GPD Win (x86 Computer / Palmtop)


@ED

They had a metal shell version of the WIN going back to the Indiegogo crowdfunding for the GPD WIN. At one point that was what they were going to release. They didn't in the end because the iniital purpose of it was to perform part of the passive cooling solution which didn't do the trick well enough. Thus the inclusoion of a fan and the reversion back to a plastic shell.

I remember people asked that question of the WiFi at the time and they said it was tested and worked fine I think.
 
Well, depends on their testing :)

I've seen so many issues with GPD Wins that were supposed to be tested, that I'm not really sure about it... they might've tried to download something via the internet and as long as the download started, it was okay ;)

The GPD Win apparently also has a lot of design issues, so hopefully they'll fix that for the GPD Pocket as well.
 
Yes that's my main concern with the Win: it looks like the HDMI port on mine is now broken (it used to work fine but now the Win turns off when I plug in my HDMI cable, I don't have another cable to check if it comes from the cable or the Win), and I once had the keyboard stop working until I changed the fan setting... I'm afraid it'll fail me sooner or later.
 
Hi,
last week I received my GPD WIN Z8750 (rev2). I didn’t purchase it from ED, but from GearBest; I know it’s a gamble to buy a Chinese product from a Chinese store and, for now, it seems I lost my bet.

This is a video of my GPD WIN randomly crashing with a BugCheck entry in Windows Event Log.

I tried all possible combinations of Windows 10 revisions, driver revisions, BIOS settings and/or BIOS revisions; sometimes it even crashes on simple desktop tasks (i.e. opening Start Menu). It seems to me that the crashes are more frequent when the unit is cold (just booted from a long shutdown period).

I’m trying to get GearBest to honour its own “3 Day Dead on Arrival (DOA) Guarantee” policy, but it’s a strenuous effort to deal with Chinese dealers: I just switched from their ticket system to PayPal message system in order to, eventually, escalate the dispute.

Do you think is there any further test I can try?

Do you think that a hardware mod, like the resistors mod, can resolve the problem?



Many thanks in advance.
Best regards.

Locutus73
 
Sounds like a hardware problem. You could try and install linux to check if it actually is. I would send it back if possible.

If you are not able to send it back and it is a hardware problem, you could think trying the heatsink mod proposed by Phawx on youtube, as they use quite a bit of metallic thermal paste which could provoke a short on some resistor.
But that is quite a far fetched possibility...
 
Further investigations lead to low grade DRAM issues.

I changed Chipset->North Bridge->Memory Configuration Options->Frequency A selection from 1600 to 1067 and the system is rock solid, even on Dram Speed Grade set to Fast.
I verified memory settings with different BIOS configurations with CPU-Z and the results were:

Frequency A selection: 1600 - Dram Speed Grade: Typical
DRAM Frequency 1599.3MHz
FSB: DRAM: 1:20
CAS# Latency (CL): 12.0 clocks
RAS# to CAS# Delay (tRCD): 15 clocks
RAS# Percharge (tRP): 20 clocks
Cycle Time (tRAS): 34 clocks
Command Rate (CR): 1T

Frequency A selection: 1600 - Dram Speed Grade: Fast
DRAM Frequency 1599.3MHz
FSB: DRAM: 1:20
CAS# Latency (CL): 12.0 clocks
RAS# to CAS# Delay (tRCD): 12 clocks
RAS# Percharge (tRP): 20 clocks
Cycle Time (tRAS): 34 clocks
Command Rate (CR): 1T

Frequency A selection: 1067 - Dram Speed Grade: Typical
DRAM Frequency 1066.2MHz
FSB: DRAM: 3:40
CAS# Latency (CL): 8.0 clocks
RAS# to CAS# Delay (tRCD): 10 clocks
RAS# Percharge (tRP): 11 clocks
Cycle Time (tRAS): 23 clocks
Command Rate (CR): 1T

Frequency A selection: 1067 - Dram Speed Grade: Fast
DRAM Frequency 1066.2MHz
FSB: DRAM: 3:40
CAS# Latency (CL): 8.0 clocks
RAS# to CAS# Delay (tRCD): 8 clocks
RAS# Percharge (tRP): 11 clocks
Cycle Time (tRAS): 23 clocks
Command Rate (CR): 1T

A simplistic math exercise suggests that at 1066MHz with CAS 8, 1066/8=133,25 so we have a latency of 1/133,25=0,0075s, while at 1600Mhz with CAS 12, 1600/12=133,33 so we have the same latency of 1/133,33=0,0075s.
So it seems that with the memory set at 1067 we should have the same latency (ignoring RAS, etc.) but less throughput. In theory, less throughput should affect the integrated GPU (that uses the main DRAM as VRAM), but doing some 3DMark benchmarks I didn’t notice any difference from 1066 to 1600: the results were interleaved (I got some statistically insignificant better results at 1066 fast). Using geekbench I got slightly better RAM benchmark results at 1600MHz.
It would be interesting to verify if memory set at 1067MHz resolves the graphic driver freezes/crashes for the GPD WIN units that don’t go blue screen. It would be interesting even to better investigate the performance consequences of the slower clocked memory.

On Dingoonity community we have 5 people with Z8750 GPD WIN units recently received from GearBest all affected by this issue.

5 people are not sufficient for scientific stats, but we are free to make guesses.
It seems that the March GearBest GPD WIN rev2 Z8750 batch is affected by low grade memories that are not able to run at 1600MHz; we can even suppose that all the widespread bugs seen in the past are memory related. If the hypothesis is right, we should concentrate on memory.
How much does 1067MHz DRAM affects performance vs 1600MHz DRAM?
Does the “resistor fix” increase DRAM voltage leading to a more stable system?



Many thanks in advance.
Best regards.

Locutus73
 
Further investigations lead to low grade DRAM issues.

I changed Chipset->North Bridge->Memory Configuration Options->Frequency A selection from 1600 to 1067 and the system is rock solid,
Good find! Looks like it easier to win in the lottery then actually get a "perfect" unit from GPD. Mine started to develop an annoying chirping sound when its charging and the hinge lost a lot of its "stiffness". But otherwise my unit seems fine, no non working keyboard or freezes. Occasionally the touchscreen stops working and WLAN sometimes does not work after several wakeups.
The only annyoing thing is that the settings reagrding shutting of the display/going to sleep while beeing connected to the A/C adapter are either ignored or the HW does not deliver the proper information to windows, my device always goes to sleep regardles of wether its connected to a wall socket or not.
Oh, and I can't connect to a WiFi network anymore since I just clicked on the "do the creators update" button yesterday, without having a second look.
 
New findings:

1) With memory set at 1067 Fast the system seems to be rock solid
2) With memory set at 1600 Fast
a. On alternate current, 3DMark IceStorm seems to be stable
b. On battery, with Intel Driver->Graphics Properties->Power->Extended Battery Life for Gaming ENABLED, 3DMark IceStorm frequently crashes as in my previous video (when it completes I get results around 15000)
c. On battery, with Intel Driver->Graphics Properties->Power->Extended Battery Life for Gaming DISABLED, 3DMark IceStorm seems to be stable (and I get higher results, around 27000 instead of 15000)
d. Prime95 seems stable
e. Geekbench 4 randomly produces vertical artifacts in Desktop as in my previous video and sometime crashes into a bluescreen
f. MemTest 86 seems stable (5 hours run with 0 errors)​
3) With memory set at 1600 Typical: work in progress…


Any ideas regarding the correlation between memory speed 1600 Fast and "Extended Battery Life for Gaming" ENABLED (so lower performances) on battery causing 3D rendering crashes?



Best regards
Locutus73
 
I have a great update!
Continuing with my investigation I found that using the following BIOS setting
Chipset Tab -> North Bridge -> Memory Configuration -> DRAM PM5 -> Disabled
the system seems to be rock solid even with memory set at 1600MHz Fast.

So, with Memory set at 1600MHz Fast and PM5 disabled I have
1) 3DMark IceStorm stable regardless of power source (battery and AC) and Intel Driver “Extended Battery Life for Gaming” mode
2) Geekbench 4 stable​
I even tried to run 3DMark and Geekbench at the same time with no problems.

So it seems that disabling DRAM PM5 leads to a perfect stable system regardless of Windows/drivers settings.

This DRAM PM5 parameter is poorly documented (the BIOS states “Enable or disable DRAM PM5 PUNIT configuration.”), but digging around PM5 seems to be an advanced power saving mode for DRAM. My hypothesis is that, disabling this parameter, the system uses higher voltages for the DDR3LP memories (or, for the sake of precision, the system avoids lower voltages for the memories) avoiding possible instability at 1600MHz. After all, when we overclock CPUs, GPUs and RAMs, we obtain stability using higher voltages. The hypothesis is compatible with the observed correlation between crashes and Intel "Extended Battery Life for Gaming" parameter: it is possible that, in this mode, all power saving technologies kick in frequently, leading to lower voltage and instability.

It would be interesting to determine if disabling “DRAM PM5” resolves the driver freezes (not the total crashes like mine) and, in general, makes the GPD WIN more (or totally) stable.



Best regards
Locutus73
 
Last edited:
Just bought the GPD Win in-person from the HQ in Shenzhen. They were very nice and I offered to proofread their releases to fix any English mistakes.

However, in my first hour of playing it crashed on Binding of Isaac, and went into a reboot loop. It would boot to the login screen and immediately reboot upon flashing the login screen for about half a second.

I fixed the reboot cycle by plugging in.

I messaged the HQ and they told me to power down and charge it for 5 hours because the battery "curve" was off. Maybe some kind of mistranslation of calibration?

Either way, I powered off and charged over the night.

Next day, I installed Tie Fighter and started playing. Beat a few levels, no issues.

Today, I tried Resident Evil 4 on lowest settings, 800x600 resolution, low shadows, lowest textures, etc.

Game played for a bit, had an unexpected shutdown and then rebooted without problems. Played some more with no issues.

I shut the game down to do some other stuff. Next time I turned on the system, reboot loop again.

I selected restore system to factory settings (maybe the crash corrupted some boot files?) and it showed the BIOS image for a few hours and and then the screen went blank.

Now I can't boot or anything. Pretty lame.

I could fix the problems on my own probably with a clean windows install from USB, but honestly, I bought this from the company in person, so I'm probably just going to take it back and make them do it.

Any thoughts?
 
Last edited:
@1313Q

I would take it back if you are still there. I wonder if you got caught in the Windoqws 10 creators update and it messed up or somehting. Either way, if your there, take it back.
 
Last edited:
Yup. Contacted them already. I'm taking it back tomorrow.

By the way, it went back to only being able to boot while plugged in. I'm doing a full factory reset right now.


I think this sheds light on why the Pyra is going to be good though- it's not too ambitious in terms of how much hardware is packed into the small form factor. Additionally the modular design will mean better trouble shooting.

Finally, linux is less of a resource hog, so there will be much less bullshit when it comes to system optimization.
 
The Pyra has a LOT more hardware packed in than the GPD Win.
Mobile internet, all kinds of sensors, GPS, multiple SD Card slots, backlit keyboard, etc.

Good point! I forgot about the mobile internet and sensors.

I guess quality control is just the real reason I'm looking forward to it. I know that the Pyra will take a longer time because you and your team are painstakingly going through the necessary steps to make a good device.
 
So in thread talking about specs people would like to see in a future GPD WIN 2, GPD posted for discussion a potential keyboard layout that could be used. Mind you , the WIN 2 is a long way off, this is all just very early what do you think of this kind of stuff.

Anyway, this is the keyboard they showed below. .

MMOlcxs.png
 
So in thread talking about specs people would like to see in a future GPD WIN 2, GPD posted for discussion a potential keyboard layout that could be used. Mind you , the WIN 2 is a long way off, this is all just very early what do you think of this kind of stuff.

Anyway, this is the keyboard they showed below. .

MMOlcxs.png
The layout is not bad at all (except that power button) but isn't this a bit too big/wide for a mobile computer?
 
@Alpibrine28

It's a similar layout to the current WIN, except the two right columns of keys are now on a new 6th row of keys on the top. So whereas the current WIN has 5 rows of keys , this layout has 6 rows
 
It is far better than the nightmare that was the GPD Win 1 keyboard.

The coloration of WASD keys and the arrow cluster seems silly. The Fn layer is not color coded - so F7-F12, which would be on the Fn layer, are displayed on the keys as if they were on the Shift layer.

Where did Insert go?

Since this is a Windows device and a thumb keyboard, it will need easy access to the most commonly chorded set of keys on a Windows unit. How does one hold Alt + Ctrl and tap Delete with that supported mid-air using two thumbs?

What is the Select, Start and X on the top-right? Seems VERY similar to the Select, Start and Pyra/Pandora key...

Screen brightness looks to be either on the shift+volume keys or Fn+volume keys. IMHO turning the display brightness up needs to be a top-layer action.

Using 3 top layer keys to control volume seems a bit silly given the constraints. One key could do it.
Volkey = Volume Up
Shift+Volkey = Volume Down
Fn+Volkey = mute/unmute

Same with brightness, currently residing on the Shift or Fn layer behind the volume keys. Doing the above would free up two keys, one of which could be the screen brightness.
ScreenKey = Increase brightness
Shift+ScreenKey = Decrease brightness

That frees up a key that could be used for getting rid of the GPD Win's silly fan switch.
FanKey = Increase fan speed
Shift+FanKey = Decrease fan speed

I see no indicators for keyboard backlight control...

Was this a Saber design? It isn't bad for a first cut. Again - far better than the original GPD Win.
 
Just wanted to post a new update I messed with Hans De Goede's custom kernel last night on the ship and got this! Battery indicator and brightness. Haven't anything else yet.
hdgkernel1.jpg
 
It is far better than the nightmare that was the GPD Win 1 keyboard.

The coloration of WASD keys and the arrow cluster seems silly. The Fn layer is not color coded - so F7-F12, which would be on the Fn layer, are displayed on the keys as if they were on the Shift layer.

Where did Insert go?

Since this is a Windows device and a thumb keyboard, it will need easy access to the most commonly chorded set of keys on a Windows unit. How does one hold Alt + Ctrl and tap Delete with that supported mid-air using two thumbs?

What is the Select, Start and X on the top-right? Seems VERY similar to the Select, Start and Pyra/Pandora key...

Screen brightness looks to be either on the shift+volume keys or Fn+volume keys. IMHO turning the display brightness up needs to be a top-layer action.

Using 3 top layer keys to control volume seems a bit silly given the constraints. One key could do it.
Volkey = Volume Up
Shift+Volkey = Volume Down
Fn+Volkey = mute/unmute

Same with brightness, currently residing on the Shift or Fn layer behind the volume keys. Doing the above would free up two keys, one of which could be the screen brightness.
ScreenKey = Increase brightness
Shift+ScreenKey = Decrease brightness

That frees up a key that could be used for getting rid of the GPD Win's silly fan switch.
FanKey = Increase fan speed
Shift+FanKey = Decrease fan speed

I see no indicators for keyboard backlight control...

Was this a Saber design? It isn't bad for a first cut. Again - far better than the original GPD Win.
Here is Saber's one :

FNDao0W.png

He put the sticks on the Edges because Wade talked about putting them there.

Here is mine :
816234keyboardlayoutCommandFN.jpg
 
Back
Top