AMD Says They'll Be Open-Sourcing More Of Their GPU Software Stack & Hardware Docs

[gnif]

There is if youŕe dual booting Windows and want to access the volume from both OSes. mdadm is not usable in Windows, and AMDRAID is not usable in Linux.

Good point, sorry, it's super rare I care to do this so I had forgot about this specific use case.

[Anux]

There is if youŕe dual booting Windows and want to access the volume from both OSes. mdadm is not usable in Windows, and AMDRAID is not usable in Linux.

Don't use a raid if you are rebooting your machine. Raid is for uninterrupted availability of storage. No use if you constantly reboot that system anyway.

dmraid is for fake raids but again don't use it if you really need raid functionality. It's more for troubleshooting to access your fake raid if windows fails.

[Panix]

LACT (Linux AMDGPU Control Application) which was announced here on Phoronix, CoreCTL and TuxClocker. LACT is a favourite because of its UI and the ability to run headless.

Right. I know. It's good that one of them, at least, have become a 'fav. pick' - AMD's gpus need such a program for their top tier gpu line - i.e. 7900 series - since, they run hot - there's no shortage of ppl trying to figure out why their hot spot is over 90 degrees. That's insane. Most of the Nvidia gpus - top tier - 4080 and above - have about high 60s top - or low 70s - which is more typical and acceptable. It's good that AMD has open sourced their gpus, obviously - and it allows for development and progress with these programs - but, AMD has the hot running gpus - and they should provide programs to allow undervolting/fan profiles and if they won't do it - they should help out the developers of these programs, imho. It's so commendable that those developers have done so well and they are an option for AMD gpu users - I wouldn't consider an AMD gpu without those.
Edit: Checked TPU and the 4080 Supers tend to be high 70s - although, mostly on the OC'ed cards - still, that's 10 degrees cooler than the average 7900 xtx?

[Panix]

Lol so far you are the only one bringing personal attacks, if you feel attacked that one is on you. Also where is this strawman you made up then? You are the first to bring up the lack of an Adrenalin tool and you don't even own an AMD GPU. And what is wrong with the multiple overclocking tools from the open source community? Like how it is a sore spot if the person bringing it up is one that doesn't own an AMD GPU at all? Talk about excuses.

Wanna talk about your forte? It is being wrong, pretending a question or response never happened, deflection and denial, you don't even apologize for being an ass while being absolutely wrong on what motoring meant, just pretend it never happened and deny you are wrong. It isn't blabbering about semantics, you were wrong, period and that would be fine if you weren't such an absolute ass about it and maybe went "My bad, I got the two mixed up since most overclocking tools have monitoring" or something like that, you know humble, open to discussion those kind of things, I would go "Alright, fair enough" and we would continue this but no instead you accuse me of not addressing the point while that is literally all I do.

Speaking of which, about that Blender use case I need to keep asking you about, what is it? This isn't question out of malice but general helpfulness, there are cases where AMD indeed falls short but the same applies to Nvidia, the landscape has shifted since the RTX 3000 and the RX 6000 releases. But your continued silence when I bring it up continues to strengthen the idea you are nothing but a troll, you don't want to discuss the topic, you aren't interested in it, you just want to shit on AMD for some selfish reason.

Have you ever? Like legitimately, I don't think I have seen you post quite in the same way about Nvidia shortcomings (or similar to AMD anything you perceive as such) inside of Nvidia oriented articles.

By not providing a GUI tool but giving the necessary resources to the community to make their own to suite their needs?
Buddy that is what FOSS people like, that they can make their own tool. And how do they neglect support, I already told you they opened up overclocking through pp_od_clk_voltage, you can write your own script or your own program can you admit that much? It's enough for FOSS people, you are the one that has a problem with it.

Okay but why does that bother you and why does it even matter?
You clearly don't seem to understand FOSS with all of this, that Nvidia provides an overclocking tool isn't a good thing towards FOSS (nvidia-settings is not even FOSS to begin with), nor is the lack of an official AMD one a bad thing as long as we can make our own. You can have your qualms with it but not everyone here will agree. You are literally the only one that has a problem with it in this thread and I can't stress enough that I know you don't own an AMD GPU so why should they bother?

Btw what is a bad thing for FOSS is blocking reclocking to even the base advertised clock on open source drivers through signed firmware making your closed source drivers the only possible way to run your GPUs, then a couple of years later suddenly embrace "open source" (still gotta make sure a lot of it is behind closed source firmware) but oops leave out your most popular and value tiered GPUs of their decade meaning that once the closed sourced drivers mark them as EOL users won't have any alternative and will have to buy new ones in order to play new games. You know that would be quite a bad thing towards FOSS, lets hope AMD doesn't do something like that.

Most of the programs include monitoring - 'thermal monitoring' - I just used a generic term.
This isn't about Nvidia. There's enough Nvidia bashing on here. I also think Nvidia has a 'pass' here ONLY because their latest gen doesn't run hot - but, if, in the future- they have overheating cards - then having one in Linux will be a major problem. I am saving up for a 4080 - I have considered a 7900 xtx - but, then I'd need to use one of the programs ppl discuss here - Corectrl, LACT, Tuxclocker - and I criticized AMD for not providing their own program since they pride themselves in being FOSS in Linux - yet, no program/utility - they don't want to invest in it and can't be bothered - so, they have the fortune of 3rd party programmers deciding to do it. I think AMD should be chastised for that. It sounds like the options for these programs are progressing - I haven't read the latest reports but the last I heard - there were options one couldn't use because AMD was not including the required info in the kernel - that was supposed to be 'fixed' in kernel 6.8/6.9.

I haven't been wrong about anything - the 'monitoring' mention was just generic comments about it - they can be called overclockers/control whatever - is that better, sir?
Why does it matter which gpu I use/own? I am researching what is out there and what ppl say who have them - not fanboys but ppl who are objective - who actually use them and good or bad - what they report.

The industry I'm interested in - tend to say 'to stay away from AMD' - that's their words and they have used them. You're ignoring that because you're an AMD fanboy. I already said before, I'd like to pick an AMD gpu for the FOSS situation but, AMD has poor support and is getting reamed by numerous ppl - someone even posted or cited reports on AMD's poor driver and support - ppl who have pretty good reputations.

[duby229]

Blah, blah, blah...they're still incompetent - bugs and software that doesn't work right - it's probably to distance themselves and allow the 'open source community' pick up the slack. AMD still doesn't have a gpu monitoring program in Linux - and they've made up excuses for that, too.

AMD already hired every developer that was making significant contributions... AMD -IS- the "open source community"

[Panix]

AMD already hired every developer that was making significant contributions... AMD -IS- the "open source community"

What about Intel?

[duby229]

What about Intel?

Same with Intel, they too have already hired everyone that was making significant contributions.

EDIT: You know years ago that seemed to be one of John Bridgmans biggest complaints, was that they were doing most of the work themselves. Developing GPU drivers and implementing hardware interfaces is hard and most people don't have the skill or knowledge to do it.

[tenchrio]

Most of the programs include monitoring - 'thermal monitoring' - I just used a generic term.

Yes, you used a generic term for monitoring which you specifically asked for but you still then changed it when we gave you specifically monitoring tools as those tools have more in depth monitoring then overclocking tools. You were still wrong, an overclocking tool can be seen as a monitoring tool (doesn't make it one, just that it can fulfill the role somewhat) but a monitoring tool is not an overclock tool.
It like the analogy I gave you before, a car is a vehicle but not every vehicle is a car, if you ask for a vehicle you can't be surprised when someone hands you a bike.

If you ask for monitoring tools we will give you monitoring tools, why would we give you an overclock tool if monitoring tools tend to for instance include the usage of resources per process and people that use monitoring tools aren't necessarily looking to overclock to begin with, tools like nvtop are available in about every repository.
Just admit you are wrong, it's not that bad, the confusion on why you mixed the two up is pretty obvious and somewhat understandable.

This isn't about Nvidia. There's enough Nvidia bashing on here. I also think Nvidia has a 'pass' here ONLY because their latest gen doesn't run hot - but, if, in the future- they have overheating cards - then having one in Linux will be a major problem.

Honestly it seems Nvidia and AMD switched places here since last gen, seems to be a bit like the tik-tok principle similar to AMD and Intel Cpu releases. One gen has performance increases, the next sees better thermals and power draw. Also Nvidia had overheating cards this gen, didn't stop me or others but still it was an issue to keep in mind.

I am saving up for a 4080 - I have considered a 7900 xtx - but, then I'd need to use one of the programs ppl discuss here - Corectrl, LACT, Tuxclocker - and I criticized AMD for not providing their own program since they pride themselves in being FOSS in Linux - yet, no program/utility - they don't want to invest in it and can't be bothered - so, they have the fortune of 3rd party programmers deciding to do it. I think AMD should be chastised for that.

Ok, you do you. But again me and others don't really see an issue here (nobody so far agrees with you about this). The entire point about FOSS is having freedom, AMD not offering a tool here is about as much a problem as Linus Torvalds or the Linux Foundation not providing their own made DE, as long as there is the freedom to make/choose your own then it is fine. You can have your personal qualms with it but from the perspective of FOSS this isn't an issue. I don't think anyone here agrees AMD should be chastised for that.

It sounds like the options for these programs are progressing - I haven't read the latest reports but the last I heard - there were options one couldn't use because AMD was not including the required info in the kernel - that was supposed to be 'fixed' in kernel 6.8/6.9.

I haven't been wrong about anything - the 'monitoring' mention was just generic comments about it - they can be called overclockers/control whatever - is that better, sir?

No, that is exactly what they are called, that is how you would find them if you were to google for them.
Even AMD calls their module in Adrenalin to overclock, surprisingly, an overclock tool. The word monitor doesn't even appear on the entire page nor does it even on the Adrenalin page, you know what set of words do appear on Performance tuning page? "Performance Tuning is AMD’s groundbreaking tweaking and overclocking tool ".

Why does it matter which gpu I use/own?

Would that be with or without your massive confirmation bias towards NVIDIA? Also simply put you are shit talking GPUs you don't own (or ever owned) on workloads you don't understand or have experience with all while praising the other side that you don't own or use either but are intend on buying, so its more about the emphasis that your opinion doesn't seem to come from the best of places.

I am researching what is out there and what ppl say who have them - not fanboys but ppl who are objective - who actually use them and good or bad - what they report.

Right so you ignore the person giving you objective reasons on AMD's strengths since you only focused on Nvidia's (all the while throwing insults at AMD only showcasing how little you actually know on the topic, nice research), then name call me on numerous occasions while ignoring all of the areas that have been laid out where AMD does exceed and you think you are being objective and in the process of "researching"?
Best joke I have heard all day.

The industry I'm interested in - tend to say 'to stay away from AMD' - that's their words and they have used them.

Considering how often you make things up, how about you back this up with a source? Give me an actual person (besides you) that says only Nvidia is an option and AMD is not (and with Blender 3.6 or later, since I told you before I held this opinion back with 3.0 but that is ~2 years ago and there is this crazy thing called time, which when it passes by, things change, don't worry it is a natural process). Hell the full quoted words you are saying "stay away from AMD", come on proof someone actually used it over just recommending Nvidia.

Because many Blender artists will repeat what I have been saying, Nvidia is faster for Cycles but the VRAM can be an issue and if you use Eevee that performance difference disappears (with a trade off for realism) but the VRAM potentially backfiring remains. I have given you countless sources before that also recommend you use Eevee for animation, or how Cycles leverages faster render time at the cost of VRAM. The opposite exists too where you can decrease performance and your memory usage with for instance Tiling (generally speaking you want to avoid this), but good luck when you run into VRAM issues with Eevee (man it is almost as if I have experience with one of the tools you criticized AMD on before and not only that but also with memory issues as well as knowing how to optimize render performance outside of the defaults most tech journalists use when benchmarking). All of that btw is something you have been unable to admit (or even contest, you just outright ignore it in the posts or sometimes the entire post with full details and links showing that you don't care about the industry or the tools, only about spouting your biased perception).

You're ignoring that because you're an AMD fanboy. I already said before, I'd like to pick an AMD gpu for the FOSS situation but, AMD has poor support and is getting reamed by numerous ppl - someone even posted or cited reports on AMD's poor driver and support - ppl who have pretty good reputations.

The guy with actual experience, who gives detailed explanations on render performance including how factors like VRAM would affect it while not being evident in benchmarks and that owns an RTX 4090 and RX 7900XTX is an AMD fanboy but you, the guy who never owned an AMD GPU, never even opened Blender yet somehow has the most stoic anti-AMD opinion on it, can't admit the difference between monitoring tools and overclocking tools, bases their buying decision on the most arbitrary bullshit I have seen in a while and even best of all claims to want to support the FOSS environment yet buys an Nvidia gpu for a non FOSS issue is supposed to be "objective" and not a fanboy?

LMAO, you are clearly an Nvidia fanboy and a troll. You don't care about FOSS, you don't care about "the industry", you don't care about Rendering performance or 3D art, you don't care about about objectivity. You are just a trolling fanboy that wants to pretend Nvidia is the only possible option for a niche that you aren't even participating in (in the comment section of an AMD compute post no less, like why are you here, can't exactly say your "contribution" has been anything but eye rolling for everyone).

[WileEPyote]

Don't use a raid if you are rebooting your machine. Raid is for uninterrupted availability of storage. No use if you constantly reboot that system anyway.

dmraid is for fake raids but again don't use it if you really need raid functionality. It's more for troubleshooting to access your fake raid if windows fails.

RAID, in my case, means RAID0 with 2 2TB nvme drives., and is only for performance. I have backup drives for preserving data. dmraid doesn´t work with dual booting either. Not accessible from Windows.

Thus my desire for AMD to open source the storage driver as well.

[Anux]

dmraid doesn´t work with dual booting either. Not accessible from Windows.

Are you sure? With dmraid you would first install Windows (and it should therefore always be accessible there) and then make the raid available in Linux to install your dualboot.

Maybe AMD uses a new raid/disk format that isn't jet supported by dmraid. But still Windows should have access either way. It's over 10 years since I've last used this, so I might be out of touch.