Topton Jasper Lake Quad i225V Mini PC Report

[sqrwv]

AIDA64 does not qualify as a real stress test, never did.

For full load stress testing please use Prime95 and small FFT for like 30-60 minutes and report back. Thou im pretty sure it will throttle much sooner.

I also did prime95 with same result and that aida test even makes the cpu hotter than prime95, just try it.
Also @burtal test was limited to 2.6ghz with aida and then even drops to 2.2Ghz, so I matched his test case, as I was replying to him. Please see his post I replied to.

 

[Immortal]

I also did prime95 with same result and that aida test even makes the cpu hotter than prime95, just try it.
Also @burtal test was limited to 2.6ghz with aida, so I matched his test case, as I was replying to him. Please see his post I replied to.

As you quoted me i thought you were replying to me.

Did you use small FFTs in Prime 95? Coz there is no way that AIDA can match it especially for longer run. As for the difference(you hitting 2,8 GHz for full load, long run), maybe you got lucky and won silicon lottery, which means your CPU can do more with lower or the same voltage. Good for you.

 

[sqrwv]

As you quoted me i thought you were replying to me.

Did you use small FFTs in Prime 95? Coz there is no way that AIDA can match it especially for longer run. As for the difference(you hitting 2,8 GHz for full load, long run), maybe you got lucky and won silicon lottery, which means your CPU can do more with lower or the same voltage. Good for you.

I was replying to you, but the aida FPU test I linked was a reply to Burtal original post, where he is limited to 2.2Ghz in that test. If you follow the link, you'll find his original post with the details.
Those posts was the first time I used aida FPU stress test, as I said to match his test.

I always use prime95 to stress test, because I can enable checksum and find if machine is stable.
I did that here with the same results, controlling temp., freq. and power, for more than an hour.

Probably silicon lottery plays a role as you say.

PS: I don't think N5105 has the AVX instructions that really load cpus and drive temps. up, so probably prime95 is just using SSE here.

 

[morini]

Ain't it fun having to re-do the entire house for lower Leccy cost

Ain't it just!

I want to go back to the time I could run all my stuff on one sturdy ex enterprise server and not worry too much about the cost. Wife won't have it at the current leccy rates though :-(

I hear what you say about headaches and downtime, but I've ordered the RAM now so I feel obliged to give opnsense a go under proxmox. I won't fight with it and if I get any pain at all I'll just put a smaller SODIMM in it and run opnsense bare metal. Seems a waste of capable (and low power) hardware though. I'm sure the issues are fixable, it seems folk are having variable experiences on newer kernel versions. Frustrating really.

Routers have to be 100% reliable at the end of the day. Unless you're happy to soak up a ton of abuse from unhappy wives and daughters!

 

[bemore]

I've just received a TopTon N5105/i226 barebones box. My NVME drive + RAM are arriving tomorrow, but I thought I'd share a few pics before I get it all setup and run it through some tests.

View attachment 27352View attachment 27353View attachment 27354View attachment 27355View attachment 27356View attachment 27357View attachment 27358View attachment 27359View attachment 27360View attachment 27361

I tried to get some decent pics of the CPU/block/heatsink
View attachment 27362View attachment 27363View attachment 27364

My Crucial P3 500GB NVME drive and Crucial 16GB RAM kit turned up today so I got started on some testing. TL;DR I don't think I have any thermal problems. I didn't tweak anything in the BIOS, nor did I fiddle with the thermal paste/heatsink, before running the Prime95 test for about 2 hours.

I gathered some temperature data (average CPU core, drive and ambient) plus some power draw stats as reported by HWiNFO and my Tapo P110 smart plug. Thought people here might be interested

Edit: Just added another chart showing CPU Core Usage vs Clock Speeds during the same Prime95 run.

 

[bemore]

Oh, and HWiNFO output...

 

[skimikes]

My Crucial P3 500GB NVME drive and Crucial 16GB RAM kit turned up today so I got started on some testing. TL;DR I don't think I have any thermal problems. I didn't tweak anything in the BIOS, nor did I fiddle with the thermal paste/heatsink, before running the Prime95 test for about 2 hours.

I gathered some temperature data (average CPU core, drive and ambient) plus some power draw stats as reported by HWiNFO and my Tapo P110 smart plug. Thought people here might be interested

Edit: Just added another chart showing CPU Core Usage vs Clock Speeds during the same Prime95 run.


View attachment 27404

The drive temp is concerning since that's throttle territory for nvme. Can you run the same test with the unit on its side? Yes, it sounds stupid, but I saw a 5C drop in drive temps just by having the unit vertical rather than horizontal. I'm curious if that will be reflected in your unit and thus the chart.

 

[fta]

Mine can sustain 2.8Ghz all cores full load with P1=P2=20W (tested more than an hour):
https://forums.servethehome.com/ind...225v-mini-pc-report.36699/page-93#post-356099

Your graphs show you're actually being throttled to 17.7W. HWINFO can tell you what you're being limited by. I'll bet it's VR TDC.

 

[sqrwv]

Your graphs show you're actually being throttled to 17.7W. HWINFO can tell you what you're being limited by. I'll bet it's VR TDC.

What you mean being throttled, if all cores are doing 2.8GHz all the time?
17.7W it's what the cpu needs at 2.8, it could go up to 20W, but doesn't need it.
The cpu can only do 2.9Ghz single core and it's not unlocked.

 

[fta]

What you mean being throttled, if all cores are doing 2.8GHz all the time?
17.7W it's what the cpu needs at 2.8, it could go up to 20W, but doesn't need it.
The cpu can only do 2.9Ghz single core and it's not unlocked.

Is the CPU only capable of hitting 17.7W? If yes, never mind. In chips that go higher, when you set PL1=PL2=20 and don’t hit it, you’re being throttled.

 

[sqrwv]

Is the CPU only capable of hitting 17.7W? If yes, never mind. In chips that go higher, when you set PL1=PL2=20 and don’t hit it, you’re being throttled.

Yes, mine does always 17.7W (in that test) with all cores at 2.8Ghz, but others can have better or worst values because of silicon lottery.

 

[bemore]

The drive temp is concerning since that's throttle territory for nvme. Can you run the same test with the unit on its side? Yes, it sounds stupid, but I saw a 5C drop in drive temps just by having the unit vertical rather than horizontal. I'm curious if that will be reflected in your unit and thus the chart.

I did two Prime95 runs, one with the chassis vertical and the other horizontal. I tried to cram all the data on to one chart but it looked like a mess. I included the ambient temperatures for reference on each chart. I forgot to stop the vertical test, so it was under load a bit longer, not that it's significant for what we're looking at.

While the chassis was placed vertically, I had a plastic box under each end, away from the underside vent, raising it ~3cm from my desk.

I've included a summary of the average temperatures where the CPU core load == 100%.

 

[romprod]

Fixed my own issue, one of the guides had the following command
echo "options i915 enable_guc=2" > /etc/modprobe.d/i915.conf
I found another thread with someone suggesting to run this to see if guc is enabled, mine wasn't
dmesg | grep -iE "huc|guc|dmc"
I edited /etc/modprobe.d/i915.conf and removed the quotations, rebooted and it's now functioning as expected, not really sure why but I'll take it as a win

Thankyou so much for this! Been trying for days to get this working.

 

[skimikes]

I did two Prime95 runs, one with the chassis vertical and the other horizontal. I tried to cram all the data on to one chart but it looked like a mess. I included the ambient temperatures for reference on each chart. I forgot to stop the vertical test, so it was under load a bit longer, not that it's significant for what we're looking at.

While the chassis was placed vertically, I had a plastic box under each end, away from the underside vent, raising it ~3cm from my desk.

I've included a summary of the average temperatures where the CPU core load == 100%.
View attachment 27432View attachment 27431View attachment 27430

Thanks for this! In short, don't place the horizontally designed unit horizontal.

 

[tusk9541]

These might benefit a bit by having some vents on the heatsink itself, but as they come there are other things that impact heat dissipation a lot more, like improper pressure/contact of the copper block, though it may be fixed in newer designs like with these 2 pressure screws near the CPU: https://forums.servethehome.com/ind...ort.36699/page-125#lg=attachment27251&slide=0

 

[T.Sharp]

I think the cooling issues are a tad overblown, we're looking at temps at 100% sustained load, where I think in many cases (at least mine with basic OPNsense) the average load will be pretty low. Even with the ~0.3mm thermal paste gap on the CPU cooler, I'm sure mine would function just as smoothly as it does with all the mods.

I'll still do the mods anyway, but I don't think it's necessarily a requirement.
Currently running vertically, idle cpu temp ~33c with no fan

 

[Immortal]

If you're doing baremetal OPNsense without any IPS/IDS or other resource intensive things then N5105 is a total overkill/kinda a waste of resources.

People who talk about 100% load etc, usually are doing something more, so it matters to them.

 

[T.Sharp]

If you're doing baremetal OPNsense without any IPS/IDS or other resource intensive things then N5105 is a total overkill/kinda a waste of resources.

People who talk about 100% load etc, usually are doing something more, so it matters to them.

Yeah the N5100 is certainly overkill with my current setup, but at $110 for 4x2.5Gb, it didn't make sense to buy an older model. I'm just saying the silicon can handle the heat. Plenty of laptops out there running 90-100c for years without issue, so if you're hitting 70-80c at 100% load, that's still well within design spec. Cooking the SSD would be my main concern, especially with the systems where the NVMe sits on top of the VRM.

Thinking of drilling some speed holes in the bottom cover plate next time I'm bored, to get some passive airflow for the SSD

 

[Stovar]

I screwed my mounting plate on my wall and went to hang the unit before realising the small included internal cooling fan and air vents then get completely blocked!

So now just kind of use the mounting plate to stop any burn or heat discolouration marks and just kind of wired and hanged it on a screw above it then rotated it onto the open vented side to expel warm air.

I noticed the nvme/ram and entire unit there is some heat coming off it, imo it does not look ideal or good for nvme/ssd or ram health so for peace of mind I think ill be ditching the internal fan which makes an slight electrical whine and strap an 120mm silent fan on the back of it.

 

[Stovar]

Does that black panel bit come off and how did you get it off?

I was going to take a flat head screwdriver and hammer to it, feels like its welded on before I damage it.