<warpme_>
this conforms with DT config about PM3 for pwrseq.
<warpme_>
i catch boot with working wifi and then set PM3 to 0. this breaks wifi so it looks like PM3 is correct for wifi
<warpme_>
still don't know why on majority boots PM3 is 0 !
<apritzel>
warpme_: can you change GPIO_ACTIVE_LOW to GPIO_ACTIVE_HIGH?
Mangy_Dog has joined #linux-sunxi
<warpme_>
huh - good catch!. with this PM3 is consistently high after every reboot. so progress. unfortunately still majority boots are with "mmc1: error -22 whilst initialising MMC card". now it started looking to me like wifi on this box uses 3 I/O: regon, reset & wake-up and we miss regon. but why gpios regs are the same for working and non-working wifi. maybe issue is with timmings? i.e. pwrseq is too soon after box power-on (so
<warpme_>
reg not yet provide correct voltage to sdio interface)?
vpeter has quit [Remote host closed the connection]
<MoeIcenowy>
this change to power sequence is needed for R329 to correctly power up secondary cores
<MoeIcenowy>
and it looks reasonable (because it's the reversal of the power down sequence)
<MoeIcenowy>
I just wonder whether this will work on A64/H5/H6/H616
<apritzel>
MoeIcenowy: thanks, will have a look
<MoeIcenowy>
the ifdef above can be ignored, because it's dealing with the change of AA32nAA64 position on R329
<apritzel>
yeah, you should just define SUNXI_CPUCFG_CLS_CTRL_REG0 differently (or rename that), and have some macro for the bit location
vpeter has joined #linux-sunxi
hlauer has joined #linux-sunxi
<MoeIcenowy>
apritzel: nope, the register still exists well, only the AA32nAA64 part is moved out
pg12_ has quit [Ping timeout: 480 seconds]
<apritzel>
oh well, our beloved SoC vendor again ... I guess we use a macro pointing to either of them then
<apritzel>
MoeIcenowy: are there more differences? How close is that to the H616 otherwise?
<Mangy_Dog>
a question for all you ARM experts here :D What is it going to take to get arm based cpus or socs to get up to the performance of current high end desktops? im talking top end ryzens and i9s. and heck even xeons and epipcs.... Whats missing? Are they currently able to crunch out the numbers as well as current CPUs if clocked at the same rate? Or is the different instruction sets a big
<Mangy_Dog>
hinderance (or gain) for it?
<Mangy_Dog>
im just wondering when ill be seeing an arm based desktop cpu (proper not the M1 PR stunt) running vray, matching or beating current desktops
<apritzel>
Mangy_Dog: the actual instruction set has only a relatively small influence on the actual performance
<Mangy_Dog>
the reason i ask... is simply. if ARM chips got onto the desktop space and managed to realistically compete with x86. it would be amazingly good for competition
<apritzel>
Mangy_Dog: it's just that no one apart from Apple wants to really build a desktop class Arm CPU
<Mangy_Dog>
well tbh
<Mangy_Dog>
even apple doesnt want to build a desktop class cpu
<Mangy_Dog>
the m1 still isnt really desktop level
<Mangy_Dog>
its still in essence a low powered mobile chip
<apritzel>
that kicks ass
<Mangy_Dog>
in some very few select benchmarks
<Mangy_Dog>
against mid range cpus
<Mangy_Dog>
im not against ARM just m1 really was a big pr stunt and extreamly cinical attempt to control production and hamper right to repair stuff
<apritzel>
the performance of a CPU is mostly set by the micro-architecture, and by its "size" (number of cores, TDP)
<Mangy_Dog>
but like i say.... can a 4ghz 12 core (is multi threading possible in arm?) 24 thread ARM chip complete with a 5900x?
<apritzel>
what Arm chip? There is no such thing
<Mangy_Dog>
well thats why im asking
<Mangy_Dog>
is there anything that can be theorhetocally made?
<apritzel>
sure
<Mangy_Dog>
or is arm limited to a niche (sort of) use case
<apritzel>
the M1 is a good example, it's just not scaled up (yet)
<Mangy_Dog>
windows 11 is apparantly meant to suport arm
<Mangy_Dog>
im just curious to see when ill be seeing high end pc based desktop software using it nativly. And performing and properly competing with x86
<apritzel>
my desktop is a dual socket, 32-core, 4 thread (=> 256 CPUs) "Arm" box (Cavium ThunderX2), that works very well
<apritzel>
I guess people don't really want Windows, they want off-the-shelf Windows applications, which are x86-64
<Mangy_Dog>
well like i say
<Mangy_Dog>
i want to see it competing in that space
<libv>
apritzel: did cavium not die or get bought up or something?
<apritzel>
they are Marvell now, yes
<libv>
or did one of the arm server companies actually make it?
<Mangy_Dog>
currently we only have 2 companies currently in the commercial desktop space
<libv>
is marvell continuing work on a server class chip?
<libv>
Mangy_Dog: and even that was almost down to 1
<Mangy_Dog>
it be nice to see arm come up and the tens + companies doing work in that space
<Mangy_Dog>
ARM are pretty much wipping intel at the moment
<Mangy_Dog>
but i dont see either company disapearing any time soon
<Mangy_Dog>
umm
<Mangy_Dog>
AMD not ARM :D
<libv>
they were not, from the second they bought ati, and ati dragged the company down big time
<Mangy_Dog>
ATI is a problem for there graphics side
<Mangy_Dog>
but the cpus are
<Mangy_Dog>
really great
<Mangy_Dog>
the whole ryzen and zen push has really done amazing things for them
<libv>
especially since they had to choose between investing in the still working graphics side which was more profitable when the financial crash of 2008 happened
psydroid[m] has joined #linux-sunxi
<libv>
and the cpu side, which was suffering from the catch-up that core was
<apritzel>
there is simply no real economical incentive or market for a company to build an Arm desktop
<Mangy_Dog>
really?
<Mangy_Dog>
there isnt?
<libv>
people hyped the arm desktop and server since back in 2011
<libv>
and it never went anywhere
<libv>
the reason why it might happen with risc-v is china.
<apritzel>
Mangy_Dog: because it means you have to compete against the big and long running x86 desktop market
<Mangy_Dog>
this is kinda what im getting at, theres tons of chip makers out there now
<libv>
it feels like it was a backup strategy in case zen hit the wall
<apritzel>
Mangy_Dog: try to convince someone to solder an Octeon 10 on a micro-ATX board
<Mangy_Dog>
they could just make a socketed chip :p
<apritzel>
that's not the problem, you would still need a board designed for that platform
<Mangy_Dog>
well clearly there would need to be some standards set
<apritzel>
standards?
<Mangy_Dog>
but thats not something thats out the question if they were willing
<Mangy_Dog>
Like set sockets types
<Mangy_Dog>
to limit fracturing
<Mangy_Dog>
if the various arm based cpu makers banded together as a consortium... to compete themselves they could work on a standard of pin outs and socket types
<apritzel>
while socketed chips would be nice, I don't think they are really needed (initially)
<Mangy_Dog>
not tooooo disimular to intel or amd sticking to a socket for a gneration or two... or three
<apritzel>
socketed CPUs from different SoC vendors will never work, that's basically against the whole idea of the Arm ecosystem
<apritzel>
for the rest of the systems there are already standard (including software), see the Honeycomb board
MangyDog has quit [Remote host closed the connection]
Mangy_Dog has joined #linux-sunxi
cmeerw has joined #linux-sunxi
tnovotny_ has quit []
megi has quit [Quit: WeeChat 3.2]
hlauer has quit [Remote host closed the connection]
megi has joined #linux-sunxi
ftg has joined #linux-sunxi
<cyrozap>
Mangy_Dog: To directly answer your question from earlier, there's no technical reason why a company couldn't produce HEDT-class ARM CPUs. e.g., AWS has their custon Graviton2 servers, which are price-competitive with their AMD and Intel instances.
<Mangy_Dog>
fair
<cyrozap>
And as for mass-adoption, it seems Apple's decided to go full-ARM, so they'll keep improving their ARM chips until they become HEDT-grade. I know you think of the M1 as a publicity stunt, but it's a lot easier to scale-up a high-efficiency CPU and turn it into a high performance CPU than it is to scale down a high-performance CPU to be high-efficiency. Which is pretty much why Intel failed to make
<cyrozap>
competitive phone/tablet SoCs, and why ARM chips are now being found in servers.
<cyrozap>
On the Windows side, Microsoft has been doing a lot with x86-64 emulation, too, so for consumers I can imagine AArch64 Windows laptops becoming the option of choice for consumers who want the cheapest devices possible.
<cyrozap>
And then eventually, if Qualcomm and Samsung actually make good ARM chips (big "if"), then maybe the Windows AArch64 SoCs might make their way to becoming HEDT-grade.
<cyrozap>
But I think at this point, for both Apple and Microsoft systems, thanks to emulation the software dependence on x86-64 is no longer the primary obstacle to adoption.
vagrantc has joined #linux-sunxi
vagrantc has quit []
vagrantc has joined #linux-sunxi
vagrantc has quit [Ping timeout: 480 seconds]
vagrantc has joined #linux-sunxi
vagrantc_ has joined #linux-sunxi
vagrantc has quit [Ping timeout: 480 seconds]
vagrantc_ is now known as vagrantc
vagrantc_ has joined #linux-sunxi
vagrantc has quit [Ping timeout: 480 seconds]
vagrantc_ is now known as vagrantc
Daanct12 has joined #linux-sunxi
Danct12 has quit [Ping timeout: 480 seconds]
Daanct12 is now known as Danct12
cmeerw has quit [Ping timeout: 480 seconds]
vagrantc_ has joined #linux-sunxi
vagrantc has quit [Ping timeout: 480 seconds]
vagrantc_ is now known as vagrantc
vagrantc has quit [Quit: leaving]
<karlp>
what's hedt?
<karlp>
(if it's soething other than hi end desktop, let me know, otherwise I'll go with that :)