Best and Worst Positions for a 360 AIO, Phase 1 Results (TPCEA)

The top 5 configurations for the CPU.

The following contains temperatures and ranks first. Temperatures are expressed NOT in Delta T Celsius. There are 3 listed for each component (VRM, Mobo, CPU & GC) in this order: 1-hour Prime 95 Small FFTs, Cinebench 2024 CPU Multi-Core & 3DMark Time Spy. Prime 95 has a more rigorous torture test for the CPU than Linpack Extreme and the other tests I tried.

The second section shows the scores from CB24 and Time Spy, including FPS. I think the results are suggestive of which configurations will work better for gaming (TS) and artistic work (CB); does that make sense?

Naturally, results will vary for each PC, especially since I used a vertical mount on the PSU shroud, and there was no direct exhaust point. More exhaust and/or intake fans might've been helpful. Phase 2 will explore the configurations with the GC in the PCIe slot.

The ambient temperature was usually between 21-23, but it got up to 25 on a couple of tests.

1. Side mount, tubes up, exhaust

Overall temps score (lower is better): 11

 5000x side top exhaust.png   316.72KB   0 downloads

2.. Side mount, tubes down, intake

Overall temps score: 26

 5000x side bottom intake.png   320.76KB   0 downloads

3. Side mount, tubes up, intake

Overall temps score: 27

 5000x side top intake.png   317.71KB   0 downloads

4. Front mount, tubes up, intake

Overall temps score: 23

 5000x front top intake.png   404.75KB   0 downloads

5. Side mount, tubes down, exhaust
Overall temps score: 10

 5000x side top exhaust.png   316.72KB   0 downloads

Edited by RevGAM, 20 March 2024 - 12:30 PM.

The top 5 configurations for the GPU (GC).

1. Side mount, tubes down, exhaust

Overall temps score: 10

 5000x side bottom exhaust.png   317.31KB   0 downloads

2.Side mount, tubes up, exhaust

Overall temps score: 11

 5000x side top exhaust.png   316.72KB   0 downloads

3. Front mount, tubes down, exhaust

Overall temps score: 20

 5000x front bottom exhaust.png   404.64KB   0 downloads

Sorry, I cannot upload any more images as there is not enough space left in my account.

4. Top mount, tubes forward, exhaust

Overall temps score: 14

5. Front mount, tubes down, intake

Overall temps score: 32

The best compromise configurations to help your CPU and GPU.

The points shown are only for the CPU & GPU. Ties were resolved by looking at VRM and mobo points, too.

1. Side mount, tubes up, exhaust (1+2=3 points)

2. Side mount, tubes down, exhaust (5+1=6 points)

3. Top mount, tubes forward, exhaust (6+4=10 points)

4. Front mount, tubes up, intake (4+6=10 points)

5. None

The best compromise configurations for all 4 components.

CPU, GPU, Mobo & VRM (heatsinks by the CPU socket)

Overall Best (cool CPU & GPU, hotter mobo and VRM)

Side mount, tubes up, exhaust (1+2+4+4=11 points)

Runner-Up (hotter CPU, cool GPU, mobo and VRM)

Side mount, tubes down, exhaust (5+1+1+3=10 points)

2nd Runner-Up (hotter CPU & GPU, cool mobo and VRM)

Top mount, tubes forward, exhaust (6+4+3+1=14 points)

If you don't have a side-mount possibility, just remember that the AIO is for cooling your CPU, not your GPU, VRM and mobo. However, that said, there are ways to improve the temperatures of the other 3.

Mobo: Honestly, the mobo is less of a concern. The highest temperature that I saw was 45^o C. Still, you can position fans to blow across the mobo or exhaust air faster. Or just laminate the whole thing with waterproof paint and sit it in a tank of water.      I suppose never turning it on is not an option?

VRM: An AIO with a VRM fan (a fan built into the block) can help cool off the VRM and some other components nearby, including the Arctic Liquid Freezer III, Enermax LiqMaxFlo and LiqMaxFlo SR, Asus ROG Ryujin 3, MSI MEG CORELIQUID S series, InWin BR series, and Silverstone IceMyst. I cannot endorse any of these. I have the LF3 and the LMF SR, but I haven't tested them.

Graphics Card (GC): The best thing you can do for this is make sure fans are blowing air to it. The easiest solution is to create ducting between the intake fan, the GC and (if needed) the exhaust fan so that it's a channel all the way through, ensuring that the card gets plenty of fresh air. Actually, you can do the same thing for the CPU! Ducting may not be your idea of beautiful (unless you make it beautiful) but, when properly done, it'll keep your GC from thermal throttling !

Edited by RevGAM, 20 March 2024 - 08:56 AM.

I'm just reiterating a note I added to the first visual aid post.

The ambient temperature was usually between 21-23, but it got up to 25 on a couple of tests.

This is important because, although we use delta T to indicate the actual temperature above ambient, the ambient is actually somewhat important. If your room's ambient is 40^oC, you're going to have a much harder time keeping your computer cool because the WHOLE THING will start at 40^oC, whereas it's much easier when your ambient is, say, 25^oC. Also, it's not necessarily true that heat will increase in a linear fashion given the influence of the ambient temperature of the other components further contributing to the heat in the case. Thus, if the temperature increase is 50^oC, that might mean 25^oC+50^oC=75^oC+increase vs 40^oC+50^oC=90^oC+increase, and now you're starting to get into thermal throttling, although certain chips are DESIGNED to drive the temperature UP to 95^oC you are still losing performance because the full potential is limited when you don't keep the processors cooler..

Edited by RevGAM, 20 March 2024 - 08:02 AM.

Very nice!

Very nice!

Thank you, sir!

I'm going to add one more thing.

The following is purely theoretical.

A high rank in Cinebench 2024 MIGHT indicate which configuration is the best for artistic types, e.g. designers, engineers, artists, videographers, 3D rendering, etc. from the perspective of the CPU (& iGPU?)

A high rank in 3DMark: Time Spy MIGHT indicate which configuration is the best for gamers from the perspective of the CPU and GPU.

Since my CPU didn't thermally throttle, having gotten no higher than 89^oC, however, we didn't see what would happen in that scenario, thus the test scores don't align with the temperature rankings. I can't do anything about this, however, because my CPU is only MARGINALLY "K" and when I tried to overclock or overvolt it, I got more heat and lost performance.

Also, please note that:

• The difference in scores for the CPU was only 728 points. and the difference in FPS was only 2.44, so it doesn't really matter for almost anyone except maybe pros.
• The difference in scores for the GPU was only 1,366 points, and the differences in FPS in tests 1 & 2 were only 8.52 and 8.18 FPS respectively.
• The overall difference in Time Spy scores was only 1,200 points.
• The overall difference in Cinebench 2024 CPU Multi-Core scores was only 64 points.

I do not feel that some of these are significant enough to warrant being guided by the following results, but it's up to you. So, the following should be taken with a handful of salt (i.e. be skeptical). Really, though, just stick with Side mount, top, exhaust, if you can! It's the only one that's all blue and green ranks (1-5). It is in the top 5 (in blue) anyways for each of the below.

The Best Configurations for CPU-Oriented Software (by Scores & FPS)

TIME SPY                                                                                                 CINEBENCH 2024 Multi-Core CPU Test

1. Front mount, tubes down, intake (#8 for the CPU temps)                     Front mount, tubes up, intake (#4 for the CPU temps)

Time Spy CPU Score: 15,607, 52.44 FPS                                                CB24 Score: 1,213

2. Top mount, tubes forward, intake (#7 for the CPU temps)                   Top mount, tubes forward, exhaust (#6 for the CPU temps)

Time Spy CPU Score: 15,577, 52.33 FPS                                               CB24 Score: 1,194

3. Side mount, tubes up, intake (#3 for the CPU temps)                         Front mount, tubes down, intake (#8 for the CPU temps)

Time Spy CPU Score: 15,463, 51.95 FPS                                               CB24 Score: 1,191

4. Side mount, tubes up, exhaust (#1 for the CPU temps)                      Front mount, tubes down, exhaust (#10 for the CPU temps)

Time Spy CPU Score: 15,413, 51.78 FPS                                               CB24 Score: 1,190

5. Front mount, tubes down, exhaust (#10 for the CPU temps)              Side mount, tubes up, exhaust (#1 for the CPU temps)

Time Spy CPU Score: 15,340, 51.54 FPS                                               CB24 Score: 1,186

The Best Configurations for GPU-Oriented Software (by Scores & FPS)

TIME SPY Graphics Score with FPS

1. Front mount, tubes up, intake (#6 for the GPU temps)

Score: 20,136: 131.99 (#2) & 114.87 (#2) FPS

2. Top mount, tubes forward, exhaust (#4 for the GPU temps)

Score: 20,130: 132.04 (#1) & 114.77 (#3)

3. Side mount, tubes up, exhaust (#2 for the GPU temps)

Score: 20,122: 131.78 (#4) & 114.89 (#1)

4. Front mount, tubes up, exhaust (#7 for the GPU temps)

Score: 20,106: 131.63 (#3) & 114.59 (#5)

5. Top mount, tubes forward, intake (#10 for the GPU temps)

Score: 20,100: 131.68 (#5) & 114.72 (#4)

So, I'm going to have to delay phase 2.

Because I have to do a different phase 2.

This will be a PITA because I'm going to do all 10 tests again, but with the fans between the rad and case instead. This will take into account differences between the results for push and pull without a gasket to minimize those differences, since most people don't use, and I'm not sure if any AIOs are sold with such a gasket. I think that the amount of difference will vary depending on various factors but, if the same fans were used on all AIOs, then the difference would boil down to how dense and thick the rad is.

I can't go back and edit the post with the test setup info, so I want to clarify a few points.

1. The case includes glass panels on the front and top that have a gap on each side. I didn't remove them since most people who would buy this case would not remove them.
2. For exhaust tests, I removed the dust filter from the top, front or side where the AIO was mounted because there shouldn't be a dust filter for exhaust.
3. I implied but didn't make patently clear that the fans were all running on fan curves (via BIOS), which is what the "turbo" and "standard" mean. The pump was at 100%, which is normal for a pump and is not a curve (it's flat, after all). ;P

Just a brief update that I completed the second half of phase 1. Everything was the same as before, EXCEPT that the fans were sandwiched between the radiator and the case this time. That is rather more labor intensive and challenging, and I ended up accidentally stripping 2 of the mounting holes in the rad.

I will start entering and processing the data today. I hope to have those results posted later this week. 

Edited by RevGAM, 15 April 2024 - 03:34 AM.

The initial analysis shows that side mount (with glass panels on) is the best. Regrettably, I no longer have space on here to upload more visual aids, so I will only be able to release the second half of phase one's results in text.

A peer of mine on LTT Forum, NorKris, who was inspired by my work, just released his results. He did many things differently, but here are what I feel are the most critical points. If you want to look at the results he released, which are quite interesting, please PM me for the link, or go there to find it yourself. I don't feel comfortable linking to a major competing forum in this instance.

1. He used a 4000D that he modified by making a hole and adding a side mount dust filtration. He called this the 4500D. The 4000D has not got the glass panels on the top and front like my 5000X.
2. He didn't test tubes up and down, probably just tubes up.
3. He didn't test the fans pushing AND pulling air through the rad. He only tested pushing.
4. He used a smaller, old GTX 680 (not sure which model)

The strangest thing about my results is that they seem largely divorced from temperatures in the CPU, GC, mobo and VRMs. This tells me that, as long as parts aren't overheating, temperatures are not the most important factor!

The top 5 configurations for the CPU.

The previously posted test results were done with the fans on the inner side of the radiator. This series of tests was done with the fans between the rad and the case. This is the second step in Phase 1. The only other difference between this and the above test results is minor, but it bears keeping in mind. ALL tests for this step were done with the block mounted such that the tubes came out the right (forward) side to maximize the reach of  the tubes in all directions. I don't foresee that the variance in the first test had any significant impact on the results, and expect that it would easily fall within the margin of error.

The following contains temperatures and ranks first. Temperatures are NOT expressed in Delta T Celsius. The room temperature was between 21-23^oC. There were a total of 10 tests covering CPU & GC, VRM, Mobo, clockspeed in this order: 1-hour Prime 95 Small FFTs, Cinebench 2024 CPU Multi-Core & 3DMark Time Spy. Prime 95 has a more rigorous torture test for the CPU than Linpack Extreme and the other tests I tried, and I have been informed that LinPack Extreme requires that customized settings be applied for the most rigorous results. I'm not going to do something that most users won't do. Temperatures, RPMs and clocks were taken from HWiNFO64.

Naturally, results will vary for each PC, especially since I used a vertical mount on the PSU shroud, and there was no direct exhaust point. More exhaust and/or intake fans might've been helpful. For example, my peer @NorKris on another forum did similar testing as I mentioned before, and the results demonstrate that when there are no glass panels blocking the airflow, the front is better than the side. You need to carefully consider the case you have to determine which position is going to provide the best airflow.

Also, how many fans you have and where you put them can sometimes make a good/bad difference. Use of airflow ducts to direct air from fans to critical components (the CPU and the graphics card) are an opportunity to improve this, and also a chance to add some superficial art to the ducts if they don't look nice. You can 3D print ducts, which is very challenging, or make them from cardboard - don't worry, the only time that cardboard would ignite is if the temperature were above 451^oF, which means your computer caught fire before the cardboard did!

In other words, there is no simple answer to the question of where to put the AIO except to say that front and side are generally better than top.

Phase 2 will explore the same 20 configurations with the GC in the PCIe slot. For that, unfortunately, I will have to use a different GC. The only other one, which I haven't tested yet to verify it works correctly, is an EVGA Geforce GTZ 760 SC (two fans).

The ambient temperature was usually between 21-23, but it got up to 25 on the final TS test for top, intake as I had to rerun it at a later date.

1. Side mount, tubes up, intake

Overall temps score (lower is better): 14

2. Side, down, intake

Overall temps score (lower is better): 23

3. Front, up, intake

Overall temps score (lower is better): 21

4. Side, up, exhaust

Overall temps score (lower is better): 14

5. Top, exhaust

Overall temps score (lower is better): 13

Top, intake was the worst for the CPU, GPU and mobo, and #3 for the VRM.

The temperature range was from 81-93^oC during Prime 95 according to HWiNFO64. This is a 4^o larger range than the first step of this phase.

Edited by RevGAM, 15 April 2024 - 06:02 PM.

The top 5 configurations for the GPU (GC).

I think I didn't mention before that the overall temps score is simply the total of the four component temperatures that were chosen for being the hottest.

1. CPU: Prime 95
2. GPU: 3DMark Time Spy
3. Mobo: Time Spy
4. VRM: Prime 95

Unlike the first step of phase 1, NONE of the highest temps were consistently from Cinebench 2024. However, for the CPU, Mobo and VRM, they were the second highest (but there was generally a difference in temperature that was quite large for the CPU, medium for the VRM, and small for the mobo. For the GPU, most of the 2nd highest scores were from Prime 95, oddly enough, except for front, down, intake and side, down, intake, for all 10 of which there was a major temperature difference.

1. Side mount, tubes up, exhaust (CPU rank #4)

Overall temps score: 12

2. Side, up, intake (CPU rank #1)
Overall temps score:14

3. Front, up, exhaust (CPU rank #8)
Overall temps score: 24

4. Front, down, exhaust (CPU rank #9)
Overall temps score: 31

5. Side, down, exhaust (CPU rank #7)
Overall temps score: 25

The top, intake position was worst (i.e. 10th place) for the CPU, GPU and mobo, and #3 for the VRM.

The temperature range was from 93-97^oC during Time Spy according to HWiNFO64. This is a 3^o smaller range than the first step of this phase.

Edited by RevGAM, 15 April 2024 - 05:28 PM.