Gentoo Forums
Gentoo Forums
Gentoo Forums
Quick Search: in
AMD Zen/Ryzen thread
View unanswered posts
View posts from last 24 hours

Goto page Previous  1, 2, 3 ... 18, 19, 20, 21, 22  Next  
This topic is locked: you cannot edit posts or make replies.    Gentoo Forums Forum Index Gentoo Chat
View previous topic :: View next topic  
Author Message
mir3x
Guru
Guru


Joined: 02 Jun 2012
Posts: 343

PostPosted: Mon Jul 31, 2017 8:09 pm    Post subject: Reply with quote

OK i did 3 things and havent seen any segfault or internal compiler error:
- switched to gcc 7.1
- added those CPU_FLAGS_X86="aes avx avx2 f16c fma3 mmx mmxext popcnt sse sse2 sse3 sse4_1 sse4_2 sse4a ssse3" to make.conf
( i had earlier those from core 2 duo)
- i reset dram timing to auto ( im quite sure that wasnt issue - it was set to values given by kingston earlier )

Since then i recompiled mesa more than 30 times, emerged -e system 2 times, did many newuse changes and a lot of recompiles.
Im not going to test anymore :D

I also measured power usage on my ASUS prime b350 + 1050Ti + ssd + magnetic disk:
grub screen: 145W
booting - windows (jumping from 70 to 140), linux - stable 65
windows 10 idle - mostly 50-80, rarely 35W
linux idle - jumping from 30 to 70

emerge mesa - 145 W
youtube on win and gentoo, on chrome - the same about - 70-80

psu - be quiet pure power 10 500W

EIDT: also noticed it takes 2W when comp is off
Asus prime B350 plus has no usb type c-connector for quick charge.
and has some useless graphics card built in - i would suggest some other motherboard probably
_________________
Installation aborted to prevent system self-destruction


Last edited by mir3x on Mon Jul 31, 2017 8:30 pm; edited 2 times in total
Back to top
View user's profile Send private message
Naib
Watchman
Watchman


Joined: 21 May 2004
Posts: 5073
Location: Removed by Neddy

PostPosted: Mon Jul 31, 2017 8:22 pm    Post subject: Reply with quote

interesting...

either there is some burn-in occuring or GCC-7 is when things start to get good with Ryzen. I to am on GCC-7, but I never really had issues with 6
_________________
The best argument against democracy is a five-minute conversation with the average voter
Great Britain is a republic, with a hereditary president, while the United States is a monarchy with an elective king
Back to top
View user's profile Send private message
Ant P.
Advocate
Advocate


Joined: 18 Apr 2009
Posts: 4439

PostPosted: Mon Jul 31, 2017 11:43 pm    Post subject: Reply with quote

mir3x wrote:
EIDT: also noticed it takes 2W when comp is off
Asus prime B350 plus has no usb type c-connector for quick charge.

Might be wake-on-lan. powertop or poking it with ethtool can disable that.
_________________
*.ebuild // /etc/service/*
Back to top
View user's profile Send private message
Tony0945
Veteran
Veteran


Joined: 25 Jul 2006
Posts: 1980
Location: Illinois, USA

PostPosted: Tue Aug 01, 2017 1:10 am    Post subject: Reply with quote

mir3x wrote:

EDIT: also noticed it takes 2W when comp is off

I much appreciate those measurements.

It's normal for a PS to have a parasitic draw of .5W-1.0W. When you shut if off with the button, many power supplies disconnect from the mains, but otherwise they stay connected for things like WOL. Then they draw a little waste current. 2W is a little high. My Antec Edge draws 3W but has a 1.3W LED on all the time.
Back to top
View user's profile Send private message
mir3x
Guru
Guru


Joined: 02 Jun 2012
Posts: 343

PostPosted: Tue Aug 01, 2017 6:25 am    Post subject: Reply with quote

OK I enabled Erp Ready
Code:
ErP/EuP (Energy Using Product) is a function to save power when the computer is in standby mode. This is a regulation so the total power of the system won't exceed 1.0W

THen I noticed nothing changed, then i noticed that my batteries are dead in power meter and its probably leeching some power, or maybe my power meter just sucks :D
_________________
Installation aborted to prevent system self-destruction
Back to top
View user's profile Send private message
Naib
Watchman
Watchman


Joined: 21 May 2004
Posts: 5073
Location: Removed by Neddy

PostPosted: Fri Aug 04, 2017 4:31 pm    Post subject: Reply with quote

FreeBSD have issued a patch w.r.t. their kernel to work around "ryzen hardware issue"

https://svnweb.freebsd.org/base?view=revision&revision=321899

Quote:
I did find what I believe to be a
hardware issue with Ryzen related to concurrent operations. In a
nutshell, for any given hyperthread pair, if one hyperthread is
in a cpu-bound loop of any kind (can be in user mode), and the
other hyperthread is returning from an interrupt via IRETQ, the
hyperthread issuing the IRETQ can stall indefinitely until the
other hyperthread with the cpu-bound loop pauses (aka HLT until
next interrupt).

_________________
The best argument against democracy is a five-minute conversation with the average voter
Great Britain is a republic, with a hereditary president, while the United States is a monarchy with an elective king
Back to top
View user's profile Send private message
Tony0945
Veteran
Veteran


Joined: 25 Jul 2006
Posts: 1980
Location: Illinois, USA

PostPosted: Fri Aug 04, 2017 10:33 pm    Post subject: Reply with quote

Naib, this would mean that the Ryzen 3's are immune to the bug because there is no SMT. I had wanted to buy an 1800X but the bug and lack of cooler deterred me. This seems that I can buy a $129 Ryzen 3 1300X, still have a performance boost over my Athlon II X3 440 and buy a Ryzen 2+ later when they fix the bug, assuming they keep the AM4 socket. I've already read that they are abandoning AM4 for the new Ryzen APU's that now are due out next year. I smell a company in trouble. Head designer leaves before launch, schedules keep getting pushed back, website just full of Madison avenue hype rather than engineering information.

I don't think that SMT adds that much to Linux anyway, unlike the fat bloated excuse for an OS called Windoze.
Back to top
View user's profile Send private message
mir3x
Guru
Guru


Joined: 02 Jun 2012
Posts: 343

PostPosted: Sat Aug 05, 2017 8:06 am    Post subject: Reply with quote

Tony0945 wrote:
Naib, this would mean that the Ryzen 3's are immune to the bug because there is no SMT. I had wanted to buy an 1800X but the bug and lack of cooler deterred me. This seems that I can buy a $129 Ryzen 3 1300X, still have a performance boost over my Athlon II X3 440 and buy a Ryzen 2+ later when they fix the bug, assuming they keep the AM4 socket. I've already read that they are abandoning AM4 for the new Ryzen APU's that now are due out next year. I smell a company in trouble. Head designer leaves before launch, schedules keep getting pushed back, website just full of Madison avenue hype rather than engineering information.

I don't think that SMT adds that much to Linux anyway, unlike the fat bloated excuse for an OS called Windoze.


AMD is always in trouble :D, ppl said it happens also with SMT disabled.
Maybe you dont new CPU at all now ? 1300X is not just big boost.
I was using core 2 for 7 years, in fact i could still use it. I had hard freezes on it when using torrent, then I just blocked whole china, afhani via iptables and freezes stopped :D. Now i still didnt blocked them yet.
Few simple rules to keep it fast:
- for office use wps office (links might be dead sometimes, but that superfast and nice office, u dont need to install it, just run from unpacked dir)
- never compile chrome or firefox
- for your custom compiling use ccache always
- for mem leaks or bug hunting dont use valgrind - use sanitizers instead
- you can try using zapcc for your custom compiles ( I remeber I changed header once and it recompiled it instantly in no time,when clang/gcc would recompile 10 extra files - they release 30 day trial version every few months, later u can run it with combo libfaketime:D, if u send them some fail compile u will get free version for opensource use), but overall probably not worth maybe,
i stucked once on some recompile bug, it took some time i need to make clean, so all time benefit was lost :D
- dont sync more than once per 2 weeks - less to update :D
- OS is as great as running software, and windows has many unique soft for taxes and bookkeeping .. and more. U can buy w10 key
for 2usd (ppl sell as 100% legal - I never had issue with such key )
_________________
Installation aborted to prevent system self-destruction
Back to top
View user's profile Send private message
Tony0945
Veteran
Veteran


Joined: 25 Jul 2006
Posts: 1980
Location: Illinois, USA

PostPosted: Tue Aug 08, 2017 2:01 pm    Post subject: Reply with quote

mir3x wrote:
Maybe you dont new CPU at all now ? 1300X is not just big boost.


After Phoronix's tests of Ryzen 3 came out. I installed phoronix-test-suite from overlay, built it and ran the test suite indicated in the article. The first test would not run because of an error in the command, leading me to wonder if the Ryzen 3 was tested against 7.2.1 (stable) or 7.4 (testing). All the other tests ran on my Athlon II X3. {Can anyone tell me how to quote from a spreadsheet without having embeded tabs which are screwing up formatting?}

I then constructed (via spreadsheet) this list of relative performance (with Athlon II X3 as 1):

Code:
FX-8370   ------  3.12   
1200        ------  2.96   
1300X      ------  3.27
1400        ------  3.37
1700        ------  5.01
1800X      ------  5.61


Added the code tags and made a mess of the formatting. -- Neddyseagoon :)

Sorry about formatting, I've spent ten minutes trying to get neat columns and am just giving up.


A tripling of performance seems worthwhile, especially if the mobo, memory, and video card can be reused with a later Zen generation. I did this a long time ago with Athlon64, replacing the chip with an Athlon64 X2. Yes, I could nearly double performance again for about $400, but I'm having a hard time figuring if the monster aftermarket coolers will fit my case and I don't want to spend that money for a CPU that segfaults! Buying the cheapest Ryzen and then recycling the DDR4 memory and video card if I have to switch to Intel seems a reasonable gamble. Especially when SMT seems to be a big part of the problem.
Back to top
View user's profile Send private message
mir3x
Guru
Guru


Joined: 02 Jun 2012
Posts: 343

PostPosted: Tue Aug 08, 2017 3:48 pm    Post subject: Reply with quote

Code:

FX-8370 --- 3.12   
1200 ------ 2.96   
1300X ----- 3.27
1400 ------ 3.37
1700 ------ 5.01
1800X ----- 5.61


U must use "code" tag - it should use monospace font then.

EDIT: heheh, I guess u r not programmer then , anyway performance per 1 core might be slower, youtube wont be faster :D
_________________
Installation aborted to prevent system self-destruction
Back to top
View user's profile Send private message
Tony0945
Veteran
Veteran


Joined: 25 Jul 2006
Posts: 1980
Location: Illinois, USA

PostPosted: Tue Aug 08, 2017 8:07 pm    Post subject: Reply with quote

Code tag didn't work for copying and pasting the spreadsheet. I'll take screenshots and post the pictures.
Back to top
View user's profile Send private message
Tony0945
Veteran
Veteran


Joined: 25 Jul 2006
Posts: 1980
Location: Illinois, USA

PostPosted: Tue Aug 08, 2017 8:25 pm    Post subject: Reply with quote

mir3x wrote:
anyway performance per 1 core might be slower

Slower than budget processor ($50) from February 2010? I doubt it!
Back to top
View user's profile Send private message
Tony0945
Veteran
Veteran


Joined: 25 Jul 2006
Posts: 1980
Location: Illinois, USA

PostPosted: Tue Aug 08, 2017 9:29 pm    Post subject: Reply with quote

Spreadsheets:

Test results https://ibb.co/jPzWMa
Relative ranking https://ibb.co/gV2bMa
Back to top
View user's profile Send private message
thumper
Guru
Guru


Joined: 06 Dec 2002
Posts: 506
Location: Venice FL

PostPosted: Wed Aug 16, 2017 9:34 pm    Post subject: Reply with quote

When this (hopefully soon) makes it into Gentoo proper it will be interesting to see how it changes if any the Ryzen game of quirks.
https://phoronix.com/scan.php?page=news_item&px=GCC-7.2-Released
Back to top
View user's profile Send private message
Bloot
Tux's lil' helper
Tux's lil' helper


Joined: 10 Mar 2006
Posts: 90
Location: Barcelona

PostPosted: Thu Aug 17, 2017 9:13 am    Post subject: Reply with quote

What compiler are you guys using? I'm on 6.3 but sometimes compiler fails, retrying again it compiles without issues though. Weird.
_________________
Portage 2.3.6 (python 3.4.5-final-0, default/linux/amd64/13.0/desktop/plasma, gcc-6.4.0, glibc-2.23-r4, 4.12.5-gentoo x86_64)
Back to top
View user's profile Send private message
Tony0945
Veteran
Veteran


Joined: 25 Jul 2006
Posts: 1980
Location: Illinois, USA

PostPosted: Thu Aug 17, 2017 12:32 pm    Post subject: Reply with quote

Bloot wrote:
What compiler are you guys using? I'm on 6.3 but sometimes compiler fails, retrying again it compiles without issues though. Weird.
I ran those benchmarks with gcc 6.4.0 which is a bugfix update to 6.3

my glibc is 2.23-r4 recompiled with gcc 6.4

https://lists.gnu.org/archive/html/info-gnu/2017-07/msg00002.html
https://www.phoronix.com/scan.php?page=news_item&px=GCC-6.4-Released
https://packages.gentoo.org/packages/sys-devel/gcc
Back to top
View user's profile Send private message
Bloot
Tux's lil' helper
Tux's lil' helper


Joined: 10 Mar 2006
Posts: 90
Location: Barcelona

PostPosted: Thu Aug 17, 2017 1:27 pm    Post subject: Reply with quote

Tony0945 wrote:
Bloot wrote:
What compiler are you guys using? I'm on 6.3 but sometimes compiler fails, retrying again it compiles without issues though. Weird.
I ran those benchmarks with gcc 6.4.0 which is a bugfix update to 6.3

my glibc is 2.23-r4 recompiled with gcc 6.4

https://lists.gnu.org/archive/html/info-gnu/2017-07/msg00002.html
https://www.phoronix.com/scan.php?page=news_item&px=GCC-6.4-Released
https://packages.gentoo.org/packages/sys-devel/gcc


Thanks, I'll try 6.4 :)
_________________
Portage 2.3.6 (python 3.4.5-final-0, default/linux/amd64/13.0/desktop/plasma, gcc-6.4.0, glibc-2.23-r4, 4.12.5-gentoo x86_64)
Back to top
View user's profile Send private message
Naib
Watchman
Watchman


Joined: 21 May 2004
Posts: 5073
Location: Removed by Neddy

PostPosted: Thu Aug 17, 2017 1:37 pm    Post subject: Reply with quote

7.1 is lovely btw
_________________
The best argument against democracy is a five-minute conversation with the average voter
Great Britain is a republic, with a hereditary president, while the United States is a monarchy with an elective king
Back to top
View user's profile Send private message
Bloot
Tux's lil' helper
Tux's lil' helper


Joined: 10 Mar 2006
Posts: 90
Location: Barcelona

PostPosted: Thu Aug 17, 2017 1:40 pm    Post subject: Reply with quote

Naib wrote:
7.1 is lovely btw


Is it safe to use it? It's hard masked :?
_________________
Portage 2.3.6 (python 3.4.5-final-0, default/linux/amd64/13.0/desktop/plasma, gcc-6.4.0, glibc-2.23-r4, 4.12.5-gentoo x86_64)
Back to top
View user's profile Send private message
Naib
Watchman
Watchman


Joined: 21 May 2004
Posts: 5073
Location: Removed by Neddy

PostPosted: Thu Aug 17, 2017 1:42 pm    Post subject: Reply with quote

Its hard masked because not everything builds with it.
The issue is gcc has become alot stricter but as long as the code is valid, the build is good
5.x to 6.x has some ABI annoyances, 6.x to 7.x does not (ie smoother transition if the packages have been updated).
Inkscape was the only one that I needed a user patch for
_________________
The best argument against democracy is a five-minute conversation with the average voter
Great Britain is a republic, with a hereditary president, while the United States is a monarchy with an elective king
Back to top
View user's profile Send private message
Bloot
Tux's lil' helper
Tux's lil' helper


Joined: 10 Mar 2006
Posts: 90
Location: Barcelona

PostPosted: Thu Aug 17, 2017 1:49 pm    Post subject: Reply with quote

Naib wrote:
Its hard masked because not everything builds with it.
The issue is gcc has become alot stricter but as long as the code is valid, the build is good
5.x to 6.x has some ABI annoyances, 6.x to 7.x does not (ie smoother transition if the packages have been updated).
Inkscape was the only one that I needed a user patch for


I'll give it a try, thanks mate :)
_________________
Portage 2.3.6 (python 3.4.5-final-0, default/linux/amd64/13.0/desktop/plasma, gcc-6.4.0, glibc-2.23-r4, 4.12.5-gentoo x86_64)
Back to top
View user's profile Send private message
thumper
Guru
Guru


Joined: 06 Dec 2002
Posts: 506
Location: Venice FL

PostPosted: Thu Aug 17, 2017 8:28 pm    Post subject: Reply with quote

GCC 7.1 has been great for me except for the firefox/thunderbird issue, but there is as manual workaround.

As for segfaults, they were non existent until the latest bios update, turning X-AMP off seems to have solved it.

Only problem with all auto settings seems to be a very rare stuck thread.

I may be one of the lucky ones (knock wood)

George
Back to top
View user's profile Send private message
Chewi
Developer
Developer


Joined: 01 Sep 2003
Posts: 856
Location: Edinburgh, Scotland

PostPosted: Mon Aug 21, 2017 9:03 pm    Post subject: Reply with quote

I was seeing a very slow but sure memory leak on 4.11, probably from the experimental AMD DC patches I am using. I figured it was time to update to 4.12 so I did that yesterday but I kept running into segfaults and ICEs while trying to build virtualbox among other things. When it happened with gcc-6.4.0, I knew something was up. I dropped back to 4.11 and virtualbox built first time.

ASLR is definitely still disabled and the kernel configuration has barely changed. Everything was working perfectly for a long time before. Has anyone else seen new issues following this update?
Back to top
View user's profile Send private message
mir3x
Guru
Guru


Joined: 02 Jun 2012
Posts: 343

PostPosted: Sun Aug 27, 2017 1:03 pm    Post subject: Reply with quote

I tried gcc 6.4 and 7.2 and im getting internal compilers errors there.
( I dont get any when using 7.1 ) ...
_________________
Installation aborted to prevent system self-destruction
Back to top
View user's profile Send private message
Chewi
Developer
Developer


Joined: 01 Sep 2003
Posts: 856
Location: Edinburgh, Scotland

PostPosted: Sun Aug 27, 2017 1:07 pm    Post subject: Reply with quote

mir3x wrote:
I tried gcc 6.4 and 7.2 and im getting internal compilers errors there.
( I dont get any when using 7.1 ) ...

Interesting, I ran into an issue that was broken with 7.1 but worked with 7.2. What a mess! Just to add to the fun, after finding issues with 4.12, I updated to the latest 4.11 and have started seeing segfaults there too. This is starting to get very annoying.

I also hear from Phoronix that newer chip revisions don't have these problems. If they don't find a way around these issues soon, they could be looking at mass replacements.
Back to top
View user's profile Send private message
Display posts from previous:   
This topic is locked: you cannot edit posts or make replies.    Gentoo Forums Forum Index Gentoo Chat All times are GMT
Goto page Previous  1, 2, 3 ... 18, 19, 20, 21, 22  Next
Page 19 of 22

 
Jump to:  
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum