Jump to content

AMD High Sierra Kernel Release and Testing


1,119 posts in this topic

Recommended Posts

i agree. there is something not right about those benchmarks. its almost impossible to reach those scores i guess?

 

iam getting cinebench at like 1290 with my ryzen 1600 clocked at 3,8ghz and from what i know those are correct numbers even under windows.

 

the kernel youre using is doing something weird... but i gotta admit the benchmark pictures look awesome though. makes intel look stupid a bit ;)

  • Like 2
Link to comment
Share on other sites

i agree. there is something not right about those benchmarks. its almost impossible to reach those scores i guess?

 

iam getting cinebench at like 1290 with my ryzen 1600 clocked at 3,8ghz and from what i know those are correct numbers even under windows.

 

the kernel youre using is doing something weird... but i gotta admit the benchmark pictures look awesome though. makes intel look stupid a bit ;)

I’m up to 6.0GHZ, sudo speed, it’s incredible because I don’t know how you can fake a Geekbench, Cinebench, Luxmark, amd Uningine. It’s openCL and GL formats. I think it’s real Ryzen is alive and learning to take over the world. I crashed Luxmark at the 6.0GHZ but the others are running fine 298,000 Vega Compute in Geekbench. Single core over 8000, multi core just under 50,000.

Link to comment
Share on other sites

I’m up to 6.0GHZ, sudo speed, it’s incredible because I don’t know how you can fake a Geekbench, Cinebench, Luxmark, amd Uningine. It’s openCL and GL formats. I think it’s real Ryzen is alive and learning to take over the world. I crashed Luxmark at the 6.0GHZ but the others are running fine 298,000 Vega Compute in Geekbench. Single core over 8000, multi core just under 50,000.

the numbers beeing displayed must be incorrect somehow. no ryzen can go that high and even if it could - the vcore needed would blow up the temps and you would need atleast water cooling or better ;)

not sure what you have done there, but it is some kind of world record i guess ;) 

Link to comment
Share on other sites

the numbers beeing displayed must be incorrect somehow. no ryzen can go that high and even if it could - the vcore needed would blow up the temps and you would need atleast water cooling or better ;)

not sure what you have done there, but it is some kind of world record i guess ;)

these are just numbers, if the base unit of time is changed, not suprising

  • Like 1
Link to comment
Share on other sites

the numbers beeing displayed must be incorrect somehow. no ryzen can go that high and even if it could - the vcore needed would blow up the temps and you would need atleast water cooling or better ;)

not sure what you have done there, but it is some kind of world record i guess ;)

 

Its all about the numbers or software, the CPU is "thinking" that its at 6.0GHz and the Graphics card seems to be following along.   OpenGl is mainly GPU intensive and its responding, FPS are FPS there's no way around it.    A CPU without integrated graphics cannot run openGl, or at least I've never seen it happen.   But the openGl is greatly improved, this has been a mystery with Ryzen it has up to noe been sub par to Intel for graphics performance.    My pentium has been able to run circles around my Ryzen in graphics capabilities, with the exception of Luxmark Ball.    This is a strange anomaly but if it turns out to be repeatable why not use it?   

 

In prior testing we went through this and most of us came to the same conclusion that the busrsatio=xx value must match your bios CPU frequency settings.   If not we saw benchmarks that were very unstable and a lot of crashing.   Freeze ups and the spinning Beach Ball of death.    This is running solid, and is nothing like previous Ryzen CPU frequency testing.   The greatest thing about this is that Clover is giving the orders and the software is responding.

Link to comment
Share on other sites

zT4oIBZ.png wanna fake 5.5ghz or even 6ghz here is how he did it have fun guys to any mod please remove his posts we don't need fake content here

 

Can you fake a Uingine Benchmark doing that?    When you do that it will show in the Geekbench and Luxmark data as well as about this Mac.  I am only adjusting busratio=xx settings in boot manager screen i'm not touching those settings in clover.  

Link to comment
Share on other sites

Its all about the numbers or software, the CPU is "thinking" that its at 6.0GHz and the Graphics card seems to be following along.   OpenGl is mainly GPU intensive and its responding, FPS are FPS there's no way around it.    A CPU without integrated graphics cannot run openGl, or at least I've never seen it happen.   But the openGl is greatly improved, this has been a mystery with Ryzen it has up to noe been sub par to Intel for graphics performance.    My pentium has been able to run circles around my Ryzen in graphics capabilities, with the exception of Luxmark Ball.    This is a strange anomaly but if it turns out to be repeatable why not use it?   

 

In prior testing we went through this and most of us came to the same conclusion that the busrsatio=xx value must match your bios CPU frequency settings.   If not we saw benchmarks that were very unstable and a lot of crashing.   Freeze ups and the spinning Beach Ball of death.    This is running solid, and is nothing like previous Ryzen CPU frequency testing.   The greatest thing about this is that Clover is giving the orders and the software is responding.

 

yeah sure, if it improves performance for you and others its a good thing. i will just live with like 50% gfx performance compared to windows, and its not that important to me.

 

by the way, my ryzen is overclocked to 3,8 and is running very stable for days now. i had usb problems (with my audio interface) when i set frequency and bus ratio in clover, so i deleted the data again and its just running smooth. for me its not necessary so set those values and ive never had a crash, freeze or spinning beach ball of death.

 

iam very happy with the current build to be honest and didnt expect it to run that smooth. ok, ive spend some time learning quite some stuff but i guess it was worth it... well it is also fun fiddling aroung a bit ;)

Link to comment
Share on other sites

yeah sure, if it improves performance for you and others its a good thing. i will just live with like 50% gfx performance compared to windows, and its not that important to me.

 

by the way, my ryzen is overclocked to 3,8 and is running very stable for days now. i had usb problems (with my audio interface) when i set frequency and bus ratio in clover, so i deleted the data again and its just running smooth. for me its not necessary so set those values and ive never had a crash, freeze or spinning beach ball of death.

 

iam very happy with the current build to be honest and didnt expect it to run that smooth. ok, ive spend some time learning quite some stuff but i guess it was worth it... well it is also fun fiddling aroung a bit ;)

 

It’s all about the fun for me as well, the 1/2 performance issue may not need to be. Maybe a clock timing adjustment to the kernel is all that is needed to run on par with windows apps on Ryzen systems. On average my AMD cards run about 10-15% lower on Ryzen versus my Intel systems. And for Nvidia it is much worse, it would be nice if we could get the web drivers working at Intel levels. A 10-15% bump would put it on par for AMD card performance, but that still wouldn’t totally explain the Nvidia performance loss.

 

When we tested Ryzen kernels in Sierra in April setting these values was tricky and I noticed that setting busratio=xx was necessary to ensure a stable setup. If it didn’t match the bios setting it would get buggy, freeze and crash. This was after the auto kernel came out which was supposed to detect the proper frequency based on bootloader detection. I remember Gils bagging on me for using busratio with auto kernel, about a few weeks later everybody came to realize it was necessary for a stable system. If you ran without a busratio value that matched your bios it would crash eventually. Are you running a boot arg busratio=xx? This High Sierra kernel is the first one I’ve worked with that is allowing me to set busratio=xx to any frequency independent of bios settings without crashes.

 

I know this is not a real frequency or overclocking adjustment as I said it’s sudo overclocking and I’ve tried this on my Intel systems but anything above 4.0gb would crash the system. Can you try setting boot arg busratio=4.2 and see if you get any sudo performance increase?

Link to comment
Share on other sites

It’s all about the fun for me as well, the 1/2 performance issue may not need to be. Maybe a clock timing adjustment to the kernel is all that is needed to run on par with windows apps on Ryzen systems. On average my AMD cards run about 10-15% lower on Ryzen versus my Intel systems. And for Nvidia it is much worse, it would be nice if we could get the web drivers working at Intel levels. A 10-15% bump would put it on par for AMD card performance, but that still wouldn’t totally explain the Nvidia performance loss.

I'm curious to see what kind of FPS you're able to attain in a game that isn't benchmark software, such as League of Legends for example.

I'm also curious to see if you get a kernel panic, when you run 16 threads of

yes > /dev/null &

with your most stable clock that is above 4.0Ghz reported.

Typically when I'm overclocking, I use Cinebench 15 to test if it crashes, if so it's unstable, if not then I run the generic yes test as well.

post-1312076-0-03319300-1507513595_thumb.png

 

After you run the 16 threads, you can stop them by using 

killall yes
Link to comment
Share on other sites

It’s all about the fun for me as well, the 1/2 performance issue may not need to be. Maybe a clock timing adjustment to the kernel is all that is needed to run on par with windows apps on Ryzen systems. On average my AMD cards run about 10-15% lower on Ryzen versus my Intel systems. And for Nvidia it is much worse, it would be nice if we could get the web drivers working at Intel levels. A 10-15% bump would put it on par for AMD card performance, but that still wouldn’t totally explain the Nvidia performance loss.

 

When we tested Ryzen kernels in Sierra in April setting these values was tricky and I noticed that setting busratio=xx was necessary to ensure a stable setup. If it didn’t match the bios setting it would get buggy, freeze and crash. This was after the auto kernel came out which was supposed to detect the proper frequency based on bootloader detection. I remember Gils bagging on me for using busratio with auto kernel, about a few weeks later everybody came to realize it was necessary for a stable system. If you ran without a busratio value that matched your bios it would crash eventually. Are you running a boot arg busratio=xx? This High Sierra kernel is the first one I’ve worked with that is allowing me to set busratio=xx to any frequency independent of bios settings without crashes.

 

I know this is not a real frequency or overclocking adjustment as I said it’s sudo overclocking and I’ve tried this on my Intel systems but anything above 4.0gb would crash the system. Can you try setting boot arg busratio=4.2 and see if you get any sudo performance increase?

i can try it tomorrow - iam on my windows laptop now and about to sleep.

 

and no, i dont set boot flag (busratio). system running stable from day one with that latest kernel.

 

msi pc mate b350 ryzen 1600 ddr4 2800 gtx 770 (it ran with ati hd5450 aswell).

 

benchmarks are the same as on windows except for graphics ofource.

 

sometimes only shutdown doesnt work and it seems its a problem with unmounting one of my drives when i have them all connected, but besides that its running smooth. iam not using onboard audio. usb audio interface.

 

now iam considering buying a threadripper since some people get it to run ;)

  • Like 2
Link to comment
Share on other sites

i can try it tomorrow - iam on my windows laptop now and about to sleep.

 

and no, i dont set boot flag (busratio). system running stable from day one with that latest kernel.

 

msi pc mate b350 ryzen 1600 ddr4 2800 gtx 770 (it ran with ati hd5450 aswell).

 

benchmarks are the same as on windows except for graphics ofource.

 

sometimes only shutdown doesnt work and it seems its a problem with unmounting one of my drives when i have them all connected, but besides that its running smooth. iam not using onboard audio. usb audio interface.

 

now iam considering buying a threadripper since some people get it to run ;)

To macOS kernel, thread ripper is the same as Ryzen.  I have intel x99 which kind of sucks because I am using almost all of the PCIE slots and at some point Thunderbolt 3 stops showing up in pci enumeration.

 

That new gigabyte x399 board looks pretty good.  Need to wait for ddr4 memory to drop.

Link to comment
Share on other sites

well increasing the cpu speed/ bus speed in CPU tab of Clover settings , more than your actual CPU speed/bus speed 
causes many issues such as slow animations and stutters in audio etc.
that will sure bump up your cinebench cpu scores but there will be no real gains in terms of actual performance .(consider it as a false score)
how do i know ? i have messed with it long back . 

 

Link to comment
Share on other sites

well increasing the cpu speed/ bus speed in CPU tab of Clover settings , more than your actual CPU speed/bus speed

causes many issues such as slow animations and stutters in audio etc.

that will sure bump up your cinebench cpu scores but there will be no real gains in terms of actual performance .(consider it as a false score)

how do i know ? i have messed with it long back .

 

I know, the odd thing is it is only affecting audio, no other bugs or crashes. With the previous kernels doing this would crash apps especially when browsing.. It is a sudo score I was able to go to 7.0 GHZ and scored a 340,000 Geekbench compute with my vega but that is impossible. Can you try quickly and see if it’s stable at higher settings with this kernel? I’m only losing sound, not actually losing it but it’s cutting in and out.

 

I am only setting busratio=xx values in boot arg. No clover CPU settings, when you do that it lists in About this Mac and is reflected in the benches, mine all have 3.0GHz listed. Like I said earlier with prior kernels when setting improper busratio=xx values the system becomes unstable and crashes, not happening for me with this kernel. Can someone else test this and confirm, just set busratio value 5 points above actual settings in bios. Run system and see if it crashes.

Link to comment
Share on other sites

I know, the odd thing is it is only affecting audio, no other bugs or crashes. With the previous kernels doing this would crash apps especially when browsing.. It is a sudo score I was able to go to 7.0 GHZ and scored a 340,000 Geekbench compute with my vega but that is impossible. Can you try quickly and see if it’s stable at higher settings with this kernel? I’m only losing sound, not actually losing it but it’s cutting in and out.

i have already did this on sierra

the more you will increase it , the more it will affect the stability (especially the bus speed entry)

also causes sync out issues .

 

Link to comment
Share on other sites

i have already did this on sierra

the more you will increase it , the more it will affect the stability (especially the bus speed entry)

also causes sync out issues .

 

This is exactly what I expecting but it is not happening, 7.0 GHZ and it’s not crashing?

I'm curious to see what kind of FPS you're able to attain in a game that isn't benchmark software, such as League of Legends for example.

I'm also curious to see if you get a kernel panic, when you run 16 threads of

 

yes > /dev/null &
with your most stable clock that is above 4.0Ghz reported.

Typically when I'm overclocking, I use Cinebench 15 to test if it crashes, if so it's unstable, if not then I run the generic yes test as well.

attachicon.gifScreen Shot 2017-10-08 at 9.43.51 PM.png

 

After you run the 16 threads, you can stop them by using

killall yes
I tried that last night a couple of times and got a flat summit as shown during the GPU test and a high peak during the CPU test. CPU % was 86% idle for most of the run. It doesn’t seem to be stressing the CPU during the test according to Activity monitor readings.

 

P.S. is that a different version of activity monitor, I’m not getting the blue load bars at the bottom of window?

Link to comment
Share on other sites

This is exactly what I expecting but it is not happening, 7.0 GHZ and it’s not crashing?

I tried that last night a couple of times and got a flat summit as shown during the GPU test and a high peak during the CPU test. CPU % was 86% idle for most of the run. It doesn’t seem to be stressing the CPU during the test according to Activity monitor readings.

 

P.S. is that a different version of activity monitor, I’m not getting the blue load bars at the bottom of window?

In the menu bar of Activity Monitor:

Window->CPU Usage

Personal preference, but if you'd like you can also use the faster update option:

View->Update Frequency->Very Often(1 sec)
Link to comment
Share on other sites

can someone help me install?

 

Watch the video in this link a few times to see the general method for install.  Start with a fresh disk and only use the kernel, relinked kernel and system kext.   don't worry about the Frameworks files.     

 

http://www.insanelymac.com/forum/topic/324392-ryzen-clover-installation-guide-macos-sierra/page-1?do=findComment&comment=2442531

 

Add Lilukext and Whatevergreen to your S/L/Extensions folder as well as Clover Kexts Other folder.    Check back when this is done and I will try to help.    You may need to tweak your config.plist or other clover settings a bit.

Link to comment
Share on other sites

Can you send me the kernel, system.kext, prelinkedkernel and immukernel? Because the ones i use, the system is stoping at rebuild prelinkedkernel.. :( (at the 3/4 loadbar)

 

remove the immutable kernel and use only the prelinked here.    The kernel should be fine.  

 

Also, try boot arg. lilubeta and radbeta along with busratio=xx and npc-0x3000 or 0x2000.

prelinkedkernel.zip

Link to comment
Share on other sites

×
×
  • Create New...