Info Comparison of the Optimus Foundation AM4 block to the Corsair XC7 RGB CPU block in the same system

[Bavor]

I did a comparison of the Optimus Foundation AM4 block to the Corsair XC7 RGB CPU block in the same system.

When I built my open loop, the Optimus Foundation AM4 block I ordered was on back order. I wasn't going to wait an unknown amount of time for the CPU block to arrive to wait to finish the loop since it was the last part I ordered, so, I bought a Corsair XC7 RGB CPU block to use until the Optimus CPU block arrived. I figured I could resell the Corsair CPU block later.

My Optimus Foundation AM4 block arrived recently. Since I now have the two CPU blocks available, I decided to compare them in the same system. My open loop has EKWB quick disconnect fittings for the GPU blocks and CPU block. That way I can remove components to add NVMe SSDs, reach wiring connectors, or change components without draining the loop. It also makes it easy to fill the new parts with coolant when I swap parts.

The loop consists of a Corsair XD5 RGB pump/res combo, EKWB EK-CoolStream XE 360 360mm 60mm thick radiator with six 120mm fans in push pull, EKWB EK-CoolStream CE 280 280mm 45mm thick radiator with 2 140mm fans, and 2 GPU water blocks. GPUs were idle during the test. I use EKWB ZMT 10/16 tubing for the loop. I use EK Cryofuel Clear coolant. The system is in a Fractal Design Define R6 case with the front door open.

The rest of the system:
CPU: AMD Ryzen R9 3950X
Motherboard: MSI MEG ACE X570
RAM: 64GB DDR4 3600 MHz Corsair Vengance RGB PRO
GPUs: 2x EVGA RTX 2080 Super with EKWB GPU blocks

I ran Intel Burn Test, Prime 95 small FFT, and AIDA64 for over 20 minutes each with the Corsair CPU block installed and a fixed pump speed and fan speed. I recorded the temperatures in an Excel spreadsheet. Then I removed the Corsair block, swapped the tubing and fittings onto the Optimus block, and filled the optimums block and attached tubing with new coolant. I installed the Optimus Foundation AM4 block and tightened the thumb nuts by hand as tight as I could with my fingers, per their instructions earlier in this thread. After installing the Optimus Foundation AM4 block and making sure I had no air bubbles in the block by having the pump run at full speed as I moved the system around in various positions then I also varied the pump speed high and low with the system sitting on the desk.

I ran the same test for the same time period with the same fixed pump speed(3700 RPM) and same fixed fan speed(1800 RPM for 120mm fans and 1600 RPM for 140mm fans). Room temperature was controlled with the home central air conditioning system and a digital thermometer on my desk monitored ambient air temperatures. I used the same thermal paste(Thermal Grizzly Kryonaut) for both blocks.

These are my results:

Temperatures are in degrees Celsius.

I thought that maybe the Optimus AM4 block has higher flow resistance than the Corsair block and requires more coolant flow, so I set the pump to the maximum speed and ran the tests again. However, there was no difference in results. I expected a larger temperature difference between the two CPU blocks considering the design differences between the two blocks.

 

[thesmokingman]

From the test data I've seen the Optimus block is on the high end for pressure drop, its a pretty restrictive block. The high end ek/aqua/hk blocks are on the low end for pressure drop. Regarding the no change when you maxed pump speed, I'd imagine that you are already saddled with a on the high side restriction for a single D5 pump. It makes sense then that the fans are running quite high rpm in relation.

The loop consists of a Corsair XD5 RGB pump/res combo, EKWB EK-CoolStream XE 360 360mm 60mm thick radiator with six 120mm fans in push pull, EKWB EK-CoolStream CE 280 280mm 45mm thick radiator with 2 140mm fans, and 2 GPU water blocks.

Also, you really need a water temp sensor. That will tell you how your loop is doing efficiency wise. On paper the Optimus should do much better but I think your flow rate is too low to make out the differences.

 

[Bavor]

From the test data I've seen the Optimus block is on the high end for pressure drop, its a pretty restrictive block. The high end ek/aqua/hk blocks are on the low end for pressure drop. Regarding the no change when you maxed pump speed, I'd imagine that you are already saddled with a on the high side restriction for a single D5 pump. It makes sense then that the fans are running quite high rpm in relation.

The EX Supremacy EVO seems to be average to above average for restriction or below average in flow rates among CPU block flow rates based on the tests I've seen depending on the version. The x99 version flowed better and had less pressure drop than the 1151 version.

I'm not sure if I agree that my loop is that restrictive or my pump is inadequate. The EKWB quick disconnects have significantly lower flow restriction and significantly lower pressure drop than any Radiator, CPU block or GPU block that I've seen tested. The GPU blocks run in parallel and aren't a significant restriction.

The CoolStream XE 360 radiator is about average in flow restriction/pressure drop for 360mm radiators and is one of the best performing 360mm radiators on the market.
https://www.xtremerigs.net/2015/05/31/ek-coolstream-xe-360mm-radiator-review/

The CoolStream CE 280 is higher restriction than the average 280mm radiator, but is still a very low restriction loop component when compared to the CPU blocks.
https://www.xtremerigs.net/2016/02/21/ek-coolstream-ce-280mm-radiator-review/
From the review "This means you could comfortably run 2 or 3 CE 280 radiators without having to be overly concerned about them reducing your flow rate by too much."

If you have reliable data that shows a single D5 pump can't handle 2 GPU blocks in parallel, a CPU block, and 2 radiators, I'd like to see it.

As, I said in the original post, I manually set the fans to that speed.

Also, you really need a water temp sensor. That will tell you how your loop is doing efficiency wise. On paper the Optimus should do much better but I think your flow rate is too low to make out the differences.

I have a coolant temperature sensor. It was included with the Corsair Pump/Res combo and gives me the temperature of the coolant in the reservoir. After 20+ minutes running Prime 95 small FFT, the coolant temperature was between 32C and 33C with the pump at 3700 RPM and fans at max speed with both CPU blocks. Running Prime 95 small FFT again and setting the pump to max speed didn't significantly change the coolant temperature. It looked like the coolant temperature stayed a little closer to 33C when the pump was at max speed. However that's probably within margin of error of the sensor.

 

[thesmokingman]

No offense but you have a D5 pump. You should know it has limited head pressure, even with an aftermarket top. Add up the pressure drops and you will not have much static pressure left. The Optimus block has over 1.5psi drop vs EK and the rest at around .75psi. Each rad, block, fitting, especially qdc adds to the pressure drop. Gpu blocks typically are 1psi too. D5's stock have just over 4psi and up to 5psi with a nice top. That's not a lot of static head to go around. You are not adding the losses up, the parts don't exist in a vaccum.

D5's are hiflow, but low head pressure pumps. The crux is all that hiflow is not needed in a pc loop and static head pressure is the most important attribute and that's a weakness of D5's. They're great for simple loops. D5's also dump their heat into your loop, so add another 30w+ to the heatload, etc. Your loop is getting to the limit of a D5 and with a super restrictive block with a greater than 1.5psi drop it can definitely affect things.

EK-Quantum Magnitude CPU Water Block Review

EK Water Blocks finally launches the EK-Quantum Magnitude, a halo CPU block with customization cues from the EK-Supremacy EVO from yesteryear that kicks things up a notch. With a clear focus on design, material choice, performance, and customization, it aims to offer a lasting CPU cooling...

www.techpowerup.com

 

[Bavor]

I've seen a single D5 pump powering systems with three radiators, a GPU block, and the Intel version of the Optimus CPU block without any issues. With my GPU blocks in parallel, their pressure drop total should be the same as a single GPU block according to every reliable source I could find on the subject. Unless there is a big difference between the pressure drop of the Intel and AMD versions of the block, I'm not sure if you are correct.

If multiple systems with more radiators than mine, single GPU block or GPU blocks in parallel, and a D5 pump aren't having issues, the only way I could see having an issue is if there was a large difference in pressure drop between the Intel and AMD blocks.

I can bypass the GPU blocks and remove the GPUs then put in an air cooled GPU to see if there is a difference in CPU cooling performance.

 

[Bavor]

I tried taking out the GPUs and putting in the GT 1030 I have in their place. Then I reconnected the the tubing with the EKWB/CLC quick disconnect fittings i had in place to make removing the GPUs easier. I ran AIDA64, Prime95 small FFT and Intel Burn Test on the Very High preset. The temperature differences were either 0C or 1C in the tests compared to running the same tests with the GPUs in the loop. It doesn't look like the D5 is insufficient in flow or head pressure if removing the GPU blocks make no significant difference.

 

[aigomorla]

A better measure of ambients would of been at the intake of the radiators.
Typically either right at the fins, or 1-2 inches away which is more ideal, but if not at the fins is fine as long as its in the same location for both testing blocks.

 

[Bavor]

I have a temperature sensor behind the dust filter on the front of the case. The sensor reports between 25.5C and 26.5C with the front door of the case open. It seems to be consistent and rarely varies when I set the fans to a fixed speed and the front door open. If I were to use that temperature instead of ambient, it doesn't seem that the results would vary much.

I also have a sensor measuring the temperature of the air exiting the top of the case. It usually reports a temperature between 1C and 2C higher than the coolant temperature depending on the fan speed.

 

[aigomorla]

I also have a sensor measuring the temperature of the air exiting the top of the case. It usually reports a temperature between 1C and 2C higher than the coolant temperature depending on the fan speed.

This is irrelevant because so many things can trip this value.

Sensor behind the dust filter is also not accurate because unless the radiator is right up against it, this value can swing, and when your dealing with LCS sometimes even that 1-2C can be a number significant enough to outrank it from the next product.

The sensor should ideally be right up against the radiator which you are getting air from.
You should also think of investing in a liquid coolant temperature probe of some kind to measure coolant temp.
This is how we used to do it to rule out of lot of people saying our tests are invalid, or trying to pull things to invalidate our tests.

I come from a time when things were a lot more scientific and not as blingy.
I used to be sponsored by Koolance / eK / Swiftech / DangerDen back during my active days, and helped a lot with MartinsLiquidLab, and SkinneeLabs.
My other alas is Naekuh, and i had earned quite the notoriety with LCS vendors back during my day.

 

[Bavor]

This is irrelevant because so many things can trip this value.

Sensor behind the dust filter is also not accurate because unless the radiator is right up against it, this value can swing, and when your dealing with LCS sometimes even that 1-2C can be a number significant enough to outrank it from the next product.

The sensor should ideally be right up against the radiator which you are getting air from.

Is this one close enough?

Its a few mm away from the radiator. Much closer and it would be touching.

You should also think of investing in a liquid coolant temperature probe of some kind to measure coolant temp.

I have one after the radiators and before the pump. I didn't think that the data would be that relevant because the pump speed, fan speed, and ambient air temperature were the same for all the tests. When I reran the tests with the GPU blocks removed and the pump output going directly to the GPU block the coolant temperature when the CPU was under load for 20+ minutes was between 31.6C and 31.9C for the three tests. Both the coolant temperature and CPU temperature stabilized before 20 minutes of testing.

Also, the CPU speeds were the same for all tests.

This is how we used to do it to rule out of lot of people saying our tests are invalid, or trying to pull things to invalidate our tests.

I come from a time when things were a lot more scientific and not as blingy.
I used to be sponsored by Koolance / eK / Swiftech / DangerDen back during my active days, and helped a lot with MartinsLiquidLab, and SkinneeLabs.
My other alas is Naekuh, and i had earned quite the notoriety with LCS vendors back during my day.

In theory the Optimus block should cool better than the blocks not designed for the Ryzen 3000 series chiplet design, but I'm not getting anything close to a significant difference. Two of the three tests with the exact same loop configuration were a 1 C difference, which is probably within margin of error. Any idea why the Optimus block isn't giving me the results that other people claimed to have? I've seen others claim 5-7C lower in Prime 95 with the Optimus Foundation AM4 block compared to EK Supremacy and Velocity CPU blocks and some other popular CPU blocks. Was their testing inaccurate? I'm beginning to think that the other people testing the difference between blocks didn't use consistent conditions and methods and that's why they saw a larger performance difference.

The other thing I thought of is maybe the big temperature difference is only with the higher voltage and higher power draw from a 4.3 GHz all core overclock on a 3950X and when the CPU is using less than 220 watts the block design doesn't make a huge difference in temperatures. I didn't see anyone claiming that the temperature differences were with an all core overclock.

 

[thesmokingman]

I have one after the radiators and before the pump. I didn't think that the data would be that relevant because the pump speed, fan speed, and ambient air temperature were the same for all the tests. When I reran the tests with the GPU blocks removed and the pump output going directly to the GPU block the coolant temperature when the CPU was under load for 20+ minutes was between 31.6C and 31.9C for the three tests. Both the coolant temperature and CPU temperature stabilized before 20 minutes of testing.

Also, the CPU speeds were the same for all tests.

Water temp is probably the most important sensor when it comes to loop performance besides the actual cpu/gpu temps. My TR 3970x build with dual 2080ti, dual 480mm rads, and a single 35x pump runs cooler water temp than your loop given average ambient of 23c -ish. Water temp on average does not exceed 28c. And for a massive 5 hour plus render job water temp will max out at around 33c. And to put that into context, I readily admit the TR loop is undersized on radiators given the rig's power draw over 1100w when maxed out. Your loop should be performing a bit better just based off of your water temp imo.

 

[aigomorla]

The other thing I thought of is maybe the big temperature difference is only with the higher voltage and higher power draw from a 4.3 GHz all core overclock on a 3950X and when the CPU is using less than 220 watts the block design doesn't make a huge difference in temperatures. I didn't see anyone claiming that the temperature differences were with an all core overclock.

there is one big reason why i do not like AMD for sampling.
Its because of a lack of core temp.
On an Intel processor, we can see the core spread, meaning what the lowest vs highest is when all the cores are loaded.
This allows us to record the cooling head spread, by approximating the swing from the lowest to highest cores.

I wish AMD would have some form of core temp for us to do this as well, but i think implementing a tjunct. on each core is something beyond TSMC is capable of.

Other then that yes, the probe is fine there.
Its also good to record Detla T values from water temp - ambient.
This would give you a standard basis reference for how overall efficient your system is outside the block to narrow where your bottleneck is, if any.
Remember this is all a big game of playing around with Thermodynamics, and there will always be a bottleneck somewhere.

Other then that, best of luck with you on your testing, it seems your probably well off now to publish without getting the right hand wing yelling at you claiming u did something to invalidate it.

 

[Bavor]

Water temp is probably the most important sensor when it comes to loop performance besides the actual cpu/gpu temps. My TR 3970x build with dual 2080ti, dual 480mm rads, and a single 35x pump runs cooler water temp than your loop given average ambient of 23c -ish. Water temp on average does not exceed 28c. And for a massive 5 hour plus render job water temp will max out at around 33c. And to put that into context, I readily admit the TR loop is undersized on radiators given the rig's power draw over 1100w when maxed out. Your loop should be performing a bit better just based off of your water temp imo.

Consumer grade mass produced temperature sensors aren't the most accurate. I've seen two sensors of the same model show a 1C difference between them when in the same location. They are usually consistent, but not accurate. So that makes it difficult to accurately compare differences.

The fans used can have an impact on water temperature. I've seen coolant temperatures vary base on using different fans with different static pressure. Also, its hard to compare performance between loops in two different systems unless you use the same radiators, the same fans, and the same sensors.

What case is your system in? The air flow of the case can have a significant influence on component temperatures and cooling. The tests of the Define R6 showing that with air cooled setups you could drop the CPU temperature 10C if you completely removed the front panel when compared to the stock case setup. The silenced focus design of the case causes higher temperatures in air cooled setups. I'm sure it does the same with radiators. I leave the front door open on my system, but I'm sure its still not the same air flow as other cases with no front door and a less restrictive filter on the intake.

https://www.gamersnexus.net/hwreviews/3177-fractal-define-r6-case-review-mid-tower

When the CPU and GPUs in my system are all running at 100%, its drawing between 900 watts and 986 watts at the wall.