Topton Jasper Lake Quad i225V Mini PC Report
- Thread starter ReturnedSword
- Start date
Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.
Number breakdown for the Ryzen CPUs. That way you don't end up getting stuck with an old microarchitecture.
Attachments
told you guys ... lap and gap
fixes 90% of thermal issues on this machines. The copper block is not plain .... mine was concave on the cpu part and convex on the case part
yeah....shims have the risck of puuting to much presure on cpu ... that is why i think carefully and slowly grinding the mobo standoffs. In the way you can control when CPU touches the copper gently, grind a bit ..check .. grind a bit check ... eventualy the cpu will touch the copper with less risc of to much pressure
bad part is that you cannot sand down equaly the case .... so you might end up screwing the plane between the copper and case even more.
also the more you grind the case ... the larger the cpu gap becomes if you dont fix it by grinding the standoffs or shims
I can confirm this unit is not the only one with the gap problem. I have a Qotom J4125 unit, it was always running a bit hot (running Opnsense), about 54° in ilde. I checked it and it has a nice gap between the aluminum block (ugh) and the CPU. Added a 0.3mm shim and now it is about 15-20° cooler, about 35° in ilde. The board is fixed with 8 screws, four around the cpu and four on the edge. Let's see how it's working...
I'd rather mill up a new copper block than grind off the stand-offs!
LOL I totally agree - it's easier to do a 0.2mm + 0.3mm shim and then move to a 0.3mm+0.4mm shim if necessary - you cannot say the same about grinding the standoffs - too much work and you cannot correct if something goes wrong
Also - because the mounting is such low tension/pressure, there's no sense in lapping the surfaces and making them mirror smooth - I just polished with 320/600/800 grit.
There's also no sense in using some exotic thermal paste - just a thin layer of anything reasonably good like NT-H1 or MX-5 is enough
I was going to try those Graphite Patches but I had some NT-H1 kicking about so I used that last time.
yeah I got the pad for 12 bucks and cut several rectangular strips and used those. The IC pad *needs* a high tension mount i.e. a lot of pressure. With low pressure the pad is not as good as paste
Each shim adds paste-copper-paste layer with some extra mess + thermal performance penalty. Also I suppose in worst case scenario standoffs can be easily replaced.
if you can mill a new block ... by all means that is the best solution, but not all of us have access to that ...
if you have access to a mill sky is the limit ... you can mill the case and make a custom bigger cooper block .
yes lapping to a mirror finish is not necessary but the plane corection is ...and to get a almost mirror look you just need about 10 passes on a 1k /2k grit sandpaper ... not much work effort ...
now im still trying to fine tune this .... in current room temps of about 22-23c on full load i see max of 63c, no fans, no nothing, for a full hour .... that is really good in my book .
what i notice is that the freq starts at 2.7-2.75 ghz all cores for 128 seconds (i set it in the bios) but then they start to drop to the 2.5ghz where they stabilize.
i put the PL1 to 20w and PL2 to 25w .... and no power or theermal limits where tripped package power never went above 20w .... but i still get a down clock....
what would be nice if it will boost to max all untill it reaches the thermal limit....
anyway bioses are crap and most of the settings do not apply anyway
if you have access to a mill sky is the limit ... you can mill the case and make a custom bigger cooper block .
yes lapping to a mirror finish is not necessary but the plane corection is ...and to get a almost mirror look you just need about 10 passes on a 1k /2k grit sandpaper ... not much work effort ...
now im still trying to fine tune this .... in current room temps of about 22-23c on full load i see max of 63c, no fans, no nothing, for a full hour .... that is really good in my book .
what i notice is that the freq starts at 2.7-2.75 ghz all cores for 128 seconds (i set it in the bios) but then they start to drop to the 2.5ghz where they stabilize.
i put the PL1 to 20w and PL2 to 25w .... and no power or theermal limits where tripped package power never went above 20w .... but i still get a down clock....
what would be nice if it will boost to max all untill it reaches the thermal limit....
anyway bioses are crap and most of the settings do not apply anyway
Last edited:
That is strange. The 4 core turbo multiplier is 28 which means you should get sustained 28 x 99.784 = 2793.952 MHz unless the SOC trips on power or thermal limits.
For example here's a TS Bench 8GB run on my system after 1 minute - just after a run with 120M and another with 960M to build up some heat. I've let the system run for quite some time and it's been consistent at the 28x multiplier for all 4 cores. (all my BIOS settings for power are default)
What are your results for the 120M and 960M benchmark runs in ThrottleStop? Does the multiplier stay constant at 28x for all four cores during those runs?
Maybe reset BIOS to default settings (default limits for PL1, PL2, PL3, PL4 and Turbo Performance) and try again?
for some reason in TS bench is not going to 28x or 29x .... is just capping at 25x .... prime95 is running 28x but the freq varies a lot
my bios is also default ... only PL1 is modified to 20w and PL1 time to 128 sec
cause if you leave the default 15w PL1 after the PL1 time is done, it will drop to 15W and in my case it will run at about 2.3-2.4 ghz
im on V103 bios .... default has only 2 PL ...and they are disabled by default)) . You can activate the PL3 and 4 ...but with them disabled default is 15/25w PL
maybe is silicon lottery
im ok with it ... 128sec for full boost all cores should be more then enough for any daily stuff this machine will encountr. so i think ill leave it to 15 w PL1 and 128 Sec PL1 limit
my bios is also default ... only PL1 is modified to 20w and PL1 time to 128 sec
cause if you leave the default 15w PL1 after the PL1 time is done, it will drop to 15W and in my case it will run at about 2.3-2.4 ghz
im on V103 bios .... default has only 2 PL ...and they are disabled by default
)) . You can activate the PL3 and 4 ...but with them disabled default is 15/25w PLmaybe is silicon lottery
im ok with it ... 128sec for full boost all cores should be more then enough for any daily stuff this machine will encountr. so i think ill leave it to 15 w PL1 and 128 Sec PL1 limit
Last edited:
That's not how the silicon lottery works. What would happen is that you would get a SoC with higher VID, meaning Vcore gets higher => more heat dissipated. As long as you can sink 30W of thermal load, you shouldn't throttle and therefore maintain the 28x multiplier regardless of whether you have a SoC with higher VID or lowermaybe is silicon lottery
- What is the power brick you are using for the PC?
- Are both chiplets on the SoC covered by the copper block? It has two screw holes - is any part of the chiplet under this?
- Does changing the Windows power plan from Balanced to High Performance make any difference?
One of the Topton 4-port i226 w/ N6005 barebones is lowed to around $220(link) and I haven't found its major differences to this one(link), except for the later one's case. But its case has no heat pipes, I don't think this case make its price difference higher to $40. And I only found the later one on CWWK's store.
By the looks of it, it appears it could be cheaper because they skimmed on beefy heatsink with fins but then there is breathing grill on either side for better air circulation. Would be interesting to know if anybody benchmarked thermal performance.
With so many people reading/writing here, I hope I have a chance to get some info about the actual power consumption of the N5105 and the J6412 one (which I just found on page 69 I think). I read something about 11W for the N5105 idle, is that correct, even after some tweaking of the Bios and/or running programs like powerd to reduce the consumption during runtime?
Anyone else has some data about theose two boards (N5105 and J6412), please?
Anyone else has some data about theose two boards (N5105 and J6412), please?
Topton just released a new two port i226 J4125 for around $130. For those wanting to separate their network controllers (recommended) and run something like Pi-hole this is a great option.
I have a $25 7" IPS display off AliExpress that I use in my closet as my modem and router are connected at the drop. It makes changing my LAN subnet a breeze.
I have a $25 7" IPS display off AliExpress that I use in my closet as my modem and router are connected at the drop. It makes changing my LAN subnet a breeze.