AMD Addresses Controversy: RDNA 3 Shader Pre-Fetching Works Fine
AMD Addresses Controversy: RDNA 3 Shader Pre-Fetching Works Fine
These aren't the bugs you're looking for.
Ryzen 7000 Testing
- Thread starter PoMpIs
- Start date
That's a relief -- now we wait for the next false alarm bell to ring!AMD Addresses Controversy: RDNA 3 Shader Pre-Fetching Works Fine
AMD Addresses Controversy: RDNA 3 Shader Pre-Fetching Works Fine
These aren't the bugs you're looking for.www.tomshardware.com
I also do not speak any Russian, but it looks like at that portion of the video the narrator is just going through currently supported Navi cards, up to Navi23. Unless I missed something, I didn't see any reference of Navi 31 or device ids for any of the 7000 series cards, unfortunately
As for the previous bits of the video where their system info shows a 7900 XT under his CPU, perhaps they just changed their current gpu model name for illustrative purposes? Just based on the title, I wonder if the video is just a discussion of whether Navi 31 will be supported? Especially since it was published 2 months ago, before any of these gpus were released haha.The more rumors I hear about Apple's struggles with their upcoming Mac Pro, the more hopeful I get that we'll get another generation of AMD gpu support. Again, just rumors, but I'll keep dreaming until my dreams are officially squashed dead!
@unforsaken
That is my understanding as well. They are just describing how to check which versions of AMD GPUs are supported. They show that the list ends with Navi23 and that we probably need to wait until a new Navi31 section is added.
That is my understanding as well. They are just describing how to check which versions of AMD GPUs are supported. They show that the list ends with Navi23 and that we probably need to wait until a new Navi31 section is added.
Last edited:
@ovenlite1 Apple is more than happy selling outdated products for full price (see MacPro6,1). Are there any rumours than just speculation? Last time we actually had developers of Mac rendering software leaking hints that they are testing beta drivers of Apple for Navi21.
True. I forgot to even check when was this video uploaded out of excitement
so thank you and @ovenlite1for clarification. I guess the GPU can go back in the box for a while or just use it on another setup Ha you're right, I suppose they have no problems leaving us hanging! I hope for the best, but it's nice to already be on 6000 series so as to soften the blow if/when support does not come!
Believe me, I was just as excited when I saw what you posted!!! No denying that it's a powerful gpu- hopefully you'll be able to make use of it somehow!True. I forgot to even check when was this video uploaded out of excitement
I updated my EFI to support AM5 from my old x570 setup, and while it boots and is stable, network performance is absolute trash.
I used the updated driver from a couple of pages ago, and while it does work, it maxes out at like 10mb/s, which is obviously much slower than full gigabit. Wifi using the Fenvi T-919 is also maxed out at the same rate, and was much faster on my old x570 setup.
Does this have to do with maybe having the wrong MMIO whitelist enabled?
I used the updated driver from a couple of pages ago, and while it does work, it maxes out at like 10mb/s, which is obviously much slower than full gigabit. Wifi using the Fenvi T-919 is also maxed out at the same rate, and was much faster on my old x570 setup.
Does this have to do with maybe having the wrong MMIO whitelist enabled?
** Experimental Version of RestrictEvents **
As discussed previously in this thread, when RestrictEvents kext is enabled on an AMD platform, we lose the ability to perform incremental macOS updates. Instead, the full 12GB installer is downloaded, which wastes network bandwidth, time and disk space.
The attached experimental version of RestrictEvents attempts to fix this problem. This is very much a work in progress and a handful of revisions will be needed until it works.
Do not expect this version to work. It is meant primarily to collect information.
Beta Testers Wanted:
Because I'm currently on vacation far from home, I don't have access to any of my systems. I'm working on my Apple Silicon 14" MacBook Pro as time permits. Volunteers wanted! Your mission, should you choose to accept, is far from impossible...
- Prerequisites:
- Must be running Ventura
- Must use MacPro7,1
- Must have an AMD AM4 or AM5 system
- Must be using RestrictEvents.kext in OpenCore
- Rename or backup your existing RestrictEvents.kext in EFI/OC/Kexts folder
- Download and copy the attached version into EFI/OC/Kexts
- Reboot
- After logging in, open Terminal and type the following:
Bash:
log show --last boot > ~/Documents/bootlog.txt- This will generate a file in Documents folder called bootlog.txt
- Search that file for Casey
- Copy and paste any log line that contains that name
- If System Settings --> General --> Software Update shows that you have an available macOS update, feel free to try installing it, otherwise just copy your original RestrictEvents kext back and reboot
Last edited:
ciao CaseySJ
great challenge
Only for your information, in my system (trx40 chipset but I think also in newer or lower ones) problem happen only from Ventura OS incremental update and only if you use MacPro7.1 SMBIOS
With iMacPro 1.1 it updates well
great challenge
Only for your information, in my system (trx40 chipset but I think also in newer or lower ones) problem happen only from Ventura OS incremental update and only if you use MacPro7.1 SMBIOS
With iMacPro 1.1 it updates well
Thanks — this is helpful because I can restrict my changes to just these conditions.ciao CaseySJ
great challenge
Only for your information, in my system (trx40 chipset but I think also in newer or lower ones) problem happen only from Ventura OS incremental update and only if you use MacPro7.1 SMBIOS
With iMacPro 1.1 it updates well
I used the same guide with my PowerColor 6900XT Ultimate on my x570/5950X build.
it works well.
I'm having trouble booting into the installer with the same card in my X670E / 7950X system but I have not tinkered with it for some time. my latest setback was that my NVMe drive died so not even windows 11 will boot so I can get the BIOS path to the graphics card. I also have to remove tghe CPU cooler to get to the NMVe drive. kill me now.
i got a new NVMe in the mail a dew days ago, i should be able to install it this weekend.
Hi,** Experimental Version of RestrictEvents **
As discussed previously in this thread, when RestrictEvents kext is enabled on an AMD platform, we lose the ability to perform incremental macOS updates. Instead, the full 12GB installer is downloaded, which wastes network bandwidth, time and disk space.
The attached experimental version of RestrictEvents attempts to fix this problem. This is very much a work in progress and a handful of revisions will be needed until it works.
Do not expect this version to work. It is meant primarily to collect information.
Beta Testers Wanted:
Because I'm currently on vacation far from home, I don't have access to any of my systems. I'm working on my Apple Silicon 14" MacBook Pro as time permits. Volunteers wanted! Your mission, should you choose to accept, is far from impossible...
- Prerequisites:
- Must be running Ventura
- Must use MacPro7,1
- Must have an AMD AM4 or AM5 system
- Must be using RestrictEvents.kext in OpenCore
- Rename or backup your existing RestrictEvents.kext in EFI/OC/Kexts folder
- Download and copy the attached version into EFI/OC/Kexts
- Reboot
- After logging in, open Terminal and type the following:
Bash:log show --last boot > ~/Documents/bootlog.txt
- This will generate a file in Documents folder called bootlog.txt
- Search that file for Casey
- Copy and paste any log line that contains that name
- If System Settings --> General --> Software Update shows that you have an available macOS update, feel free to try installing it, otherwise just copy your original RestrictEvents kext back and reboot
first of all great job: you already saved my life some months ago with the Titan ridge now working in my x570/ 3700X system and I'm pretty sure I will get the ACQ113cs working on my 5950X / x570e (ProArt Creator) leveraging on your great progress here!
For the latter I'm currently limited to Big Sur since I'm stuck with a Gt710 waiting for a 6900Xt coming tomorrow (hopefully not having the XTXh chip...) and as I'm getting incrementally frustrated by the ACQ113cs not even being recognized by the OS, I'm planning to switch to Ventura to make it work and with the occasion, I could try to test the Kext ...
Here is the log, CPU name spoofing and memory modules fixes are working, will try to OTA update it right now.** Experimental Version of RestrictEvents **
As discussed previously in this thread, when RestrictEvents kext is enabled on an AMD platform, we lose the ability to perform incremental macOS updates. Instead, the full 12GB installer is downloaded, which wastes network bandwidth, time and disk space.
The attached experimental version of RestrictEvents attempts to fix this problem. This is very much a work in progress and a handful of revisions will be needed until it works.
Do not expect this version to work. It is meant primarily to collect information.
Beta Testers Wanted:
Because I'm currently on vacation far from home, I don't have access to any of my systems. I'm working on my Apple Silicon 14" MacBook Pro as time permits. Volunteers wanted! Your mission, should you choose to accept, is far from impossible...
- Prerequisites:
- Must be running Ventura
- Must use MacPro7,1
- Must have an AMD AM4 or AM5 system
- Must be using RestrictEvents.kext in OpenCore
- Rename or backup your existing RestrictEvents.kext in EFI/OC/Kexts folder
- Download and copy the attached version into EFI/OC/Kexts
- Reboot
- After logging in, open Terminal and type the following:
Bash:log show --last boot > ~/Documents/bootlog.txt
- This will generate a file in Documents folder called bootlog.txt
- Search that file for Casey
- Copy and paste any log line that contains that name
- If System Settings --> General --> Software Update shows that you have an available macOS update, feel free to try installing it, otherwise just copy your original RestrictEvents kext back and reboot
Edit: It failed unfortunately.
Last edited:
Anyway I have noticed tonight something odd while moving date from my "Old" X570 (3700X, Asus AQC107 PCI card) on Catalina and the "New" setup with he 5950X/x570e on Big sur with a AQC107 connected via Thunderbolt 3 to the system.
Both are connecting at 10GB.
Now the oddity:
If i move the very same file from "Old" to "New". I get almost full 10GB/s speed, If I do the viceversa I only move the file at around 1GB/s speed...
Both cards are configurate as Full Duplex, MTU 1500.
Any idea?
As you can imagine I unfortunately need to move data from "Old" to "New"...
Both are connecting at 10GB.
Now the oddity:
If i move the very same file from "Old" to "New". I get almost full 10GB/s speed, If I do the viceversa I only move the file at around 1GB/s speed...
Both cards are configurate as Full Duplex, MTU 1500.
Any idea?
As you can imagine I unfortunately need to move data from "Old" to "New"...
Thank you for trying it out.
Couple of notes:
- I didn't find any log entry containing "Casey" so perhaps "SYSLOG" in RestrictEvents writes to dmesg instead? This log cycles quickly, which means we have to capture it immediately after booting:
sudo dmesg > dmesg.txt - Based on what the log says I'll know what modification to make next
- Because incremental updates work when RestrictEvents is disabled, I'm trying to keep Memory and PCI notifications suppressed while fixing just incremental updates; this means either forcing-on or forcing-off a specific part of this kext
I responded to something similar a couple of posts above:
Ryzen 7000 Testing
About what? Can you send a link? AMD Addresses Controversy: RDNA 3 Shader Pre-Fetching Works Fine https://www.tomshardware.com/news/amd-addresses-controversy-rdna-3-shader-pre-fetching-works-fine
forum.amd-osx.com
Let's see if this helps.
Thanks. The problem is actually the "direction" of the link: the link itself (meaning the speed both cards negotiate while connecting) can run at full speed, otherwise I won't be getting full speed from "new" to "old".I responded to something similar a couple of posts above:
Ryzen 7000 Testing
About what? Can you send a link? AMD Addresses Controversy: RDNA 3 Shader Pre-Fetching Works Fine https://www.tomshardware.com/news/amd-addresses-controversy-rdna-3-shader-pre-fetching-works-fine
Let's see if this helps.
Anyway very curious to test it out as soon as I got home.
Strangely enough, the same ACQ107 Tb3 when connected to a M1 Air can "receive" date from old at full link speed...