Gateway Intros its First 11.6" Netbook

Yeah, I put the decimals in the wrong spot. Don't know what I was thinking.

 

I swapped in a TL-52 and didn't have any luck. I tried all of the tricks you mentioned to no avail. Did you happen to try a 90nm processor, and if so, did it work?

The power draw during POST is probably too much for this thing. Mine came with a 30W brick, and the TDP of the TL-52 alone is 31W...

That's good to know. It looks like linux-PHC can't handle VID changes on the L110 at the moment.

EDIT: I got it to POST once, but everything was disconnected. Thankfully, the TL-52 was from the previously mentioned Gateway MX6453, so it didn't cost me anything.

 

I looked more at the gateway heatsink and don't know if it would work but here is my solution.

we could but a copper plate that goes from the cpu to the chipset with thermal paste transferring the heat on a larger surface and by the same time getting a little help from the fan casing that will spread the heat from the cpu trough the copper plate and chipset heatsink and with the fan getting is air from outside the netbook. That should help( i think).

 

I ended up purchasing a 1.6GHz Athlon 64 TF-20 on eBay. It's no powerhouse, but it only has a TDP of 15W. I doubt I'll be able to get my hands on a TK-42 anytime soon.

 

I didn't have anything to take SSDT from, so I wrote it manually. Downloaded ACPI DSDT table. Disassembled. Fixed 2 errors and 1 warning. Then I added custom _PCT table at the end with 5 P-states. On windows I use RMClock so I know it's stable at 800MHz, 0.7V.
Unfortunately I don't know how to calculate VID for voltages lower than 0.8V (perhaps extended vid should be used) so I ended up with P-States [email protected] (P4 fid=00, vid=1e), [email protected] (P3 fid=01, vid=1e), [email protected] (P2 fid=02, vid=1e), [email protected] (P1 fid=03, vid=1c), [email protected] (P0 fid=04, vid=1a).
Compiled DSDT into kernel (2.6.30-gentoo-r4) and rebooted. Unfortunately this was not enough, because everything below 1.6GHz is considered low by Linux powernow-k8 module and it allows only 1 P-State in low freq table. If there are more (as in my case) it bails out with an error "powernow-k8: Too many lo freq table entries" which you can see if you enable cpufreq debugging in kernel.
I fixed this problem by lowering low frequency table boundary in powernow-k8.h:

Code:

#define LO_FID_TABLE_TOP 0
#define HI_FID_TABLE_BOTTOM 1

#define LO_VCOFREQ_TABLE_TOP 800
#define LO_VCOFREQ_TABLE_BOTTOM 900

Works with ondemand governor. At least /sys/devices/system/cpu/cpu0/cpufreq changes. Only problem is that it jumps by 200MHz. I suppose it can be fixed by changing frequncy resolution in powernow-k8.h to:

Code:

#define MIN_FREQ_RESOLUTION 100

Also I noticed that sometimes it fails, dmesg shows error "powernow-k8: fid trans failed, fid 0xa, curr 0x4". I have no Idea why it wants to change to such high frequency, because highest it has in table is 0x4. Hopefully this can be fixed by changing some defines in powernow-k8.h...

 

Wow, that's a lot more progress than I've made. I also got those iasl compile errors...not sure if I fixed them correctly but they went away I tried to modify the tables I got from the SSDT, but I don't really know what I'm doing and all I ended up with was powernow-k8 complaining. Any chance you could post that _PCT table you made somewhere so I could play with it? I'm pretty sure the voltages are wrong in the SSDT I used. Also, I only get two P-states (800MHz/1600MHz), and one of them is obviously wrong. Fortunately, linux-PHC lets me change them.

 

Certainly:
Source (dsdt.dsl)
Compiled (dsdt.hex)

Use at your own risk. fid and vid values are probably wrong. I'll fix it when I get time.

Keep in mind this notice in powernow-k8.h when changing the file:

Code:

/*
 * There are restrictions frequencies have to follow:
 * - only 1 entry in the low fid table ( <=1.4GHz )
 * - lowest entry in the high fid table must be >= 2 * the entry in the
 *   low fid table
 * - lowest entry in the high fid table must be a <= 200MHz + 2 * the entry
 *   in the low fid table
 * - the parts can only step at <= 200 MHz intervals, odd fid values are
 *   supported in revision G and later revisions.
 * - lowest frequency must be >= interprocessor hypertransport link speed
 *   (only applies to MP systems obviously)
 */

This probably answers my question why it sometimes wants to change to frequency higher than in table - "lowest entry in the high fid table must be >= 2 * the entry in the low fid table". There's no way we can guarrantee that.
I wonder why there are these restrictions. This means that even if Gateway fixed BIOS powernow-k8 would not work.

 

so is a turion x2 @ 800mhz faster then the athlon 64 @ 1.2 Ghz and is the battery life much worse?Because i was thinking of putting in an x2 in my packard bell dot m/a.

 

Does anyone have a review on this?

Is the battery removable'?

 

Fixed bug in powernow-k8 module that prevented voltages lower than 0.8V from working. Also fixed "powernow-k8: fid trans failed, fid 0xa, curr 0x4" error. Updated DSDT table with lower voltages - [email protected], [email protected], [email protected], [email protected], [email protected].
Files:
dsdt.dsl - LT3103u DSDT source
dsdt.hex - LT3103u DSDT binary
powernow-k8 low frequency, low voltage patch

This might fry your laptop. Use at your own risk.

 

I think it's a great little mini notebook. The screen is the perfect size where it's not too small (I came from a 8.9" Acer) or too big (I normally hate regular sized laptops).

It's a little bit more beefy than Atom-based netbooks, as it runs the Athlon cpu and is upgradeable to 4gb of ram.

It's built solidly, isn't too bulky or anything for a 11" netbook, and the keyboard is full size, too!

And yes the battery is removable.

 

Thank you very much for your work! I would never have figured this out on my own. Your DSDT modifications are working great on my system. It sure runs a lot cooler. linux-PHC is even working with your powernow-k8 modifications applied. It appears to have some problems with low vid values when using the "direct_transitions" parameter, but when using your DSDT, that is no longer needed.

EDIT: I'm going to be upgrading to a 1.6GHz chip soon, and I would like to add/modify some of the PSS entries. I understand how to add each frequency, but how are you calculating the last two entries of each "package" (the fid/vid parts)? How does...

Code:

0xe8201[b]78[/b]0,
0x00000[b]78[/b]0

...correlate to "voltage 0.800000 [V] TDP 0 [mW] fid 0x0 vid 0x1e"?

Pardon my ignorance.

 

fid = (frequency - 800) / 100
vid = (1550 - voltage) / 25
status = (vid << 6) | (fid & 0x3f)
control = (3 << 30) | (2 << 28) | (1 << 27) | (2 << 20) | (0 << 18) | (2 << 11) | status
(IRT 80us, RVO 50mV, EXT_TYPE, PLL_LOCK_TIME 2us, MVS 25mv, VST 40us)
Note that this will only work with revF chips.

 

it's their second netbook, unless you mean AMD's first netbook. What's different between Atom and AMD???

 

According to http://patchwork.kernel.org/patch/32584/, we can have 2 low frequency p-states. I don't want to recompile my kernel. Moreover I can't use iasl to compile your code. Can you please modify your source code such that it has only [email protected] and 800MHz at some lower voltage, and then compile it with -ta output option? So that I can inject it during boot time without recompiling the kernel. Your time and effort will be greatly appliciated.

 

That is Linux git tree. It still has a bug that prevents voltages lower than 0.8V. That means you still need my patch and since you use kernel from git tree I can't see how recompiling your kernel is difficult for you.
I talked with Mark Langsdorf (author of the patch) and sent him updated version of my patch that supports more than 2 low frequency p-states and fixes low voltage bug. It should appear in git repository sometime soon. Then hopefully it will be included into next Linux kernel release.
Until then you should patch your kernel and recompile.
powernow-k8 low frequency, low voltage patch (against linux-2.6.git)
powernow-k8 low frequency, low voltage patch (against linux-2.6.30) - this will also bring your powernow-k8 module to git revision.
This is new and updated version with no LO_FID_TABLE_TOP, HI_FID_TABLE_BOTTOM hacks. Whoever was using previous version of my patch should update to this.

 

If you are planing to stay at 1.2 GHz with the L110 or not overclock with setfsb above 240 MHz then the Cas3 Kinston 2 Gig 267/DDR2 533 chips is great. I tried 250 on setfsb to no avail, the stock memory chips as no issue with 250 but it is cas5. Also the Kingston chip is Command Rate 1T.

Carefull with replacing faster CPU's here as the slower Kingston memory may force CPU/6 or more where as the stock memory will always be CPU/5 unless you can get over 2.0 GHz.............

 

That's really good news!

I am currently dual booting Windows 7 and Ubuntu 9.04. On Windows 7 I get close to 5 hours (~4:50 mins actually) at minimum settings and underclocked ([email protected]) using RMclock. I am expecting better results on Linux using fluxbox or no (!!) X server at all (for some long good ol' vim/ emacs development sessions).

I'm a noob at kernel recompiling though (never done it manually). I am currently reading the https://help.ubuntu.com/community/Kernel/Compile wiki entry. I haven't tried any other advanced distro (Gentoo etc.) but am willing to install Arch linux for its minimalism and speed.

I have two queries:
1. Can someone please guide me with steps on enabling PowerNow! for my LT3103u on Ubuntu and/ or Arch linux? :confused2:
2. Is it possible to underclock this CPU to an even lower unofficial P-state (say 3x@600MHz) ?

(I registered to this forum to thank you guys for this :smile

 

I'm getting 5 hours 55 minutes in Windows XP Battery Eater idle test with display brightness at 50% and WiFi connected and idle. Make sure you have drivers for all USB devices installed. Idle power usage should be around 9W.
Btw, your voltage is very close to unstable voltage

I wouldn't be so optimistic. Linux actually does worse even with PowerNow enabled. Idle power draw is around 12W. That is with tickless kernel, USB devices power down enabled and X with radeon driver (DynamicClocks option on).

Lowest frequency for any PowerNow enabled chip is 800MHz. That's FID 0. You can't go lower than 0.


Btw, for anyone interested. I tried my dsdt table in Windows XP with amdk8.sys driver. I used Microsoft ASL compiler to substitute the DSDT table in registry (asl.exe /loadtable dsdt.aml). It worked. Only problem is that amd8k.sys driver will use just 2 P-States - 800MHz and 1200MHz.
If RMClock is not an option for some reason, you can try this method.

 

I tried this method with Vista several days back. It turns out that Vista has native support for PowerNow!, and it utilizes all five P-States.

 

Can you explain this battery eater test that got you 5 hours and 55 minutes? I have xp installed to but its just a basic home edition (i dont know if it matters or not)
When you say iddle to you mean the battery test basically just tested to see how long the computer would run with wifi and 50% screen but not actually doing anything just sitting at desktop, or do you mean the 'battery eater" test actually worked the computer to see how long the battery would last under normal/heavy usage?

 

Link. It's just an application to measure battery life.

Yes.

 

I believing the LT31 is an AWESOME mini laptop aka netbook!

 

Ans 1. Step 1 Find "Howto compile custom Ubuntu kernel on the web". Make sure you follow the one that is newer than 8.04.
Ans1. Step 2 Patch your powernow-k8.c and powernow-k8.h. I have to manually cut and place Dr_oldschool's diff file since it is for kernel v2.6.30. If you are interest you can use mine powernow-k8.*. (post upon request)
Ans 1. Step 3. copy dsdt.hex to linux-2.6.28/include, run make oldconfig, manually edit linux-2.6.30/.config so that it uses your dsdt.hex. Make sure "CONFIG_ACPI_CUSTOM_DSDT=y" and sCONFIG_ACPI_CUSTOM_DSDT_FILE="dsdt.hex"
Ans 1. Step 4. Modify your kernel revision.
Ans 1. Step 5. run a script to get linux-image-.*.deb, install it. Reboot. Enjoy.

 

dr_oldschool,

Have you been having problems with your LT3103 locking up during frequency transitions? I've been noticing this when using the ondemand cpufreq governor. It doesn't seem to like sudden frequency and/or voltage changes of more than one step. The conservative governor has been working fine as long as the freq_step tunable is not set too high. I've done some searching and found that others are encountering the same problem. Any ideas?

For what it's worth, I've managed to get power usage down to 10.7W with Linux (according to powertop). That's with wireless connected and the backlight at it's lowest setting.

 

I have some instability but I traced it to bad x.org drivers. I experience screen corruption when I run extreme-tuxracer game. I've tried radeon and radeonhd opensource drivers with all settings I could think of. radeon seems better than radeonhd i.e. less screen corruption, but still it's far from good. No hardware acceleration fixes the corruption but it's unusable that way.
I had a crash in sata driver, resulting in remounting of root file system read only but system was still usable otherwise. I think this was because I was experimenting with sata power saving settings suggested by powertop.
Haven't had any crashes when running in console mode only doing kernel compile. I'm using ondemand governor. If it's really cpu scaling that's causing instability, it should be easy to reproduce with userspace governor. Just write a shell script that changes frequency randomly. I'll try that when I get some time.

What did you do?
IIRC the best I could get was around 12W.
Backlight at 50%.
All power saving settings suggested by powertop.
Wifi interface down. ath9k "power on" iwconfig setting did not work - it could not associate with my ap.
Webcam (usb 1-4 device) in autosuspend mode. It had to be done manually because it's always on by default.

 

I'm using Ubuntu 9.04 "Jaunty" for the moment, as I slowly get a stable baseline for a Gentoo install. I also experience screen corruption when running the etuxracer game. I haven't tried the radeonhd driver yet, and I probably won't after hearing that. However, I plan on trying the "stable" git versions of libdrm, mesa, xorg-server, and xf86-video-radeon once I get Gentoo up and running in order to see if that makes any difference.

Hmm, I haven't been using console mode much with this install. All I can say is that it always appears to hard lock during sudden high CPU usage when using the ondemand governor. Even though I now have a 1.6GHz TF-20 in my LT31, it was happening with the L110 as well. I doubt it's the voltages since I've tested each step with mprime for approximately two hours.

I basically did everything powertop suggested, plus a little undervolting, along with enabling PCIE ASPM's "powersave" mode and removing any unused modules (mainly r8169 and uvcvideo). If I use a wired connection, I can get it down to 10.4W. I could probably get it even lower if I compiled CONFIG_USB_OHCI_HCD as a module and loaded it only when I needed it.

Now if only I could find a way to control the fan... This program works for me in Windows. Because of this, I was hoping that the acerhdf module might work in Linux, but I haven't had any luck with it yet.

 

I tried git x11-drm and git xf86-video-radeonhd and it didn't make any difference. git radeonhd power saving mode did not work. It could read the frequencies but could not set them. Complained about AtomBIOS errors in X log file.

I could not reproduce this.

Code:

cd /sys/devices/system/cpu/cpu0/cpufreq
echo userspace > scaling_governor
while true; do A=$[800000 + RANDOM % 5 * 100000]; echo $A > scaling_setspeed; cat cpuinfo_cur_freq; done

Tried this with kernel compile running in background, with and without x running. No lockups. Can you try this?

I did some testing. Here's what I've found:
enable powernow-k8 and ondemand governor -1.4W
set power level to auto for usb webcam -0.5W
enable pci_aspm powersave -0.4W
enable min_power scsi lpm -0.4W
backlight fom +4 to 0 -1.0W
stop wlan -0.4W
With all these power savings I could get usage from 15W (backlight +4, wifi on) to 10.9W. hdparm -B1 /dev/sata will yield some 0.2W, spinning down disk will yield some more but I don't like to do it.
I get 8W power draw in WindowsXP with backlight at min and wifi connected.

enable laptop_mode - no changes
rmmod r8169 - no changes
rmmod uvcvideo videodev - no changes
rmmod ehci_hcd ohci_hcd usbcore - no changes
stop alsa, remove sound modules +0.4W

As you can see, removing modules will not yield anything, it can only make things worse as in alsa case.

I experimented with acer_ec.pl a bit. Here's what I've found:
register 0x55:
READ - 0x00 fan off, 0x01 stage1, 0x02 stage2, maybe there's stage3 when cpu gets very hot, haven't seen it.
WRITE - 0x00 set to auto, 0x10 turn off
I would be more happy if there was a way to change temperature thresholds for various fan speeds or at least set fan speed, but it only allows to either turn it off completely or set to auto.

 

That's an easy way to reproduce it on my machine...
I tried running this in a console with and without X running, and within X itself. It locks up within a minute. SysRq doesn't work. I brought the voltages back up to stock just in case. My LT31 doesn't seem to like rapid frequency and/or voltage changes for some reason.

In my case, I can see why removing uvcvideo affected my power usage. Even though I'm booting with the 'usbcore.autosuspend=1' parameter, it doesn't seem to have any effect.

I suspect that you're running the snd_hda_intel module with the 'power_save' parameter. Perhaps this is why power usage increases when it is unloaded?

Well, that's a start. Thanks for the information.

 

Is this with stock L110 processor? Do you use my dsdt table? I updated it on Aug 27. There was a bug in how I calculated control register, basically it ended up using Ramp Voltage Offset of 75mV instead of 50mV for extended VIDs (800MHz and 900MHz P-States).

To correct myself:
control = (3 << 30) | (2 << 28) | (1 << 27) | (2 << 20) | (0 << 18) | (2 << 11) | ((vid & 0x1f) << 6) | (fid & 0x3f).
Also vid calculation formula is not correct for extended vids.
See AMD Hammer Family Processor BIOS and Kernel Developer's Guide Table 71 for VID code Voltages.

 

Hey thanks, but does this also mean that with *every* subsequent kernel update on my linux distro I will need to get the source and recompile manually? Also please post your latest kernel fixes and dsdt file (one that dos not cause any freezing issues)

 

Aaw! Whenever I find a quite good laptop with a very nice design the monitor is always to small. I want 17" or at least a 15" monitor.
It will take a long time before a decide which laptop I want to buy. And I do mean long!

 

Here is a review of a new AMD netbook class processor laptop. I wonder if one could find one of those and shoe horn it into this laptop

http://www.trustedreviews.com/lapto...-Acer-Ferrari-One--AMD-Congo-Based-Netbook/p1

 

More of a preview than a review. Sincce someone else was able to shoehorn a TK series cpu the L310 should be doable as well. Problem then is getting a CPU to do this with.

 

No, this is not with the L110 processor, but it is with a modified version of your newest DSDT table (several frequencies removed and two higher frequencies added). I was having this same problem with the L110, albeit with an older DSDT table.

All of the steps work fine if selected manually, but if I leave it to cpufreq to change them (or run something along the lines of those commands you posted earlier), I'll get a lockup. Perhaps the transition latency is not high enough? I'm running out of ideas.

Could the clock generator have anything to do with this? I know there are at least two versions out there (according to setfsb); the ICS932S421BGLF & the ICS9248BF-96. My LT31 has the latter. Which one does yours have?

 

I have the ICS932S421BGLF clock generator.

 

That doesn't look good for me.

I wish I got the ICS932S421BGLF, for multiple reasons. Too bad I've long since voided my warranty.

 

I must've missed it, but how do I check?

 

The easiest way to check is with the Windows setfsb program. First, try selecting the ICS932S421BGLF. If you get a "PLL ID Error" message, you most likely have the ICS9248BF-96 (which in reality is a SLG84605).

Alternatively, you could take the Gateway apart and read the part number on the chip itself.

The lockups I encountered with frequency/voltage scaling were most likely linked to the SLG84605. I recently picked up another LT31 with the ICS932S421BGLF and haven't had any problems with it.

 

I don't know why they, Gateway, decided to use boards with the different PLL's. I guess to most it makes no difference but consistency would have been nice. It appears the SLG one causes apparent multiple issues except if you just use it as all 100% stock.

 

At you this cpu http://www.cpu-world.com/CPUs/K8/AMD-Athlon 64 X2 TK-53 - AMDTK53HAX4DC.html ?

 

I could not start AMD X2 tk-53(
In what there can be a problem?

 

Turmlos, i am very interested about your experiment with the TL-52. Since TechGadget09 got it to work with TL-56 but with problems, i figure TL-52 is probably going to be more stable with less heat. I am surprised that you are not even able to get it to boot. Is it possible yours was overheating? have you tried again with thermo compound and make sure the hs is contacted properly?

i just ordered the lt3103 and TL-52 don't have either on hand yet.

 

my tl-52 would run, but it does crash if you let it heat up... currently looking into setfsb to underclock it. what clockgen do you guys use?

 

Sorry, I don't know the correct setfsb settings to
underclock this machine with a different processor.

However i do know you can use "RmClock" to underclock the processor,
and it works great.

My upgraded Lt3103u... has a Amd Athlon TK 42 dual core processor, so im not sure if are settings are going to be the same.

If you want to try my settings here is a link to where you can get the Info.

http://www.gatewaynetbooks.com/index.php/forums/9-internal-hardware/515-lucky-me-got-to-install-a-qtk-42q-in-my-lt3103u#535

(Left click on the pictures twice and you will be able to see the pictures clearly)


I recommend trying the Performace on damand settings on RmClock... It works great whenever my machine is not stressed it runs at .8ghz at .800v which allows it to cool of alot. I been using it where it trottles up to 1.2ghz
(it can go up to 1.6 but after a while it would get to hot)

Anyway good luck i Hope this Helps.

 

I am in the market for a netbook when I chance upon the gateway in a retail shop. After playing with both Acer Aspire One 751 and a few others I decided on Gateway LT31. The vista basic included runs well especially after the bloatware was removed.

Anyway, i loaded win 7 pro 32bit then installed the following programs, office 2007/office live, visual studio 2008, expression studio 3, Windows Live Messenger and Mesh. Windows 7 full codec and VideoLAN) After the system is fully loaded, the response is still snappy, feels the same as my notebook Compaq Q40.

I am able playback the included win 7 HD video with 0% lag.

Youtube standard is 100% playback with 0% lag, Youtube 720p has 5% lag at time. (battery life is 4 hours normal use)

I am able to overclock my cpu using setFSB to 1470Mhz [email protected] 100% stable (5 times passes on IntelBurnTest), battery life can still maintain 3 hours 50min.

With OC CPU, Windows 7 get 2.8 on the CPU, graphic Aero is 3.0, Gaming 3D is 3.2, Memory is 4.7 and Hdd is 5.7.

With this setup, Youtube 720p HD video has no noticeable lag.

The best is for you try a real Gateway LT31 for yourself. I think you will like its for it usability and price.

I also think L110 is a competent 64bit CPU with 128K L1 and 512K L2 cache like my old Athlon 64 Desktop oldie but goodies

[/URL][/IMG]

 

Sorry for the necropost.

Wanch I would be very much interested in your powernow-k8 files

*crosses fingers that they're for 8.10*

 

Has anyone managed to get a TL-52 working with linux?

I am wondering about adding the P-states to the DSDT file. Is it possible to add arbitrary frequency and voltage combinations? I was thinking to underclock the CPU and keep the frequencies the same as the stock CPU to reduce heat.

Thats if the cpu will even work. I should be receiving it soon, so we will see. I drilled some holes on the backside of the laptop below the heatsink fan and replaced the thermal compound in anticipation.

If we can get this to work in linux, TL-52's are about $10 shipped on ebay. So it is a very cheap upgrade.

 

Is this available in any brick & mortor store...??

 

joey_sfb:
how did you get the WEI score for graphics up to 3.0?
mine is reading a sad sad 2.4