Log SFFTIME U-ITX : 7 Liters to tame the R7 9800X3D and the RTX 4080 FE!

[MarcParis]

-------------------------------------------------------------------------------------------------------------------------------
SFFTIME U-ITX : A new Odyssey starts : 7 Liters to tame the AMD R7 9800X3D and the NVIDIA RTX 4080 FE (29th December 2024)
-------------------------------------------------------------------------------------------------------------------------------

My interest in this case project from @riba2233 started when I noticed & tried the Minisforum BD790i, at the very beginning of 2024!
Finally I went with the FormD T1 for this build, a good one but not the mesmerizing one I was looking for.

A little bit later, I built again inside my beloved Cerberus-X a new setup based on AM5 and the excellent AMD R7 9800X3D!

The Cerberus-X demonstrated, once again, its polyvalence as I spent a lot of time benchmarking this setup. The overall conclusion was simple : the R7 9800X3D is very fast but tends to be power hungry (and distracting loud) if it is not tamed. Thus I tested several ways to undervolt/underclock it, aiming for its best efficiency. I even tested the ID Cooling IS-47-XT 47mm tall cpu cooler, versus the Noctua C14S and D12L!

That’s when @riba2233 informed me that the U-ITX was finally available and already on its way!
The stars were aligned!

NON FUNCTIONAL PROTOTYPE BUILD

Before investing into a new motherboard I decided to build a first prototype based on my only remaining working mini ITX MB using an ultra modern A10-5700K coupled with a legendary prolimatech samuel 17..







Comparison versus my N-ATX v2






Here are shorts remarks based on the initial prototype

• The U-ITX assembly was smooth and straight forward. I noted several improvements versus the N-ATX v2 :
  • The U-ITX is offering fully independent side panels with its stand! On the N-ATX, if you want to use the stand, you have to use both side panels.
  • The bracket to secure the GPU on the U-ITX is independent of the GPU. On the N-ATX you have to assemble a small bracket on the GPU itself
  • The PSU assembly is way easier in the U-ITX compared to the N-ATX
• The U-ITX seems not compatible with the motherboard rear i/o shield
• The U-ITX is very light (less than 1kg). However all parts are reinforcing its structure.
• To fit my RTX 4080 FE I had to disassemble the screws of the front panel, located at the bottom (x2)/middle (x1), but at the top. The front panel offers very good flexibility to allow enough offset to insert the RTX 4080 FE.
• Dislike the FormD T1, the U-ITX handling manual is accurate & effective. The GPU insertion requires to couple the GPU with its riser and only then to assemble it inside the case.
• There is tight clearance between the bottom/right motherboard screw and the GPU power connector. Be careful on some mini ITX motherboard compatibility like the Aorus X870i (PCIe unlocking system)
• The U-ITX, like the N-ATX, offers 2 setups to fix the PSU extension cable : internal plug or external pigtail (like on my Sliger SM550). I’m not fond of the external solution.
• Cable management promises to be challenging..

Based on this initial test, I decided to buy an AM5 mini ITX motherboard. After some headaches, I went to a second hand Asus B650E-I model : good price, not too bulky, PCIe gen 5, 10 & 20GBps type-C at the rear.

FUNCTIONAL TEST BUILD

Here is the summary of the part I selected for this build :

• CPU : AMD Ryzen 7 9800X3D PBO PPT 95W, Curve Shaper :
  • Min Frequencies : -30
  • Low Frequencies : -30
  • Med Frequencies : -30 (Max Threads performance, like Cinebench)
  • High Frequencies : -10 (Gaming)
  • Max Frequencies : -10
• Motherboard : ROG STRIX B650E-I GAMING WIFI (without rear i/o shield and plastic cover)
• CPU Cooling :
  • Heatsink : ID Cooling IS-47-XT, black (2 orientations for AM5 is a great feature. Way better mounting system compared to Thermalright). Vertical fin orientation selected.
  • Cooling fan : Fan replaced by a Noctua NF-A9x14 HS-PWM chromax.black.swap (2500 rpm)
  • Thermal paste : Arctic MX-6
  • Contact frame : Not yet, but Thermal Grizzly will be considered
• RAM : Corsair Vengeance 2x24Go 7600 MTS (XMP profile worked well with this motherboard versus my MSI X870 Tomahawk). Booting stime is also excellent, especially versus my MSI X870 Tomahawk (pleasant surprise)
• GPU : NVIDIA RTX 4080 FE
• PSU : Corsair SF1000 with its standard cables. (Not using custom cables remains a very nice & unexpected surprise!). Cherry on the top : I can use the internal power plug (no external pigtail)
• SSD#1 : AData SX8200 Pro 2To (Gen 3)
• SSD#2 : To come
• UPS : USB connection to UPS is fully functional versus with my MSI X870 Tomahawk

Here are the photos!
I love this serial number


With the i/o shield & plastic cover, the motherboard does not align properly with fixing points



Fortunately the rear i/o shield and the plastic cover are easy to remove (2+3 screws). I kept fan grid in the final assembly



Ready to receive the GPU!


Surprisingly easily done!




With the back & front panels..and the stand assembled, time to test!


Testing potential air turbulences


When all tests were positive, I decided to measure the PSU & GPU cables I needed to order…but, in the end, it was not necessary! I managed to improve my cable management significantly to be able to close the side panel! I even have full access to the CPU cooler and the SSDs!




Here is the final result!






CONCLUSION
This U-ITX setup reminds me in many ways of my beloved NFC S4 Mini one, back to 2016!
Hopefully, since the 2016 NFC S4 mini, the U-ITX can take advantage of several major improvements :

• No need of external PSU (nor custom cables)
• Compatible with big GPU, like the NVIDIA RTX 4080 FE
• CPU are now way more efficient and easier to tame (thanks to PBO) compared to 2016
• Overall way more powerful & quiet

Put it in a simple way, I consider the U-ITX as a worthy heir of the NFC S4 mini!
I’m loving this SFFTIME U-ITX!

I’ll continue my testing inside the U-ITX and let you know!

Thanks again for @riba2233 !

 

[rfarmer]

Great build, the picture showing your serial number isn't loading. I recently did a build in the FormD T1 and while I love the case I agree the manual is crap.

 

[MarcParis]

Great build, the picture showing your serial number isn't loading. I recently did a build in the FormD T1 and while I love the case I agree the manual is crap.

2xThanks..link fixed..

 

[riba2233]

Hi,

Thanks for the detailed write-up, I really appreciate it! Glad you like the case

Build looks great and good job with that cable management, I know that it is not easy haha.

I designed clearance for the IO shield same as on P-ATX but tolerances might be a bit different so I will do one small modification for next batches, thanks for noticing it.

 

[MarcParis]

Hi,

Thanks for the detailed write-up, I really appreciate it! Glad you like the case

Build looks great and good job with that cable management, I know that it is not easy haha.

I designed clearance for the IO shield same as on P-ATX but tolerances might be a bit different so I will do one small modification for next batches, thanks for noticing it.

Hi @riba2233

My pleasure!
Np.

Let me know if you would have updated parts to test.

I should be able to share the performance level by this week end.

 

[MarcParis]

-------------------------------------------------------------------------------------------------------------------------------
SFFTIME U-ITX : Cooling & noise performance (04th January 2025)
-------------------------------------------------------------------------------------------------------------------------------

Time to test the U-ITX cooling capabilities!

AMD RYZEN 7 9800X3D Cooling

I must admit I was not expecting so many gaps between the two motherboards I’m using :

• ATX - MSI MAG X870 TOMAHAWK WIFI
• MINI ITX - ROG STRIX B650E-I GAMING WIFI

Definitively the Asus Bios is richer compared to the MSI

• Better RAM compatibility : my DDR5 Corsair Vengeance kit (2x24Go 7600MTS) worked fine (tweaked EXPO profile) and it is noticeably faster to boot
• VRM fan : can be disabled (done it)
• USB power : By default, USB ports are powered when the system is shut down. You can adjust it on the Asus. On the MSI, USB ports are off when system is shut down…and I did not find a way to toggle it
• Asus bios is saving properly PBO settings within profile, not MSI (most probably a current bug)
• My UPS through an usb connection is well supported by the Asus board, but not by the MSI, leading to repeated & temporary freezes

Overall I prefer my experience with the Asus B650E-I compared to the MSI X870 Tomahawk, however, it is not a deal breaker. Both motherboards are good.

What was not expected was their different behaviours on the PBO management. Even with identical setup I get different results on IBT 2.54 or Cinebench 2024.




Basically, the difference is on the Curve shaper application. As a reminder I’m applying -30 offset for min/low & medium frequencies and -10 on high/max frequencies. On the MSI I’m able to benefit from the -30 offset on CPU intensive workload, however the Asus board seems to apply only the -10 offset.

I would need to redo all my testing on the Asus motherboard, but it would be for another day..

As a conclusion, on pure CPU workload, the SFFTIME U-ITX paired with the ID Cooling IS-47-XT heatsink + the Noctua NF-A9x14 HS-PWM chromax.black.swap is capable to cool the AMD Ryzen 7 9800 X3D at the advertised TDP of 95W without any thermal throttling. BUT it is really noisy as the fan is spinning at around 2700 rpm.

For acceptable noise level, my recommendation is to limit the TDP to 75W, still no thermal throttle but the fan is sitting at around 1500 rpm and the CPU temperature around 85°C.

Please note I’ve applied the Thermal Grizzly AM5 contact frame with no issue.




NVIDIA RTX 4080 FE Cooling

In fact cooling the RTX 4080 FE is a non-issue...however the main issue remains the noise.

Basically the SFFTIME U-ITX is less effective on the GPU cooling due to the following points :

• Compared to the Cerberus-X & N-ATX v2, the rear fan of my RTX 4080 FE is approx 40% obstructed.
  
• Due to its small size, the GPU is receiving additional heat to dissipate from :
  • The CPU exhaust
  • The case itself as it is warming up

Overall this is leading to a 3-5°C GPU temperature increase versus Cerberus-X/N-ATX but, most importantly, GPU fan speed increase of +300rpm/+27% versus the Cerberus-X.

At the end, the RTX 4080 FE fans are reaching 1400 rpm, leading to a strong distracting noise.

Here is an example : R7 9800X3D @95W and Stock RTX 4080 FE vs Dragon Age : the Veilguard

• The AMD R7 9800 X3D @95W is almost permanently thermal throttling (~93-95°C) with a fan speed of 2000 rpm⇒non effective and too noisy
• The NVIDIA RTX 4080 FE is reaching up to 65°C with a fan speed of 1400 rpm

Here is another example : R7 9800X3D @65W and Stock RTX 4080 FE vs Dragon Age : the Veilguard

• The AMD R7 9800 X3D @65W is not thermal throttling (~81°C) with a fan speed of 1300 rpm ⇒ very effective cooling & quiet
• The NVIDIA RTX 4080 FE is reaching up to 63°C with a fan speed of 1400 rpm, still

Thus, I decided to control my RTX 4080 FE fans speed with MSI afterburner (4.6.6).
Target was to keep fanless mode on idle and 1100 rpm during load. Here is the final setup


Here is the final setup : R7 9800X3D @75W and RTX 4080 FE @1100RPM vs Dragon Age : the Veilguard

• The AMD R7 9800 X3D @75W is not thermal throttling (~88°C) with a fan speed of 1600 rpm ⇒ effective cooling & quiet
• The NVIDIA RTX 4080 FE is reaching up to 73°C (~260W) with a locked fan speed of 1100 rpm⇒ This is quiet gaming but the GPU temperature took a +12°C hit versus the Cerberus-X

Locking the GPU fan speed at 1100rpm is very effective due to the oversized heatsink of the RTX 4080 FE. But I wanted to know if the cooling potential was enough for higher GPU load.

Another test : R7 9800X3D @75W and RTX 4080 FE @1100RPM vs Star Wars : Outlaws (300W GPU load)

• The AMD R7 9800 X3D @75W is not used completely, sitting at 60W / 81°C with a fan speed of 1300 rpm ⇒ very effective cooling & quiet
• The NVIDIA RTX 4080 FE is reaching up to 75°C (~300W) with a locked fan speed of 1100 rpm ⇒ Successful

Last test : R7 9800X3D @75W and RTX 4080 FE @1100RPM vs Dragon’s Dogma 2

• The AMD R7 9800 X3D @75W could be used completely, 75W / 86°C with a fan speed of 1550 rpm ⇒ effective cooling & quiet
• The NVIDIA RTX 4080 FE is reaching up to 66°C (~230W) with no noticeable CPU bound issues with a locked fan speed of 1100 rpm ⇒ Successful

CONCLUSION

Each game has different CPU & GPU workloads, thus it is not so easy to state that my setup of an AMD R7 9800 X3d @75W + NVIDIA RTX 4080 FE @1100rpm will handle all of them without any issue.
Overall I’m really glad of the compromise.

Still I have some homeworks to do, especially on the back SSD temperature, i.e. 57°C idle and 70°C during gaming load as it is taking some exhaust hot air from the GPU.

PS : Please note that during idle/low GPU load, fanless mode is not eternal. From time to time, fans are starting up but are shutting down quickly after. The main difference versus the Cerberus-X is that there are no case fans permanently pointed towards the GPU. Even if those cases are spinning at very low RPM (500rpm) they are creating an airflow, enough to sustain GPU fanless mode forever.

 

[hrh_ginsterbusch]

As a reminder I’m applying -30 offset for min/low & medium frequencies and -10 on high/max frequencies. On the MSI I’m able to benefit from the -30 offset on CPU intensive workload, however the Asus board seems to apply only the -10 offset.

So its "ASUSified" (q. Techno Jesus), as usual?

cu, w0lf.

 

[MarcParis]

So its "ASUSified" (q. Techno Jesus), as usual?

cu, w0lf.

Well no. This Curve shape was found using my MSI board. I just carried over the same parameters on my Asus board. But behaviour is slightly different.

 

[MarcParis]

-------------------------------------------------------------------------------------------------------------------------------
SFFTIME U-ITX : SSD Cooling Improvements (14th January 2025)
-------------------------------------------------------------------------------------------------------------------------------
Here is a quick update today!
As my SSD (990 Pro 4To) located at the back of the MB was pretty hot, I decided to apply a slim (cheap) heatsink.

The installation was pretty straight forward, especially if I’ve known that the metallic brackets were not compatible with the back panel of the U-ITX..

Here is with the metallic brackets :


Here is the final version :


With the U-ITX back panel case :


In terms of cooling performance, I noticed an improvement of -5°C approx.

Dragon’s Dogma II without SSD heatsink : approx 67°C-68°C

Dragon’s Dogma II with SSD heatsink : approx 62°C-63°C

Please note both SSD temperatures are not great, especially for a SSD that is not under heavy use. In fact, this back SSD is heated by GPU exhaust air. Hopefully U-ITX is pretty an open air case, thus those temperatures are still ok.

Enjoy!

 

[MarcParis]

-------------------------------------------------------------------------------------------------------------------------------
SFFTIME U-ITX : Another Look (25th February 2025)
-------------------------------------------------------------------------------------------------------------------------------
After my successful trial with my N-ITX (N-ATX paired with mini ITX motherboard), I’ve decided to give another try to my U-ITX.

New Changes

• CPU Cooler Fan : I’m testing the Noctua NF-A9x14 PWM (2200rpm) instead of the Noctua NF-A9x14 Chromax (2500rpm).
  The main difference is their motor. The 2200rpm version is using only 1.32A vs 2.52A for the 2500rpm. At the same RPM, the 2500rpm is way noisier than the 2200rpm. I noticed that back to my Node 202 experience.
  
• Fan Duct : As ID-Cooling IS-47-XT is also using M3 thread, it is compatible directly with the Noctua NA-FD1.
  
  
  The side panel is flexible enough to accept this small bump, all their screw holes are aligned nicely. However I must admit that the screw threads are fragile. Even before this addition of the NA-FD1, 1 screw thread was damaged. After the installation of the NA-FD1 I lost a second one, both on the rear side. An easy fix was to put a cable link.
  
  
• Improve stand holding force : I’ve added textile tape where the stand is slides. It is working great!
  
• Setting my monitor to 120Hz instead of 240Hz : This action is impacting GPU power consumption at idle. At 240Hz, RTX 4080 FE power consumption is around 24w, whereas at 120Hz, it is using only 11w. I’m trying to keep fanless mode longer at idle.

Early Trials

As expected, the Noctua NF-A9x14 PWM is(way) quieter than the Chromax version. Even at full speed, its noise is still within tolerable range (audible but at the border of quietness).
In terms of performance, I noticed some negative impact at iso TDP. However, the CPU fan is way quieter so I can increase the TDP to get better overall performance.

Finally I’ll continue testing the TDP 95W & 88W on my R7 9800X3D.


In gaming, I have quickly tested one session on Dragon Age the Veilguard at TDP of 95W.


Interesting result but to be optimized!


More to come!

 

[MarcParis]

-------------------------------------------------------------------------------------------------------------------------------
SFFTIME U-ITX : GPU Assembly guide (26th February 2025)
-------------------------------------------------------------------------------------------------------------------------------
Here is a simple guide to assemble the GPU inside the U-ITX for 3 slots GPU, like my RTX 4080 FE.

Step 1 Once the motherboard & the PSU are assembled, detach completely the bottom panel.

Make sure riser is well connected to the bottom panel

Step 2 Connect the GPU to the riser and assemble it to the rest of the case


This method is less stressful when you have big & long GPU that barely fits the U-ITX.

 

[MarcParis]

-------------------------------------------------------------------------------------------------------------------------------
SFFTIME U-ITX : Best Noise/Thermals/Performance balance for the R7 9800X3D (09th March 2025)
-------------------------------------------------------------------------------------------------------------------------------

Finding the best balance between the noise, the thermals and the performance of my R7 9800X3d inside the U-ITX is a long process with many trials/errors.

From initial/naive point of view it could seem very easy :

• You want maximum performance? Simple answer : take the most powerful fan, like my Noctua A9x14 Chromax that can spin up to 2700rpm and job is done! The only issue remains the noise. It is ok while benchmarking the CPU, but it is unbearable in any other CPU intensive task.
• You want maximum quietness? Simple answer : just limit the R7 9800X3D PPT to 65W or even lower (like 45W). Like that you are reaching the performance of mobile chips, but in a quieter scenario. However, unlike mobile computers, my R7 9800X3D could be paired with more powerful/power hungry GPUs that a R7 9800X3D, limited to 65W/45W PPT, will have significant impact.

As written before, moving to a better balance fan, like the Noctua NF-A9x14 PWM equipped with the foam duct Noctua NA-FD1, is very good step in the right direction.


However I decided to explore more options, offered by the AMD Precision Boost Overdrive (PBO).

As a quick summary here are the main parameters : Original Link

• PPT = Package Power Tracking - The PPT threshold is the allowed socket power consumption permitted across the voltage rails supplying the socket. Applications with high thread counts, and/or “heavy” threads, can encounter PPT limits that can be alleviated with a raised PPT limit.
  Default for Socket AM4/AM5 is at least 160W on motherboards rated for 120W TDP processors. (Default for the R7 9800X3D)
  Default for Socket AM4/AM5 is at least 142W on motherboards rated for 105W TDP processors.
  Default for Socket AM4/AM5 is at least 88W on motherboards rated for 65W TDP processors.
• TDC = Thermal Design Current - The maximum current (amps) that can be delivered by a specific motherboard’s voltage regulator configuration in thermally constrained scenarios.
  Default for socket AM4/AM5 is at least 95A on motherboards rated for 105W TDP processors.
  Default for socket AM4/AM5 is at least 60A on motherboards rated for 65W TDP processors.
• EDC = Electrical Design Current - The maximum current (amps) that can be delivered by a specific motherboard’s voltage regulator configuration in a peak (“spike”) condition for a short period of time.
  Default for socket AM4/AM5 is 140A on motherboards rated for 105W TDP processors.
  Default for socket AM4/AM5 is 90A on motherboards rated for 65W TDP processors.

I tried several setups, but I was, either reaching thermal throttling too quickly or constraining too much the CPU, pretty frustrating! In fact, depending on the workload, the constraints (W or A) are totally different.

Finally I tried a different approach :

• In the motherboard Bios, I used the PBO/Thermal limit to aim for 90°C maximum temperature
• I setup the CPU fan curve to reach a tolerable fan speed at 90°C

The achieved results were pretty unexpected as the R7 9800 X3D was able to extract maximum power from itself.
In terms of cpu benchmarks, here are the outcome :

• Maximum sustainable wattage for the combo : R7 9800X3D + IS-47-XT + Noctua fan 2200rpm is ~103W before reaching the default thermal throttle of 95°C
• In terms of pure heavy CPU workload performance, limiting the temperature to 90°C is leading to a similar level of limiting the PPT of 95W, ie -12% versus stock R7 9800X3D while using ~35% less power!
• Compared to a setup limiting the PPT to 95W, PBO is able to squeeze the maximum performance in the allocated thermal/noise limit, in all the possible workload types. It is for sure more convenient & polyvalent.

As a good example, let’s launch Dragon Age : The Veilguard

• Step 1 : Compiling shaders : During this step, the R7 9800X3D starts using 103W, then progressively decreasing the amount to 95W
• Step 2 : While gaming, as the cpu load continues, the R7 9800X3D power usage is decreasing gently till reaching a balance at around 88W. Temperature is exceeding the limit of 90°C by 1-3°C.

At last, here is the fan curve I’m using for this Noctua NF-A9x14 PWM (2200rpm) :
CPU Temp °C / % Fan PWM : 60/30:80/50:85/80:95/100

 

[MarcParis]

-------------------------------------------------------------------------------------------------------------------------------
SFFTIME U-ITX : Testing the MSI RTX 5080 Ventus OC & RIP (19th March 2025)
-------------------------------------------------------------------------------------------------------------------------------

Following good results from the MSI RTX 5080 Ventus OC inside my Cerberus-X I decided to test it inside the U-ITX.
I was worried as this MSI RTX 5080 Ventus (like all the other AIB models) is diffusing generated heat everywhere instead of focusing it somewhere, like my RTX 4080 FE is doing.
It is working well within the Cerberus-X as it is imposing a very good airflow (intake from the bottom and exhaust and on the side), thus the GPU is breathing very well.

Step 1 : Positioning the GPU in the 3-slots position
Easy assembly, but I still need to disassemble the front panel of the U-ITX.





As a reminder this MSI RTX 5080 Ventus OC dimensions are 303 x 121 x 49 mm.
Thus, compared to my RTX 4080 FE, it is 11mm less thick, leaving some gap between the side panel and the GPU fans.

Not optimal but it should work.

Let’s test it!
Test #1 : using the same fan profile as inside the Cerberus-X, OC of 300/1000 (core/GDDR7)

From this first test, while having a load of ~240W, the RTX 5080 is around 73°-75°C while fans are spinning at around 1650rpm…not bad at all.
However when the GPU load is reaching ~300-330W, the RTX 5080 temperature exploded to 82-85°C while fans were spinning at ~1800rpm, moderating partially the customized fan curve…not brilliant.

Test #2 : using default fan profile, OC of 300/1000 (core/GDDR7), use of DLSS4 performance to save ~10% of power usage

The cooling strategy is totally different now. Fans are quickly at their maximum speed at 2200rpm while the GPU is using ~240-280W but the GPU temperature remains under control up to 75°C maximum. Not great, especially regarding the noise.

The only positive report was during lighter scenes (210W) the GPU fan speed & temperature were reducing significantly, meaning it is not an issue of hot air recycling.

Another important point is SSD temperatures. Compared to my RTX 4080 FE, where only the back SSD was getting hot (~64°C with heatsink), now, both SSDs are getting hot, above 60°C, despite having both heatsink!
It is even the recto SSD that is warmer!

As a conclusion, my MSI RTX 5080 Ventus OC is heating the full case and we are hitting heat dissipation saturation at ~ 280W (but noisy). Unfortunately we are in winter now with an ambient temperature of 21°C, but in summer it can reach ~28°C ambient, leading to even less heat dissipation. Comfortable power range would be around 210-240W.


Step 2 : Trying to position the GPU in a 2.5-slots position

By using standoffs of M3 5mm+3mm (thread) I tried to ideally position the MSI RTX 5080 Ventus OC in the U-ITX.

The card was sitting nicely but it failed.


The card was sitting nicely but it failed due to 1 critical & 1 major details :
Critical Issue : The PCIe Bracket “teeth” have no space at the bottom to fit. Thus I was unable to sit the GPU completely inside the riser! Simply a job stopper!

Major Issue : The PCIe locking mechanism is only compatible with cards of 3 or 2 slots. No variation possible between the two. I compensated by putting some foam behind the GPU…without them, the GPU is very very flimsy.

I came to the conclusion that the U-ITX is only compatible with 2-slots & 3-slots GPU. 2.5 slots GPU have to be positioned in the 3 slots position with no simple way to optimized the airflow. The ultimate option would be to “deshroud” (remove the original fans) the GPU and put 25mm thick fans instead.


Step 3 : RIP U-ITX
One by one, I got some screw threads being destroyed (I’m pretty careful during the assembly). I got 1 + 1 on the front/main side panel, 1 on the rear panel but now I got 1 on the frame itself… U-ITX, made of very thin aluminium, is way more fragile than the N-ATX, made of thicker aluminum. In comparison to both, my Cerberus-X, made of steel, is a tank.

Thus I decided to decommission my U-ITX with regrets. RIP for now.

 

[vinnyoflegend]

I came to the conclusion that the U-ITX is only compatible with 2-slots & 3-slots GPU. 2.5 slots GPU have to be positioned in the 3 slots position with no simple way to optimized the airflow. The ultimate option would be to “deshroud” (remove the original fans) the GPU and put 25mm thick fans instead.

Interesting that you mention this. I didn't even consider it with my Zotac 2080 Ti Amp. It has been deshrouded with 25mm fans zip tied on for as long as I've had it. So I had to mount it as though it were 3 slots. I have bought some blank slot covers to occupy the third slot.

 

[MarcParis]

Interesting that you mention this. I didn't even consider it with my Zotac 2080 Ti Amp. It has been deshrouded with 25mm fans zip tied on for as long as I've had it. So I had to mount it as though it were 3 slots. I have bought some blank slot covers to occupy the third slot.

Thanks for sharing.

Deshroud is my plan B, very easy to do on the Ventus (without removing warranty) but I would like to have 3D print like this one, mostly to keep the backplate in place.

But my biggest fear is that the thermals won't be improved that much as my experimentation proved that the U-ITX to dissipate a specific amount of heat (210-240W) with this kind of GPU coolers. Best GPU for the U-ITX are the FE models of RTX 4080/4090 but they are deprecated now..

The killing point is to compare how well is performing this same GPU inside the Cerberus-X vs the U-ITX. Night vs day in terms of thermals & noise.

 

[vinnyoflegend]

My thermals did not change much (same or small decrease), the biggest change was the noise dropped significantly at load. The stock fans went up to almost 5k rpm

 

[MarcParis]

My thermals did not change much (same or small decrease), the biggest change was the noise dropped significantly at load. The stock fans went up to almost 5k rpm

Thanks!
Yeah noise benefit is the obvious one as Noctua are way better than stock fans.

 

[riba2233]

Yep threads are a bit weaker than on the N-ATX and that is why new batches have these panels made out of steel so there shouldn't be any issues. Although I never had such issues with multiple disassembles tbh but maybe you are just a bit more heavy handed than me

 

[vinnyoflegend]

Yep threads are a bit weaker than on the N-ATX and that is why new batches have these panels made out of steel so there shouldn't be any issues. Although I never had such issues with multiple disassembles tbh but maybe you are just a bit more heavy handed than me

All pieces will be steel in the new batch or just the ones with threads?

 

[riba2233]

All pieces will be steel in the new batch or just the ones with threads?

Just ones with threads yeah