Intel Atom N270 boot warnings

W

wetcoast

Intel Atom N270, FreeBSD v8.1 i386, dmseg:

Code: 
Copyright (c) 1992-2010 The FreeBSD Project.
Copyright (c) 1979, 1980, 1983, 1986, 1988, 1989, 1991, 1992, 1993, 1994
        The Regents of the University of California. All rights reserved.
FreeBSD is a registered trademark of The FreeBSD Foundation.
FreeBSD 8.1-RELEASE #0: Mon Jul 19 02:55:53 UTC 2010
    root@almeida.cse.buffalo.edu:/usr/obj/usr/src/sys/GENERIC i386
Timecounter "i8254" frequency 1193182 Hz quality 0
CPU: Intel(R) Atom(TM) CPU N270   @ 1.60GHz (1596.01-MHz 686-class CPU)
  Origin = "GenuineIntel"  Id = 0x106c2  Family = 6  Model = 1c  Stepping = 2
  Features=0xbfe9fbff<FPU,VME,DE,PSE,TSC,MSR,PAE,MCE,CX8,APIC,SEP,MTRR,PGE,MCA,CMOV,PAT,CLFLUSH,DTS,ACPI,MMX,FXSR,SSE,SSE2,SS,HTT,TM,PBE>
  Features2=0x40c39d<SSE3,DTES64,MON,DS_CPL,EST,TM2,SSSE3,xTPR,PDCM,MOVBE>
  AMD Features=0x100000<NX>
  AMD Features2=0x1<LAHF>
  TSC: P-state invariant
real memory  = 2147483648 (2048 MB)
avail memory = 2081800192 (1985 MB)
ACPI Error: A valid RSDP was not found (20100331/tbxfroot-309)
MPTable: <OEM00000 PROD00000000>
WARNING: Non-uniform processors.
WARNING: Using suboptimal topology.

The system is slower than expected, much slower than an old P4 system keeping the workbench warm. Is the ACPI error and the subsequent warnings something I should be concerned about?
 
Not sure about the warning or error, but I wouldn't expect a N270 to beat or even approach a P4. The TDP is 2.5W, so while it's amazingly efficient, I wouldn't hope for amazing speed. Did you compare it to any benchmarks, or FWIW, benchmark it with a different OS installed and compare it to FBSD that way?
 
wetcoast said:
The system is slower than expected, much slower than an old P4 system keeping the workbench warm. Is the ACPI error and the subsequent warnings something I should be concerned about?
Click to expand...

Atom CPU: 2 instructions per clock cycle
Pentium 4 CPU: 3 instructions per clock cycle

Atom CPU: out of order instruction support (DOES NOT SUPPORT)
Pentium 4 CPU: out of order instruction support

Atom CPU: about 1.6GHz frequency
Pentium 4 CPU: mostly more then 2.6GHz frequency

... so:

2 * 1.00 * 1600 = 3200 (Atom CPU)
3 * 1.25 * 2600 = 9750 (Pentium 4 CPU)

Both Atom and Pentium 4 CPUs are single core CPUs with HTT, so when You will compare 1.6GHz Atom CPU, to 2.6GHz Pentium 4 CPU, then the last one would be about 3 times faster, which means that You Atom CPU will be about 3 times slower then an averge Pentium 4 CPU based system. Both CPUs support HTT, so I did not added it into calculations.
 
Which P4? The Atom has 512 L2 cache. Some P4s have more than that, and it can make a big difference in some things, like buildworld/buildkernel.
 
No one buys an Atom for speed. Their appeal is their low cost and low power requirements.

You probably have to be a minimalist at heart to make things like that your main priority. On the other hand, if you must have Compiz, HD video, or the ability to play the latest first-person shooter game, an Atom probably isn't the CPU for you.
 
ckester said:
No one buys an Atom for speed.
Click to expand...
Not those in the know, no. I do feel that they're a bit over hyped - JMHO.

Even for people wanting a small power footprint - know that a Core i3 can achieve power consumption figures very close to an Atom when at idle and plugged into a good motherboard (Intel DH55TC is one example). The difference with a Core i3 is that it'll hit 100W at full tap, but give you 100W worth of performance to boot. An Atom is like a Core i3 with its EIST/P4TCC stuck in the minimum setting.
 
Thanks for the feedback. I expected about 1/4 the performance from the Atom, but it seems about 1/10 that of the P4. I selected the Atom for a low power embedded system. The Atom has since shipped, but when the next one comes in I could run some benchmarks.

The primary concern is the lack of Intel Hyperthreading (HT) detection and those warning messages. I remember reading a while ago that FreeBSD's kernel doesn't make the greatest use of HT, is that still true? In the context of this Atom N270, I wonder how much it would help and if I should pursue getting HT properly detected and operational.
 
aragon said:
Is it enabled in the BIOS?
Click to expand...
Yes.

aragon said:
IIRC, it can be enabled/disabled from the OS with the machdep.hlt_logical_cpus sysctl(8) OIDs. See smp(4) too.
Click to expand...
Neither logical CPU bitflag is set to disabled, and from the boot messages it appears that FreeBSD is not pleased with the MP table, implying a BIOS error or incompatibility. The BIOS has settings for APIC mode ("enabled") and for MPS Version (set to 1.4).

Next time I'm at the system, I'll double-check the BIOS settings and the sysctl vars, do a verbose bootup, and have a look at mptable(1).
 
Fixed, sortof.

It turns out that the BIOS was not enabling ACPI, nor was there a setting for this. The manufacturer has released an updated BIOS with a setting to enable or disable ACPI. This results in hyperthreading getting properly recognized, so two virtual CPUs. Unfortunately I'm now getting ACPI errors with respect to querying the current temperature. I'll address those errors separately in the ACPI mailing list.

Click here for the AAEON FAQ where I figured out the connection to the problem, and where the newer BIOS resides.

Contact the manufacturer's tech support for instructions and the required utilities to update the AEC-6831 (or the GENE-9455 motherboard). They'll send you a little utility and associated files to create a bootable USB stick as well as a utility to update the BIOS. If your system appears to hang when running the update utility then figure out the command line switches so it can run unprompted, something like this:

Code: 
C:\>awd891 6831_A10.bin 6831_old.bin /py/sy/cc/cp/R

Yes, that was a DOS prompt. On one hand you can't avoid MS, on the other hand they've been relegated to BIOS support.
 
You must log in or register to reply here.
Back
Top