First water cooled pc in two decades

[wNotyarD]

I'll try loosening it up and go back with less pressure to see how that effects it.

Any result?

 

[67Elco]

Any result?

Just now finished...had to drain the loop and remove the tubing. Looked like it may have not been seated properly...the tolerance between the studs and holes is very close and I feel the block may have been binding while tightening and giving a false sense it was secure. I'll refill the loop shortly and give it a whirl.

 

[Vya Domus]

Temperatures look fine, I was hitting about 90C on the same CPU with 2x360 rads and it wasn't until I added another 360 and a 240 that they went down to about 85C.

 

[67Elco]

Temperatures look fine, I was hitting about 90C on the same CPU with 2x360 rads and it wasn't until I added another 360 and a 240 that they went down to about 85C.

It wasn't the max temps I was concerned about, but rather as ThrashZone pointed out the 20c difference between cores under load that was puzzling. I had never experienced that wide a spread before. It may just be a quirk with this particular cpu, but I tend to doubt it.

 

[Vya Domus]

It wasn't the max temps I was concerned about, but rather as ThrashZone pointed out the 20c difference between cores under load that was puzzling. I had never experienced that wide a spread before. It may just be a quirk with this particular cpu, but I tend to doubt it.

I have the same delta between the coolest and hottest cores, one CCD always runs hotter and there is always one core which runs hotter than the others in that CCD. It's not a quick, all 7000 series CPUs behave like this.

 

[67Elco]

I have the same delta between the coolest and hottest cores, one CCD always runs hotter and there is always one core which runs hotter than the others in that CCD. It's not a quick, all 7000 series CPUs behave like this.

Can you run a quick stress test with cpu-z so I can see your's? Not seeing much change here with cores #2, #3 and #5 being the biggest offenders.

 

[BoggledBeagle]

What is up with the offending air cooled graphic card there?

 

[67Elco]

Don't think I could do any better mounting this waterblock. Near impossible to do anything with it with the hardlines routed as they are. I just checked with a digital depth caliper and the block is about as parallel as you can get one. There are slots in the bracket which allowed me to measure the distance of the block corners from the face of the socket retention frame. I've never had a bad mount in my life, but not saying it could not happen. Three times with the same result suggests otherwise and something else is at play here. Domas says his/her cpu acts the same so there must be something to it.

 

[Bobaganoosh]

It's certainly possible. I think when you put as much work into something like this as you have, there's obviously a tendency to chase down every bit of performance possible, but at a certain point, you have to just use the machine and enjoy it lol.

The main thing that seemed odd (and because it's a common occurrence we probably all hoped it was simply a troubled mount) was that the performance was not as good as expected when compared to your D15, and possibly compared to a previous motherboard. The other thing you could do if you believe you've really debugged any issues with the loop is to work on a curve offset under-volt. I haven't done that with 7900, but it seemed the best way on the previous gen to get the most out of PBO and Ryzen CPUs. I'm guessing your other motherboard just did a better job with those settings by default.

 

[67Elco]

It's certainly possible. I think when you put as much work into something like this as you have, there's obviously a tendency to chase down every bit of performance possible, but at a certain point, you have to just use the machine and enjoy it lol.

The main thing that seemed odd (and because it's a common occurrence we probably all hoped it was simply a troubled mount) was that the performance was not as good as expected when compared to your D15, and possibly compared to a previous motherboard. The other thing you could do if you believe you've really debugged any issues with the loop is to work on a curve offset under-volt. I haven't done that with 7900, but it seemed the best way on the previous gen to get the most out of PBO and Ryzen CPUs. I'm guessing your other motherboard just did a better job with those settings by default.

I'm satisfied enough with the build overall, but that thermal issue bugs me not knowing if it's already baked in to the mix so to speak. Nothing I have done has affected any change in it's behavior.

I feel fairly certain the slightly degraded performance is entirely related to the motherboard. The ASRock mb's and bios just simply work better. I already have the PBO settings and other tweaks applied just as I did for the ASRock mb's. Kicking myself in the ass for not recording with HWinfo while the cpu was in two other different mb's...I could answer my own question then. Perhaps others with a 7900/7900X will chime in with their core temp differences under load... if any. We already have one who reports their 7900 acts just the same as this one.

 

[ThrashZone]

Hi,
Yep
Just comparing the mating surfaces can tell you a lot

I've never used a amd chip all I've noticed is they are a lot larger than intil even hedt chips "2066 socket" so I've never seen if they were flat corner to corner or concave like a cup like intel's are.
Then you get to the cold plate on the water block if it's bowed down enough to mate properly to the chip.

Pity 20c+- difference is normal between core 0 and 11 I'd find that hard to live with hehe

Love hard piping but this is a main reason I use soft tubing
Easier to workout the bugs find out just how good products are which is 2nd/ 3rd guessing items it's part of the fun we call water cooling

 

[Vya Domus]

Can you run a quick stress test with cpu-z so I can see your's? Not seeing much change here with cores #2, #3 and #5 being the biggest offenders.

Core 5 (CCD1) is the hottest, exactly 20C between it and the coolest one Core 7 (CCD2), just as I explained. No amount of tinkering with it will change anything, it's just how these CPU behave. You can try lapping if you really want better temps but as long as it's not hitting tjmax there really isn't much to be gained.

 

[67Elco]

View attachment 333903

Core 5 (CCD1) is the hottest, exactly 20C between it and the coolest one Core 7 (CCD2), just as I explained. No amount of tinkering with it will change anything, it's just how these CPU behave. You can try lapping if you really want better temps but as long as it's not hitting tjmax there really isn't much to be gained.

Thanks for that confirmation! My hottest core is #5 as well. I won't be lapping or anything else for that matter...it's a done deal as far as I'm concerned.

 

[67Elco]

With that out of the way I'll mention what I have learned so far using the copper tubing.

The adhesive "Just for Copper" works extremely well. In less than 30 seconds you can't loosen a joint. I've seen videos after 5 minutes no mechanical means could break the bond...the copper tubing simply collapses. Guess I will be the guinea pig for it's longevity and resistance to anti-freeze coolant.

During all of this draining the loop and checking the mount it gave me the opportunity to rectify some minor mistakes with my plumbing and I finally have it perfectly straight and level. Using the copper glue I found myself dry fitting the tubing/connectors and using a sharpie on the pieces to indicate specific depth and alignment...which works "fairly" well, but not exactly perfect every time. It is extremely easy to end up with a ever so slight misalignment once bonded and I found myself having to fudge entrance into the 16mm terminating fittings. This made me a little uncomfortable because there are only a couple of o-rings that are creating a seal...tension to one side is not ideal.

What I needed was a adhesive with a slightly longer working time so I could fit the already configured and glued tube into one connector while the other end was yet to be bonded. This way I could apply glue and adjust the length of the assembly and also the approach angle entering the other fitting. (Hope that made sense lol) What I used for this was J-B Weld ClearWeld https://www.jbweld.com/product/clearweld-syringe This worked perfect for me and I was able to go ahead and fit the piece in place and leave for an hour to completely harden. I have no idea what any interaction with anti-freeze might be with this product, but in my past experiences with J-B Weld nothing fazes it. There really is no pressure to speak of in these systems, so it would take a total break down of any joint adhesive for there to be any leakage.

I did not go the soldering route simply because I knew what a labor/time intensive project that would be. The only fool proof method I could see soldering would be dry fitting all in place and then soldering in place...inside the case. No thank you. Too little to be gained and way too much effort. (He says just before every joint in his loop begins dripping... )

 

[ThrashZone]

Hi,
Not so sure that stuff you used was very toxic but that is why I'd cut it with distilled water even though it's 50-50.
15% is still fine you just get a lighter color.
Water cools better than antifreeze anyway.

 

[67Elco]

Hi,
Not so sure that stuff you used was very toxic but that is why I'd cut it with distilled water even though it's 50-50.
15% is still fine you just get a lighter color.
Water cools better than antifreeze anyway.

You do the tests and tell us what the "c" difference is between 15% and 50%... I'm all tested out.

 

[ThrashZone]

You do the tests and tell us what the "c" difference is between 15% and 50%... I'm all tested out.

What you getting rid of your rig
Everyday is a test of will

 

[dgianstefani]

Thanks for that confirmation! My hottest core is #5 as well. I won't be lapping or anything else for that matter...it's a done deal as far as I'm concerned.

As I've mentioned, the issue is the AM5 IHS. If you delid and use a custom IHS like I do, or run bare die, your temps and core deltas are much better.

 

[67Elco]

I'm not going there lol...

 

[dgianstefani]

I'm not going there lol...

Sure, but don't expect to see results other than what you have. With stock IHS, the WC advantage is noise and heat directed out of case, not necessarily temps.

 

[67Elco]

I'm well aware of the limitations and have no elevated expectations.

 

[67Elco]

I may have to RMA this motherboard. I swapped memory between two pc's today and now this one will not boot with Expo enabled...was working just fine before. I had four sticks of PC6000 (4x16) and replaced them with two sticks of PC6400 (2x32). The PC6400 set was working normally in the other identical mb. Tried swapping them back and had a third set of PC6000 (2x16) I had on hand, but it still refuses to boot with Expo. Any ideas before I take this apart? The other mb still boots with Expo regardless of memory used. I'm going to go back with 2x16 in this pc regardless of the outcome. Both mb's have the same bios version as well.

 

[67Elco]

Fresh OS was not useful...about to re-flash the bios for effect. After that I'm out of ideas.

 

[ThrashZone]

Hi,
Yeah bios flash can clear cmos pretty well
Hope it works out.
Shit can happen swapping memory gremlins wake up hehe

 

[BoggledBeagle]

.... now this one will not boot with Expo enabled.

Do you know about memory training, which can cause huge delays when booting?