CPU [24/03/2025]ASRock DeskMini X600 BIOS "4.06" & "4.08" & "4.10" - with new AMD AGESA ComboAM5Pi v."1.2.0.2a" & DDR5 SO-DIMM Voltage Settings available
- Thread starter HydrAxx747
- Start date
Hi Gents!
Thanks to this forum and this thread in particular I have cooked me an x600 mini instead of and old HP Elitedesk 800 G1 mini
I have basically followed the recipe from this thread and here is what I have:
X600 mini stock bios 2.03
Ryzen 8700G
2x16Gb Kingston Fury Impact KF564S38IB-16 that run on it's native 6400 from XMP profile selected in BIOS
AXP90-X47 FULL with stock fan
Arctic Cooling MX-4 thermal interface
Intel WiFi 9260 AC
Kingston A2000 SSD
Stock AsRock PSU
All stock i.e. no overclocking except CPU Curve optimizer set all cores -40 in BIOS.
+USB2.0 board and rear audio jack from Ali
And in reality I'm fully happy with the build but 2 points below so I would be more than grateful for comments and directions on where to dig:
1. Temps. It keeps around 40C idle and when I run a stress test it gets to the 90C in around 1 minute and 20 seconds and stays at this temp forever. It does drop the CPU power from around 86W to 81W and that is it. Fully stable here, no throttling according to the monitoring tools (OCCT).
I double checked the CPU-to-Radiator footprint and it is absolutely OK (20+ years of experience with computers), the fan profile is to Standard so I do hear it spins up and the monitoring confirms this.
So the question here i have - is this normal behavior? I expected lower temps on this CPU-cooler pair or am I wrong?
2. Fan noise. The Thermalright stock fan is noisy... and still noisy at CPU temp 40C where BIOS keeps it at 1700rpm. Do i need to update BIOS for more aggressive FAN speed adjustments or something else?
My future plans are:
1. Install memory heatsinks (came from Ali, same as mentioned in this thread), btw, those who istalled ram heatsinks, did you use stock thermal adhesive or replaced with something, why I am asking is because the chip in a middle (SPD I believe) is around 0,5 mm higher that memory chips so the heatsink will not go straight forward..
2. memory overclock, to maximizethe iGPU performance
3. Install 120mm fan, I've bought the Silverstone Air Slimmer 120 and it fits basically same way as the Thermalrigt stock fan, but thinking of some nice way to mount it. I remember somebody mentioned about 120mm fan in x600 mini but can not find the post.
4. Install 2x40mm fans to improve hot air exhaustto the perforated side of the mini's case, ordered 2 noctua 40x10mm PWM fans.
So i apologize for long post, but it's kinda introduction + questions: 2 in one
While you wait for someone with an AXP90-X47 to reply to your post I'll share my guesses based on my experience with my builds (you can find them in my posts).
Since you have plenty of experience with computers, I would assume that the thermal paste is not the problem and that 80W is probably a bit on the edge of what that cooler can dissipate inside a small case like the X600 DeskMini. I would recommend you limiting the power consumption to 65W (including boosting behaviour) since it should be plenty unless you really need to squeeze as much performance as possible. I've done tests between 65W and 85W and the difference is too small to be worthwhile, even on gaming with OCed memory.
Your stock power supply would also appreciate not going beyond 65W on the APU side since it is not recommended to push beyond 75% of their power ratings and 65W APU + drives + memory + VRMs losses is probably around that.
Although a 120mm fan sounds like a good idea, the top of the case doesn't have a grid big enough to accommodate one which could result in no cooling improvement, specially over the ram slots. Also, placing a fan very close to a grid makes it much louder than it actually is.
Regarding the fan speed curve, the deskmini x600 BIOS has a decent fan curve editor, allowing pretty good control over the CPU fan speed based on CPU temperature.
Last edited:
Thanks for your advise. Actually I have already played with the limits in BIOS and also found that even 54W +Curve Optimizer (all cores -47) gives nearly zero performance drop in games and real app use-cases, but keep the temps much lower like ~70-75C at stress tests while fully stable. So i'll leave this as is for now. Why i made my post is i first thought my cooler might be defective or something, but now i see this should be ok.
Yes, now i managed to get it work (it's not obvious on how to adjust it) and have it silent at office loads, will make some further tests on 120mm fan so will see, thanks anyway.
So besides the real experience of the AXP90-X47 FULL users in x600 Mini, I am really interested on howto mount RAM heatsinks, maybe somebody could share some experience. What i think now is to have 1mm thermal pads for memory chips and 0,5mm for that high SPD chip for one side and 0,5mm thermal pads for the backside of the modules since it's flat and fix them with a kind of black paperclip (aka binder clip).
The Thermaltight AXP90-X47 FULL and it's current performance in the mini doesn't want to leave my head alone. The idea i have now is that it's fins are in horizontal position, when the mini is vertical, so the airflow between the fins should be struggling as hot air tries to go up while we need to spend fan power to move it sideways.. I remember some tests from the past then the first tower-type CPU coolers appeared the fins position and fan position to the fins made noticeable difference in temps.
So since unfortunately we can not rotate the AXP90-X47 90 degrees left or right with stock mount kit, why don't make a handmade mount for it. It's not a mount actually but 2 straight 9x100mm 1,5-2mm thickness rectangular metal bars (maybe wrong word, English is not my native, apologize) with 2 threaded 2 non-threaded holes in each. The mod seems to be fairly simple as to made those custom mounts you will just need 1,5 or 2mm thickness metal pieces, angle grinder or Dremel to cut it, 2 drills and one tap to make thread.
I have made a quick sketch (all measurements are subject to double check) on the AXP90-X47 FULL drawing taken from the Thermalrigh website to demonstrate how it should look like (barely visible but marked with green) and a simple sketch of the new mounts on the top-left corner of the picture.
What do you think, is it worth rotating the cooler's heatsink to simplify airflow and lower the temps? Or maybe somebody already made those kind of experiment...
PS. I will definitely give it a try and post results here, just need to find some time.
So since unfortunately we can not rotate the AXP90-X47 90 degrees left or right with stock mount kit, why don't make a handmade mount for it. It's not a mount actually but 2 straight 9x100mm 1,5-2mm thickness rectangular metal bars (maybe wrong word, English is not my native, apologize) with 2 threaded 2 non-threaded holes in each. The mod seems to be fairly simple as to made those custom mounts you will just need 1,5 or 2mm thickness metal pieces, angle grinder or Dremel to cut it, 2 drills and one tap to make thread.
I have made a quick sketch (all measurements are subject to double check) on the AXP90-X47 FULL drawing taken from the Thermalrigh website to demonstrate how it should look like (barely visible but marked with green) and a simple sketch of the new mounts on the top-left corner of the picture.
What do you think, is it worth rotating the cooler's heatsink to simplify airflow and lower the temps? Or maybe somebody already made those kind of experiment...
PS. I will definitely give it a try and post results here, just need to find some time.
I would try to avoid using thermal pads between the ram and the heatsinks. The SPD chip doesn't need to be cooled so I would drill a hole in the ram heatsinks where the SPD chip would collide so you can fully seat them using the stock thermal tape. That would transfer heat much better than any thermal pad.
I have no idea if rotating the CPU heatsink by 90° would improve cooling performance, it is one of those things that you only know when you try, but the mounting mechanism is very important. You would need to make it exactly right, this is, perfectly leveled while applying the right amount of mounting pressure otherwise you can easily loose cooling performance.
To me there is nothing better than open case cooling, but I love any type of mods.
One more thing: before overclocking the memory, remove any type of overclock from all components (curve optimizer, etc), once you get a stable memory overclock start overclocking the rest of the system (GPU and CPU).
Last edited:
just wanted to chip in on the x47 conversation. I know it's not an actual solution for most people but I'm just finishing up a custom case for the x600 (just some handcut carbon fiber panels bolted together), size should be identical if not smaller then the original case but since I mostly got rid of the bottom tray for the sata ssd's I managed to fit an x53. I'll be adding some 40mm fans since with some general testing they did help out, but have yet to decide the exact setup.
I'm running a 9800x3d (180w psu, completely unlimited in bios, -30 all cores) and while the x47 was perfectly capable of cooling it in most daily tasks, during different stress tests I saw a 15/20w reduction in power draw (from 125/130 to 110/115W) and a slow down of a about 250mhz. with the x53 I have zero issues of this sort, anything that isn't a proper load and the mini pc is dead quiet (I use Fan Control).
I'm running a 9800x3d (180w psu, completely unlimited in bios, -30 all cores) and while the x47 was perfectly capable of cooling it in most daily tasks, during different stress tests I saw a 15/20w reduction in power draw (from 125/130 to 110/115W) and a slow down of a about 250mhz. with the x53 I have zero issues of this sort, anything that isn't a proper load and the mini pc is dead quiet (I use Fan Control).
more then welcome to help test things out. I'm using the stock thermalright fan, blowing down on the cooler as it came out of fhe box. the fans I'll be adding should be pulling air out, haven't decided yet whether to add 4 pull fans, 1 push and 2 pull or just 2 pull. probably down the line I'll be testing the noctua swap, by consulting some db-nornalized tests the thermalright is superior to the noctua in terms of cooling, but the noctua fan is better if you temp-normalize the chart.
So i did some tests comparing vertical vs horizontal positioning of the cooler. I noticed that the number of holes on the top panel of x600 is nearly same to back side, so run OCCT stress test and after temps got stable, i just rotated the whole mini 90degrees to make it stand on front panel. I made this test many times actually having compared 2 FANs: stock and Silverstone Air Slimmer 120 in two positions (blowing in (stock position) and out - the Slimmer was not tested to blow out as fan is blocked from spinning in this position). All tests made with Silent fan preset in BIOS, 54w and 65w power limit and CO -45 all cores.
As a result,
Stock fan:
- blow in vs blow out : -2 degrees when blows out
- blow in, stock oriented (heatsink fins horizontal) vs face down (heatsing fins vertival) : -2 degrees when fins vertical
- blow out, stock oriented (heatsink fins horizontal) vs face down (heatsing fins vertival) : no difference
- RAM (naked, no additional heatsinks) temp - +-1 degree, same temp
- AS120 vs stock fan blows out : +5-6 degrees with AS120 that was my hope to improve cooling.
- AS120 blows in horizontal vs vertical : -3 degrees when fins vertical
- RAM (naked, no additional heatsinks) temp AS120 blows in vs stock fan : -4-5 degrees with AS120
It's pity that the 120mm Silverstone fan didn't give any reduction in CPU temp (it's actually more noisy compared to stock one), being at the same time the best slim 120mm fan in terms of perfomance (CFM and pressure).
My next test would be to put 2 40mm noctua fans (arrived today) on the x600 top panel, but it looks like that to do this i will need to do something with VRM heatsink as otherwise fans will not fit. And will probably try to find a bettter 92mm slim fan, but according to the SFF Master list the stock fan of the x47 is not too bad and fan swap will probably not give any CPU temp drop or acoustic confort at high load.
I will leave it alone for now with stock fan, blows out, 54W power limit and CurveOptimizer -45 all cores with FAN curve adjusted to have 72-73C at stress (OCCT) test and around 45C idle but silent.
Just embrace open case cooling and go bigSo i did some tests comparing vertical vs horizontal positioning of the cooler. I noticed that the number of holes on the top panel of x600 is nearly same to back side, so run OCCT stress test and after temps got stable, i just rotated the whole mini 90degrees to make it stand on front panel. I made this test many times actually having compared 2 FANs: stock and Silverstone Air Slimmer 120 in two positions (blowing in (stock position) and out - the Slimmer was not tested to blow out as fan is blocked from spinning in this position). All tests made with Silent fan preset in BIOS, 54w and 65w power limit and CO -45 all cores.
As a result,
Stock fan:
Silverstone Air Slimmer 120 (AS120), blows in:
- blow in vs blow out : -2 degrees when blows out
- blow in, stock oriented (heatsink fins horizontal) vs face down (heatsing fins vertival) : -2 degrees when fins vertical
- blow out, stock oriented (heatsink fins horizontal) vs face down (heatsing fins vertival) : no difference
- RAM (naked, no additional heatsinks) temp - +-1 degree, same temp
So from this test I decided that untill I stay in stock case I will stay with stock fan blowing out, with stock CPU cooler heatsing orintation.
- AS120 vs stock fan blows out : +5-6 degrees with AS120
- AS120 blows in horizontal vs vertical : -3 degrees when fins vertical
- RAM (naked, no additional heatsinks) temp AS120 blows in vs stock fan : -4-5 degrees with AS120
It's pity that the 120mm Silverstone fan didn't give any reduction in CPU temp (it's actually more noisy compared to stock one), being at the same time the best slim 120mm fan in terms of perfomance (CFM and pressure).
My next test would be to put 2 40mm noctua fans (arrived today) on the x600 top panel, but it looks like that to do this i will need to do something with VRM heatsink as otherwise fans will not fit. And will probably try to find a bettter 92mm slim fan, but according to the SFF Master list the stock fan of the x47 is not too bad and fan swap will probably not give any CPU temp drop or acoustic confort at high load.
I will leave it alone for now with stock fan, blows out, 54W power limit and CurveOptimizer -45 all cores with FAN curve adjusted to have 72-73C at stress (OCCT) test and around 45C idle but silent.
you'll never go back
(more pictures: https://smallformfactor.net/forum/threads/asrock’s-x600-deskmini-–-finally-–-again.19619/post-290101)
Stress test at 65W plus 6W each RAM dimm (2x32GB 7600 MT/s double sided) with fan spinning at 600 RPM = 60ºC CPU, 60ºC VRM, 53ºC RAM dead silent
Last edited:
You're lucky to have that kind of free space and can leave mini on the desk. Mine hangs behind the monitor and goes to storage when i finish work together with it
btw, to save some 0.5mm on cooler height I replaced stock fan mount to plastic fasteners:
I have a question, when you installed the RAM heatsinks, have you pealed off stock ram stickers to apply heatsink straight to the chip? And how's the 3m tape, holds good so far? If you took them away, was it hard, are they glued well i.e. what's the best way to remove them without damaging ram chips?
Tks
btw, to save some 0.5mm on cooler height I replaced stock fan mount to plastic fasteners:
I have a question, when you installed the RAM heatsinks, have you pealed off stock ram stickers to apply heatsink straight to the chip? And how's the 3m tape, holds good so far? If you took them away, was it hard, are they glued well i.e. what's the best way to remove them without damaging ram chips?
Tks
I didn't peal off the stock ram stickers because I'd void the warranty otherwise, but they're so thin that I guess temperatures aren't affected more than a couple of degrees based on what I've been able to measure with a thermocouple.
The 3M tape is holding as new, I guess it's because it is specifically designed for this application (heat transfer tape instead of heat resistant tape) and because the heatsinks are very light compared to what they can hold. I even had to remove them a couple of times and I've been able to reuse the tape without loosing cooling performance or stickiness. To do that the best way I found is to push and twist instead of pulling to prevent ripping off the chips. It requires quite a lot of force but since you're not pulling I'd assume it is safe to do it, didn't have any problem so far.
Last edited:
I understand better now
why you get higher RAM temperatures than me, because of course these stickers have a significant impact on the thermal transfer to your heatsinks (I didn't ask the question at the time because I logically thought that you would have removed them to maximize the efficiency of your heatsinks), so I put in mind your high temperatures on the fact that your heatsinks are made of aluminum unlike mine which are made of copper (I also told myself that the thermal glue of your adhesive was perhaps less good too), but sincerely if you removed the original stickers, even if it is at the expense of a loss of warranty (the last time I saw RAM being damaged and dying little by little because of overclocking, and especially overvoltage was at the time of the DDR1 generation) then you could reduce your maximum temperatures by between 10 and 15°C easily, personally I am on Kingston 32GB (2x16) 6400MHz CL38 overclocked to 8000MHz CL38 (and optimized subtimings) under 1.43v in a 3D printed case with a 120mm fan (15mm thick) that can go up to 72cfm (with a pretty good static pressure) and a Thermalright AXP90-X53 Full Cooper CPU heatsink that barely hinders the airflow to the RAM and they have never exceeded 45°C so far.I have the feeling that the difference in temperatures comes from me using 2x32GB dimms vs your 2x16GB dimms, which should result in more power consumption. Could you please check the power consumption of each of your dimms using hwinfo while running for example Furmark? I've always wondered how much 16GB dimms at 8000 MT/s consume. I also run my sticks at 1.43v. Thank you in advance!I understand better now
I ran a little over 30 minutes of Furmark2 (the latest version is v2.7.0.0) in CPUBurner mode. I got 6.625 Watts (at maximum) on each stick in "Total Power." The VDD (SWA) of each varied between 1.390V (min) and 1.450V (max) depending on the load level, for an average of 1.434V (measured from the start to the end of the test). I reached a maximum of 44°C (SPD Hub Temp) for the hottest stick (closest to the CPU). Afterwards, I was in a room where the ambient temperature varied between 18 and 20°C depending on the time of day (during the test, the ambient temperature varied between 19 and 19.5°C).
There you go, I hope you have the information you wanted, and if not, I apologize for forgetting that you were using Dual-Sided on your two 32GB RAM sticks. It's much harder to cool than Single-Sided sticks, I agree. But I maintain that I'm sure that you would lose at least 5°C if you had removed the original stickers beforehand (the heat transfer to your heat sinks would be much better).
There you go, I hope you have the information you wanted, and if not, I apologize for forgetting that you were using Dual-Sided on your two 32GB RAM sticks. It's much harder to cool than Single-Sided sticks, I agree. But I maintain that I'm sure that you would lose at least 5°C if you had removed the original stickers beforehand (the heat transfer to your heat sinks would be much better).
For some reason when I run CPU Burner on Furmark 2.7.0.0 my RAM doesn't consume more than 1.0~1.5W, but with the regular benckmark they hover at around 6W on Hwinfo. The maximum I've been able to push them is around 7.5W during some TM5 tests.
It is interesting to see that your power consumption is around 10% more than mine. I wonder if the SPD Hub power consumption report is just for one side of the dimm, which would explain the 10% power difference between my 7600 MT/s vs your 8000 MT/s at the same voltages.
This Der8auer video suggests that the power consumption is proportional to the number of chips in the dimm (which also makes sense)
All I can say is that I have been unable to measure a difference bigger than a couple of degrees between chips on the same side of each dimm with the thermocouple (I was only able to measure chip edge temperature). The base of the heatsink was also at almost the same temperature. The SPD hub reported temperature were also in the same range, which makes me think that they whole dimm was saturated with heat at roughly the same temperature. It is hard to say how much temps would improve without the stickers though
Gents, your tests are very useful I believe, and very interesting to read. I can just add that i have tested my ram (2х16 Impact fury 6400 cl38 stock) with no heatsinks with OCCT ram stress test and temps easily reach out 75c fairly quickly and stay there +-2C. So it would be very interesting to see the results if you could test your ram with heatsinks with this tool, as to me it makes more intensive load than furmark.
Last edited:
Similar threads
