Best gpu that works with free software?
- Vous devez vous identifier ou créer un compte pour écrire des commentaires
i am currently using integrated graphics in my pc and i am wanting
to upgrade to a graphics card(for video editing)
i have seen thinkpenguins GeForce 8400GS 1GB PCI Express 2.0 Video Card
but this seems to be a buget card and i am looking for somthing more powerfull.
dose anyone know of a card like this:
http://www.maplin.co.uk/p/asus-nvidia-strix-gtx-970-oc-4gb-gddr5-graphics-card-a26rr
that will work with free software?
Every GPU that isn't an Intel integrated GPU only gets libre software support through reverse-engineering, if at all. Some more recent Nvidia cards do work well in Nouveau; check the Nouveau project's website to start off.[1][2] Just make sure your research is thorough.
[1] http://nouveau.freedesktop.org/wiki/CodeNames/
[2] http://nouveau.freedesktop.org/wiki/FeatureMatrix/
I am a translator!
based on h-node i think i have found the card i am looking for
https://h-node.org/videocards/view/en/1001/NVIDIA-Corporation-Device-1183--rev-a1-/1/1/undef/undef/works_with_3D/PCI-E/undef/undef
http://www.maplin.co.uk/p/palit-geforce-gtx-660-ti-2gb-980mhz-gddr5-2gb-6008mhz-graphics-card-n83nn
but how reliable is the info on h-node?
is there anyway of ensuring that the card will work well?
I am a translator!
as far as i know they test all the hardware listed there on several libre distros - so yes you can be pretty sure that it is gonna work.
thanks for the help!
I am a member!
The issue with picking out a graphics card is a lot more complicated than can be easily represented by a “works/doesn't work” graph or by “supported/not supported” list such as with h-node- particularly where the free nouveau driver is concerned. The nouveau driver has better support for the older graphics chips and I'd caution the use of newer cards. There is basic support from what I understand on the newer cards, but it has some way to go yet to being adequate.
There are a variety of factors and simply saying "it works" isn't necessarily going to be sufficient to telling you how well it works or doesn't work. Does working mean that there is video acceleration support? Does working mean that there is support for controlling the fan- or will it overheat should you actually try and use it? Does working mean the fan is set to full blast or is there support for controlling it in software? Does it have to be manually set?
There are also other things to be concerned about like regressions. We have switched cards a few times for reasons of improving the support situation. Not all cards are created equal. Even among 8400GS cards there are some cards that are better than others. We're switching 8400GS cards right now because we realized that the support is not quite what it could be and a slight change would improve the support. It's not a critical issue, but the improved support should make some people happy.
so we cant rely get a modern gpu that works well if we want to use free software?
I am a member!
Yes- spot on. It's either you get something older, you go integrated, or you go without. Quite frankly the integrated stuff works just fine for the majority. The people who need something more are those who play 3d intense games. Despite what a lot of gamers- and even people in general think 3d isn't that critical for most people. If your circle of friends are young and into gaming... well... your view is going to be a little different. However the large majority of people do not play games. Is gaming a growing market segment? Yea. However it is still not the majority.
I hear you on this one. I have a machine with a GeForce GT 325M, h Node says it is compatible but in use it is very unstable and performance is dire. Shame really.
I am a translator!
yes. you can.
yes we can !
:)
so how can we use modern gpu's
you mean develope the drivers?
i am not a gpu expert.
The GTX 760 works with 3D acceleration: https://h-node.org/videocards/catalogue/en/1/1/undef/2013/works_with_3D/PCI-E
This is a card from 2013 though... makes you wonder if/when the Nvidia GTX 970 and 980 cards make this list.
I am a member!
There isn't enough information to go on though in that post. There just are too many problems with the newer cards. If your system crashes every half hour, but it works with 3D, and performs better, is that acceptable?
Simply running glxgears isn't really adequate and yet thats the kind of test results you get with h-node and similar databases. This is why I'm not a fan of such databases. They're too unreliable and the vast majority do not want to have to order a part to begin with let alone return and repeat until they find something that works.
I would much rather see a more thorough set of standards developed by an independent body. There aren't that many people I'd consider technically competent enough to do it though. Quite frankly there is one person I can think of who I'd really consider pretty knowledgeable within the community and he already does hardware testing and reviews (magazine e-zine format). He does not seem to consider free software though a high priority unfortunately.
I think the free distros should come with a program that would tell the user whether their hardware works in freedom or not ( https://trisquel.info/en/issues/2124 ) and it could send the same report to h-node automagically. This would really, really help with the sample size which would make the data so much more reliable and increase vastly the set of devices listed. (I don't know whether it can be programmatically detected whether for example 3D works though but I hope so.)
I get an "Access denied" error when trying to view that link :(
same
> I get an "Access denied" error when trying to view that link :(
This link? https://trisquel.info/en/issues/2124
o_O
I am a member!
It won't work well. Any kind of automated test is going to have a lot of problem areas.
3D Examples:
Reports back, “WORKING”, but crashes every 30 minutes.
Reports back, “WORKING”, but is extremely loud, because the fan doesn't work properly (auto-sets to high)
Reports back, “WORKING”, but only because the user installed a non-free blob manually
Reports back, "WORKING", but overheats if you actually use it
Wireless Examples:
Reports back, “WORKING”, but you can't transfer more than 500MB before crashing
Reports back, “WORKING”, but the bluetooth portion of the card is dependent on non-free software
Reports back, “WORKING”, but the bluetooth only worked because the user has MS Windows/Ubuntu/or some other OS w proprietary software loading it
Reports back, “WORKING”, but you can't connect due to digital restriction or on/off button issue
56K Dial-up Modems:
Reports back, “WORKING”, but doesn't mention the fact fax doesn't work
Reports back, “WORKING”, but fails to mention a variety of other features don't work
Reports back, “NOT WORKING”, but actually would if only the distribution had included the right permissions so the modem was accessible to the test application
Reports back, “NOT WORKING”, but only because it didn't get the port right, auto-detect actually picks it up, but hey- we're just a simple test app
Printers:
Reports back, “WORKING”, but the machine is a multi-function device without support for scanning despite having scanning capability
Reports back, “NOT WORKING”, but actually does, your just missing some piece of software which has to be installed from the repository
Those are just a handful that come to mind. There are lots of other issues I probably couldn't even begin to think up. While you can certainly automate testing this stuff needs more testing that is likely going to be done by a simple test application.
This certainly is true and a good point. However many of those are also applicable to manual reporting as we see with the AMD cards that supposedly can do 3D. The data is simply not reliable. The automated reports would have to be kept strictly separate from the manual reports. However I think it would still provide useful clues as to what hardware to look at closer, especially to people like you Chris. It certainly would not be a simple app to build. It's just that hardware seems like one of the biggest Achilles' heels for free software so it might be worth the trouble.
I am a member!
Yes- these are problems that would be true of manual reporting. This is why we need people with the expertise focused on these issues. Even when you have experts focusing on the issue it's not a guarantee that there won't be any problems. However it can significantly reduce the likelihood and difficulty of your average user (and technical user alike) running into said problems- or being able to get the part- or being able to get support for the part under the OS they use.
The reporting databases can be useful, but they are more or less useful in helping people with the expertise, time, energy, and resources to to narrow down and/or put something better together. If your working for a huge company with lots of money and time to do your own in-house testing they work reasonably well as a tool in helping even if its not a total solution. You can overnight parts (depending on where they are coming from, ie could be 5-10 days to expedite from China to the US for instance) from different companies, test, etc. It's not a guarantee that those parts won't change though by the time you need to get 1000 of part X or that part X will actually be the same throughout the stock you do get (as stocks are mixed @ warehouses so you could test 10 of part X and then order a 1000, but end up getting 300 of the part you tested, and 700 of some part you didn't, and won't even work for you).
We get frustrated all the time by companies who change products on us. They don't change the model, but they do change whats inside. Right now I'm trying to secure stocks of one particular small part that hasn't changed in several years- but now all-of-a-sudden has. I'm still unsure if we're going to be successful because of the mixed-stock issue. There is nothing that identifies the part as being different so you can't just ask “does the stock list v2” on the box somewhere. Unfortunately there really aren't any other good options for this particular part (that said half the time nobody can see the physical stock- unless-maybe your ordering 1000 of something- and even then sometimes not). We've already gone and done a lot of research, testing, etc. and not much has really changed. If we are forced to change parts it'll probably result in a significant increase in the cost which would be really bad as whats a relatively cheap item in our catalog right now would become significantly more expensive without any real benefits. I'm not even sure if after all that it would be a good solution either compared to what we have now. What we have now is excellent- but most other parts which might work are junk. At least within the same price range.
- Vous devez vous identifier ou créer un compte pour écrire des commentaires