Around 80° on some other sff case. Sould be fine with undervolting. Coil wine is pretty random thought...
Prototype SFFtime presents: U-ITX - 7L lightweight console case that fits 3slot GPUs
- Thread starter riba2233
- Start date
Around 80° on some other sff case. Sould be fine with undervolting. Coil wine is pretty random thought...
Based on my personal experience, I would disagree on many point on this video :
- Alpenfohn Black ridge, with only one 92mm fan, is not performing significantly better than an AXP90-47 or IS-47-XT...but it is imposing way too many constraints on other components
- If you let the R7 9800X3D (even with PBO -20) at stock power with a 47mm it is very loud under load! He is testing only with games with low CPU use (F1 2024 with 77W....waouh). In fact, you need to limit power usage (PPT in PBO menu) to 75W to keep the R7 9800X3 under quiet control. This will limit specific application, even some games (Starfield ~ 120W, Dragon Age the Veilguard ~ 110W) especially paired with a RTX 5090!
- Regarding the RTX 5090 FE, setting the powerlimit is pretty effective...but the card remains audible (1650rpm vs 1550 rpm). His +250Mhz OC on the GPU core is pretty useless as the GPU core speed is adjusting itself with other components like GDDR7 . That's why he noticed very ppor impact on some games, like God of War.
- Good luck to him to put a SDD on the motherboard backplate, just in front of the RTX 5090 FE exhaust...
- I will not be surprise he will switched some parts to watercooling on his setup
Those PCIe power connector temps are wild man.
I have a 5800X3D under a copper AXP-90 and it's also super duper toasty without PBO cranked to -30. the CO -30 setting makes a huge difference for it. I used to have liquid metal under there. It was a bad choice especially due to the pure copper cooler, it wasn't working all that great and then became mildly welded (ended up ripping the CPU out of the socket and had to use a stiff razor blade to detach the heatsink, but luckily no parts were damaged!). Since then running normal paste, the difference between paste and LM is probably less than 5% of the difference that -30 curve optimizer makes, so let me tell ya i was glad they added it to the bios for this CPU.
I want to also note I think I did liquid metal (damn I need to cool it with that stuff) on my NH-D14 on my 5950X, even though this cooler does have the nickel plating it's still pretty fused on there and it's also being awkward to remove. It stopped me actually rearranging my machines the other week, I will just have to remove it after taking the whole motherboard out of the case to do it later.
I think if we add a layer of low profile fans to take up the remaining slot in the U-ITX to force feed it or force pull exhaust, the 5090 won't necessarily be loud especially if we pair it with a tasteful underclock. Ordinarily the extra volume could have been subject to claiming by auxiliary components (fiber NIC in my case). I may still attempt it, but will need to play it by ear.
My 3080ti really struggles on its own inside the velka 7, plenty of airflow obstructions, but a pair of NF-A12x25 bolted on the external side of the side panel to force feed it with fresh air do a phenomenal job keeping it a good 15 degrees cooler than without, so i would comfortably be able to run the GPU with nearly 100W more power with the fans added on. Or equivalently, the noise could be reduced by a noticeable amount.
Obviously if they could have kept a 3 slot cooler on the 5090 it would be more ideal specifically for a case like the U-ITX, but the flow through airflow here should win over any sandwich design case. My understanding is that the U-ITX is strictly only able to host 3.0 slot GPUs so any of these AIB cards are too large, in literally all 3 dimensions.
Actually makes me worry a bit about the 5090's 12V port fitting, but between it being angled already and flexy silicone custom cables, probably will be alright.
I have a 5800X3D under a copper AXP-90 and it's also super duper toasty without PBO cranked to -30. the CO -30 setting makes a huge difference for it. I used to have liquid metal under there. It was a bad choice especially due to the pure copper cooler, it wasn't working all that great and then became mildly welded (ended up ripping the CPU out of the socket and had to use a stiff razor blade to detach the heatsink, but luckily no parts were damaged!). Since then running normal paste, the difference between paste and LM is probably less than 5% of the difference that -30 curve optimizer makes, so let me tell ya i was glad they added it to the bios for this CPU.
I want to also note I think I did liquid metal (damn I need to cool it with that stuff) on my NH-D14 on my 5950X, even though this cooler does have the nickel plating it's still pretty fused on there and it's also being awkward to remove. It stopped me actually rearranging my machines the other week, I will just have to remove it after taking the whole motherboard out of the case to do it later.
I think if we add a layer of low profile fans to take up the remaining slot in the U-ITX to force feed it or force pull exhaust, the 5090 won't necessarily be loud especially if we pair it with a tasteful underclock. Ordinarily the extra volume could have been subject to claiming by auxiliary components (fiber NIC in my case). I may still attempt it, but will need to play it by ear.
My 3080ti really struggles on its own inside the velka 7, plenty of airflow obstructions, but a pair of NF-A12x25 bolted on the external side of the side panel to force feed it with fresh air do a phenomenal job keeping it a good 15 degrees cooler than without, so i would comfortably be able to run the GPU with nearly 100W more power with the fans added on. Or equivalently, the noise could be reduced by a noticeable amount.
Obviously if they could have kept a 3 slot cooler on the 5090 it would be more ideal specifically for a case like the U-ITX, but the flow through airflow here should win over any sandwich design case. My understanding is that the U-ITX is strictly only able to host 3.0 slot GPUs so any of these AIB cards are too large, in literally all 3 dimensions.
Actually makes me worry a bit about the 5090's 12V port fitting, but between it being angled already and flexy silicone custom cables, probably will be alright.
Last edited:
Keep in mind that U-ITX is "shocking" GPU temperature versus bigger case (U-ITX is covering approx of 40% exhaust fan, GPU is receiving additional heat from CPU cooler and from the case itself). For example on my RTX 4080 FE experience, GPU temperature is increased by +5°C with also a +300 RPM fan speed compared to my Cerberus-X (almost open test bench cooling performance on the GPU). At 1500RPM my RTX 4080 FE noise is becoming very distracting!
You can read more in my U-ITX thread.
Back to the RTX 5090 FE, I'm sure it will fit mechanically and thermally inside the U-ITX : you need to put the GPU in 2-slots location to create gap between the back of the GPU and the PCIe riser & you can put a slim 120mm on the back front location (just below the PSU)
My only concern (and this is a big one) is the noise (fan+coil whine) of the RTX 5090 FE. Of course the power target adjustement (even -30%) will help significantly in terms of thermal & coil whine but not that much on the fan speed.
In fact, testing the RTX 5090 inside the U-ITX remains a 2000$ trial. If not succesful I can still send it back (or sell) or move my setup back to bigger cases like the N-ATX, S620 or even Cerberus-X. Anyway, most probably I won't be able to get my hands on a RTX 5090 before months due to low stock/high demand (and scalpers.
)I'll see by Thursday.
It is a none issue. I'm able to connect my RTX 4080 FE in the U-ITX with standard SF1000 cable.
Last edited:
Thank you for confirming. I am just really hesitant to compromise much on portability coming from a Velka 7 packed to the gills, good noise levels are desirable of course, but I should be able to achieve silence with really agressive underclocking if I want. I may also attempt the very same fan-bolton technique (maybe with the new noctua 140mm static pressure fans) to force feed the GPU with air, the pair of huge 5090 stock fans look plenty capable of this on their own but adding more fans allows all of them to run at a lower speed and helps force it more effectively through the case perforations.
I believe some clever ducting could reduce heat interchange between GPU and other components but I expect such attempts to yield very small returns.
First let's see if we can grab one of these RTX 5090 FE...I guess it will be pretty challenging!
The AXP90-X53 certainly isnt doing any better; it improved a bit when using the 25 mm fan of the SS110, but the noise level at high loads was .. ugh. Tested with a 7950X in 105W Eco mode (ie. 142W).
What I dont understand is: The FormD T1 would also allow for AIO / liquid cooling. There are quite a few AIOs with rather small CPU blocks, eg. the Silverstone Vida 240 or the Cooler Master MasterLiquid Lite / Flux 240. Why didnt Optimum test that instead?
Or maybe he did test it, but it turned out horrible?
cu, w0lf.
Picking it up on Wednesday, so if everything is ok we are talking about a week or two at most
I know, I know , I'm the biggest spammer of this thread
. Any news?
In fact it is redundant as for the RTX 3090 or RTX 4090. Optimum will switch to custom watercooling, at least on the CPU.The AXP90-X53 certainly isnt doing any better; it improved a bit when using the 25 mm fan of the SS110, but the noise level at high loads was .. ugh. Tested with a 7950X in 105W Eco mode (ie. 142W).
What I dont understand is: The FormD T1 would also allow for AIO / liquid cooling. There are quite a few AIOs with rather small CPU blocks, eg. the Silverstone Vida 240 or the Cooler Master MasterLiquid Lite / Flux 240. Why didnt Optimum test that instead?
Or maybe he did test it, but it turned out horrible?
cu, w0lf.
He always act like this since 2 generations of GPU...
I found new review with Fractal Ridge and Thorzone Nanoq.
In both cases, the RTX 5090 FE is reaching thermal throttle, ie 83°C.
Basically undervolting remains mandatory but noise was not so bad.
PS : Also I'm very upset that all those journalists (Hardware Canucks, Optimum, Devyn Johnston) "pretends" to face some "unforeseen" issues on riser compatibility (Gen 4 vs Gen 5)...it is just a fake as they faced the exact same issue with RTX 30 series (Gen 3 vs Gen 4)...pfff they are bad actors lol OR they simply forgot...but it would raise me some concerns on their technical skills...
Agreed, they act surprised they are having issues with PCIe 5.0 GPUs and motherboards with PCIe 4.0 riser cables.
Just packing them up, so we are talking about days nowI know, I know , I'm the biggest spammer of this thread
Any P-ATX coming back in stock too? Almost picked one up last time with the U-ITX but shipping makes me sad every timeJust packing them up, so we are talking about days now
Hopefully less than 365 days ..Just packing them up, so we are talking about days now
Any P-ATX coming back in stock too? Almost picked one up last time with the U-ITX but shipping makes me sad every time
I only have a few white units left for now, there will probably me more in Q2.
Hopefully less than 365 days ..
Haha yeah, certainly
I have a hunch that Nvidia will pull a fast one and release a 5080 Super next year with 24GB of VRAM. But still wouldn’t be worth the upgrade I don’t think in your case.Based on first leaks from RTX 5080 review, I would summarize very quickly :
- Switching from an RTX 4080 to an RTX 5080 is non sense
- If I want to replace my RTX 4080 FE it would be by a RTX 5090 FE or I skip RTX 50 completely
Definitely not interestin enough fo rme.
What is worrying me is that monolithic GPU are more & more difficult to produce, even with smaller node size.
Meanwhile chiplet seems to be a dead end for GPU due to the higher latency (even next gen AMD GPU, after the 9070XT, would be still monolithic. Basically AMD tried and failed chiplet approach on GPU)
Let's see...
Are you kidding, you can buy a "picture" of a 5090 FE for the low price of $1500 on ebay. Seriously are these guys smoking crack? Sad part is it's not the only one.RTX 50 GPU is a paper launch..
No GPU available
Nice decoy! Scalpers are one thing, but no stock remains the issue from Nvidia.
I would say it is the worst launch of Nvidia GPU since, at least RTX30 in late 2020.
