GIGABYTE GA-D525TUD Intel Atom D525
- BigJohnT
-
Topic Author
- Offline
- Administrator
-
- Posts: 7000
- Thank you received: 1175
Well, and desk space!
I just ordered a GA-D525TUD and a ASUS E35M1-M to test
Stack them up like cord wood!
John
Please Log in or Create an account to join the conversation.
- AlpineKyle
- Offline
- New Member
-
- Posts: 2
- Thank you received: 0
Please Log in or Create an account to join the conversation.
- ArcEye
- Offline
- Junior Member
-
- Posts: 24
- Thank you received: 758
You should be able to get the latency lower just by using the isolcpus boot time parameter.I haven't had much time to tweak it but with stock bios settings and a fresh install of linuxcnc, I'm getting latency max of 13835
wiki.linuxcnc.org/cgi-bin/wiki.pl?The_Is..._Parameter_And_GRUB2
The Intel version of this board is reported to run sub 10K max jitter after this.
regards
Please Log in or Create an account to join the conversation.
- PCW
-
- Offline
- Moderator
-
- Posts: 18567
- Thank you received: 5101
running flash videos and multiple glxgears, I had as bad as 49000 base thread latency. This is in the range that I can measure with the scope on all the Atom MBs (the latency test is not to be completely trusted as it does no hardware I/O that may be blocked) The isolcpus trick seems to make the average jitter better but the worst case stays about the same
The ASUS E35M1-M was significantly better with 21000, 26000 (servo/base) thread max jitters so seems like a fairly nice fanless LinuxCNC system
Please Log in or Create an account to join the conversation.
- BigJohnT
-
Topic Author
- Offline
- Administrator
-
- Posts: 7000
- Thank you received: 1175
The ASUS E35M1-M was significantly better with 21000, 26000 (servo/base) thread max jitters so seems like a fairly nice fanless LinuxCNC system
I just checked NewEgg and wouldn't you know it this board is out of stock. Also it had quite a few DOA reviews... which don't matter if you can't buy one.
How did you test the latency with your scope?
John
Please Log in or Create an account to join the conversation.
- PCW
-
- Offline
- Moderator
-
- Posts: 18567
- Thank you received: 5101
chipset upgrade (e450 instead of e350 APU) so its a tiny bit faster.
I checked the latency with a scope by toggling an I/O bit in the servo thread
(I originally did this to see why, even though they often show good latency test results,
the Atom motheboards wont run a servo thread above about 1.5 KHz without error)
Please Log in or Create an account to join the conversation.
- PCW
-
- Offline
- Moderator
-
- Posts: 18567
- Thank you received: 5101
Please Log in or Create an account to join the conversation.
- dab77
- Offline
- Elite Member
-
- Posts: 194
- Thank you received: 2
Please Log in or Create an account to join the conversation.
- ArcEye
- Offline
- Junior Member
-
- Posts: 24
- Thank you received: 758
You mustn't confuse notifications with occurrences.
Linuxcnc only produces one realtime error warning at the first instance, no matter how many further instances there are.
If you get one straight away after starting axis, it generally means you have a serious enough problem for it not to run properly.
regards
Please Log in or Create an account to join the conversation.
- BigJohnT
-
Topic Author
- Offline
- Administrator
-
- Posts: 7000
- Thank you received: 1175
I've always wanted to know.. can one live with the initial realtime error, if then it doesn't appear anymore during use?
If you read the real time error carefully it will tell you that you only get one. Imagine if you have a real time latency every 100mS that was bad enough to throw an error and got an error screen for each one... I think that Axis should not even run after a real time error... or even better each time you acknowledge the error another one pops up.
John
Please Log in or Create an account to join the conversation.