From mboxrd@z Thu Jan 1 00:00:00 1970 From: =?UTF-8?B?U3TDqXBoYW5lIEFOQ0VMT1Q=?= Subject: INTEL ATOM E3826 Feedback from an industrial customer. Date: Fri, 13 Feb 2015 11:04:40 +0100 Message-ID: <54DDCC38.9030605@free.fr> Mime-Version: 1.0 Content-Type: multipart/mixed; boundary="===============2071726730==" Return-path: Received: from smtp6-g21.free.fr (smtp6-g21.free.fr [212.27.42.6]) by gabe.freedesktop.org (Postfix) with ESMTP id 2C5956E47E for ; Fri, 13 Feb 2015 02:04:46 -0800 (PST) Received: from [192.168.1.242] (unknown [62.212.97.179]) (Authenticated sender: sancelot) by smtp6-g21.free.fr (Postfix) with ESMTPSA id D1E2382342 for ; Fri, 13 Feb 2015 11:04:38 +0100 (CET) List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: intel-gfx-bounces@lists.freedesktop.org Sender: "Intel-gfx" To: intel-gfx@lists.freedesktop.org List-Id: intel-gfx@lists.freedesktop.org This is a multi-part message in MIME format. --===============2071726730== Content-Type: multipart/alternative; boundary="------------090706060502010808020108" This is a multi-part message in MIME format. --------------090706060502010808020108 Content-Type: text/plain; charset=utf-8; format=flowed Content-Transfer-Encoding: quoted-printable Hi, My name is St=C3=A9phane ANCELOT, I am working at Numalliance R&D Team in= =20 France (http://www.numalliance.com). We are making our own wire bending CNC platform, linux based using INTEL=20 PC platforms (automation and GUI in the same PC). That may be the wrong place, but I think it is important to report my=20 experience, regarding intel graphics performance , when benchmarking INTEL ATOM platforms for usage in our CNC.You may be=20 able to report to the right persons in INTEL group Our application need realtime performance to run automation tasks. This=20 is done using Realtime patches against standard linux kernels. This means we can not use the more recent kernels, but stabilised kernel=20 releases versions (at time of writing, RT preempt : kernel 3.14,=20 xenomai API: kernel 3.16...) I used a kernel 3.16.2. We are using a 19 inch vertical display at 1280x1024 resolution. We faced following problems with GFX driver : a/console flickering console at screen bottom in kernel 3.16.2 . The problem=20 increased when there was CPU/disk/network activity.This problem does not=20 appear from kernel 3.18.2 release. Unfortunately in our environment, we can not use 3.18 kernel , because=20 it is not ready with realtime patches. b/ 2D performance Poor 2D performance, looks like we have not had 2D acceleration. Visually poor performance visible when raising/lowering fullscreen window= . When moving object in paint application (inkscape) , the object does not=20 follow efficiently the mouse. c/ 3D performances In our application,we are making heavy usage of 3D for CNC simulation=20 (some screenshots available on request only). We have seen lot better performances than ATOM D2550 , we tried in the=20 past. That seems a good thing. Conclusion Although there is a wish from Intel to provide ATOM platforms ready for=20 industry, it is not ready regarding ATOM platforms. Because we can not change kernel releases versions, when validating a=20 product. This requirement should be considered. In the same way, we can not change the PC platform every year, because=20 of processor obsolescence. In our case, we are dependant on Ethernet realtime driver, Realtime=20 patches, graphic 2D and 3D performance. We think too, that since ATOM platforms is not very spreaded and so=20 common as Desktop platforms, BayTrail drivers are not so efficients. I=20 am sure They will be... but in may be 2 years... For these reasons, we will stop benchmarking ATOM platforms, and will=20 benchmark Core Ix platforms, since we think the GFX chipsets is better=20 supported regarding drivers . Am I right ? I am an open minded guy, so feel free to give your positive or negative=20 opinion ! ;-) I can give more details if needed. Have a look at what we are doing=20 with an INTEL platform : https://www.youtube.com/watch?v=3Dwj30CeAFwuk Regards Stephane ANCELOT sancelot@numalliance.com --------------090706060502010808020108 Content-Type: text/html; charset=utf-8 Content-Transfer-Encoding: quoted-printable Hi,

My name is St=C3=A9phane ANCELOT, I am working at Numalliance R&D Team in France (http://www.numalliance.com).
We are making our own wire bending CNC platform, linux based using INTEL PC platforms (automation and GUI in the same PC).
That may be the wrong place, but I think it is important to report my experience, regarding intel graphics performance , when
=C2=A0benchmarking=C2=A0 INTEL ATOM platforms for usage in our CNC.Yo= u may be able to report to the right persons in INTEL group


Our application need realtime performance to run automation tasks. This is done using Realtime patches against standard linux kernels. This means we can not use the more recent kernels, but stabilised kernel releases versions (at=C2=A0 time of writing, RT preempt : kern= el 3.14, xenomai API: kernel 3.16...)
I used a kernel 3.16.2.


We are using a 19 inch vertical display at 1280x1024 resolution.
We faced following problems with GFX driver :

a/console
flickering console at screen bottom in kernel 3.16.2 . The problem increased when there was CPU/disk/network activity.This problem does not appear from kernel 3.18.2 release.
Unfortunately in our environment, we can not use 3.18 kernel , because it is not ready with realtime patches.


b/ 2D performance
Poor 2D performance, looks like we have not had 2D acceleration.
Visually poor performance visible when raising/lowering fullscreen window.
When moving object in paint application (inkscape) , the object does not follow efficiently the mouse.

c/ 3D performances
In our application,we are making heavy usage of 3D for CNC simulation (some screenshots available on request only).
We have seen lot better performances than ATOM D2550 , we tried in the past. That seems a good thing.
=C2=A0

Conclusion
Although there is a wish from Intel to provide ATOM platforms ready for industry, it is not ready regarding ATOM platforms.
Because we can not change kernel releases versions, when validating a product. This requirement should be considered.
In the same way, we can not change the PC platform every year, because of processor obsolescence.
=C2=A0In our case, we are dependant on Ethernet realtime driver, Real= time patches, graphic 2D and 3D performance.
We think too, that since ATOM platforms is not very spreaded and so common as Desktop platforms, BayTrail drivers are not so efficients. I am sure They will be... but in may be 2 years...


For these reasons, we will stop benchmarking ATOM platforms, and will benchmark Core Ix platforms, since we think the GFX chipsets is better supported regarding drivers .
Am I right ?

I am an open minded guy, so feel free to give your positive or negative opinion ! ;-)

I can give more details if needed.=C2=A0 Have a look at what we are d= oing with an INTEL platform=C2=A0 :
https://www= .youtube.com/watch?v=3Dwj30CeAFwuk

Regards
Stephane ANCELOT
sancelot@numalliance.com


--------------090706060502010808020108-- --===============2071726730== Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: base64 Content-Disposition: inline X19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX18KSW50ZWwtZ2Z4 IG1haWxpbmcgbGlzdApJbnRlbC1nZnhAbGlzdHMuZnJlZWRlc2t0b3Aub3JnCmh0dHA6Ly9saXN0 cy5mcmVlZGVza3RvcC5vcmcvbWFpbG1hbi9saXN0aW5mby9pbnRlbC1nZngK --===============2071726730==--