S4MAX: Brickless S4M w/ 3090 FE and R9 5950x - 800W, 5l, water cooled
- Thread starter petricor
- Start date
Where did you get the 3.5 and 5.5 bullet connector? Link please Im trying to making the same PSU
Post from China!
In the box there's the long-awaited ADT PCIe extension...
...and a set of port extensions for the GPU.
Before getting to work I give the riser a quick test, and the length appears to be spot-on:
...and BAM: Massive setback!
The GPU doesn't fire up - and upon close inspection, the conductive layer on the ADT adapter's A1 pin ripped off.
To make matters worse, some of the metal ended up in the slot, apparently shorting the x16 port on my motherboard.
Clearing out the port doesn't help.
At this point, I don't know whether I just fried the board or whether the CPU is dead, too...
Also, it's mildly mysterious to me how this could have happened. Either there is a layer adhesion issue at the adapter's PCB, or a piece of metal debris found its way into the x16 port mechanically damaging things upon insertion of the adapter - in any case, there is mechanical damage to the connector, and apparent damage to the system's components, too, as I cannot get things to work with a different GPU and riser.
On the upside, the 3090 still works in a different rig.
As I don't have another AM4 board for testing, I have just hit the order button on a Gigabyte B550 Aorus Pro Wifi, and this is where the build may take a different turn if it is indeed (just) the board that's dead.
My initial reaction was just to buy another ASRock m-ITX tb/3 for a 1:1 swap, but just before hitting the "buy" button, luckily, I remembered that this is one of the few boards without a quick flash option, which is a real pain when using 4th gen Ryzens. In lieu of my old 3950x (sold), I would have had to organise a gen 3 Ryzen just to flash the BIOS - not an attractive proposition, and not really the way going forward.
So the choice was now between a different x570 board, which would have required me to laboriously re-design my active thermal solution for the power-hungry x570 chipset (different board geometry), or considering a B550 with a much thermal-friendlier design.
B550 wasn't around when I started this build, and having read into the matter, I expect a good one with proper power delivery to do exactly what I need, without any performance penalty: With the ASRock's stand-out thunderbolt feature off the table (no one else does it), all that X570 appears to have going for it over B550 is more PCIe4 lanes, which I will not make use of with a single x16 slot, and no plans to use more than one SSD - nb that the ASRock board only has one M.2 slot.
The B550 will require a different water block mount though as my current ASRock m/itx has an Intel mounting pattern - I have ordered an AM4 mount for EK's Supremacy blocks which appear to have the same footprint as the Annihilator, so possibly that's a fit.
We shall see!
In the box there's the long-awaited ADT PCIe extension...
...and a set of port extensions for the GPU.
Before getting to work I give the riser a quick test, and the length appears to be spot-on:
...and BAM: Massive setback!
The GPU doesn't fire up - and upon close inspection, the conductive layer on the ADT adapter's A1 pin ripped off.
To make matters worse, some of the metal ended up in the slot, apparently shorting the x16 port on my motherboard.
Clearing out the port doesn't help.
At this point, I don't know whether I just fried the board or whether the CPU is dead, too...
Also, it's mildly mysterious to me how this could have happened. Either there is a layer adhesion issue at the adapter's PCB, or a piece of metal debris found its way into the x16 port mechanically damaging things upon insertion of the adapter - in any case, there is mechanical damage to the connector, and apparent damage to the system's components, too, as I cannot get things to work with a different GPU and riser.
On the upside, the 3090 still works in a different rig.
As I don't have another AM4 board for testing, I have just hit the order button on a Gigabyte B550 Aorus Pro Wifi, and this is where the build may take a different turn if it is indeed (just) the board that's dead.
My initial reaction was just to buy another ASRock m-ITX tb/3 for a 1:1 swap, but just before hitting the "buy" button, luckily, I remembered that this is one of the few boards without a quick flash option, which is a real pain when using 4th gen Ryzens. In lieu of my old 3950x (sold), I would have had to organise a gen 3 Ryzen just to flash the BIOS - not an attractive proposition, and not really the way going forward.
So the choice was now between a different x570 board, which would have required me to laboriously re-design my active thermal solution for the power-hungry x570 chipset (different board geometry), or considering a B550 with a much thermal-friendlier design.
B550 wasn't around when I started this build, and having read into the matter, I expect a good one with proper power delivery to do exactly what I need, without any performance penalty: With the ASRock's stand-out thunderbolt feature off the table (no one else does it), all that X570 appears to have going for it over B550 is more PCIe4 lanes, which I will not make use of with a single x16 slot, and no plans to use more than one SSD - nb that the ASRock board only has one M.2 slot.
The B550 will require a different water block mount though as my current ASRock m/itx has an Intel mounting pattern - I have ordered an AM4 mount for EK's Supremacy blocks which appear to have the same footprint as the Annihilator, so possibly that's a fit.
We shall see!
Link to the 3.5mm is further up in this post, the 5.5mm ones are these.
There are different manufacturers offering them, I'd try a different one next time as I have not been too happy with how the gold plating held up on the male ends when soldering.
RC / Drone forums are excellent sources for all kinds of high current connectors, it's where I did my research.
B550I Aorus is a very solid board. Interesting to see how you will handle the VRM cooling. The factory heat sink by Gigabyte seems to be incompatible with your layout due to its height...
Yeah that one will have to lose a bit of mass indeed- but not everything. Assume I can compensate that with a few added copper fins; possibly I can connect it to the chassis with a heat pipe. Bigger challenge will be redesigning the water loop as the AM4 mounting pattern differs quite significantly from what I had before- I already have a plan in mind… could be much neater than the previous layout as the board has pretty generous clearances.
Sad news
![:\](data:image/gif;base64,R0lGODlhAQABAIAAAAAAAP///yH5BAEAAAAALAAAAAABAAEAAAIBRAA7 "Unsure :\")
A GPU still working and you still motivated, looks good, nothing is lost!
After some lengthy testing, I found the problem.
This is the culprit - or, to be more precise, what I have used it for:
The 1:1 PCIe extension I used has reversed my pin-out. So, rather counter-intuitively, for an extension, I need a "signal swap" (TX/RX) cable, as, when looking at the male connectors, the pin-out indeed needs to be reversed on one end: Pin A-1 needs to be connected to B-1 and so on. The TX-TX variant without signal swap (R33SS) replicates the pinout on each side to be exactly as per PCIe spec (ironically, to do so, the signal lanes need to cross from one side to the other), and by using it as an extension I managed to short pretty much everything. The contacts coming off the connector's PCB were the 12V pins, and they have molten as shorted to GND.
I guess I can count myself VERY lucky that the 3090 has survived this - neither board nor CPU did, though, so I have a nice paperweight reading "5950x" now. At least there is no shortage of them anymore so getting my hands onto a new one has been pretty straightforward.
Also, I have ordered the ADT-Link R33NS with TX-RX signal swap which should solve my problem - should be on the boat from China now.
So, moving on:
Enter my fresh Gigabyte Aorus B550 Gaming!
It comes with the usual accessories: SATA cables, Wifi antennae, and some decoration in the form of a sticker and a driver CD. Not sure what it is good for - it's not really a great mirror and does not stick to anything. I think it's a circle ruler - if anyone wants it let me know.
The board looks amazing - 10 out of 10 for blackness...
...and reassuringly blast-proof with a full metal back cover.
In all likelihood, this will have to go to enable the SFFness commanded by this build - but a pretty nice design.
I start by transplanting my M.2 to the top slot - this is the one with PCIe4 support - and when taking things apart, it would appear that the top metal block is for decoration only and that the part actually doing the work is the smaller black metal heat sink you can see.
For the SSD to fit, I need to remove my self made contact plate that I used with the ASRock mITX/TB3 to establish heat transfer between the case and SSD...
...and the copper label coming with the SSD comes off right with it. No problem - the less material I have between SSD and Heat Sink, the better.
The screw coming with the board is actually a three-part assembly (that's not really made clear in the manual): the bottom part goes onto the board to support the SSD at the right height, the second one onto it to secure the SSD and form a spacer to the heatsink, and the small screw on top is to fixes the heatsink on top.
If your SSD is not horizontal, you got it wrong!
The big metal block over the heatsink will not be used going ahead: No space for decoration!
Enter my second R5950X - having tried my original one I only get a blue screen so it appears to be a write-off - and some A4M mounting kit:
Together with my G.Skill RAM, I should now be looking at something I can power up, cooling provided.
To get this going, I'll need to tweak my EK Annihilator for use with an A4M pattern:
To the left, you see the adapter plate, to the right the EK Annihilator with an Intel mounting pattern as used by my previous ASRock board.
Taking things apart presents a good opportunity to clean the block (as you can see there is hardly any gunk in there)...
...and to have a look at the block that has some pressure marks from the CPU's lid. The scratches you see come from twisting off the block from the CPU - the paste was still pretty liquid and had caused a rather impressive adhesion between block and CPU.
...and re-assemble the cooler and the re-polished copper block.
This should now make for a nice fit onto the Aorus B550:
...and everything works, using a GTX1080 for a test drive!
The board is incredibly tidy and has a lot of space compared to the ASRock mITX/TB3.
A few conclusions from this test assembly:
- I'll rotate the water block's outlets by 90 degrees to get more space between block and PCIe slot. It will require a slightly different routing of my water loop, but this should buy me more space for my PCIe cable sandwich, and power delivery from PSU to GPU which will become the pinch point in this build - I still may have to de-lid my RAM to make this fit.
- I'll have to trim back the heat sink on the board to the red line indicated - it will be interesting to see how this affects chipset temps. As the sink is pure mass without fins it should not contribute much. Should there be a thermal issue, I may be able to compensate this by just adding some copper fins to the trimmed cooling stack; the worst case is re-instating the heatpipe solution of my previous board, but ideally I can run this build without any active chipset or VRM cooling.
This is the culprit - or, to be more precise, what I have used it for:
The 1:1 PCIe extension I used has reversed my pin-out. So, rather counter-intuitively, for an extension, I need a "signal swap" (TX/RX) cable, as, when looking at the male connectors, the pin-out indeed needs to be reversed on one end: Pin A-1 needs to be connected to B-1 and so on. The TX-TX variant without signal swap (R33SS) replicates the pinout on each side to be exactly as per PCIe spec (ironically, to do so, the signal lanes need to cross from one side to the other), and by using it as an extension I managed to short pretty much everything. The contacts coming off the connector's PCB were the 12V pins, and they have molten as shorted to GND.
I guess I can count myself VERY lucky that the 3090 has survived this - neither board nor CPU did, though, so I have a nice paperweight reading "5950x" now. At least there is no shortage of them anymore so getting my hands onto a new one has been pretty straightforward.
Also, I have ordered the ADT-Link R33NS with TX-RX signal swap which should solve my problem - should be on the boat from China now.
So, moving on:
Enter my fresh Gigabyte Aorus B550 Gaming!
It comes with the usual accessories: SATA cables, Wifi antennae, and some decoration in the form of a sticker and a driver CD. Not sure what it is good for - it's not really a great mirror and does not stick to anything. I think it's a circle ruler - if anyone wants it let me know.
The board looks amazing - 10 out of 10 for blackness...
...and reassuringly blast-proof with a full metal back cover.
In all likelihood, this will have to go to enable the SFFness commanded by this build - but a pretty nice design.
I start by transplanting my M.2 to the top slot - this is the one with PCIe4 support - and when taking things apart, it would appear that the top metal block is for decoration only and that the part actually doing the work is the smaller black metal heat sink you can see.
For the SSD to fit, I need to remove my self made contact plate that I used with the ASRock mITX/TB3 to establish heat transfer between the case and SSD...
...and the copper label coming with the SSD comes off right with it. No problem - the less material I have between SSD and Heat Sink, the better.
The screw coming with the board is actually a three-part assembly (that's not really made clear in the manual): the bottom part goes onto the board to support the SSD at the right height, the second one onto it to secure the SSD and form a spacer to the heatsink, and the small screw on top is to fixes the heatsink on top.
If your SSD is not horizontal, you got it wrong!
The big metal block over the heatsink will not be used going ahead: No space for decoration!
Enter my second R5950X - having tried my original one I only get a blue screen so it appears to be a write-off - and some A4M mounting kit:
Together with my G.Skill RAM, I should now be looking at something I can power up, cooling provided.
To get this going, I'll need to tweak my EK Annihilator for use with an A4M pattern:
To the left, you see the adapter plate, to the right the EK Annihilator with an Intel mounting pattern as used by my previous ASRock board.
Taking things apart presents a good opportunity to clean the block (as you can see there is hardly any gunk in there)...
...and to have a look at the block that has some pressure marks from the CPU's lid. The scratches you see come from twisting off the block from the CPU - the paste was still pretty liquid and had caused a rather impressive adhesion between block and CPU.
...and re-assemble the cooler and the re-polished copper block.
This should now make for a nice fit onto the Aorus B550:
...and everything works, using a GTX1080 for a test drive!
The board is incredibly tidy and has a lot of space compared to the ASRock mITX/TB3.
A few conclusions from this test assembly:
- I'll rotate the water block's outlets by 90 degrees to get more space between block and PCIe slot. It will require a slightly different routing of my water loop, but this should buy me more space for my PCIe cable sandwich, and power delivery from PSU to GPU which will become the pinch point in this build - I still may have to de-lid my RAM to make this fit.
- I'll have to trim back the heat sink on the board to the red line indicated - it will be interesting to see how this affects chipset temps. As the sink is pure mass without fins it should not contribute much. Should there be a thermal issue, I may be able to compensate this by just adding some copper fins to the trimmed cooling stack; the worst case is re-instating the heatpipe solution of my previous board, but ideally I can run this build without any active chipset or VRM cooling.
Time to cut things to measure:
First up is the not precisely SFF 3090. After giving it a final spin to check that it works with the new board...
...I start by preparing the water block:
The areas x-ed out on the plexi block need to go to allow the 90 degree Festo fitting to rotate towards the block (left side)...
.... and to make space for the PCIe cable (bottom edge). It looks like I can retain the RGB strip - perhaps it will finally grow on me!
After taking the block apart...
...I use a 4mm milling head to cut the plexi block.
After some sanding and polishing, I get to something surprisingly presentable...
...that I can mount back onto the nickel block, making sure to apply a good coat of oil-free silicone lubricant to the seal...
...resulting in a reasonably compact version of the Corsair block that should avoid clashes with other components.
This photo shows...
...how the notch milled into the Plexi block allows the L-fitting to swivel to about 30 degrees off the GPU plane as required for my water loop to fit:
Next point of order is dealing with the slight mismatch in scale of these two components:
A few screws and tiny connectors later, the 3090 lost its fat suit and looks far less intimidating (don't do this before skilling up on youtube - this one here is pretty helpful)
Next up is ridding the Corsair cooler from it's pre-applied (!) thermal paste...
...and readying it for some MX-4 to go between water block and GPU.
That's what it should require to deal with the "A"-side of the GPU...
...whereas the rear requires some modifications:
As the Corsair block I use is designed for a 3080 (it comes with a smaller footprint whilst cooling all relevant components - see my comparison a few pages back), I need to cut two little notches into the backplate's edge upstand to accomodate for the larger PCB...
...and build up some thermal pads on the rear to establish contact between the backplate and rear side RAM modules - which are not required for the 3080 which this plate has been designed for: 2mm over the RAM, and 1mm over the POSCAPS make for reliable contact.
With that sorted, and some insulation tape so mask the exposed parts of the PCB, the backplate can go on...
...making for a nice and compact sandwich.
Bon appetit!
First up is the not precisely SFF 3090. After giving it a final spin to check that it works with the new board...
...I start by preparing the water block:
The areas x-ed out on the plexi block need to go to allow the 90 degree Festo fitting to rotate towards the block (left side)...
.... and to make space for the PCIe cable (bottom edge). It looks like I can retain the RGB strip - perhaps it will finally grow on me!
After taking the block apart...
...I use a 4mm milling head to cut the plexi block.
After some sanding and polishing, I get to something surprisingly presentable...
...that I can mount back onto the nickel block, making sure to apply a good coat of oil-free silicone lubricant to the seal...
...resulting in a reasonably compact version of the Corsair block that should avoid clashes with other components.
This photo shows...
...how the notch milled into the Plexi block allows the L-fitting to swivel to about 30 degrees off the GPU plane as required for my water loop to fit:
Next point of order is dealing with the slight mismatch in scale of these two components:
A few screws and tiny connectors later, the 3090 lost its fat suit and looks far less intimidating (don't do this before skilling up on youtube - this one here is pretty helpful)
Next up is ridding the Corsair cooler from it's pre-applied (!) thermal paste...
...and readying it for some MX-4 to go between water block and GPU.
That's what it should require to deal with the "A"-side of the GPU...
...whereas the rear requires some modifications:
As the Corsair block I use is designed for a 3080 (it comes with a smaller footprint whilst cooling all relevant components - see my comparison a few pages back), I need to cut two little notches into the backplate's edge upstand to accomodate for the larger PCB...
...and build up some thermal pads on the rear to establish contact between the backplate and rear side RAM modules - which are not required for the 3080 which this plate has been designed for: 2mm over the RAM, and 1mm over the POSCAPS make for reliable contact.
With that sorted, and some insulation tape so mask the exposed parts of the PCB, the backplate can go on...
...making for a nice and compact sandwich.
Bon appetit!
I used this Gigabyte board a few months ago to build a PC for Chia plotting (cryptocurrency). This is significantly more taxing on the SSD than anything else I've ever done, with the system running through over 25TB of the SSDs write endurance each day.
I quickly found out that the included SSD heatsink was trash, and the decorative top portion is not being used in my build. Not for space reasons, but because the SSD temps are lower without it in place.
I also reduced the SSD temperature a few more degrees by replacing the included black thermal pad with a Gelid Extreme one, although I see they now have an even better "Ultimate" version.
If you're looking for pads to fill up larger gaps, and to avoid stacking thermal pads, which isn't ideal, then Thermalright's pads are a good choice.
Good to see the project is back on track after your riser incident. I think quite a few people would have thrown in the towel at the thought of having to replace so much and redesign parts of the build again.
10/10 for tenacity!
I quickly found out that the included SSD heatsink was trash, and the decorative top portion is not being used in my build. Not for space reasons, but because the SSD temps are lower without it in place.
I also reduced the SSD temperature a few more degrees by replacing the included black thermal pad with a Gelid Extreme one, although I see they now have an even better "Ultimate" version.
If you're looking for pads to fill up larger gaps, and to avoid stacking thermal pads, which isn't ideal, then Thermalright's pads are a good choice.
Good to see the project is back on track after your riser incident. I think quite a few people would have thrown in the towel at the thought of having to replace so much and redesign parts of the build again.
10/10 for tenacity!
Next up: Fitting the board!
Placing it into the S4M's frame, it broadly seems to fit, but for it to be packed as densely as I need it to, both board and frame will require some modifications.
Stripping the board of its add-ons is pretty straightforward...
...and the first thing to lose some volume is the heat sink - it projects too high and needs to come down a little to make space for the GPU sandwiched on top of the board.
The heat sink is on the beefy end of things...
...and I had to go through two cutting discs to trim off the upper portion.
Next up: Preparing frame and backplate.
As the backplate has been mounted to the board's bottom plate, I need to drill a pair of holes into the S4M's frame to fix the backplate to it with a pair of scews.
Drilling two slightly messy holes (they need to be a bit larger to allow for some tolerance when fitting the backplate)...
...and counter-sinking them...
or a flush fit of the screws.
Next, the two flanges on the backplate x-ed out on the image below need to go: As I have shortened the board spacers on the S4M's frame by 3 mm to achieve the vertical density I need for the build to pan out (see this post earlier in the thread), board and backplate will sit lower in the frame and would clash.
A little cutting...
...and grinding later...
...I have something that securely fits into the frame and works with the board's lower position.
That said, I now have a clash between the A4M mounting bracket and the frame:
As you can see in the photo above, the shorter spacers do not leave enough clearance between the board and frame for the bracket to fit, and tightening the screws results in a slightly worrying bending of the motherboard.
In order to resolve this, I'll need to cut a pocket into the frame for it to stay clear of the AM4 bracket:
This should do the trick...
...and gets me a flush fit of the bracket and the S4M's bottom frame. This also means that I can establish direct contact between bracket and case for added heat dissipation: Every little helps!
After dressing up the cut edge of the heat sink with some insulation tape...
...I can now re-assemble the board...
...and fit the re-configured EK Annihilator, with the 8mm fittings now rotated by 90 degrees for an improved tubing layout.
A little MX-4 on the CPU later (not shown), I get to a fairly compact package...
...that should just about fit into the S4Ms frame: Stacking GPU and board illustrates the vertical density (note the folded PCIe cable between CPU- and GPU block). The two visible 8mm fittings facing the camera will be linked with a U-fitting, and with a little luck and using @Josh | NFC 's low profile 12-pin power plug solution, I may even be able to retain the housing of the RAM modules.
The board now fits into the frame without any clashes...
...and I can test-fit the GPU to mark an extension to the PCIe-notch I have introduced for the 2080ti in the previous version of this build.
Also it becomes apparent that the plexi block will need to lose some more material along the upper edge to make some more space for the riser cable not to clash with the board's wifi module.
A little milling later...
...I get to a version of the water block with a notch for the cable (I have milled down the block to the level of the screw head's bottom face to not risk cracking the plexi under the screw's pressure)...
...allowing me to stack things tight enough...
...for 3090 and riser cable to fit flush into the frame.
Case fits: Major box ticked!
Loosely fitting the radiator / pump module gives an idea of the final packaging - it will need an upgrade to be able to cope with the 3090 / 5950x combo though.
So, next up: Tripling the radiator's airflow for some extra chilling!
Placing it into the S4M's frame, it broadly seems to fit, but for it to be packed as densely as I need it to, both board and frame will require some modifications.
Stripping the board of its add-ons is pretty straightforward...
...and the first thing to lose some volume is the heat sink - it projects too high and needs to come down a little to make space for the GPU sandwiched on top of the board.
The heat sink is on the beefy end of things...
...and I had to go through two cutting discs to trim off the upper portion.
Next up: Preparing frame and backplate.
As the backplate has been mounted to the board's bottom plate, I need to drill a pair of holes into the S4M's frame to fix the backplate to it with a pair of scews.
Drilling two slightly messy holes (they need to be a bit larger to allow for some tolerance when fitting the backplate)...
...and counter-sinking them...
or a flush fit of the screws.
Next, the two flanges on the backplate x-ed out on the image below need to go: As I have shortened the board spacers on the S4M's frame by 3 mm to achieve the vertical density I need for the build to pan out (see this post earlier in the thread), board and backplate will sit lower in the frame and would clash.
A little cutting...
...and grinding later...
...I have something that securely fits into the frame and works with the board's lower position.
That said, I now have a clash between the A4M mounting bracket and the frame:
As you can see in the photo above, the shorter spacers do not leave enough clearance between the board and frame for the bracket to fit, and tightening the screws results in a slightly worrying bending of the motherboard.
In order to resolve this, I'll need to cut a pocket into the frame for it to stay clear of the AM4 bracket:
This should do the trick...
...and gets me a flush fit of the bracket and the S4M's bottom frame. This also means that I can establish direct contact between bracket and case for added heat dissipation: Every little helps!
After dressing up the cut edge of the heat sink with some insulation tape...
...I can now re-assemble the board...
...and fit the re-configured EK Annihilator, with the 8mm fittings now rotated by 90 degrees for an improved tubing layout.
A little MX-4 on the CPU later (not shown), I get to a fairly compact package...
...that should just about fit into the S4Ms frame: Stacking GPU and board illustrates the vertical density (note the folded PCIe cable between CPU- and GPU block). The two visible 8mm fittings facing the camera will be linked with a U-fitting, and with a little luck and using @Josh | NFC 's low profile 12-pin power plug solution, I may even be able to retain the housing of the RAM modules.
The board now fits into the frame without any clashes...
...and I can test-fit the GPU to mark an extension to the PCIe-notch I have introduced for the 2080ti in the previous version of this build.
Also it becomes apparent that the plexi block will need to lose some more material along the upper edge to make some more space for the riser cable not to clash with the board's wifi module.
A little milling later...
...I get to a version of the water block with a notch for the cable (I have milled down the block to the level of the screw head's bottom face to not risk cracking the plexi under the screw's pressure)...
...allowing me to stack things tight enough...
...for 3090 and riser cable to fit flush into the frame.
Case fits: Major box ticked!
Loosely fitting the radiator / pump module gives an idea of the final packaging - it will need an upgrade to be able to cope with the 3090 / 5950x combo though.
So, next up: Tripling the radiator's airflow for some extra chilling!
Once again an amazing read. Thanks for documenting and sharing this processNext up: Fitting the board!
Placing it into the S4M's frame, it broadly seems to fit, but for it to be packed as densely as I need it to, both board and frame will require some modifications.
Stripping the board of its add-ons is pretty straightforward...
...and the first thing to lose some volume is the heat sink - it projects too high and needs to come down a little to make space for the GPU sandwiched on top of the board.
The heat sink is on the beefy end of things...
...and I had to go through two cutting discs to trim off the upper portion.
Next up: Preparing frame and backplate.
As the backplate has been mounted to the board's bottom plate, I need to drill a pair of holes into the S4M's frame to fix the backplate to it with a pair of scews.
Drilling two slightly messy holes (they need to be a bit larger to allow for some tolerance when fitting the backplate)...
...and counter-sinking them...
or a flush fit of the screws.
Next, the two flanges on the backplate x-ed out on the image below need to go: As I have shortened the board spacers on the S4M's frame by 3 mm to achieve the vertical density I need for the build to pan out (see this post earlier in the thread), board and backplate will sit lower in the frame and would clash.
A little cutting...
...and grinding later...
...I have something that securely fits into the frame and works with the board's lower position.
That said, I now have a clash between the A4M mounting bracket and the frame:
As you can see in the photo above, the shorter spacers do not leave enough clearance between the board and frame for the bracket to fit, and tightening the screws results in a slightly worrying bending of the motherboard.
In order to resolve this, I'll need to cut a pocket into the frame for it to stay clear of the AM4 bracket:
This should do the trick...
...and gets me a flush fit of the bracket and the S4M's bottom frame. This also means that I can establish direct contact between bracket and case for added heat dissipation: Every little helps!
After dressing up the cut edge of the heat sink with some insulation tape...
...I can now re-assemble the board...
...and fit the re-configured EK Annihilator, with the 8mm fittings now rotated by 90 degrees for an improved tubing layout.
A little MX-4 on the CPU later (not shown), I get to a fairly compact package...
...that should just about fit into the S4Ms frame: Stacking GPU and board illustrates the vertical density (note the folded PCIe cable between CPU- and GPU block). The two visible 8mm fittings facing the camera will be linked with a U-fitting, and with a little luck and using @Josh | NFC 's low profile 12-pin power plug solution, I may even be able to retain the housing of the RAM modules.
The board now fits into the frame without any clashes...
...and I can test-fit the GPU to mark an extension to the PCIe-notch I have introduced for the 2080ti in the previous version of this build.
Also it becomes apparent that the plexi block will need to lose some more material along the upper edge to make some more space for the riser cable not to clash with the board's wifi module.
A little milling later...
...I get to a version of the water block with a notch for the cable (I have milled down the block to the level of the screw head's bottom face to not risk cracking the plexi under the screw's pressure)...
...allowing me to stack things tight enough...
...for 3090 and riser cable to fit flush into the frame.
Case fits: Major box ticked!
Loosely fitting the radiator / pump module gives an idea of the final packaging - it will need an upgrade to be able to cope with the 3090 / 5950x combo though.
So, next up: Tripling the radiator's airflow for some extra chilling!
Definitely looking forward to whatever crazy fan mod you're up to Btw, how are your experiences milling things with a rotary tool like that? I assume it's only suitable for soft materials and smaller jobs? Still, do you feel you have sufficient control (along all axes) compared to a proper router?
I was also surprised to see such milling was doable with what looks like a Dremel (4000?) and a Dremel Workstation!
I guess that such a static tool must imply to manually move the piece you want to machine, thus the use of some guide and doing one direction at a time. Right?
"it will need an upgrade to be able to cope with the 3090 / 5950x combo though"
me: with that space available soon he will start exploring ways around Newton's thermodynamics laws
me: with that space available soon he will start exploring ways around Newton's thermodynamics laws
Yeah, you wouldn't want to do anything harder than Plexi Glass (with a good milling bit that works surprisingly well though). Metals are out of the question. For the original modification of the case (about 12 pages back in this thread) I had a friend helping me out with his pretty well-equipped metal workshop.Once again an amazing read. Thanks for documenting and sharing this process
Btw, how are your experiences milling things with a rotary tool like that? I assume it's only suitable for soft materials and smaller jobs? Still, do you feel you have sufficient control (along all axes) compared to a proper router?
It's a Dremel 8200 with Dremel 220 workstation (that's a stand, really) - incredible value for money. Next time i'll probably go for a wired one - since I have the stand, I'm not really making much use of the "portable" feature, and I run into the odd charging break every now and then...
Similar threads
