Any update on 5700 Xt support? #887
Comments
|
There are different levels of support in ROCm, since it's an entire stack, which makes this a difficult question to answer. 1-Kernel support. This usually comes pretty quickly from upstream, and is already there in its infancy. The basic code is there, and there is work going on to thoroughly test the kernel functionality So there may be some partial support for it right now, or it might work almost perfectly. It's all a bit of a crapshoot until we update the documentation, as we haven't tested everything to fulfill our exit criteria for supporting a GPU. You can always try it out, and post PRs to help to support them in the interim, though. We always like help from the community! |
|
Hi, in my experience GPUs listed as "supported" have no guarantee to work properly. I have opend a few issues here and the speed at which theses issues are fixed is extremely disappointing. The most annoying thing is missing OpenCL 2.0 support for hardware that is still beeing sold (gfx803). I have zero confidence that the latest hardware will have OpenCL 2.0 support in ROCm, therefore I am not buying new hardware. There would be more business for AMD if the drivers delivered better OpenCL support. |
|
@kentrussell Could you create a new issue for us to track Navi integration? You could create a task list to mark the process:
|
|
I'm new here, and don't know anything about anything, so it's perfectly fine that I chip in. ;) I'm interested in ROCm since I'm currently buying the 5700 XT for myself and wanted, aside from gaming, to play around with PyTorch and Tensorflow. Seems like kernel support has been merged in on Sep 23rd, Thunk has support for Navi 10 since July, so we're getting there. |
|
@aak-amd @zhang2amd @Rmalavally @kentrussell Any updates about Navi/RDNA support? |
|
Bump. |
|
It is kind of sad that AMD has forgotten Navi users. I have found a temporary solution to use my RX 5700XT for deep learning with rather astonishing results. I thought I should share it here. Using Linux kernel 5.6rc and AMDGPU PRO driver I was able to set up OpenCL 2.0 on Manjaro Linux. Then I installed PlaidML which supports opencl devices. After that, I set Keras backend to PlaidML and just used Keras. I ran some benchmarks and the results are just amazing. It outperforms my 12C/24T Ryzen 3900 CPU by a massive margin. While it takes more than 12 minutes to train MobileNet on Ryzen 3900X, it takes less than a minute on Radeon 5700XT. Here are the Inference latency and Time/FPS comparison in mobilenet benchmark:
And for those who prefer to see a working example, here is a CNN trained on MNIST in Keras using RX 5700XT:
|
|
ROCm-OpenCL-Runtime 3.1 changes introduced ROC_GFX10, that's Navi! |
|
The Navi support seems to be a work in progress. There are no clear instructions on how to compile ROCm from scratch anyway. I downloaded all their sources codes using Google Repo tool and there are at least 40 different projects to compile separately! I couldn't find any easy way to automate the building process. There is a project that is supposed to automate the build process but has not been updated in the past few years. If the guys at AMD can illuminate us on how to automate ROCm compilation, it would be greatly appreciated. |
|
There is no or no complete Navi support in ROCM 3.1 . ` Generating training model |
|
Any update on the status of this issue? |
|
I'm not a ROCm user, but I tried installing it just to get OpenCL working because as I found somewhere that it's how ROCm works on AMD devices: through OpenCL. So ROCm supposed to set OpenCL up, but failed in my case. Eventually, I got OpenCL working on Ubuntu 18.04, but I think it should work on any distro (only kernel version matters): https://askubuntu.com/questions/1209725/how-to-get-opencl-support-for-navi10-gpus-from-amd <-- that's how I managed to achieve it. |
|
@SlavMFM |
|
Interesting. Was you able to run some raw OpenCL examples to verify it's working properly? I once had "fake" OpenCL setup where |
|
@SlavMFM As far as I know, AMD provides two seperate OpenCL 2.0 implementations. One in ROCm and another in AMDGPU PRO driver. I have tested the OpenCL 2.0 in AMDGPU PRO and can confirm it works fine (see my post above). Although, getting OpenCL to work does not guarantee that ROCm would work too. |
|
@jemzipx cool! I'm surprised amdgpu-pro drivers can be actually installed ^^ - you just get kernel 5.6 and install amdgpu-pro on top of it? |
|
@SlavMFM that is right. I installed amdgpu-pro on top of kernel 5.6rc in Manjaro (should work in Arch too). I tried Mesa before (actually that was the first thing I tried) but Mesa's OpenCL version was 1.2 which is quite old and kind of outdated. As far as I can tell, OpenCL 2.0 support for Navi can only be found in AMDGPU-PRO and ROCm. |
|
@SlavMFM It should work on all distros. I have only tested Manjaro personally. I'm also curious to know how it plays on Ubuntu. BTW, have you tested Keras/PlaidML with your Mesa OpenCL on Ubuntu? Please give it a go and let us know the result. |
|
[Update] |
|
AMD ROCm is validated for GPU compute hardware such as AMD Radeon Instinct GPUs. Other AMD cards may work, however, they are not officially supported at this time. We appreciate your feedback and we will consider it for future versions of ROCm. Regards, |
|
I don't think anyone here is asking for validation or even official support. All that Navi owners and potential buyers would like is for it to somehow work, now several months after it's release, just like it would work for them from day zero if they have got, say, a RTX 2070 using CUDA. |
|
let's just call it AMD on the compute market is a joke at the moment and the only option for the consumer is to go nvidia. |
|
@Rmalavally That is a very disappointing answer from an AMD's representative. The thing is that, Nvidia's success was due to the fact that every college gamer kid with their consumer-based GTX/RTX card could run deep learning algorithms as well. There are many of us that do not want to rely on Nvidia's propriety CUDA and would prefer to use open standards like OpenCL etc. This is a great business opportunity for AMD and is the one that matters more than designing graphic cards for next-gen xbox and playstation. We hope AMD change its approach to consumer-grade cards. |
This is not the first time we get disappointing answers from AMD, We ALL hope AMD change its approach to consumer-grade cards, and its approach to Radeon Instinct MI50 selling strategy for consumers too, not only for businesses. |
|
Well let's face it: If AMD decides to put all their effort on fancy 7nm tech, but does not want to spend the money for software developers, we just have to swallow that pill and move on. |
Repeating: Just to remember to everybody discussing here: if you are not a ROCm developer or github admin, it is not your task to reprimand people here on whatever you might disagree with. Let the admins do their work, and stay at your place. If you don't want to be notified it is as simple as to mute notifications or change your notification preferences to something less invasive. |
This statement has confused me so much that I'm implored to ask: So, are you a dev/admin and making this statement from the position as such? (I honestly can't tell, as I fail to see why somebody not in the position would make such a comment) If so, sure, agreed. If not, I do still agree but that sounds oddly hypocritical. Especially from a person who recommended another person to apologize. (Is that not reprimanding?) I also fail to see how "unsubscribe or change your notification settings" is a solution. I - and I presume others too - want to be notified, that is, for stuff I consider relevant to the issue of "Any update on 5700 Xt support?". But yes, I'll follow your advice and unsubscribe for a day or two or three. |
No I am not a dev/admin, only a long time ROCm user and app contributor. My reply was in defense of free speaking and personal liberty to express opinion. The alleged recommendation to apologize, was merely given as suggestion (would and not must) for the sake of community respect. I didn't start this flame and my attempt to help make peace failed miserably. For penitence I will skip dinner. |
|
@PeterNjeim You wrong in saying that complaining does not work and it does not add to the conversation. It works very very well if there is enough numbers. AMD is completely ignorant on |
I agree with this, AMD timelines have been optimistic for at least 3 years now, often they release software past the timeline where the issues are not yet resolved, and they push a subsequent extension to timeline. This gives the impression of, at best, poor estimation of workload to do, or excessive estimation of developer capacity. Professionally this makes a loss to credibility of AMD estimation capabilities. |
What a disgustingly condescending message and downright malicious. Several people have already said to stop talking about this subject (including the person right above this message, ironically). I hate to repeat myself, but I will again: I never said you couldn't complain. Want me to say it again? I will: I never said you couldn't complain. Again? The point is simple yet you and others complicate it so much: your opinion has been stated so often, that it is no longer useful, and unnecessarily notifies every one in this thread about sometime that isn't new. That is not the point of this forum, and the difference between 100 and 101 complaints doesn't do anything. The fact that ROCm has already closed this issue and stated they are working on it should tell you that ROCm has already heard the complaints, and have acted upon them. Want me say another thing that I already said? I will: the complaints were useful the first time around. I never said you can't complain, I never said complaints were not acceptable here, I said that the same complaints, repeated over and over again even after action has been taken are so unnecessary and notify people for utterly no reason. This is the specific point I have made throughout this thread. This is probably the 4th time someone has strawmanned me (misrepresented my argument to make it easier to attack). If you have a problem with my specific reasoning, please, tell me why you disagree. If you continue to make up arguments that I never made, then I guess we'll continue going back forth, with me telling you how you aren't even responding to anything I have said. I do have one insult to make though: there is an absurd amount of passive aggressiveness in your message, which is why you probably saw some passive aggressiveness in this message too. You don't decide who "understands the point" or not, it's incredibly ironic how you don't understand my point yet continue to talk as if you do. Please, I ask for this nonsensical discussion to end, there is nothing more to talk about. |
|
Guys, it is very simple: do not buy AMD GPUs at least this year if you need GPU for work. If you do have an opportunity to buy a GPU these days) |
I still cannot understand why you think that it is your right to decide that one opinion is not useful, you are only one voice in the whole, and I think after all to conclude this unfortunate thread, this should be put to vote for a democratic exit. |
That would be an appeal to popularity logical fallacy. It is extremely strange the way you're acting here. You don't respond to anything I have said, you just make remarks telling me to shut up. I won't. I will logically argue against those who fallaciously misrepresent me. I have a right to defend myself, and they have a right to defend themselves. There are several people in this discussion, you are "one brick" as well, and you are talking. Do you not see the hypocrisy? "Let the wall speak", I am. I will repeat myself again: If you disagree with anything I have said, please, tell me why. In what world is this not letting others speak? Stop strawmanning me. You are correct, it's very unfortunate that this thread devolved into logically fallacious arguments. Remember, anyone who argues fallaciously is intellectually dishonest, with complete disregard of the truth and progress. Not a single person has argued against my specific point I made. I await the disagreeing comment, or, if the person I was talking to agrees with me, I am awaiting their message of agreement. Edit: not to mention, you replied to someone earlier saying that you agreed with them, when their whole message was about not allowing me to voice my opinion. You are a hypocrite. You cannot in one message agree with those who want to silence others, then in another message, which explicitly asks for more discussion (while admitting that is a nonsensical one), you call "not letting the wall speak." This is entirely disrespectful. You also said to stop talking, you even spammed the same message twice telling others to stop talking, yet you came back later to talk. You are allowed to, of course, but I do not understand, are you trying to cause a commotion? Be consistent, I beg of you. |
I can't speak for the person you replied to, but I suspect a lot of people subscribed to this issue have a checklist of priorities where AMD is the preferred GPU choice in all metrics except for the "it works" metric. There's nothing wrong with keeping an eye on things from that perspective. |
It is obvious that you have misinterpreted my words all along. I am not against anybody speak, but as yra-wtag said, you gave the impression to be trying to educate someone else in this context about when and how often to express their opinion, which is something that I truly disagree with, and it isn't with insults that you will resolve this issue.
|
In my last message to you, I listed all the instances in which you were insulting, disrespectful, fallacious, and malicious. In this reply, you baselessly claim that I insulted you (which I did not, it is you who is insulting me). I asked for someone to reply to the specific claim I have made, as not a single person in this thread has replied to it. One person, Sage, is the only one who even mentioned my argument, and they agreed with it. Everyone else created strawmen. I can't "give an impression", as I am explicit in my messaging. Isn't it ironic, how I even bold some of my words to make it entirely clear the point I am making, and then you conveniently ignore that point and baselessly claim that I "gave an impression of trying to educate" others? I have given no such impression. My point is so clear that Sage already understood it 2 messages in. It's a shame how you continue to ignore my point and continue this nonsensical tangent. I love how you are asking 3 questions, the first being one with a false premise (another logical fallacy), and the second being a question that's already been answered (my point has been made, onus is on you to respond to it, Sage already has), and your third question is quite simply the most disrespectful thing in this whole thread. Your third question is quite literally asking me to repeat my point again. I will not. You have seen what I have said. My point is entirely clear, and that is not my opinion. The fact Sage understood it early on is proof that my point is clear. I am going to ask a third time: if you disagree with the specific point I have made, please tell me why. If not, then we can end this discussion. You continue not to respond the claim being made. Why is it that you ignore it? Your next reply should be this: "Hi Peter, now that I understand your point by rereading your message closely, I agree that repeating an opinion that ROCm already acted upon, which notifies people in the thread, is unnecessary and should be kept to a minimum. I agree that opinions are valuable and should be expressed, even repeatedly, but that after ROCm has acted upon that opinion, then only new opinions shall be expressed, not the same opinion that has already been accounted for." or "Hi Peter, after understanding your point clearly, I have to say I still disagree. I believe that repeating the same opinion, even after ROCm has already heard that opinion and acted upon it, does indeed make a difference and makes ROCm think more thoroughly about this issue, despite them already giving a timeframe, because that timeframe can be wrong, as demonstrated in the past." I would then respond to either of these messages with: "Great, thank you for reading my message and responding to my point in specificity. I thoroughly believe that ROCm is currently working on the issue based on their wording of recent messages, and so I believe the current timeframe is accurate, peace." |
You are being verbose, a lot in my view, and uselessly, and I think the only way to end this is to block you. Thus I block you. Stay well. |
|
We can now confirm that @valeriob01 is a troll! Imagine being a troll in a thread which people are subscribed to. This is not social media, I hope we all agree. Good riddance. |
this is your hypocritical and malevolent behavior that you translate on others, how dishonest you are exceptionally gifted |
Excuse me? I refrained from calling out your sheer hypocrisy in your message, and instead just called you a troll. Now that you have come back, as a hypocrite once again, I will indeed reply to your other message properly. You said I was "uselessly being verbose". Isn't this hypocritcal of you? You said that I cannot tell others when an opinion can be useless, yet you are here calling mine useless. This isn't a question for you to answer, the answer is "yes, you are being hypocritcal". (This is the reason I called you a troll. I cannot fathom how someone can contradict themselves in every message, lobbing insults in every message, and not be a troll, your behvaiour just doesn't make any sense). Edit: Not to mention, I even gave steps on how to resolve the issue, and you ignored them. I honestly thought the commotions would end and my message was gonna reach a conclusion, that you were actually going to respond the claim I made. I thought this was the end, but instead you just insulted me again and again. I am fed up with your antics. Respond to the claim I have made (that is, create a reply with substance), or don't say anything at all. In fact, this was my initial point in this whole discussion! My point is clear, it's your turn to respond to that point, not to just insult my way of speaking, or create yet more logical fallacies against me. I beg of you, in all honesty, to please respond to my point, I even gave a template of how to do so if you don't want to put the effort in (sincerely). Edit 2: This is a serious point. I don't baselessly call others hypocritcal, insulting, or malicious. I only do so with evidence. This is a standard I hold myself to. I honestly do not like it when I have to call others that, but when they are persistent, it is a must. In your latest message, you have literally called me "hypocritical" and "malevolant" and that I cannot call others that or else I am "dishonest", and then in an insulting manner called me "gifted". Ok, if I am hypocritical, show me what contradictions I have made. Please, I would like to see. I have pointed out every time explicitly when you were being hypocritical, that is fair. You have provided nothing, that is unfair, that is wrong. I don't call people hypocrites for fun, I hope you don't either. Ok, if I am malevolent, please show me how I am not trying to logically argue with others, or furthering the discussion by giving examples of how we should proceed. Please, I would like to see. I have pointed out every time explicitly when you have been logically fallacious (which is intellectual dishonesty), this is fair. You have provided nothing, that is unfair, that is wrong. I don't call people logically fallacious or malicious for fun, I hope you don't either. Do you see what I'm getting at? I am sincererly defending myself when people make strawmen against me, and sincerely calling out the insults that are made. You, however, have just insulted me but you didn't say why. I just, don't, get it. You have the balls to insult me without evidence, but not to actually continue the discussion? Let's put all this behind us, this is a nonsensical discussion. In my "verbose" reply I made to you, I explained how this discussion can end. Please, choose one of those options, and we're done, that'll be the end of it. Forgive and forget is my motto, if we can end this, I won't care about anything that happened in this discussion. Edit 3: So instead of apologizing or backing up your baseless insults, you deflect with a "workaround" that was shared here over a year ago. You cannot just randomly insult people, this is not social media. You are despicable. |
Going back to origin of this issue thread. |
This is the officially supported way for OpenCL since 2019. The question you are citing is also from 2019. Why bring it up now? |
How did you manage to install |
Mesa is not needed, amdgpu-pro installs without problems, versions needed Ubuntu 20.04 and amdgpu-pro 20.10 . Instructions: https://amdgpu-install.readthedocs.io/en/latest/install-installing.html#installing-the-pro-variant |
|
https://gitlab.freedesktop.org/agd5f/linux/-/commit/78d62bd72a18891e0851f95c7ed19fc317c9d02a <- might be a fix that hints at upcoming ROCm Support for Navi 10 :D but not sure. |
|
At the end, navi10 run successfully with patches on ROCm-5.2.0. |
|
@xuhuisheng do you build and run all the tests for the mathlibs? |
|
@cgmb Not yet, I just confirmed if we need pytorch and tensorflow run with navi10, we need rebuild 5 components at least. and i think you are right, i will find time to run testcases. |
and others
Just curious if you have any plans on supporting the 5700 Xt card in the near future. As of now I haven't seen it in the supported GPUs list. Would be nice to also have it included!
The text was updated successfully, but these errors were encountered: