The NRZ N2 Waterblock Beta Thread

[thenrz]

You ahould wait until you reach your established idle temperature. Or when you get your temp sensors, when the water goes back to the starting temp...


I think you need to move past your 'liking' and adhere to what the specifications tell you is safe. 80C for the MOSFETs on that board are fine.

As far as heat soak... welcome to modern big boy(read: many cores and threads) computing! All blocks deal with that...the level varies by board chip and load.

I still think you need to set a static voltage and clockspeed just due to how AMD chips work.

While 80C, 90C, 100C might be within operating parameters, it doesn't mean I shouldn't attempt to mitigate the high temperatures. I'm not sure what the upper limit that will be reached running a long 200W test. Stressing the components when there are a lot of options to offset the issue doesn't make sense to me. I've seen MOSFETs and PCBs melt thermal adhesive and catch on fire on h-bridge motor controllers, and they worked right up to the point where some physical transformation rendered them fatally damaged. This is logic and power components married together, and I can't just let it go and send it at this point. I am going to try a fan first, and I think the airflow across the heatsink and the air movement up and under the heatsink that removes the stale air will get me a solid 5-10C off of that top number. The exhaust fans above will do the rest. At that point I can just send it and perform my tests to my planned specifications.

I'm working on messing with OC options to dial in something specific. Right now the 200W benchmark is something that I can comfortably say is repeatable within the specified timescales. I can work into something even more solid as I continue to deepen my understanding of Ryzen... which is a pretty incredible thing. I'm still a novice in all aspects.


Edit: Have to work some early overtime tomorrow, so it's time to shut this thing down or I'll be up all night again... Thanks for all the input!

 

[thenrz]

Ran to Microcenter after work, donning a mask and gloves, to try and grab some more 1/2"ID primoflex tubing since it's on sale for just a buck or two above Amazon et al. Well, they were out of stock... I browsed around the liquid cooling section because who doesn't need to be able to see all the cool stuff in person? In the one small aisle that was heatsinks + fans on the other side, they had kind of the overflow section for liquid cooling stuff. There were some refurbished radiators, hard line tubing, and a clearance "pile" of fittings. Being one who can't pass on the prospect of a good deal, I began to sift through it. I found all kinds of great stuff and some of the prices were pretty incredible. The best part? 90% of it was Bitspower. They also started carrying EK fittings as well in the last week, and it was surprising to see that they were all a good 30% less expensive compared to Bitspower.

Clearance purchases:
Bitspower 20mm extender: $1.96
Bitspower Male to Male with 6 random o-rings: $1.96
Swiftech 4-way adapter block: $3.96
Bitspower 45deg rotary: $2.96
Bitspower hard 90 M->F: $1.96
Bitspower 8mm Extender: $0.96

Was pretty surprised to be able to spend under $15 for all 6 items. Picked up a couple of 1/2"ID barbs (in addition to the 6-pack coming on Monday via Amazon), a double rotary 45, an inline ball valve, and 2 plugs. Spent less than $45 for a satisfying bag of great quality fittings. This hobby isn't cheap with things like this, and it felt great to get a deal. Amazon seems to be returning back to near normalcy as things I ordered on Thursday started showing up today: 4x rotary 90s, usb 2.0 header splitter, aquaero 5, and the flow meter cord. It's too bad that the inline temp sensors and the flow meter don't come until Tuesday and Wednesday considering that's the point of buying pretty much everything else... The aquaero came with 4 temperature probes, so that will be fantastic to measure temps around the radiator. Might try to slip one between the acrylic frame and the coldplate on the block to see how things differ between the surface and the liquid. Would be interesting an interesting data point.

Hey, so I was browsing some builds, and I noticed something that I might have been missing here. Is it better to plumb the return line to the reservoir on the bottom rather than the top? I am going to be making some changes to the loop layout to put in a fill port setup to the top of the case with a ball valve, and I am wondering if I need to flip the reservoir over and plumb to that orientation. There's a last little bit of air that I can't seem to purge, and I am wondering if this is due to the small and unavoidable air gap that I have at the top of the res is being constantly mixed into the liquid and pumped through. I use a fill syringe to get the res as full as possible, but I have the res so close to the top to keep it at the highest point that I cannot access the plug on the cap end. I will be drilling out some holes for it about an inch and a half down from it's current position, and I want to be sure that my own assumptions aren't getting in the way of the ideal layout.

Thanks to all again, and have a very happy Easter! I hope everyone gets a chance to spend some time with loved ones, or if you can't due to the pandemic, at least relax and take a day to do something fun and enjoyable.



I will never not get excited to have so many possibilities in front of me...

View attachment 209545

 

[trents]

Any air in the system will rise to the highest point so if possible, I would make the res the highest component in the loop. Once you get all the air out and the res completely filled (all the air displaced by liquid) you can relocate the reservoir to a more convenient place. Of course, this may be difficult to do in system with hard tubes. You could also do this by putting the res at the lowest point in the loop and turning the computer upside down to temporarily make the res the highest component in the loop until you got in filled and sealed off.

Another option might be to plumb a fill tube extension into the res that is long enough to have it's open end higher than any other component in the loop, even sticking out the top of the case, and cap it off after all the air in the loop has been evacuated and displaced by liquid. I always wondered how companies fill their AIO loops completely with no res. Maybe they don't and there is always some air in an AIO. Excuse my rambling tonight. Been roto tilling the garden all afternoon and carrying bags of compost up a hill and kind of beat.

Once you get all, or essentially all, of the air out of a loop then it makes no difference what is the order or placement of any of the components since there is no risk of the pump running dry. Neither does the sequence of components in a loop have any bearing on performance.

 

[thenrz]

Any air in the system will rise to the highest point so if possible, I would make the res the highest component in the loop. Once you get all the air out and the res completely filled (all the air displaced by liquid) you can relocate the reservoir to a more convenient place. Of course, this may be difficult to do in system with hard tubes. You could also do this by putting the res at the lowest point in the loop and turning the computer upside down to temporarily make the res the highest component in the loop until you got in filled and sealed off.

Another option might be to plumb a fill tube extension into the res that is long enough to have it's open end higher than any other component in the loop, even sticking out the top of the case, and cap it off after all the air in the loop has been evacuated and displaced by liquid. I always wondered how companies fill their AIO loops completely with no res. Maybe they don't and there is always some air in an AIO. Excuse my rambling tonight. Been roto tilling the garden all afternoon and carrying bags of compost up a hill and kind of beat.

Once you get all, or essentially all, of the air out of a loop then it makes no difference what is the order or placement of any of the components since there is no risk of the pump running dry. Neither does the sequence of components in a loop have any bearing on performance.

The res is the highest component in the loop just for this reason. I think the fact that I can't fill it completely is my problem. I need to flip the end tubes around and use the end cap with the single G1/4" on the end face on the top and the other end cap with 3 G1/4" ports on the bottom. I am going to install a fill port once the rest of my parts get here early this week so I only have to disassemble the entire thing once.


Here's what I mean:

View attachment 209546


I have the res essentially installed upside down because I didn't leave enough space above it to properly fill the system and just used the side oriented ports. The res inlet from the loop is on top where it needs to be on the bottom. It all makes sense now... I am going to lower the res roughly an inch and a half, install my fill port with ball valve which will have a fill port drilled on top of the case, and then re-plumb my reservoir inlet from the loop on the side port on the bottom.

I want to completely review my cable management and all of that as well. This started out as something simple, and I ended up jamming a whole lot of stuff in the case. There's three fan/RGB controllers on the back as well as everything else. Since I am adding all the inline temp sensors, the aquaero, the flow meter, the regular air temp probes, and the internal USB 2 header splitter, I am going to have even more to deal with. At least this time I will have a much better understanding of what exactly I am working with so I can make some better decisions rather than bundling and zip tying to other bundled and zip tied cable runs... I think the adhesive backed zip tie blocks are going to come in really handy this build so I can keep things running parallel rather than rely on the existing punch tabs for zip tie locations. With the additional 10 or 11 wires, the cable routing itself is going to take probably more than several hours!

 

[trents]

Yes, that hose being at the top is your problem. That's were the air bubbles will collect.

 

[thenrz]

Yes, that hose being at the top is your problem. That's were the air bubbles will collect.

That's exactly what's happening. Rather than do my research or ask, I just assuming my thinking was correct which it is not. I'm hesitant to rip everything apart until the flow meter and temperature probes come as I'm going to have to take it all apart again at that point anyway. But... looking at the incorrect plumbing is bugging me now. I guess this is the whole point of having a drain and fill port is the option to quickly remove the liquid, make changes, and quickly fill and purge again. Eh... There isn't a whole lot going on with family today even though it's Easter considering the pandemic... so... I guess I will do what I can later this afternoon. Once I get the aquaero software going, I want to see what the sensor reading part of the software looks like and if it could be resized to fit say a 5in LCD panel that I could mount in the case via HDMI. Would be really cool to have all of the sensor data on a small screen mounted in the case. They aren't pricey for quality ones on Amazon, and I think the look and the convenient access to the data would be worth the $40 or $50.

Thanks for the troubleshooting trents! Really appreciate being pushed in the right direction. Have a happy Easter!


Edit: Went and grabbed a couple packs of small zip ties and some organizational stuff. Going to go ahead and get everything torn apart today and just try to plan for the additional wires from the aquaero and sensors. Does anyone have some input on the best place for the flow meter? I am thinking either after the pump or right before the return to the reservoir. I know the aqua computer flow meter isn't restrictive compared to a lot of the other off the shelf meters, so I am hoping that I don't have to end up increasing the pump power that much.

 

[trents]

Happy Easter to you too! He is risen!

 

[thenrz]

So, I got the loop replumbed, the fill port installed, the aquaero installed, and some things generally cleaned up. For whatever reason, my idle is running 4-5C warmer. My only idea for the reason is the addition of a double 45 between the reservoir and the pump. I think I have introduced a restriction... There's no air at all in the system now, and it's incredibly quiet. The increase in my idle is really bugging me, as 28C I felt was fantastic. I am feeling a lot less thrilled with 33C. Would this be logical to think that a smaller line in this important part of the loop would be responsible for the rather large increase? I guess either that or the aquasuite software. Seems prudent to uninstall and see what happens.

 

[EarthDog]

Would this be logical to think that a smaller line in this important part of the loop would be responsible for the rather large increase?

Nope.

You mention the aquero software, did you take the reading from that instead of coretemp/hwmonitor?

 

[thenrz]

Nope.

You mention the aquero software, did you take the reading from that instead of coretemp/hwmonitor?

No, I have been using hwinfo exclusively to keep things homogeneous. I ran a 5 minute 200W test and was averaging 2.3C higher. The whole thing isn't making much sense to me. I think tomorrow I will return some things back to the way they were, test, and then uninstall the aquasuite software and see where I'm at. It would be really disappointing as the whole point was to be able to record as much data as possible, but if it works against the integrity of the system to that degree, then it would be illogical to keep it. Good thing Amazon's return process is painless. The flow meter and inline temp probes are still on the way, and those wouldn't do much good without a way to log the data. I think I could use an Arduino to log the inline probes and possible figure out the flow meter's signaling.


Here's the modified plumbing and the potential restriction:
View attachment 209562

 

[EarthDog]

I'm cant imagine flow rates changed that much to effect temps as it did. Some other variable(s) is(are) different. Be it higher ambient, or something. It's possible the software may be using some cpu cycles so idle is a bit higher. But check it out and see. Look at ie cpu use with...then uninstall and without.

The data now isn't terribly relevant anyway. If when you get everything installed that is the new normal, it is what it is. At least then you have data like flow rates and temps on top of it.

 

[thenrz]

I'm cant imagine flow rates changed that much to effect temps as it did. Some other variable(s) is(are) different. Be it higher ambient, or something. It's possible the software may be using some cpu cycles so idle is a bit higher. But check it out and see. Look at ie cpu use with...then uninstall and without.

The data now isn't terribly relevant anyway. If when you get everything installed that is the new normal, it is what it is. At least then you have data like flow rates and temps on top of it.

I will change one thing, test, and repeat tomorrow. I guess since I have four in-air temp probes now via the aquaero, I can keep track of ambient in and out of the case while I'm testing and have my data all in one place rather than have to try and keep looking at my multimeter. I also have a probe on both sides of the radiator. During the 200W test from the photo below, there was a 4.8C difference from baseline to the final moments before I stopped the test. That was interesting data

Losing the idle temps in the 20s I think is probably more psychological than anything, but it still is a large enough deviation to be of a concern. I would hate to see a device like the aquaero be an issue moving forward, as once I got the issue with my install figured out (the internal USB header splitter was acting up), it has been really easy to configure. I updated the device firmware and exposed even more options which was really cool. Once I have the data from the liquid and some idea of flow rates, I will be set for testing whatever block I could want in this loop. Tomorrow's round of problem hunting will answer the question... Either way, having a fill port and drain port with super convenient ball valves is certainly one of my better ideas I've ever implemented in a custom loop. It was really nice not to have to reach into the system to fill or drain it whatsoever.



~13 minute test:

View attachment 209563

Just about 1.8C warmer. I guess not a huge deal, but it does render moot all of my previous data. I think until the remainder of the measurement devices arrive and are installed, performing tests isn't going to generate anything really useful. Setting up logging within aquasuite only made the CPU usage worse which is not ideal at all. It's almost like I need a second system to measure the test system. A small ITX secondary seems really enticing right about now... but that's just money I don't feel like spending right now. My fiancee, as amazing as she is about all of this, would probably begin to dislike the copious cash I have been dropping on the little things that have added up quite a bit.

 

[trents]

I'm cant imagine flow rates changed that much to effect temps as it did. Some other variable(s) is(are) different. Be it higher ambient, or something. It's possible the software may be using some cpu cycles so idle is a bit higher. But check it out and see. Look at ie cpu use with...then uninstall and without.

The data now isn't terribly relevant anyway. If when you get everything installed that is the new normal, it is what it is. At least then you have data like flow rates and temps on top of it.

I agree with that. Don't lose track of the fact that how your custom block performs in relation to other blocks on the market is the goal here. Don't let yourself be side tracked.

But speaking of he higher idle temps, in making changes to the loop did you happen to remove, repaste and reseat the block? Reseating a cooler will often give different results because of the differences in how the TIM spreads from one to another. If you did remove and reseat the block, that is the first place I would look to account for the changes in idle temp.

And what about ambient temps? Did they change?

But who cares about idle temps anyway as long as they are obviously out of whack? Have the load temps changed by that same amount?

 

[thenrz]

I agree with that. Don't lose track of the fact that how your custom block performs in relation to other blocks on the market is the goal here. Don't let yourself be side tracked.

But speaking of he higher idle temps, in making changes to the loop did you happen to remove, repaste and reseat the block? Reseating a cooler will often give different results because of the differences in how the TIM spreads from one to another. If you did remove and reseat the block, that is the first place I would look to account for the changes in idle temp.

And what about ambient temps? Did they change?

But who cares about idle temps anyway as long as they are obviously out of whack? Have the load temps changed by that same amount?

The block was not reseated. The ambient was about 1C warmer, but not enough to account for the +5C idle change or the +2C 200W load change. I think by adding the fittings, I have reduced the overall efficiency of the system. Back in the day, we used to avoid 90s and even 45s to a degree like the plague just for that reason. Smooth bends was the name of the game back when...

 

[trents]

But again, isn't it pretty irrelevant to the overall goal?

Have you thought of using a hot plate?: https://linuxreviews.org/Thermal_design_power

 

[Blaylock]

Are you running multiple programs at the same time for measuring temps? I seem to remember someone stating that running more than one will effect the results.

 

[thenrz]

But again, isn't it pretty irrelevant to the overall goal?

Have you thought of using a hot plate?: https://linuxreviews.org/Thermal_design_power

That was my initial plan was to use power resistors to setup a constant load. This was changed once I realized that I could target specific wattage points in a real world situation. My data is significantly more relevant being in situ than based purely a heat parameter. Mounting, CPU geometries, heat soak (thanks earthdog)... All of these things plus more are giving me numbers that I can desicively say can be replicated in a retail product.

And Blaylock, I am only running hwinfo64 for temps, and I have disabled hardware monitoring in aquasuite in an attempt to partially mitigate the problem. Turning off the aquasuite service and closing the software gets be back to an idle of 30-31C. I am assuming that removing the restrictive fitting would get me back to 28-29C. I have contacted Aqua Computer and asked if there is a way to reduce the amount of resources drawn by their software, and I hope to get an answer from them sooner rather than later.

I need to find a way to pick up some competing AM4 blocks without spending a lot. Used would be best. Also, once the stuff for a very well known company is done cutting this week, I will have machine time. Finally. I could get all the data in the world, but I am only testing one setup. Even though I am giving my best effort to collect data points objectively, there are still too many variables that are intrinsically linked to my specific set of hardware. Really looking forward to getting some more eyes and hands on a beta N2.

 

[EarthDog]

The block was not reseated. The ambient was about 1C warmer, but not enough to account for the +5C idle change or the +2C 200W load change. I think by adding the fittings, I have reduced the overall efficiency of the system. Back in the day, we used to avoid 90s and even 45s to a degree like the plague just for that reason. Smooth bends was the name of the game back when...

2.3C on load and ambient changed by 1C....that leaves ~1C....or within margin of error.

Again, not terribly concerned with idle temps...you'll want to move past that snooner than later IMO.

I also highly doubt a 45 elbow or two caused any significant difference. I know what you are saying as more bends can reduce flow, but 'avoiding like the plague' was never a part of it. With such a small difference at load, it's impossible to track considering the resolution of the sensors (typically ~1C). Again, it is what it is when you start testing. Just make sure you have 1-1.5 GPM.

Keep your eye in the prize...feels like you are losing site of the goal and lost in the minutia (idle temps, heat soak...). Get all of your parts... set up the loop and use that data. It doesn't matter one bit, idle cpu temps. Just work with load temps and deltas over ambient...

As I said earlier... it is what it is when you add the software. You're barking up the wrong tree asking them about it. First, I doubt there is anything they can do...its likely just using cycles delievering your data. Second, the data you are hung on is largely irrelevant. Think about it. Your wattage was the same, right? So how would software change load temps? Also, your testing feels all over.. 13 mins here.. 5 there... an hour...etc. you've got to figure out what is going to be what. I know you are just putzing around now... youve said that, but yet are still hung up on this data. Get it all together and start from there.

 

[thenrz]

2.3C on load and ambient changed by 1C....that leaves ~1C....or within margin of error.

Again, not terribly concerned with idle temps...you'll want to move past that snooner than later IMO.

I also highly doubt a 45 elbow or two caused any significant difference. I know what you are saying as more bends can reduce flow, but 'avoiding like the plague' was never a part of it. With such a small difference at load, it's impossible to track considering the resolution of the sensors (typically ~1C). Again, it is what it is when you start testing. Just make sure you have 1-1.5 GPM.

Keep your eye in the prize...feels like you are losing site of the goal and lost in the minutia (idle temps, heat soak...). Get all of your parts... set up the loop and use that data. It doesn't matter one bit, idle cpu temps. Just work with load temps and deltas over ambient...

As I said earlier... it is what it is when you add the software. You're barking up the wrong tree asking them about it. First, I doubt there is anything they can do...its likely just using cycles delievering your data. Second, the data you are hung on is largely irrelevant. Think about it. Your wattage was the same, right? So how would software change load temps? Also, your testing feels all over.. 13 mins here.. 5 there... an hour...etc. you've got to figure out what is going to be what. I know you are just putzing around now... youve said that, but yet are still hung up on this data. Get it all together and start from there.

You are quite correct in just about everything. Since I am waiting for the remainder of the testing hardware, getting things as homogeneous as possible now saves time later. I get hung up on things, as I am really detail oriented. This fact is what makes me good at my job and at engineering in general. Not so much when it comes to the inane hyperfocus on the irrelevant. While idle temps aren't nearly as important, it also implies that the system is now operating at a warmer baseline, and this does affect final results. The difference essentially goes away when figuring deltas, but having constructed the block to function at a high ability to transfer heat away from the source, the new norm here does in fact augment my final result. The extra CPU cycles will exist throughout every moment of testing and will affect the numbers.

As for testing duration being irregular, you are correct that I am not collecting final data and am for sure "putzing" around. Keeping the system in a state where it can churn out 200W+ tests for a long duration whenever I ask of it is a must. Having to worry about VRM temps and other extraneous factors then would greatly reduce the ability I have to pull relevant data. My eventual testing lengths will be 30m, 1h, and 2h at 200W, and up to 8h at 145W (which is the stock maximum fused power limit). Good thing I pay a fixed rate monthly with my city's public utility company I will attempt to increase this limit with a defined OC at some point, but that will definitely require cooling the VRM. I am hoping to have some time this coming weekend to take the MB apart and get some measurements for a waterblock. It will be great experience for manufacturing board specific VRM blocks in the future.

Ans as for contacting the manufacturer, it doesn't hurt to ask. Perhaps they have run into this issue in the past and have a fix or some internal build that is more in line with my needs. All I need is temp sensor data, flow data, and a single RPM (pump) input. I am not even worried about the fans right now considering they are finely controlled elsewhere. If not, all I've lost is 3 minutes when sending an email. If so then I have been able to negate a part of my problem.

 

[EarthDog]

While idle temps aren't nearly as important, it also implies that the system is now operating at a warmer baseline, and this does affect final results.

Respectfuly, so what . It is idle and largely not relevant. CPU temperature sensors are known to be less accurate here and more accurate at load in the first place. Also, a 200W from your stress test app is a 200W load. Idle doesn't come into play there. I'd use the delta and eliminate idle, personally. Seriously... it's nothing to get hung up about. Just use a delta... problem solved... and is a more accurate way of displaying that information in the first place (everyone's temps will be different, but deltas will not change!!!!!).

the new norm here does in fact augment my final result

Does it? I thought I explained the 1C ambient increase which leaves 1C difference... margin of error of your parts....but this isn't the real test in the first place. In reality, you don't have a first result, none the less anything final. You don't have a baseline testing setup, all of the hardware, etc. So, again, who cares about these values currently? You want the values with your system in its final state in the first place. I understand hammering out the process, but you are holding on to the current results as if they matter.

Good thing I pay a fixed rate monthly with my city's public utility company

You should do the math and see how little difference it makes $$ wise. At 10c /KW hr... the difference is change you'll find in the couch over a month.

but that will definitely require cooling the VRM.

Very few will cool the VRMs... why not 'run as you brung'? I understand the need to isolate as best you can, but very few will be doing so and heat sink is a factor... why not test the block in its most likely environment? You'll never eliminate the effect of heat soak (even with block on them)... I wouldn't bother as every chassis and situation will be different. If you really want to isolate the block with nothing else, use a heat plate.

Ans as for contacting the manufacturer, it doesn't hurt to ask...If not, all I've lost is 3 minutes when sending an email. If so then I have been able to negate a part of my problem.

Again.. that isn't a problem. Idle isn't terribly relevant! A 200W load is a 200W load regardless of the monitoring software running or not. As you can see from your results, you had a 200W load again...


So, my best advice is to stop thinking about these values you gathered before your test rig is completely built. I get refining the process, but your head feels like it is stuck in the minutia and missing the bigger picture of it all. A warmer idle doesn't change the delta.

EDIT: Some info about power use on your 3900x... - https://www.anandtech.com/show/14605/the-and-ryzen-3700x-3900x-review-raising-the-bar/19