Jump to content
About Just Joined group Read more... ×
jnolla

Vega Frontier Edition on High Sierra

746 posts in this topic

Recommended Posts

Hi Chris.... similar observations. For some reason, the drivers ignore most of the fan table info that would further control fan speeds.

My hunch is that since discrete Vega cards are used only for Egpu, the drivers are for that reason not optimised at all in terms of fan control / noise.

For the iMac Pro Vegas, Apple does use a different fan control method, since the Vega is directly implemented on the Mainbord and so they use their proprietary cooling system.

I wonder how the Vega 64 would behave in a EGpu Setup through TB3.

 

So my interpretation: accoustic limit is a desirable nominal speed, its the rpm that will be established after heavy load.

min and max fan have no effect on fan speed control, they just help calculate percentage of fan speed for display reasons.

 

 

 

Regards, Michael

Edited by Mike Ranger

Share this post


Link to post
Share on other sites
Advertisement

It'll be interesting to see if a new Mac Pro with no igpu and probably shipping with the Vega will impact on the OS and the functionality of the drivers.

The smbios 6,1 is still the best in some ways because of the lack of integrated GPU, but a refresh of this with a new Mac Pro release will def be welcome to the hackintosh community...

Share this post


Link to post
Share on other sites
4 hours ago, surfinchina said:

It'll be interesting to see if a new Mac Pro with no igpu and probably shipping with the Vega will impact on the OS and the functionality of the drivers.

The smbios 6,1 is still the best in some ways because of the lack of integrated GPU, but a refresh of this with a new Mac Pro release will def be welcome to the hackintosh community...

Curious.... does your card spin up after heavy load as well?

The easiest way to reproduce this is Luxmark benchmark.

Again....my biggest interest at the moment is Fan-Control.

Regards, Michael

Share this post


Link to post
Share on other sites

Hi

 

the new version of Whatevergreen seems to bring an overheating pb for me. (Sapphire Nitro+ vega 64). Mojave 10.14.1. Nothing strange for you? 

Share this post


Link to post
Share on other sites
14 hours ago, CMMChris said:

 PP_Tables can look different depending on the card. The one inside my generator is extracted from my V64 Gaming OC BIOS.

 

Hey Chris, 

 

I'm trying to modify the PP_Table of a liquid cooled Vega64, but the table is shorter and has different header, as far as I can see. How can I change your excel file to accommodate this?

VGTab seems to generate the same length (shorter) as  hellm's linked files.

 

Thanks!

 

PS. I inspected my PP_Table, extracted from my original bios, it's also shorter. offset 37254 : 669 bytes long

Edited by Pval2

Share this post


Link to post
Share on other sites

Another new version of my Excel generator with ready to copy output to place inside the .plist files. (FILE DELETED, NEW VERSION HERE!)

 

 

@Mike Ranger It's normal. Does the same on my card on Windows when controlling it with a PPT instead of Wattman.

@Pval2 Try using the PPT generated by my Excel generator. Might still work. If it doesn't, you need to reverse engineer the PPT of your card and create your own generator for it.

 

Edited by CMMChris

Share this post


Link to post
Share on other sites
1 hour ago, CMMChris said:

Another new version of my Excel generator with ready to copy output to place inside the .plist files.

 

@Mike Ranger It's normal. Does the same on my card on Windows when controlling it with a PPT instead of Wattman.

@Pval2 Try using the PPT generated by my Excel generator. Might still work. If it doesn't, you need to reverse engineer the PPT of your card and create your own generator for it.

vega64-mac-pp_table-gen-V1.0.xlsx.zip

 

Thanks again for your work.

 

I think since Apple claims to officially support Vega descrete GPUs, we should not have to use PPT in the first place.

And yet.... there are some funky things going on even with the PPTs in place..... why do the fans spin down with PPTs once you start a h.265 video in Quicktime?

This action seems to overrule even the PPTs to some degree.

I really wonder if this is all related to non standard bios files being different from reference design cards bios. In the end, the PPT is just one portion of the bios.

In the end....PPTs for the time being are certainly acceptable but certainly not what a decent driver could achieve.

 

Regards, Mike

 

 

Share this post


Link to post
Share on other sites

You don't need a PPT for the cards officially supported by Apple. Dunno what's different with them. Would be interesting to extract a PPT from their VGA BIOS.

 

Regarding the H.265: Can't confirm that on my card. Fans stay up if they are spinning while starting a video. Strange.

Share this post


Link to post
Share on other sites
15 minutes ago, CMMChris said:

You don't need a PPT for the cards officially supported by Apple. Dunno what's different with them. Would be interesting to extract a PPT from their VGA BIOS.

 

Regarding the H.265: Can't confirm that on my card. Fans stay up if they are spinning while starting a video. Strange.

I will post a sample video later.... when playing it in QT, Fan will go back to nominal (idle) speed quicker.

 

Share this post


Link to post
Share on other sites
1 hour ago, CMMChris said:

You don't need a PPT for the cards officially supported by Apple. Dunno what's different with them. Would be interesting to extract a PPT from their VGA BIOS.

 

You can simply get the BIOS with ioreg: 

ioreg -l -n display@0 -r | grep ATY,bin_image

Might be at another display (@0 / @1 / etc), but you get the idea. From looking at mine, the PPT starts at offset 37254. 

The PP_PhmSoftPowerPlayTable, if you're using one is there as well, as is a PP_PhmSoftWTTable?

 

I've adapted your excel-sheet, I can confirm my Accoustic Limit (RPM) is set to 1500, which is exactly the rpm I'm seeing in ioreg. And Throttling (RPM) is set to 2300. 

I have seen another issue though, my fan RPM drops under load, by up to 33%, tested with Luxmark, Furmark, etc. Only when I quit; remove the load, will the fan RPM rise again, even above 1500 RPM, until the card is cool again. I have a bug report open with Apple, but they haven't told me anything yet.

 

Monitor fanspeed / other PerformanceStatistics: 

# Grab the column headers from ioreg's output: 
ioreg -l -n display@0 -r |grep \"PerformanceStatistics\" | cut -d '{' -f 2 | tr '|' ',' | tr -d '}' | tr ',' '\n'|grep 'Temp\|Fan\|%\|(W)\|Hz'|cut -d\" -f2|tr '\n' ',' && printf '\n'

# Grab the values every second: 
while true
do 
    ioreg -l -n display@0 -r |grep \"PerformanceStatistics\" | cut -d '{' -f 2 | tr '|' ',' | tr -d '}' | tr ',' '\n'|grep 'Temp\|Fan\|%\|(W)\|Hz'|cut -d\" -f3|tr -d '='|tr '\n' ',' && printf '\n'
    sleep 1
done

 

Edited by Pval2

Share this post


Link to post
Share on other sites
1 hour ago, CMMChris said:

Yes it does but not when it is still under load.

yes..... it just makes the card go quiet again quicker

Share this post


Link to post
Share on other sites
11 hours ago, Mike Ranger said:

Curious.... does your card spin up after heavy load as well?

The easiest way to reproduce this is Luxmark benchmark.

Again....my biggest interest at the moment is Fan-Control.

Regards, Michael

I've got an EK block on my card, so get no spin up.

And just for the record, it's a Vega FE and works OOB on HS and Mojave - excepting I need the fixup kext to allow for 2 monitors on DP.

What I was inferring with the original post is that Apple hasn't yet optimised the drivers for use without igpu because at the moment there isn't such an Apple machine. When there is such a machine (the new Mac Pro), they will make changes.

I noticed this back in the day when I had an X58 on SMBIOS 6,1 and after that an X99 and finally my current X299.

This might all only be great news for those of us without igpu on our motherboards.

Share this post


Link to post
Share on other sites

Uh? The iMac Pro doesn't have an IGPU. I am currently using my machine without IGPU on iMacPro1,1 SMBIOS, works fine. Gives a slight performance boost compared to the iMac18,3 SMBIOS.

Share this post


Link to post
Share on other sites
2 hours ago, CMMChris said:

Uh? The iMac Pro doesn't have an IGPU. I am currently using my machine without IGPU on iMacPro1,1 SMBIOS, works fine. Gives a slight performance boost compared to the iMac18,3 SMBIOS.

Would be interesting to see someone do a 10-core Xeon-W build based on that SMBIOS, a true iMac Pro hackmac attempt. 

 

Back to the topic at hand: If I'm able to choose a new card specifically for Mojave, I assume I should go with the XFX Vega 64 or an FE since they won't require PPT tweaking? 

 

A used 16GB FE looks temping!

Share this post


Link to post
Share on other sites
4 minutes ago, Beav said:

Would be interesting to see someone do a 10-core Xeon-W build based on that SMBIOS, a true iMac Pro hackmac attempt. 

 

Plenty of us already have! Do a search for KPG's X299 guide here on the forums.

Share this post


Link to post
Share on other sites
5 minutes ago, Tony Arnold said:

 

Plenty of us already have! Do a search for KPG's X299 guide here on the forums.

Thanks Tony. Just realized my past posts here were all from early 2006, can't believe I've been running hacks for that long. :)

Share this post


Link to post
Share on other sites
4 hours ago, CMMChris said:

Uh? The iMac Pro doesn't have an IGPU. I am currently using my machine without IGPU on iMacPro1,1 SMBIOS, works fine. Gives a slight performance boost compared to the iMac18,3 SMBIOS.

Yes true, but the root cause of the fan issues (in my opinion) is that it's not using a discrete GPU as such, and uses a propriety cooling system with small and high speed fans. 

Share this post


Link to post
Share on other sites

I have now opened a bug report with Apple as well, didn't hide that I am using the Vega 64 inside a Hackintosh but pointed out that eGPU and Mac Pro users face the same issue. I recommend everyone doing the same since only that way there is a chance they get their drivers right.

Share this post


Link to post
Share on other sites
On 10/31/2018 at 11:57 PM, CMMChris said:

Most Vega cards use Min and Max Fan Speed to control their fan. MacOS on the other hand uses the Throttling RPM value as the Max Fan speed and Accoustic Limit for the Min Fan for some reason.

It seems there are two versions of the Fan_Table, in V1 what is now the acoustic limit was a RPM value, the description says it's the default RPM value. In V2 it became acoustic limit, the highest allowable RPM. FanMin and FanMax RPM didn't exist in V1, but exist in V2. This might explain the behavior, if Apple is not interpreting the table correctly. I've updated my Apple bug with this.

 

Graphics clock table also has two versions, the Powertune table even has three versions: https://github.com/torvalds/linux/blob/master/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_pptable.h

 

However, I'm checking table versions and PPT versions of some of the problem and fixed bios and I'm only finding PPT version: ucTableFormatRevision 8, ucTableContentRevision 1, ucTableRevision 0. And all PPT have Fan_table revision 0B (11). I've looked at versions 016.001.001.000.000000(Vega64), 016.001.001.000.008733 (64LC my own), 016.001.001.000.008737(Vega64), 016.001.001.000.008774(64LC) and 016.001.001.000.009360(Vega56) from Techpowerup.

I'll try to put together a more detailed comparisson, including Gfxclk_Table and PowerTune_Table.

 

Share this post


Link to post
Share on other sites

Another update for my table. Changes:
- Optimized Layout
- Added SoC clock configuration
- Added some more markup of the PPT including fields for optional manual editing
- Added a short guide for using the generated data and a warning for the noobs ( ;) )
- Preset settings optimized for cards that suffer from coil whine, reduces coil whine to near zero for me while retaining great performance and low heat

vega64-mac-pp_table-gen-V1.5.xlsx

Share this post


Link to post
Share on other sites

Edit: Doesn't work reliable unfortunately.

To everyone who has the issue with the fans spinning up for a while after load: I've got a fix for that. The secret ingredient is to load the VBios of the MSI Vega 64 through Clover. Simply put the ROM file into "EFI/Clover/ROM/" and in Clover Configurator select "Load vBIOS" under "Graphics". Now reboot and test. Fans will spin down immediately after load. Please note you still have to use a custom PowerPlayTable for it to work.

1002_687f.rom

Edited by CMMChris

Share this post


Link to post
Share on other sites

  • Recently Browsing   0 members

    No registered users viewing this page.

Announcements

×