hoo
|
|
Summary
Got the new motherboard on 2008-01-18, bought from Hanobox. The old Gentoo installation booted with no fuss, except for the HDMI. VGA is fine for now though :)
The mobo stopped working on 2008-02-21, and I've returned it for refund. I'm getting a different model for replacement, so this page shall stay as a reference only.
Linux drivers
| Component | Kernel
module(s) | Notes |
| Ethernet | via_velocity | |
| IDE/SATA | pata_via | |
| USB 2.0 | uhci_hcd, ehci_hcd | |
| Firewire | ieee1394,
ohci1394 | |
| Graphics / X | via-agp, via, drm | Xorg driver: Openchrome |
| Graphics / console | vesafb | |
| Sound / ALSA | snd_hda_intel | Enable Azalia HDA in BIOS |
| CPU frequency scaling | acpi_cpufreq (also try e_powersaver) | |
| CPU temperature monitor | thermal (part of ACPI) | acpitool -t |
Files
External links
Quirks
-
Both the IDE and SATA channels use the same PATA driver. SATA hotplug doesn't seem to work — these things are probably related.
- SMART (hard drive monitoring stuff) must be enabled from BIOS. Otherwise smartctl shows strange errors.
Problems / TODO
-
The Openchrome Xorg driver doesn't currently deal with HDMI at all. I can get a picture via HDMI into my DVI-capable monitor, by setting CRT+HDMI in the BIOS. Xorg only recognizes VGA, and the display is copied into HDMI, but it's output with the weird resolution of 1632x1026, which looks funny on the 1280x1024 monitor.
- With one IDE channel and two IDE drives, activities like CD ripping quicky turn into nasty master/slave sessions ;) This is of course not a defect in the motherboard (with SATA ports too), but rather in my cheap-ass recycling attitude.
- The fansink is not half bad — at 7 V it's practically inaudible behind hard drive noise. However, it still represents a cheap and dumb design, considering passive cooling would be perfectly VIAble.
Fortunately, it looks like standard socket 479 (not the actual CPU socket though :) so there are some half-sensible alternatives.
- The CPU frequency scaling currently works with acpi_cpufreq, with the only alternatives being 800 and 2000 MHz. AFAIK, the C7 has exceptionally fast switching between frequencies, so it makes sense to omit the middle ground. (It has two PLLs, which brings to mind dual clutch transmissions :) However, the usual Linux driver is based on the assumption of slow and expensive switching. Which is why I'd like to get e_powersaver working.
Improvements over VIA EPIA Mii10000N
- VGA display quality is slightly crisper
- More powerful processor, with double the frequency and SSE2/3 instructions, but no increase in power consumption. Frequency scaling also works to some extent.
- Xvideo is smooth, which is good since the alternative of Xvidix doesn't currently work.
- DDR2 533 is probably not a significant improvement over DDR 266. They have the same latency, for starters.
- Gigabit Ethernet (not much help in my 100mbit home network)
- HDMI (DVI) if/when it works
- Solid state capacitors :D
Experiences
- [2008-01-22] The performance limit for video playing (MPlayer) seems to be 1280x720 H.264. I
have one file that plays fine, with about 90% CPU utilization, which
is a big improvement in just a few days :) This
is of course without any hardware acceleration (besides Xvideo), so
future graphics drivers might change the situation. Perhaps even to
enable full 1080p playback?-) (For a back-to-the-reality comparison,
the same improvement applies to my laptop
with a Pentium M. It now runs the same video at about 1 GHz. The 2:1
performance ratio per clock is not surprising though.)
- [2008-01-23] SSH acceleration works, CPU load reduced to 23% of the vanilla version. However, sshd isn't accelerated for some reason.
- Acpitool shows CPU temperature when the 'thermal' module is loaded :) This is nice after struggling to get temperature info via lm_sensors in this and the previous motherboard. Should help setting up passive cooling some day.
Temperature measurements
Because I'm bored on 2008-01-29, having just discovered the ACPI CPU temperature sensor :) These are with the stock heatsink and fan, running full CPU load (mainly distributed.net) for several minutes until the temperature settles. Room temperature nearby (on the keyboard, not used for a long time) is about 22...23 °C.
| Stock HS, 7 V fan | Stock HS, 12 V fan | Nexus HS, 7 V fan |
| 800 Mhz | 44...45 °C | 37...38 °C | 35 °C |
| 2000 MHz | | 60...61 °C | 58..60 °C |
At 2000 MHz and 7 V, the experiment was stopped at 67 °C with the temperature still increasing. Using the idea that the temperature difference is proportional to heat transfer power, which in turn is roughly proportional to the CPU frequency (not exactly true, but happens to work out nicely for the 12-volt case), I predict a final temperature around 77 °C.
I got the Nexus PSM-5000 copper heatsink on 2008-02-20, the above measurements are from the same day. It's a little unfortunate that the finely spaced fins don't lend themselves to passive cooling; the Reynolds number that identifies similar flow conditions has velocity*distance as a constant. However, with the stock fan on the side at a quiet 7 V, the heatsink really shows off its cool.
So far, the CPU voltage has been at the default 1196 / 956 mV. The BIOS does have pretty flexible clock and voltage settings, but no CPU voltages. They might be adjustable with the e_powersaver driver, similarly to what I've done with my laptop.
Risto A. Paju
