Assertion: Using AC (Charge:100%) 28 secs
2019-05-11 10:04:05 -0400 Sleep Entering DarkWake state due to 'Clamshell Sleep':TCPKeepAlive=active Using AC (Charge:100%) 7 secs
2019-05-11 10:04:12 -0400 Wake DarkWake to FullWake from Deep Idle [CDNVA] due to HID Activity: Using AC (Charge:100%)

pmset -g assertions
2019-05-11 10:08:10 -0400
Assertion status system-wide:
BackgroundTask 0
ApplePushServiceTask 0
UserIsActive 1
PreventUserIdleDisplaySleep 0
PreventSystemSleep 0
ExternalMedia 1
PreventUserIdleSystemSleep 1
NetworkClientActive 0
Listed by owning process:
pid 378(sharingd): [0x0000017400018385] 00:03:54 PreventUserIdleSystemSleep named: "Handoff"
pid 66(powerd): [0x0000000800088000] 00:10:53 ExternalMedia named: "com.apple.powermanagement.externalmediamounted"
pid 108(hidd): [0x0000004400098185] 00:00:00 UserIsActive named: "com.apple.iohideventsystem.queue.tickle.4294968933.3"
Timeout will fire in 600 secs Action=TimeoutActionRelease
Kernel Assertions: 0x124=USB,THNDR,MAGICWAKE
id=505 level=255 0x20=THNDR mod=5/11/19, 10:12 AM description=NHI0 owner=AppleThunderboltHAL
id=507 level=255 0x4=USB mod=12/31/69, 7:00 PM description=com.apple.usb.externaldevice.03100000 owner=AX88179
id=508 level=255 0x4=USB mod=12/31/69, 7:00 PM description=com.apple.usb.externaldevice.00200000 owner=ASM235CM
id=509 level=255 0x4=USB mod=12/31/69, 7:00 PM description=com.apple.usb.externaldevice.01100000 owner=USB3.0 Hub
id=511 level=255 0x4=USB mod=12/31/69, 7:00 PM description=com.apple.usb.externaldevice.01300000 owner=USB2.0 Hub
id=513 level=255 0x4=USB mod=12/31/69, 7:00 PM description=com.apple.usb.externaldevice.03400000 owner=Razer Core
id=515 level=255 0x4=USB mod=12/31/69, 7:00 PM description=com.apple.usb.externaldevice.02300000 owner=Corsair STRAFE RGB Gaming Keyboard
id=516 level=255 0x4=USB mod=12/31/69, 7:00 PM description=com.apple.usb.externaldevice.01140000 owner=ADATA USB Flash Drive
id=519 level=255 0x100=MAGICWAKE mod=12/31/69, 7:00 PM description=en0 owner=en0
id=520 level=255 0x100=MAGICWAKE mod=5/11/19, 10:16 AM description=en7 owner=en7
id=521 level=255 0x4=USB mod=12/31/69, 7:00 PM description=com.apple.usb.externaldevice.01330000 owner=IOUSBHostDevice
id=523 level=255 0x4=USB mod=12/31/69, 7:00 PM description=com.apple.usb.externaldevice.02400000 owner=Razer Naga Chroma
Idle sleep preventers: IODisplayWrangler

Having multiples devices attached to the enclosure did change its sleep behavior?

I mean, having a keyboard, external hard drive, random recognized hardware connected to the enclosure, override the sleep features of the sole gpu plugged in?

Maybe using the sleep feature of a known hardware, like a GbEth, could fix the sleep issue of the whole enclosure?

@ningauble77 looks like sleep is functioning perfectly! I have no idea what to make of it - mantiz venus + RVII + USB (Logitech G900, Drevo Blademaster Pro) sleeps with full fan blast. I have not dug out the chroma yet it's still in the box, waiting for new pump/res to try and shove the whole WC loop inside the enclosure.

it could be that there are different PCB versions.

@juan_c_yunis I won't be able to try and solve sleep anymore as i'm on WC with the fan and pump controlled by an Aquacomputer Quadro board, hence that takes over the pump and fan regulation.

Looks like 10.14.5 was released. Anyone have a chance to run the VII on it? I was hoping to hear people’s comments before I opening my box. Thanks!

Fans still rev up on sleep with 10.14.5.  Running a 2017 full spec iMac with Akitio Node Pro and AMD VII.

Not had much time with it so far, but the issues with my HP Spectre 32 not being immediately detected when I boot with the egpu attached and often not detected after waking from sleep or hotplugged persist.  Fan on sleep is for me, as before working properly.  I unenrolled this Mac from the beta program for now also, glad I didn't apparently make a mess of things.

GOOD NEWS:
with MacOS 10.14.5 (update last night) the RadeonVII now works as eGPU ... but seems not 100% optimized... a little bit lower Metal Geekbench score than a Vega64 (at least on my System)

Posted by: FCPXUserGermany

GOOD NEWS:
with MacOS 10.14.5 (update last night) the RadeonVII now works as eGPU ... but seems not 100% optimized... 

Optimized or bust! 

MAN I love using my RVII as a paperweight. Looking forward to using it for a performance boost over my V64 LC in like..... September.

It works. 
Seems like performance could be tweaked a bit, and it still has a few issues with display output when coming back from sleep. 
Hopefully it will be fixed. But at least it is supported.

@fcpxusergermany It won't be fully optimized until Apple lists it in in the support article

https://support.apple.com/en-gb/HT208544

Don't worry about Geekbench. The Radeon scores in the macOS version of the app have been dropping with each update. The drop in performance isn't seen in any other test or app. Put Geekbench in the trash and never think about it again.

Very happy so far with the WC Radeon VII. No crashes, no problems of any kind, and dead quiet. Dremel-ed a cutout for the WC tubes to extend out to the pump/res at the top, I might or might not use the Chroma with the smaller EKWB XRES 100 SPC-60 pump/res as i'm quite happy with this setup for now, no USB issues too.

I'm not sure why are Geekbench numbers falling for some of you, even prior to watercooling on b5, geekbench went to 171652, and after watercooling it went up to 179843. You can see FCPX_Germany's scores up there too, where he has gotten above 170k too.

Finally got to a stable working setup with great temps, gonna change the coolant to AMD Red, and i'm not touching this setup for the next two years.

EDIT: Blender Benchmark scores. It takes a while to run - and I don't really have a comparison frame:

{
"benchmark_client": {
"client_version": "1.0b2"
},
"blender_version": {
"build_commit_date": "2018-03-22",
"build_commit_time": "14:10",
"build_date": "2018-03-22",
"build_hash": "f4dc9f9",
"build_time": "21:48:25",
"version": "2.79 (sub 0)"
},
"device_info": {
"compute_devices": [
"AMD Radeon VII Compute Engine"
],
"device_type": "OPENCL",
"num_cpu_threads": 12
},
"scenes": [
{
"name": "barbershop_interior",
"stats": {
"device_memory_usage": 2417.85,
"device_peak_memory": 2417.85,
"pipeline_render_time": 176.87,
"render_time_no_sync": 136.736,
"result": "OK",
"total_render_time": 175.856
}
},
{
"name": "bmw27",
"stats": {
"device_memory_usage": 149.48,
"device_peak_memory": 149.48,
"pipeline_render_time": 422.56,
"render_time_no_sync": 417.978,
"result": "OK",
"total_render_time": 421.224
}
},
{
"name": "classroom",
"stats": {
"device_memory_usage": 352.31,
"device_peak_memory": 352.31,
"pipeline_render_time": 612.86,
"render_time_no_sync": 610.658,
"result": "OK",
"total_render_time": 611.856
}
},
{
"name": "fishy_cat",
"stats": {
"device_memory_usage": 470.97,
"device_peak_memory": 470.97,
"pipeline_render_time": 612.52,
"render_time_no_sync": 602.673,
"result": "OK",
"total_render_time": 611.181
}
},
{
"name": "koro",
"stats": {
"device_memory_usage": 457.7,
"device_peak_memory": 468.64,
"pipeline_render_time": 974.77,
"render_time_no_sync": 963.103,
"result": "OK",
"total_render_time": 974.088
}
},
{
"name": "pavillon_barcelona",
"stats": {
"device_memory_usage": 163.32,
"device_peak_memory": 163.32,
"pipeline_render_time": 1097.77,
"render_time_no_sync": 1094.79,
"result": "OK",
"total_render_time": 1095.78
}
}
],
"system_info": {
"bitness": "64bit",
"cpu_brand": "Intel(R) Core(TM) i7-8850H CPU @ 2.60GHz",
"devices": [
{
"name": "Intel Core i7-8850H CPU @ 2.60GHz",
"type": "CPU"
},
{
"name": "Intel UHD Graphics 630",
"type": "OPENCL"
},
{
"name": "AMD Radeon Pro 560X Compute Engine",
"type": "OPENCL"
},
{
"name": "AMD Radeon VII Compute Engine",
"type": "OPENCL"
}
],
"machine": "x86_64",
"num_cpu_cores": 6,
"num_cpu_sockets": 1,
"num_cpu_threads": 12,
"system": "Darwin"
},
"timestamp": "2019-05-15T04:04:58.061207+00:00"
}

BTW, anyone else has the issue of output when returning from sleep?
I have the macbook in clamshell on a stand, connected to LG 5K and the eGPU is connected to asus xg35vq.
The asus shows empty screen (even though it shows as if there is input). It is connected via HDMI.
My "fix" is also connect the DP from the VII to the asus panel. Then it shows as both xg35vq (1 and 2), and it shows the screen. When I disconnect the DP, it only shows the HDMI and everything is normal and working.

so yes I did a few more tests ... also some more Metal and OpenCL run's with Geekbench and got better results, not sure what happened on that 1st test 

Metal performance:

Radeon VII - 162.127 (MacOS 10.14.5, 14.05.2019)
Vega 64 - 140.247 (MacOS 10.14.5, 14.05.2019)

OpenCL performance:

Radeon VII - 179.052 (MacOS 10.14.5, 14.05.2019)
Vega 64 - 145.701 (MacOS 10.14.5, 14.05.2019)

Engine Heaven Benchmark

Vega 64 - 1000, 39.7/71.1/8.0
Radeon VII - 1211, 48.1/85.6/8.3

Engine Valley Benchmark

Vega 64 - 1689, 40.4/71.3/21.7
Radeon VII - 2147, 51.3/91.1/25.1

... testing FCPX real world work export times now

It's now AMD red!

If you look at my Geekbench run in Windows, there is only a short burst where the Geekbench computation finishes, it's not even long enough to register on the wattman polling, only the memory spike where it goes to 1200mhz. This is what @craftsman mean when we say Geekbench is a trashy piece of benchmark for measuring performance on full-fledged desktops/laptops.

"Linus Torvalds, who is the creator and principal developer of the Linux kernel, criticized Geekbench versions 3 and earlier for showing partiality for ARM64 devices and being a bad performance measure of x86-based systems. He called Geekbench as bad a benchmark as Dhrystone, which he and others have ridiculed for years.^[3] When version 4.0 of Geekbench was released his comment was, "Looks much better." - Geekbench on Wikipedia

You will see that Geekbench is not a very good benchmark, because it only tests a short single burst (see the peak - Geekbench was designed to benchmark iPhones and mobile devices, at their peak burst performance, not long-term sustained. I remembered it was then geared towards benchmarking mobile devices, and only later on in v4 people started to take it semi-seriously for benchmarking full-fledged computers.

The problem with MacOS is there's no really good apple-to-apple benchmarking software for GPUs. Maybe the new Blender benchmark but even that on Nvidia cards it runs on CUDA, and on AMD it runs on OpenCL. Maybe the metal benchmarking tool would come close to being the best benchmarking tool, since it looks like Apple is dead-set on pushing the Metal/Metal2 API.

Let's lay out what I think are the facts as of now, and you are free to correct any of them.

1. Radeon VII is the best (performance wise) consumer-oriented GPU that is natively supported by the latest MacOS (10.14.5) currently.
2. It's not good value, it's for those who do not have a performance-to-cost ratio mindset, and just want the best without fuss.
3. It provides a seamless experience for eGPUs operationally, except for the fan issue (which some folks say is solved).
4. Nvidia cards are only supported till the 10xx series with the web driver, and it only works up till High Sierra.
5. Even in High Sierra, Nvidia cards are not integrated well - need to shutdown before ejecting, etc. For desktops and mac minis users not an issue, but for laptops some people just prefer not to shutdown to move to another location.
6. Nvidia cards are more performant than the AMD cards, if you don't mind  working around the quirks and staying on HS.

Edit: Radeon VII (2000core/[email protected]) in Bootcamp with Userbench, which is a holistic benchmark we commonly use.

https://www.userbenchmark.com/UserRun/16943653

That said, I’m quite disappointed with the Radeon VII’s performance in MacOS even now, but my primary tasks are coding and office work so it’s not really important. Just that occasionally it’s nice to boot into Bootcamp for gaming ultrawide @1440p60.e

i think Apple is going to push for Navi instead, but as of now general egpu usage, for AMD cards, I think an rx580 or Vega 56 will suffice and is best value for money.

here’s hoping to have improved performance for AMD cards, and drivers for Nvidia, on MacOS in a couple of weeks time! I would love to run my 1080ti in the egpu.

@kelvin_tan i'm really hoping that they push this card to the max, it is a very powerful card that could bring great performance to Mac, i think LuxMark is the one that shows the true colors of this card, but yeah if Apple were to release Nvidia drivers for Mac, i would seriously consider moving to Nvidia.

Navi, on the other hand, would be based in GCN (would be last one hopefully) so not sure if they would be able to compete at all with Nvidia based on the current cards in the market, Nvidia is not even 7nm.

@juan_c_yunis I think it’s not just MacOS, but every single individual pro app, starting with of course FCP X, adobe suite, resolve.

however programmers are lazy, and coding for gpgpu compute in software is magnitudes harder than expending cpu cycles, also not everyone has a beefy gpu. Resolve is leading the way - resolve 16 is reported to have significant gains in gpu usage vs resolve 15, thus pulling even further ahead in this aspect compared to fcp x and premiere pro.

so, provis apps in windows already have pro gpu support (solidworks is quite a bit faster with the Radeon vii compared to Vega 56, and even faster with the wx7100 (which I use in my office desktop for CAD/CAM work).

then you throw in all the nasty proprietary optimisations like brand specific (cuda, mantle which is now dead), and os specific (DirectX, metal, vulkan), and you can see why no matter what gpu you choose, you will win some and lose some.

I prioritise stability, ease-of-use, aesthetics, and os integration highly, which is why I use MacOS mainly (Windows and Linux is actually the more performant OSes now). I use CentOS for work, and Windows 10 with the new WSL and a full fledged linux kernel in win10(!!!) might just tempt me to switch entirely.

Don’t worry about performance, since you bought it just know it’s the best natively (MacOS) supported consumer card you can have, and go do work and game on it. The best coder in my company refuses to move from a ThinkPad x230, and the big video production house which we recently engaged for corporate work are all on trashcan mac pros. Their work is still outstanding, and their mentality is “these are just tools”. The only ones really living on the bleeding edge are gamers and overclocking enthusiasts.

As posted several weeks ago, non-Radeon VII on a Mac discussions don't belong to this thread. There's no censorship. We don't have time to filter through every off-topic post.

Posted by: itsage

As posted several weeks ago, non-Radeon VII on a Mac discussions don't belong to this thread. There's no censorship. We don't have time to filter through every off-topic post.

Yeah... while it may be reasonable (it was off-topic of course) it still is censorship to delete people's posts, by definition. Is there a category where such discussions can be held. Why not move unwanted posts there instead of deleting them?

@OliverB

It would have been also reasonable that you open a thread where discuss freely about a topic, already yourself.
To be honest nowadays we are used to scream when moderators do their job.
Back in my days, you would have been already banned after being warned at least three times about your Forum posting habit and/or behavior.

There is no censorship; if you can't post anymore! 😎 

Hope springs eternal: 10.14.6 is out for developer beta. Any good news there?

Found that eGpu icon doesn't work for me in 10.14.6B1 :).  So I can't safely disconnect GPU.

Posted by: alex_ai

Found that eGpu icon doesn't work for me in 10.14.6B1 :).  So I can't safely disconnect GPU.

@alex_ai
Could you please elaborate more about your setup? Which GPU, which enclosure? Did it work with 10.14.5?

Posted by: OliverB

@alex_ai
Could you please elaborate more about your setup? Which GPU, which enclosure?

@oliverb it's in their signature.

Posted by: mac_editor

Posted by: OliverB

@alex_ai
Could you please elaborate more about your setup? Which GPU, which enclosure?

@oliverb it's in their signature.

thank you @mac_editor.
The question about if the eject works in 10.14.5 remains.

@oliverb which is why I didn't quote it 🙂

There are new benchmarks on Barefeats.

How is video performance in Premiere and Resolve?

@alex_ai I am noticing the same problem on 10.14.6 B1. While the eGPU icon is present, it is not selectable and does not show a dropdown of connected GPUs. Likely isn't GPU specific, since I am running a Vega 64. Future betas will likely fix the problem (perhaps we will also see minor refinements to RVII drivers).

Edit: A quick look via the command line mechanism shows that essentially there is a (rather silly) bug with the SafeEjectGPU system. Hopefully should be addressed in the next beta.