Topton Jasper Lake Quad i225V Mini PC Report

Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.

lukasig

Member
Oct 31, 2022
32
16
8
Up until now I've been using the router provided by my ISP. I've been fiddling with pfsense for a while and I've made up my mind and got a 4x 2.5G nic mini PC. Since the whole personal router is a new experience, my question is if I should consider opnsense instead of pfsense, and if so, why?
There seems to be a major update coming soon to PFsense with version 2.7.
 

bndr

New Member
Oct 28, 2022
16
15
3
The i3-1115G4 system I ordered from Topton just arrived today. This was the listing: https://www.aliexpress.com/item/1005004514115494.html

I ordered barebones, and have put a Samsung 512GB NVMe drive in. I also have a PNY 2x8GB DDR4 3200 kit which I tried to install, but found that only the bottom memory slot is working (either DIMM works fine if installed on the bottom slot, but system fails to post with no video output if I try to install memory in both slots or just the top slot).

I installed Win10 Pro 22H2, fully patched up. Drivers installed up to date, no yellow boxes in Device Manager.

Without a fan I was noticing frequent crashing under light load. I haven't disassembled and repasted it, but I did take a side plate off and verified the CPU is making good contact with the case. I stood the case on its side and put it next to a box fan blowing on the top of it. That made it stable enough for simple browsing, but it is still crashing very quickly if I try to run Prime95. HWINFO shows maximum package power around 31W and maximum core temperature around 62C before crash.

There also doesn't appear to be any sort of BIOS option to enable TPM, in spite of the listing saying that Win11 would be installed if you didn't choose a barebones option.

So basically, it looks pretty unusable, unfortunately.

The motherboard I have is a blue PCB marked with "GF-1449NP-12 ver:1.1" under the BIOS battery connector.

Gratuitous FLiR thermal imagery. This is just sitting idle on the Windows desktop, before adding a fan.

IMG_9728.JPGIMG_9729.JPGIMG_9731.JPGIMG_9732.JPG
 
Last edited:

bndr

New Member
Oct 28, 2022
16
15
3
And a snap of the power brick label they sent me.
Supposedly... LITEON, Model PA-1480-5M04, Part No 524475-063, Efficiency Level V
LITEON 4A 12V 48W PA-1480-5M04 524475-063 2022-11-08.jpg
 

bndr

New Member
Oct 28, 2022
16
15
3
Some power data. This is the CPU Package Power from HWiNFO, not power at the wall. I made sure C states are enabled in BIOS, profile is set to "Maximum Non-Turbo Performance" (which was the default). Haven't really touched the BIOS settings otherwise.

Sitting idle at the desktop with only HWiNFO Sensors page open:
4W avg
After opening Edge default home page, alt-tabbing back and forth between it and HWiNFO Sensors:
5W avg
Loading Youtube homepage:
10W avg
Playing a 4K@60 vp09 Youtube video full screen (
):
14W avg
Running Prime95 Torture Test (best run I've managed was less than 5 minutes):
~24W avg
 
Last edited:
  • Like
Reactions: Stovar and dums

sqrwv

Member
Oct 8, 2022
58
45
18
If you have any other ram, for example from a laptop, maybe try it.
I would run some RAM test like Memtest86+, to discard compatibility issues.
I don't understand why using a FAN makes it more stable, if problem is not (cpu) temp.
HWINFO says the cpu is an Eng. Sample.
 
Last edited:
  • Like
Reactions: bndr

bndr

New Member
Oct 28, 2022
16
15
3
It is an Engineering Sample. Higher temperature will always lead to less stability, all else being equal. Instability with stock settings below 70C is certainly out of spec, of course.

I have it crunching Metmtest86+ now, will report. So far it has one pass.
 
Last edited:
  • Like
Reactions: Stovar

bloodyskullz

New Member
Nov 7, 2022
13
0
1
Is anyone here running one of these boxes as a firewall only and not just a virtualization platform with proxmox? Curious to know what your temps are like and what power brick came with this unit. I know the first initial ones were just horrible.

Not going to bite the bullet just yet on these units yet as the 6 port variations haven't shown a description of v5 yet so I am guessing it is v4 or lower??? Any clarification on that would be great.

Thanks
 

dums

Member
Aug 14, 2022
48
27
18
Not going to bite the bullet just yet on these units yet as the 6 port variations haven't shown a description of v5 yet so I am guessing it is v4 or lower??? Any clarification on that would be great.
Not sure about V1 and V2, but V3 is easy to tell:
1. It has USB-C video port on front panel.
front.PNG
2. Motherboard has very distinctive optional NMVMe adapater (green in picture, switched to black recently):
mb.PNG
V4 and V5 are improvements on V3 (i226 ethernet chip upgrade for V4, clock chip and boot from SD card for V5).
I'm yet to see 6 port version that would be based on V3, so for now I presume that 6 port motherboards are different branch of development and version numbering on it isn't compatible with 4 port version numbering.
 

Snk B

Member
Jul 19, 2022
61
54
18
The CPU is engineering sample, sadly. You should contact Topton and see if they will send you another one.
They won't, friend.
They even warn in the ad that it is an ES/QS processor:

Parâmetro apenas para referência
Processadores
Núcleos & linhas
Cache
Frequência base
Max frequência turbo
Gráficos
Celeron n5105​
4 núcleos, 4 fios​
4m​
2.00 ghz​
2.90ghz​
Intel uhd gráficos 450mhz-750mhz​
7505 de ouro pentium​
2 núcleos, 4 fios​
4m​
2.00 ghz​
3.50 ghz​
Intel uhd gráficos 1.25 ghz​
I3-1115G4 do núcleo​
2 núcleos, 4 fios​
6m​
3.00 ghz​
4.10 ghz​
Intel uhd gráficos 1.25 ghz​
I5-1135G7 do núcleo​
4 núcleos, 8 tópicos​
8m​
2.40 ghz​
4.20 ghz​
Intel iris xe gráficos 1.30ghz​
I7-1165G7 do núcleo​
4 núcleos, 8 tópicos​
12m​
2.80 ghz​
4.70 ghz​
Intel iris xe gráficos 1.30ghz​
Notice: i3-1115G4 is ES/QS version with same performance, if you mind, please don't buy. Another cpu is commercial version.
 

Fury

New Member
Oct 22, 2022
3
2
3
The i3-1115G4 system I ordered from Topton just arrived today. This was the listing: https://www.aliexpress.com/item/1005004514115494.html

I ordered barebones, and have put a Samsung 512GB NVMe drive in. I also have a PNY 2x8GB DDR4 3200 kit which I tried to install, but found that only the bottom memory slot is working (either DIMM works fine if installed on the bottom slot, but system fails to post with no video output if I try to install memory in both slots or just the top slot).

I installed Win10 Pro 22H2, fully patched up. Drivers installed up to date, no yellow boxes in Device Manager.

Without a fan I was noticing frequent crashing under light load. I haven't disassembled and repasted it, but I did take a side plate off and verified the CPU is making good contact with the case. I stood the case on its side and put it next to a box fan blowing on the top of it. That made it stable enough for simple browsing, but it is still crashing very quickly if I try to run Prime95. HWINFO shows maximum package power around 31W and maximum core temperature around 62C before crash.

There also doesn't appear to be any sort of BIOS option to enable TPM, in spite of the listing saying that Win11 would be installed if you didn't choose a barebones option.

So basically, it looks pretty unusable, unfortunately.
I wonder if you’re crashing because of the power supply?
31w package power
Add 5w foe the rest of the system
80% efficiency (if that)
and you’re close to the 48w PSU.
I’d try a bigger PSU and see how it goes.
 
  • Like
Reactions: dums

andrea87

Member
Oct 15, 2022
62
86
18
36
North-east Italy
Regarding your mini PC, even if they warn about the ES cpu, they say "with same performance". I would call BS on them and argue up torwards aliexpress support. Your device is clearly defective, given the crashes and other issues.

I've uploaded the STL file for the fan adapter I've printed for my mini pc. Spacing between the mounting tabs is 138mm.

.

It should fit on any of those 4 port devices like the pic @dums posted a couple post higher.

I wonder if you’re crashing because of the power supply?
31w package power
I have the very same PSU. it crashes at ~3.95A output, I've tested it with a variable power load in this post.
The psu is clearly a "LiteON" fake, but still suffices for normal usage. Keep in mind that you would measure above 55W at the plug when the psu fails. He could try to put a power meter before it or a meter on the 12V side, and see if that's failing under load.
 
Last edited:

Becks0815

Well-Known Member
Oct 15, 2022
216
277
63
With the base VGA install of opnsense, and PCI-E port ASPM enabled in bios, it's idling at 7W with 1 NIC connected. Doesn't seem to have the reboot issue with port 8 ASPM like Windows does.
[...]
EDIT: I found a guide for enabling C3 states in opnsense, and now the power is down to 4.8W idle. The last comment in the thread states " Be aware that using lower C-states may introduce spurious instability - my Topton N5105-based unit showed that behaviour once I went lower than C1."
Well, I can confirm the approach works on my end on an older J3160 4 port "Protectli" clone. Everything stable, even after running 40 days or longer. That's the reason why i wrote this short guide. However, as I only have one system for testing, I can't guarantee it works all the time.

What also works is to use powerdxx. It shows better results on my machine. To find the possible frequency settings you can use, run the command
% sysctl dev.cpu.0
dev.cpu.0.freq_levels: 1601/2000 1600/2000 1520/1900 1440/1800 1360/1700 1280/1600 1200/1500 1120/1400 1040/1300 960/1200 880/1100 800/1000 720/900 640/800 560/700 480/600

-> I can use frequencies between 480 and 1600 MHz on my J3160, so adding
powerdxx_flags="-a adp -n adp -m 480 -M 1600" # set adaptive mode, min frequency to 480 and max freq to 1600 MHz

in /etc/rc.conf will do the job

Btw.: my J3160 opnsense FW runs at 6-8W, depending on the amount of traffic (20 MB/sec max)
 

Becks0815

Well-Known Member
Oct 15, 2022
216
277
63
Sounds like pfsense is disabling lower C-states too. I don't think PowerD did anything to reduce power draw until C3 states were manually enabled. I'm actually not sure if PowerD makes a difference even with C3 is enabled, I just had it set to minimum and left it there.

It does. The C3 states put the CPU to sleep, but at C1 they are running at whatever frequency you set (= full spoeed). Powerd now identifies low load situations and redices the frequency of the CPU, and this also saves a bit.
 

Mike9474593

Member
Aug 5, 2022
99
65
18
I am looking for the 6port variation and it doesn't seem like v5 is in that flavour I am looking for. What did they change with the 5th version?
v1 was with 2 USB Ports, v2 probably 4 USB Ports (not sure if that's the real v2, somtimes v2 is also the Bluetec Board), v3 had DP, USB-C and dual NVME. See @dums answer. Just make sure you got the right version with the connectors you like, some are with mSATA, some are with NVME, some have i225 some i226, different display outputs, serial connector.

What did you polish with? Usually with CPU's I remove paste with rubbing alcohol and apply one that does the job more efficiently.
I did remove the paste and polished the plate with very fine sanding paper (special polish cloth witch is used in high precision producing environments for metal parts, eg nozzes, if you do it right the hight difference is only a few µm. And no, the cooper plate was not flat...). I used to work in that industry, so I'm equipped.

Really? Noctua fans are usually really good but I guess 40mm wouldn't cut it. Would be nice if there was a method to mount a larger fan inside the chasis for better cooling but then again it would defeat the purpose of the small footprint. How do you ask Changwang to do this? Send an email?
40mm simply did not have not enough airflow... I will try with some standoffs and for having more space for a 80x80x25mm fan. 80x80x10mm is not that common and good fans are very rare.
You can just write them on Aliexpress, they answer really fast.

Also, should one be worrying about BIOS updates for something like this?
Changwang is offering BIOS update, installing is a breeze (just burn the iso to a usb drive, remove nvme and let it do it's duty). You just need google translate to find the correct file. But of course the availability differs on the board version, v3/v4 got several updates, other boards did not get one. But not every update is fine, i read here it run worse with the new BIOS. So far I have the latest BIOS and no issues (knock on wood).
 
Last edited:

andrea87

Member
Oct 15, 2022
62
86
18
36
North-east Italy
For those of us using an hypervisor (proxmox) on those machines, there's a few things we can change to get better performance / lower idle loads.

Under your VMs, options, for stuff running just a shell (also PFsense / OPNsense), disable "use tablet for pointer". This absolutely drops the average cpu load, for me it went from 7-8% idle to 3%. Also Netdata shows a significant change in cpu usage, one of the two cpus always stays down at 400MHz and the second one at barely 1GHz. 85%+ of the time the cpu are in C3 acpi low power state. Netdata itself is also a bit cpu hungry, I've set it to poll every 5 seconds to keep cpu usage to a reasonable minimum.

1667933789194.png

There is also This post on Proxmox forums.

Install cpufrequtils and sysfsutils. This is what my /etc/sysfs.conf shows for my dual core mini pc:

Code:
devices/system/cpu/cpu0/cpufreq/scaling_governor = powersave
devices/system/cpu/cpu1/cpufreq/scaling_governor = powersave
devices/system/cpu/cpu2/cpufreq/scaling_governor = powersave
devices/system/cpu/cpu3/cpufreq/scaling_governor = powersave

devices/system/cpu/cpu0/cpufreq/scaling_max_freq = 3500000
devices/system/cpu/cpu1/cpufreq/scaling_max_freq = 3500000
devices/system/cpu/cpu2/cpufreq/scaling_max_freq = 3500000
devices/system/cpu/cpu3/cpufreq/scaling_max_freq = 3500000

devices/system/cpu/cpu0/cpufreq/scaling_min_freq = 400000
devices/system/cpu/cpu1/cpufreq/scaling_min_freq = 400000
devices/system/cpu/cpu2/cpufreq/scaling_min_freq = 400000
devices/system/cpu/cpu3/cpufreq/scaling_min_freq = 400000
Max boost frequency on mine is 3.5 GHz, min I've set it to 400MHz (you'll need to adjust this in the bios under advanced cpu configuration - cpu flex ratio override - set to 4 or whatever you like). If you have a quad core cpu you'll have to do the same for cpu 0 to 7.
 
Last edited:

T.Sharp

Member
Oct 22, 2022
54
91
18
Well, I can confirm the approach works on my end on an older J3160 4 port "Protectli" clone. Everything stable, even after running 40 days or longer. That's the reason why i wrote this short guide. However, as I only have one system for testing, I can't guarantee it works all the time.

What also works is to use powerdxx. It shows better results on my machine. To find the possible frequency settings you can use, run the command
% sysctl dev.cpu.0
dev.cpu.0.freq_levels: 1601/2000 1600/2000 1520/1900 1440/1800 1360/1700 1280/1600 1200/1500 1120/1400 1040/1300 960/1200 880/1100 800/1000 720/900 640/800 560/700 480/600
Glad I found your guide, and nice to see you here!

Strangely It appears my N5100 box isn't supported with either powerd or powerd++.
I'm not sure what's missing but it doesn't list supported freq levels, just "1113/-1"

1667940128358.png

Trying to start powerd++, it says the driver isn't supported :

1667941597750.png

Just polling the cpu freq, it doesn't go below 2100MHz unless I enable C2 or C3 in Tuneables, then it will idle down to 800MHz, regardless of powerd settings. I found no difference in power draw between C2 and C3 also.

EDIT : This thread explains the issue. Newer SpeedShift equipped CPUs are supposed to handle downclocking / power saving by the CPU itself, making powerD obsolete.

I checked the efficiency setting using
Code:
sysctl dev.hwpstate_intel.0.epp
and it showed 50 by default.

Manually setting each core to 100 allowed the cores to downclock to 800Mhz :
Code:
sysctl dev.hwpstate_intel.0.epp=100
sysctl dev.hwpstate_intel.1.epp=100
sysctl dev.hwpstate_intel.2.epp=100
sysctl dev.hwpstate_intel.3.epp=100
(To make the epp=100 stick after reboot, you have to add them to /etc/sysctl.conf)

EDIT: Just set these inside the Tunables menu from the GUI.

With just those hwpstate settings (the C-state tuneables removed), it idles at 800MHz and draws 7.6W, vs. 8W with C-states disabled and default epp=50.

So it looks like manually enabling C2/C3 in opnsense also makes SpeedShift function correctly, without the need for modifying pstate rules. It's also slightly more efficient, idling at 7W.
 
Last edited:

T.Sharp

Member
Oct 22, 2022
54
91
18
So playing around with the p-state settings in opnsense a bit more :

epp=100 : CPU locked at 800MHz, 7.6W at idle, 9-10W during 430Mb/s speedtest
99 : CPU ranges from 800 - 1600MHz, a bit more power draw, no difference in performance with my setup
75 : 1400 - 2800MHz
50 : 2100 - 2800MHz
25 : CPU locked at 2800MHz, idle power draw is 8W, speedtest is 13.5W

C-states were set to default C1 while testing.
 

Becks0815

Well-Known Member
Oct 15, 2022
216
277
63
Glad I found your guide, and nice to see you here!

Strangely It appears my N5100 box isn't supported with either powerd or powerd++.
[...]
EDIT : This thread explains the issue. Newer SpeedShift equipped CPUs are supposed to handle downclocking / power saving by the CPU itself, making powerD obsolete.
Yeah, sorry about the lost time, but I had no newer system at the moment to see how they work and if my changes still work. That's why I am also here - looking for a potential upgrade of the box.

At least the C-states are settings I would enable. I haven't noticed any real delays during all the time I use them on the FW.
 
  • Like
Reactions: T.Sharp