Clovis559
 
That’s why I got the measuring tool, because right now I’m eyeballing it with a ruler. Hopefully I’ll be able to measure everything on the back, share, we can and go from there! Then design what ever craziness we’re going to design.

I might pick one of those up myself depending on how sketchy things are once my thermal pads come in which will hopefully all be here by Tuesday.

Dabadger84

Clovis559
 
That’s why I got the measuring tool, because right now I’m eyeballing it with a ruler. Hopefully I’ll be able to measure everything on the back, share, we can and go from there! Then design what ever craziness we’re going to design.

I might pick one of those up myself depending on how sketchy things are once my thermal pads come in which will hopefully all be here by Tuesday.

FTW3 Ultra got here early, VRAM junction temp is much higher than EVGA's readout

anchang233
 
FTW3 Ultra got here early, VRAM junction temp is much higher than EVGA's readout

Unfortunately the Junction Temp readout wasn't available in HWInfo yet when I last had my 3090 FTW3 Hybrid, so I can't really compare notes on that aspect.
But for the Kingpin, pre heatsinks, it was only 4C higher than the highest iCX Memory temperature... so I imagine in more regular airflow situations on the FTW3 Ultra it'd be... 8-10C higher perhaps?  Keep in mind, the iCX sensors are most likely edge temps on the given things they're reading the temperature of - the Memory Junction Temperature is (we assume, and mostly is) an internal TJunction from one of the memory chips - so that's more of what they're actually running at internally... but also keep in mind GDDR6X can run up to something like 110C internally before anything "bad" happens.

I think this also applies to other 3090 cards.. but I've only got the memory stuck on for now and I'm seeing a 6 degree drop in junction, and 2-4 degree drop for memory.  Most importantly, gpu boost clocks are almost a solid 2050+ (hybrid).
 
In the Pic I have heatsinks for the VRMs, but I haven't attached it yet since I haven't put thermal pads underneath the back plate.
 
Unfortunately, I had to pick low profile heatsinks (4mm) to clear the vertical GPU mount bracket.  I'm waiting for my 90mm fan to come so I can put it directly behind the card to cool these better.  Right now I just stuck a fan behind the computer blowing into the case, as the case fans on the bottom seem to be blocked by the vertical mount (running at higher RPMs don't change the temperatures much).
 
I would probably advise against the ICEPC brand, since they have their logo printed on the block.  I would have much rather had the plain ones.
 

MiztahSparklez
I think this also applies to other 3090 cards.. but I've only got the memory stuck on for now and I'm seeing a 6 degree drop in junction, and 2-4 degree drop for memory.  Most importantly, gpu boost clocks are almost a solid 2050+ (hybrid).
 
In the Pic I have heatsinks for the VRMs, but I haven't attached it yet since I haven't put thermal pads underneath the back plate.
 
Unfortunately, I had to pick low profile heatsinks (4mm) to clear the vertical GPU mount bracket.  I'm waiting for my 90mm fan to come so I can put it directly behind the card to cool these better.  Right now I just stuck a fan behind the computer blowing into the case, as the case fans on the bottom seem to be blocked by the vertical mount (running at higher RPMs don't change the temperatures much).
 
I would probably advise against the ICEPC brand, since they have their logo printed on the block.  I would have much rather had the plain ones.

Yes, this definitely applies to all 3090s.  I plan to add layouts for the 3090 FTW3 & 3090 XC3 soon, I just have to find quality shots of the PCB on both sides to know where the "target areas" are for the back side in terms of adding thermal pads & heatsinks optimally.

Marked up PCB pics of the eVGA RTX 3090 FTW3 Ultra/Hybrid PCB for those wanting to thermal pad/heatsink their backplate area - do not forget the lil' bastard VRM that's near one of the mount holes, it's important because it's a memory VRM:
 

 

So the Thermal Pads shipped todae, looks like one set of them isn't slated to be here til Wednesday... the wait continues.  Looking forward to it though!
 
@Clovis since you merced your card already, how many screws were involved in removing the backplate, or do you remember?

I've noticed that the messaging about pad size is inconsistent from EVGA.  I am starting to think that sometimes they're just giving us the clearance between parts and we should size up appropriately for decent pad compression.  Or sometimes it's for the pads they use which will likely have different firmness/density than what we'll purchase aftermarket.

_Gir_
I've noticed that the messaging about pad size is inconsistent from EVGA.  I am starting to think that sometimes they're just giving us the clearance between parts and we should size up appropriately for decent pad compression.  Or sometimes it's for the pads they use which will likely have different firmness/density than what we'll purchase aftermarket.

Indeed, that's why I got a smattering of everything.  1mm, 1.5mm, 2mm & 3mm incoming - got the most of the 2mm I think, cuz I'm pretty sure that's what I'll need for most of what I want to "cover" with thermal pads.  The 3mm is probably not going to get much use, but I got it anyway just in case.  If I don't use it at all on the video card I'll find another use for it, like redoing some ancient 2012/2014 internals on the old laptops I use for "CS:GO Trust Factor repairs" lol

_Gir_
I've noticed that the messaging about pad size is inconsistent from EVGA.  I am starting to think that sometimes they're just giving us the clearance between parts and we should size up appropriately for decent pad compression.  Or sometimes it's for the pads they use which will likely have different firmness/density than what we'll purchase aftermarket.

The back vram on my ftw3 ultra was more like 2.5 instead of 2 like they said... I was able to use 3mm for my rear vrm (bare circuit board on the ftw3 ultra) and the stock 2.25/2.5 pads from the front vram side for the rear vram and back plate when I changed out my hybrid cooler. 2mm gelid extreme didn’t have enough compression, it barely touched and 2+1mm pads stacked was way to high. 3mm gelid has about the same give as the stock putty.

One thing I did notice that dropped my hybrid temps was to use new thermal grease on the gpu chip and between the AIO pump and copper hot plate that attaches to the vram. With adequate airflow around all the heat sinks , I was able to get around -8 junction temp. -4 or so for everything else. If I was able to horizontal mount, I’d probably get even more.