Log Dual 280GTS Meshy-boi (5900X, 3080 FE, 14.6L) build log

[biopunk]

In 2020 I got into SFF and water-cooling and by the end of Q4 I built a dual-rad custom loop in T1 v1.1. While it performs quite well I'm still not completely satisfied with it.

• Even though I got lucky to get the case a few months earlier than expected, duo-tone was not the color that I wanted.
• Aquanaut's performance turned out to be a bit disappointing. After I added GPU to the loop, CPU reaches 79℃ under load.
• The noise level under load is quite high for a fully water-cooled system.
• RGB doesn't look good through the mesh and at 40-45 degree angle it gets completely blocked by the mesh.
• There is no space inside for a reservoir so passive air bleeding is pretty much impossible with side panels on.
• Making nice looking bends at such a short distance from each other was extremely hard. I wasted probably 2m of tubing before I made something usable but it was still far from perfect.

This build made me realize that if I want to achieve good thermal performance, noise level, and aesthetics at the same time I might have to upsize a bit. So I decided to do my next build in Mesh-Licious which has a lot more room for water cooling, can fit 2x GTS280, and comes with a TG panel.

My wishlist for this build:

• Standalone CPU block for better thermal performance and some OC headroom
• Reservoir for passive air bleeding
• Fill and drain ports
• Flow meter
• Second temperature sensor
• Two-stage inline filter to prevent water blocks from gunking
• Tempered glass panel to show-off the components and tubing
• RGB fans and strips

Part list

Parts to buy:

1. Front radiator: Black Ice Nemesis 280GTS XFlow. Performs within a margin of error of 280GTS while being much less restrictive.
2. Side radiator: Black Ice Nemesis 280GTS. Best performing slim radiator but very restrictive.
3. Fans: EK-Vardar 140ER RGB. They don't look as nice as Corsairs but have higher airflow and static pressure, and use standard D-RGB connectors.
4. Pump/res combo: Barrow Acrylic DDC top with 60mm tall 50mm ⌀ reservoir. A wider or taller reservoir would be problematic to fit between PSU and front radiator.
5. Flow meter: High Flow NEXT or LT
6. Temperature sensor: Aqua Computer inline temperature sensor MF
7. Additional fittings
8. Acrylic tube

I thought that 2x U-flow GTS280 being very restrictive might be too much for DDC-1T Plus to handle at 20-25% RPM so I decided to use 1x U-flow and 1x X-Flow which perform about the same.

Sketch:

Notes on the sketch:

1. Not 100% sure that 280GTS X-flow will fit at the front due to its length.
2. There is only ~8mm space for cables coming out of PSU so I'll have to use unsleeved cables.
3. I don't know yet how inlets on the reservoir cap will end up being oriented. I'll probably use a snake fitting to connect the tube.
4. I might be able to move the drain port to the lowest point when I'll know the relevant measurements of the case.
5. The pump housing will be either screwed directly to the bottom panel if hole patterns match or it will be screwed to a custom pump bracket.
6. The side radiator will be mounted to a custom bracket.
7. I'll have to rely on some RNG when choosing fittings to place the filter between the radiators.

For now I'll be waiting for Mesh-Licious pre-order and restocks on Highflow NL, and thinking how to mount Quadro and how to connect RGB stuff.

 

[thelaughingman]

I'm in a similar boat - planned for dual rad T1 and got the Aquanaut but now moving out to larger cases because of the Crosshair VIII Impact. Really wanted dual 280 rad in the C4 to work but not possible due to the SO.DIMM card. So now I'm using the NR200 whilst waiting for the right case to come out (either dual 240 rad at 12L or less or dual 280 rad at 15L or less).

For the parts the only suggestion I have is for you to consider the Uni Fan SL140. Look forward to future updates

 

[eedev]

I will follow your build log, I am interested!

 

[EndingCredits]

Whenever I see this thread title I think of the old Nvidia 280 GTX (they never did do a GTS version actually)

But 2x280 rads seems to be perfect for the Meshlicious; I wonder if they will even come out with a custom side panel for mounting the 2nd one. Looking forwards to seeing this build!

Also, for your PSU cables, could you just partially sleeve them (i.e. sleeve them all the way up to a few cm from the PSU)?

Also (v2.0), is there a reason you're using 1 xflow radiator and one regular? I imagine it would be much easier to go with both regular, or go both xflow with the tube routing.

 

[biopunk]

Also (v2.0), is there a reason you're using 1 xflow radiator and one regular? I imagine it would be much easier to go with both regular, or go both xflow with the tube routing.

280 GTS is the most restrictive slim radiator. I'm afraid that the flow rate will be really bad with 2 of them and DDC 3.2 @ 20-25% RPM.

The X-Flow model is ~3 times less restrictive and with fans at low RPM it even slightly outperforms the U-flow model:

At medium RPM the X-flow model performs about the same as the U-flow:

 

[biopunk]

Received 90-degree DP 1.4 adapter today. No issues with 4K HDR10 @ 60Hz.

 

[Kallus]

280 GTS is the most restrictive slim radiator. I'm afraid that the flow rate will be really bad with 2 of them and DDC 3.2 @ 20-25% RPM.

The X-Flow model is ~3 times less restrictive and with fans at low RPM it even slightly outperforms the U-flow model:

At medium RPM the X-flow model performs about the same as the U-flow:

Another possible way to get around the restrictiveness of the radiators is to hook them up in parallel. This works like resistors in an electrical circuit. So if they are identical it halves the resistance. The two radiators then also share the same temperature gradient along with a faster flow rate for better cooling(that is my hypothesis anyways)

 

[Koxx5D]

280 GTS is the most restrictive slim radiator. I'm afraid that the flow rate will be really bad with 2 of them and DDC 3.2 @ 20-25% RPM.

I have the same concerns for my W1 and its 2 x HWL 280 GTS. 2 DDC pumps in series can help, but you have to be careful with the excess pressure in my opinion. Possible leaks.
Have you looked here or here ?

I know the informations are very old, but they are very interesting.

 

[biopunk]

I have the same concerns for my W1 and its 2 x HWL 280 GTS. 2 DDC pumps in series can help, but you have to be careful with the excess pressure in my opinion. Possible leaks.
Have you looked here or here ?

2x DDC would probably be overkill for your use case, just get the most powerful D5 or DDC pump you can fit and find a good balance between the flow rate and noise!
D5 has the highest flow rate (~1500L/h @ 100%) and runs quieter. The next best thing is probably EK-DDC 3.2 PWM with a maximum flow rate of ~1000L/h. it should do just fine with 2x GTS 280 while providing a good flow rate and being almost silent @ 25-40% RPM.
I wish I got that one instead of Laing DDC-1T Plus which is rated at 500L/h @ 100%.

 

[Koxx5D]

2x DDC would probably be overkill for your use case,

Certainly, and I'm not going to install 2 pumps in the W1 but it will be necessary to take into a consideration these problems of constraints especially with the waterblocks containing a large number of micro channels .

 

[biopunk]

Certainly, and I'm not going to install 2 pumps in the W1 but it will be necessary to take into a consideration these problems of constraints especially with the waterblocks containing a large number of micro channels .

I'll receive my 280 GTS next week and then I'll test how much it affects the flow rate with a DDC pump. Stay tuned!

 

[biopunk]

Custom USB cable for Quadro​

Brexit messed up deliveries from the EU and while I'm still waiting for the rest of the parts I decided to replace the original 100cm USB cable which comes with Quadro by a shorter one.

Parts:

1. PicoBlade 5-pin cable assembly. These come in lengths up to 60cm and use 28 AWG wire.
2. or PicoBlade 5-pin receptacle housing and 5x PicoBlade female pre-crimped leads. Pre-crimped leads come in lengths up to 45cm, and use 26 or 28 AWG wire.
3. Dupont 2.54mm 5-pin female housing and 5x Dupont crimp terminals.

As for the tools, I used CTX3 crimping tool, some cheap wire stripper, and side cutters from Amazon.

Result:


I recommend using 26 AWG pre-crimped leads as 28 AWG wire that Molex uses seems a bit too thin for Dupont crimp terminals.

As a side note, I tried crimping PicoBlade myself, and even though I figured out a technique that works, it's PITA and absolutely not worth the hassle.

 

[Koxx5D]

I'll receive my 280 GTS next week and then I'll test how much it affects the flow rate with a DDC pump. Stay tuned!

Hello,
thanks to you, I also received my quadro and I'm going to do the same thing you did for the W1. I'm impatient for your tests on the GTS.

 

[biopunk]

Hello,
thanks to you, I also received my quadro and I'm going to do the same thing you did for the W1. I'm impatient for your tests on the GTS.

Gonna wait for D5 pump/res combo I ordered at AliExpress before testing. Want to compare D5 (high flow rate, lower head pressure) vs DDC (lower flow rate, higher head pressure).
Just crimped a cable to connect Alphacool Eisfluegel Aurora to Quadro and tested it, seems to work fine.

 

[kodek]

I have a temporary build in mid tower with double 280 GTS and Barrow SPB17-TM pump (waiting for Winter One). There's almost no difference (0.5-1C) in cpu and gpu temps between pump at 34% (2700rpm) and 60% (3700rpm). I checked it playing CoD Warzone at 3840x1600 and running 10 min Cinebench R23. Tested on 5900x and rtx 3080.

 

[biopunk]

I have a temporary build in mid tower with double 280 GTS and Barrow SPB17-TM pump (waiting for Winter One). There's almost no difference (0.5-1C) in cpu and gpu temps between pump at 34% (2700rpm) and 60% (3700rpm). I checked it playing CoD Warzone at 3840x1600 and running 10 min Cinebench R23. Tested on 5900x and rtx 3080.

Good to know. Does that pump make any noticeable noise at 34%?

Would you be able to perform a similar test but between 20% and 34% instead? I believe 1% increase in that range should have much more impact on cooling performance than 1% increase between 34% and 60%. You can see in this post that going from 23% to 34% improves CPU temps with Aquanaut by 6℃.

 

[kodek]

Good to know. Does that pump make any noticeable noise at 34%?

Would you be able to perform a similar test but between 20% and 34% instead? I believe 1% increase in that range should have much more impact on cooling performance than 1% increase between 34% and 60%. You can see in this post that going from 23% to 34% improves CPU temps with Aquanaut by 6℃.

So in addition to the 60% and 34% I've tested it on 23% (2300rpm) and the slowest possible speed (10% = 1840rpm). In 10 min Cinebench R23 there's only 1℃ difference even between 10% and 60% so there's definetely not enough heat with only CPU stress test for this amount of cooling capacity. In Warzone in GPU temp there's 1℃ diff between each tested pump speed (so 3℃ diff between 10% and 60%). The CPU temp is the same on 34% and 60% and is hotter by 1.5℃ on both 20% and 10% (no difference between these two). As for the sound... 60% is pretty loud, you can hear 34% a bit but when fans are spinning it's not that annoying, 23% is silent but if you can hear it if you're very close and focus on it and the 10% is pretty much inaudible. I think I will use 23% from now on. Hope that helps.

 

[biopunk]

So in addition to the 60% and 34% I've tested it on 23% (2300rpm) and the slowest possible speed (10% = 1840rpm). In 10 min Cinebench R23 there's only 1℃ difference even between 10% and 60% so there's definetely not enough heat with only CPU stress test for this amount of cooling capacity. In Warzone in GPU temp there's 1℃ diff between each tested pump speed (so 3℃ diff between 10% and 60%). The CPU temp is the same on 34% and 60% and is hotter by 1.5℃ on both 20% and 10% (no difference between these two). As for the sound... 60% is pretty loud, you can hear 34% a bit but when fans are spinning it's not that annoying, 23% is silent but if you can hear it if you're very close and focus on it and the 10% is pretty much inaudible. I think I will use 23% from now on. Hope that helps.

Thanks! Interesting results. So it seems that flow rate is less important for a standalone CPU block with forward flow and jetplate compared to a CPU/pump block combo with reverse flow. Good to know that 2x U-flow GTS 280 + DDC is a feasible config even at a low pump speed.

I believe 20% is the lower limit for most DDC pumps, this could be the reason why you didn't see a difference between 10% and 20%.

 

[biopunk]

GTS 280 flow restriction test​

Loop setup:​

DDC pump: Laing DDC-1T 10W DC with Alphacool Eisdecke Plexi top
D5 pump: Barrow D5 18W PWM
Flow meter: Alphacool Eisflugel Aurora
Phobya PWM to DC transformer


Added inline filter because the rads still had some rubbish stuck in them after multiple flushes with distilled water.

When I started testing I realized that the Phobya PWM to DC transformer is very imprecise and I can't go below ~2500 RPM. That's much higher and thus noisier than I would run a DDC pump in a real system. I recorded the results at 2651 RPM which is roughly equivalent to 37% RPM of PWM-controlled DDC-1T Plus that I have in my current build and keep at 25% RPM.
With that said you can expect flow restriction to be more significant at lower RPM as head pressure would also be lower.

Results:​

Config​	Flow rate 1​	RPM 1​	Flow rate 2​	RPM 2​
DDC-1T (baseline)	276 L/h	2651 RPM	438 L/h	3712 RPM
DDC-1T + GTS 280	192 L/h (70% of baseline)	2651 RPM	326 L/h (74% of baseline)	3712 RPM
DDC-1T + GTS 280 X-flow	250 L/h (91% of baseline)	2651 RPM	398 L/h (91% of baseline)	3712 RPM
Barrow D5 (baseline)	221 L/h	3303 RPM	377 L/h	4850 RPM
Barrow D5 + GTS 280	170 L/h (77% of baseline)	3303 RPM	310 L/h (82% of baseline)	4850 RPM
Barrow D5 + GTS 280 X-flow	207 L/h (94% of baseline)	3303 RPM	353 L/h (94% of baseline)	4850 RPM

This confirms that GTS 280 is a very restrictive radiator and it's much more restrictive than GTS 280 X-flow. With DDC 3.2/DDC 3.25 at 20-30% RPM which is typical for SFF builds, we can expect GTS 280 to reduce the flow rate by at least 30%.

I think anyone who intends to use two or more GTS 280 in a custom loop should consider how its restrictiveness might affect cooling performance. Read excellent reviews of GTS 280 and GTS 280 X-flow on Xtremerigs, check your pump's spec sheet, read CPU and GPU block reviews, specifically how they perform at different flow rates, and judge for yourself.

Lastly, a quote from Xtremerigs:

The fantastic thermal performance comes with a price though, it is very restrictive, almost excessively restrictive even, and this is the only real downside to the Nemesis 280 GTS. If the port locations are suitable or perhaps even more convenient for your build we suggest considering the Nemesis 280 GTS X-Flow which is virtually equal in thermal performance but is much less restrictive due to it being a single pass radiator.

 

[Koxx5D]

I transmit the link of these results on the Winter One thread. Thank you, it's perfect.

I'm not going to be able to use the 280 Xflow, but by using 240, it could be a good way ... My waterblock is very restrictive (165 microchannels of 0.1mm)