On 06/13/2019 02:56 AM, BorgLabs - Kate Draven wrote:
HI =20 I would like everyone's opinion on this. =20 I'm trying figure out the benefits of either staying with the LTS kerne=
l or=20
with the lastest kernel. The machines are every day use and stability i=
s=20
important.=20 =20 Am I tossing away any benefits, of the latest kernel, if I use the 4.8x=
/9x=20
kernel. Or do the benefits of the 5.1x kernel out weigh any instability=
?=20
=20 I'd like all schools of thought. =20 Thanks in advance, =20 Kate
Kate,
Unless you have super-new bleeding-edge hardware that needs a new featu= re added in 5.1 that is not available in previous versions -- then 5.1 provi= des absolutely no benefit. Any tweak that 5.1 provided to help with Spectre performance mitigation, etc.. will likely be backported and in a LTS kern= el.
I have Arch (that always runs the current upstream version of the kerne= l, 5.1.9 currently), and Arch also provides an LTS kernel using 4.19. I have= a SuSE leap 42.3 install running the 4.4 kernel, SuSE leap 15.0/15.1 instal= ls with the 4.12 version, I have a Pi running Debian/jessie with the 4.9 ARM kernel, and from a general computing/feature/functionality standpoint, it makes no difference.
Now if you have bleeding-edge hardware that is only supported in the la= test greatest kernel -- then yes, there is a difference, otherwise you won't k= now the difference.
HTH
--=20 David C. Rankin, J.D.,P.E.
Thanks David,
This is my thinking as well. I have no real bleeding edge tech, I tend to stay away from it. Just wanted to challenge me decision , in case I was wrong.
Kate
On Monday 17 June 2019 12:10:55 am BorgLabs - Kate Draven wrote:
On 06/13/2019 02:56 AM, BorgLabs - Kate Draven wrote:
HI =20 I would like everyone's opinion on this. =20 I'm trying figure out the benefits of either staying with the LTS kerne=
l or=20
with the lastest kernel. The machines are every day use and stability i=
s=20
important.=20 =20 Am I tossing away any benefits, of the latest kernel, if I use the 4.8x=
/9x=20
kernel. Or do the benefits of the 5.1x kernel out weigh any instability=
?=20
=20 I'd like all schools of thought. =20 Thanks in advance, =20 Kate
Kate,
Unless you have super-new bleeding-edge hardware that needs a new featu= re added in 5.1 that is not available in previous versions -- then 5.1 provi= des absolutely no benefit. Any tweak that 5.1 provided to help with Spectre performance mitigation, etc.. will likely be backported and in a LTS kern= el.
I have Arch (that always runs the current upstream version of the kerne= l, 5.1.9 currently), and Arch also provides an LTS kernel using 4.19. I have= a SuSE leap 42.3 install running the 4.4 kernel, SuSE leap 15.0/15.1 instal= ls with the 4.12 version, I have a Pi running Debian/jessie with the 4.9 ARM kernel, and from a general computing/feature/functionality standpoint, it makes no difference.
Now if you have bleeding-edge hardware that is only supported in the la= test greatest kernel -- then yes, there is a difference, otherwise you won't k= now the difference.
HTH
--=20 David C. Rankin, J.D.,P.E.
Thanks David,
This is my thinking as well. I have no real bleeding edge tech, I tend to stay away from it. Just wanted to challenge me decision , in case I was wrong.
The only time it really counts is where you might need millisecond control of a valve. For your stuff, which sounds like a big distributed farming operation, I suspect one second, up to 20 seconds to open/close a valve is essentially a never mind as long as it can be done at whatever temp might be ambient for the valve at the time.
Then there are occasionally preempt-rt kernels.
This install from a testing version of the LCNC iso, is debian stretch based and has a 4.9.0-9-rt-amd64 kernel, packaged as 4.9.168-1+deb9u2 (2019-05-13) Except for the kernel substitution, its stretch 9.8.
It is the lowest latency kernel by at least a magnitude I've every ran a latency-test on, under 20 microseconds, which for this old slow phenom, is downright amazing. I could even run software stepping on it if I wasn't in a hurry. But as far as a routine file copy, its no faster at moving gigabytes around than a stock kernel. But as has been said, unless you have bleeding edge hardware, you will not see a diff. And bleeding edge today, means its some variation of an arm cpu. Sure, there's now a 64 core threadripper rizen cpu's out there from amd, but at the price per, around 3G's a socket, I suspect only going into supercomputers paid for with taxpayer sheckles. So that's not a concern to you or I.
Kate
To unsubscribe, e-mail: trinity-users-unsubscribe@lists.pearsoncomputing.net For additional commands, e-mail: trinity-users-help@lists.pearsoncomputing.net Read list messages on the web archive: http://trinity-users.pearsoncomputing.net/ Please remember not to top-post: http://trinity.pearsoncomputing.net/mailing_lists/#top-posting
Cheers, Gene Heskett
On Mon, 17 Jun 2019 04:58:16 -0400 Gene Heskett gheskett@shentel.net wrote:
On Monday 17 June 2019 12:10:55 am BorgLabs - Kate Draven wrote:
And bleeding edge today, means its some variation of an arm cpu. Sure, there's now a 64 core threadripper rizen cpu's out there from amd, but at the price per, around 3G's a socket, I suspect only going into supercomputers paid for with taxpayer sheckles. So that's not a concern to you or I.
$2300, actually--and that's in Canadian dollars, so the US price would be <$2000 (granted, that's a sale price). I could theoretically afford one, but can't justify the cost for what I'd be using it for. However, it isn't out of range for the serious hobbyist who would rather spend money on this than, say, a motorboat or a European vacation or a really good set of golf clubs, although it isn't a casual purchase, either.
I'm curious about what's going to come out of the RISC-V architecture, though, now that chips are starting to show up on the market. That's bleeding edge and non-ARM.
E. Liddell
On Monday 17 June 2019 08:54:43 am E. Liddell wrote:
On Mon, 17 Jun 2019 04:58:16 -0400
Gene Heskett gheskett@shentel.net wrote:
On Monday 17 June 2019 12:10:55 am BorgLabs - Kate Draven wrote:
And bleeding edge today, means its some variation of an arm cpu. Sure, there's now a 64 core threadripper rizen cpu's out there from amd, but at the price per, around 3G's a socket, I suspect only going into supercomputers paid for with taxpayer sheckles. So that's not a concern to you or I.
$2300, actually--and that's in Canadian dollars, so the US price would be <$2000 (granted, that's a sale price). I could theoretically afford one, but can't justify the cost for what I'd be using it for. However, it isn't out of range for the serious hobbyist who would rather spend money on this than, say, a motorboat or a European vacation or a really good set of golf clubs, although it isn't a casual purchase, either.
I'm curious about what's going to come out of the RISC-V architecture, though, now that chips are starting to show up on the market. That's bleeding edge and non-ARM.
E. Liddell
I've only just heard about RISC-V thru the rumor mills. STM has some interesting things too, an apparently very inexpensive 32 bit processor intended for stepper motor controls that are cheap enough to include on the end cap of the motor, and which add considerably to the ability of the motor to self correct. But with PCW (Mesa) doing new fpga based designs as time goes by, I use his stuff in my machine building. 5 machines converted to be run by linuxcnc, 4 of which survive today. So I'll likely get some experience with them at some point before I miss morning roll call, barring cerebral accidents. One never knows about those.
Cheers, Gene Heskett
-
BorgLabs - Kate Draven -
E. Liddell -
Gene Heskett