Cooling the NSX: Fixing a Rare OEM Oversight

[Honcho]

I expect that the radiator inlet and outlet in a forced flow system is more a function of packaging considerations. There may have been some other design consideration that dictated the radiator connections in the NSX.

The main reason for the NSX bottom flow design is air bubbles. The mid-engine layout results in a long coolant loop, where the coolant goes down under the car. The engineers were super nervous about trapped air and designed the radiator path to encourage any bubbles that got pushed under the car to rise to the bleed plug. Flowing from the bottom up helps those bubbles make it to the top of the rad. This is a relatively unique requirement compared to the simple and short coolant path of the front engine cars Madhatter mentioned he has experience with. It's a mid-engine thing.

But, several NSX race cars and race radiators have swapped the inlet and outlet, likely for perceived improved cooling function. I remember reading install instructions that said you needed to fab up different hoses to accommodate the top-feed inlet.

I would love to see water temp log data between the OEM layout and Madhatter's modified path on a road course for say five laps each in full boost. I'm curious if there would be a difference.

 

[Madhatter10-6]

I would love to see water temp log data between the OEM layout and Madhatter's modified path on a road course for say five laps each in full boost. I'm curious if there would be a difference.

That would be interesting to see! I’d also like to mount additional sensors to monitor temperatures at multiple points throughout the system for a better understanding of what’s happening in both scenarios.

 

[TANTO]

Another Engineer (retired) who has designed quite a few systems with heat exchangers where absolute performance was imperative, i agree that unless the fluid flow is say less than 2 feet per second (which i'm sure ours is higher) convection is not a factor even a little bit. And of course as Honcho said on a car the layout and possibility of getting air into the system is likely paramount. And you don't want to go up and down more than needed. For your HP level - i think Honda's safety factors are like 20% so you certainly have to "really" up your radiator big time via double/triple pass copper radiator like now. Since it's not likely possible or desired to uprate our water pumps, the electric pumps thing (even variable speed units are possible) with added sensors all over the place sounds really interesting for your HP levels. Over analysis is a real thing. Long ago a senior Engineer at Rockwell working on the space shuttle (him - i was a facilities engineer) once said to me "one actual test is equal to 100 engineer opinions"- so there is that. Not to pile on you Madhatter - this is one learned crowd!

 

[Madhatter10-6]

Not to pile on you Madhatter - this is one learned crowd!

The community can "pile on" as much as it likes. Picture yourself in an airplane, with a crowd (forum) on the ground shouting up at you, “Your ideas are flawed! It’ll never fly!”—how much weight would their words really carry?

The key point everyone seems to have overlooked is: I am doing this in the real world on my dyno. You have feelings, I have data logs. Forum members sharing their opinions and credentials just to tuck themselves into bed at night, proud of having "won the internet" means nothing to me.

Back to the radiator flow discussion. Since I haven't implemented that component yet, I genuinely don’t know if it will perform better in the real world—or to what degree. I went ahead and ordered the dual-pass PWR last night, but it’s 4–5 weeks out. When it arrives, I plan to plumb it top-down. I do expect better temps, but it won’t be an apples-to-apples comparison since it’s drastically superior to my old setup (dual-pass, thicker core, twin SPAL fans, etc.).

If I’m feeling motivated, I may add sensors and do some testing. Measuring inlet vs. outlet temps with only the flow configuration changed should provide a solid comparison? I also wonder if it would be more accurate to rely on the TW sensor temperatures and continue using it as my true north?

 

[RyanRA]

We're entitled to our own opinions, but not our own facts : )

So be motivated, add the sensors, test, and collect the data. It'll be cool to see the results. Pun intentional.

 

[Honcho]

The community can "pile on" as much as it likes. Picture yourself in an airplane, with a crowd (forum) on the ground shouting up at you, “Your ideas are flawed! It’ll never fly!”—how much weight would their words really carry?

The key point everyone seems to have overlooked is: I am doing this in the real world on my dyno. You have feelings, I have data logs. Forum members sharing their opinions and credentials just to tuck themselves into bed at night, proud of having "won the internet" means nothing to me.

Back to the radiator flow discussion. Since I haven't implemented that component yet, I genuinely don’t know if it will perform better in the real world—or to what degree. I went ahead and ordered the dual-pass PWR last night, but it’s 4–5 weeks out. When it arrives, I plan to plumb it top-down. I do expect better temps, but it won’t be an apples-to-apples comparison since it’s drastically superior to my old setup (dual-pass, thicker core, twin SPAL fans, etc.).

If I’m feeling motivated, I may add sensors and do some testing. Measuring inlet vs. outlet temps with only the flow configuration changed should provide a solid comparison? I also wonder if it would be more accurate to rely on the TW sensor temperatures and continue using it as my true north?

I think an inlet/outlet sensor log would probably tell you about how the system is doing and I'm not sure the flow direction will make much difference at all, honestly. It's interesting to think about using TW as your primary metric since better cooling performance should translate to lower overall water temp even after passing through the engine (assuming the turbo components stay the same). I'm not sure the dyno is the best place to measure cooling performance- you'd probably need to take it to a road course and run several laps to understand any improvements due to airflow and aero considerations that only can exist in that environment (or a wind tunnel if you have access). It's what most race teams do when developing their platforms.

People have been racing turbocharged NSXs for a long time. Honda was the first with their works NSX GT1 cars at LeMans in the mid 90's. That car used a very heavy duty cooling system since it was designed to run in full boost at 650hp for 24 hours in the summer in France. If you look at the car, the radiator is nearly hermetically sealed so that all airflow goes through the core. There is a massive hood vent too, so that hot air exits without disturbing the bottom low-pressure zone under the car. Also, the radiator is mounted at an angle to the front duct, aligning it with the hood vent. None of those design choices are accidents. Honda had access to wind tunnels and as much track testing time as they wanted. That's why you're hearing from several folks here that "Step 1" for a turbo NSX is a big radiator and hood vent.

Check out the GT1 turbo setup. Honda flipped the C30 90 degrees, likely to make turbo routing and cooling easier. Look at those big intercoolers on each side.



Here's screen grab from the LeMans class-winner GT2 car, which used the same cooling layout as the GT1 turbo. Notice the angle and mounting of that giant radiator. It's also a bottom-in / bottom-out setup.



You can see the coolant lines here, showing the bottom inlet and output. It looks like they ran the oil cooler on the hot inlet side of the radiator.



So...big hood vent, big radiator, oil cooler right before radiator. Radiator flow doesn't appear to matter- at least not in a twin turbo LeMans car.

 

[Madhatter10-6]

I think an inlet/outlet sensor log would probably tell you about how the system is doing and I'm not sure the flow direction will make much difference at all, honestly. It's interesting to think about using TW as your primary metric since better cooling performance should translate to lower overall water temp even after passing through the engine (assuming the turbo components stay the same). I'm not sure the dyno is the best place to measure cooling performance- you'd probably need to take it to a road course and run several laps to understand any improvements due to airflow and aero considerations that only can exist in that environment (or a wind tunnel if you have access). It's what most race teams do when developing their platforms.

Agreed. While the dyno is a useful tool, it doesn’t perfectly replicate real-world conditions. I don’t think it has zero correlation, but I strongly believe in testing systems in the actual environment they’ll be used in.

People have been racing turbocharged NSXs for a long time. Honda was the first with their works NSX GT1 cars at LeMans in the mid 90's. That car used a very heavy duty cooling system since it was designed to run in full boost at 650hp for 24 hours in the summer in France. If you look at the car, the radiator is nearly hermetically sealed so that all airflow goes through the core. There is a massive hood vent too, so that hot air exits without disturbing the bottom low-pressure zone under the car. Also, the radiator is mounted at an angle to the front duct, aligning it with the hood vent. None of those design choices are accidents. Honda had access to wind tunnels and as much track testing time as they wanted. That's why you're hearing from several folks here that "Step 1" for a turbo NSX is a big radiator and hood vent.

Also Agreed. My '91 Turbo has a vented hood, but on my 2005, I’ll likely keep the exterior OEM for aesthetic reasons. That said, the vented hood is undeniably better for cooling. Since this car is being built as a fun street setup and likely won’t see any track time, I don’t see the need for it.

Check out the GT1 turbo setup. Honda flipped the C30 90 degrees, likely to make turbo routing and cooling easier. Look at those big intercoolers on each side.

That’s wild! Did they run the same transmission in this setup? I assume the trunk was gutted?

 

[Madhatter10-6]

Continuing the discussion on my proposed changes, there are two categories: Fluid and Electrical. For now, let’s set aside the radiator flow direction topic. I still believe top-down is better, but I’m not certain. Even if it proves to be better, I concede it will likely only offer a small margin of improvement.

So, let’s keep the conversation focused on the first two changes—please refrain from bringing up personal feelings toward me, your background, or degrees. I’m looking for pushback on the actual logic of the following points, with relevant counterpoints.

Topic: Placing the thermostat immediately before the radiator.

Reasons that support this being superior:

• The OEM thermostat’s location is sensing two temps; circulated fluid in the engine and some of the pass by from the hole in the thermostat before it opens. Once it opens the thermostats main temp reference becomes the coldest fluid in the entire chain going against modern automotive designs, where the engine’s temperature directly controls the coolant system, offering more precise cooling. The inline thermostat responds purely to engine heat output, not a mix of engine and radiator return temps. This decouples radiator performance from flow control, letting the engine dictate cooling needs without interference from cold coolant feedback.
• It will be higher flowing. The hollowing out (and eventual removal when I make a ring) of the OEM thermostat with the addition of a higher flowing less restrictive inline thermostat will allow for better coolant circulation. This will let the system respond faster and more uniformly to temp changes.
• Easier service / Fault detection. With the thermostat being moved inline pre radiator changing it becomes much simpler and troubleshooting it does as well. By simple touching the line pre and post thermostat you will be able to quickly easily tell if the thermostat is doing its job.

The OEM’s cold coolant influencing the thermostat is a limitation, not a feature. It muddies engine temp control, delays flow response, and ties cooling to radiator whims. My mod cuts through that noise—constant engine flow, thermostat driven by raw engine heat, radiator doing its job downstream. It’s a cleaner, more logical setup for performance, reliability and service.


Topic: Changing the high-speed fan to reference TW sensor

Reasons that support this being superior:

• The OEM system uses a delayed temperature reference point, arming the highest output part of the cooling system based on the coldest, least accurate measure of the engine’s actual operating temperature. This approach prioritizes quiet operation and gradual cooling, but it lags behind the engine’s cooling needs, especially under load—such as during spirited driving or in traffic on hot days. My modification responds in real time to the actual engine temperature, implementing cooling strategies immediately when needed.
• Mine is simpler. My system uses just one sensor for the dash and another to link directly to the most current engine temperature to control the fans. This eliminates the need for complex dual-stage, delayed triggers and multiple temperature points to process and adjust. Fewer components mean fewer points of failure and a more straightforward, reliable system.
• By basing fan operation on real-time engine temperature data, my system ensures the highest level of cooling when the engine needs it most, reducing the risk of thermal stress. This ensures the engine operates within its optimal temperature range, protecting components and enhancing performance.