Graphics Welder

[Mysterae_]

After getting everything back again, refilling with water and testing it, as I mentioned in my previous post one of the memory modules was not detected again.

However, stripping the rig down to find the fault was a waste of time. It turns out that one of the RAM sticks has failed, I don't know how I missed that when originally faulting the system or maybe the symptoms were different. I should have tested each one individually and in each slot, then I would have found it. Perhaps I jumped to the conclusion that it was the CPU waterblock and or the contact frame mounting pressure that was to blame because I had heard that was a known problem and the RAM was working previously before.

I used Corsair iCUE to turn the only RAM module that it could detect to green and tried it in both DIMM slots and it worked fine.

The module that's RGB could not be changed was tried in both slots and the motherboard just hung with it's RAM fault light being lit.

Getting to this stage was a bit of a joyride with this MSI Z790i EDGE Wifi motherboard with it's 'quirks and features'. You see, once you clear the CMOS (with the power lead removed) the motherboard won't boot, the PSU just clicks continuously with the motherboard doing nowt. I could be my particular hardware (Thermaltake Toughpower SFX 1000W and SteelSeries APEX Pro keyboard, I'm looking at you). The PSU has two types of clicks, a hard click and a soft click is how I can best describe it. To get it to boot, I remove the power lead, turn on the computer to drain it of power, listen for the hard click and then upon the soft click put the power lead back in - then the motherboard boots. What a palaver! I'm sure it's one particular setting that messes everything up and makes it almost impossible to get a clean boot to the BIOS.

That setting is in Settings\Advanced\Power Management Setup\USB Standby Power at S4/S5. The BIOS sets this as Disabled by default and in my case it must be set to Enabled for the motherboard to boot properly.

With that out of the way, I tried everything to get the motherboard to recognise the suspect memory module:

- updated the motherboard's bios to the latest version
- checked the contacts of the RAM and the memory slots
- set it to non-XMP settings
- disabled and enable any RAM training options in the BIOS
- tried both DIMM slots, with and without it's sibling

Nada.

It's dead Jim.

I don't have any spare DDR5 RAM modules so depending on how the RMA of these goes the rig may be out of action for a bit as the working memory stick will have to be returned with the faulty one.

 

[Mysterae_]

Amazing display on the front of the case. I didnt even know one could do that!

Thanks. I don't know if you've seen it already but you have two options when it comes to putting a screen on the front:

- The OEM route as Phanteks now do a replacement add-on front module display and is available here at Overclocks:

https://www.overclockers.co.uk/phanteks-evolv-shift-xt-hi-res-display-black-acc-phk-01392.html

The front module display from Phanteks didn't exist when I started this project so I went with the DIY option:

- Get a separate screen and with a bit of modification to the screen to reposition the HDMI and USB connectors and taking the metal discs from the case's top front panel to attach them to the screen which provides the mounting method to the case. It's all been shown previously in this tread . Route the HDMI and USB cables out the back:

 

[Mysterae_]

It's been a while since the last update, been using the system putting it through it's paces.

Once the faulty DDR5 memory module was replaced the next issue I had was thermal paste "pump-out" which saw GPU temperatures way too high for a watercooled system.

There's a thread in the Watercooling subforum if you are interested. It required a full disassembly and rebuild which isn't that bad.

Here's what I suspect thermal paste "pump-out" looks like on a GPU die:

The recommendation is to use Honeywell PTM7950 thermal interface material instead of MX-6, so if temperatures increase again I'll be sure to use that instead. I didn't know this at time of reassembly so used MX-6 again, doh!

I'm working on getting some final vanity shots of the completed system but in the meantime here's a short clip of the loop filling, gurgling the air out of the loop. I think this was at 70% pump speed.

 

[Mysterae_]

Lets hope this is the last time you have to do this, what a bleeding nightmare it's been for you. One thing after another!

It hasn't been all that bad, or has time faded (as well as broken) my memory .

With the exception of the bad memory module I've not had any bad components, the thermal paste pump-out was unforeseen and unfortunate. Time and distraction were my biggest problems!

 

[Mysterae_]

Anymore issues with thermals? Been looking at using Thermal pad sheets in place of paste, results vary but seems a good way to do it without the mess and pump out.

None at the moment. It did take a month or so to occur so it is due to happen soon .

I was even pushing the 4090FE beyond spec by putting +200 on the core and +400 on the memory (probably go higher) with a voltage of 1.070V and doing some benchmarking.

Considering there's more than 500W going through the 4090FE and it's running over 3GHz I think 53C is pretty good. If I increased the water flow rate and the fans of the system I'm sure I could get it lower but it's a good balance regarding sound.

 

[Mysterae_]

I finally read through (most of) the comments (sorry, I'm also guilty of just looking at the pictures) and I have to say that I really like the little individual things you've done, such as the power switch for the PC. Ingenious.

I'm not sure that I really click with the separate cooling solution, tending to think that small PC's should be just that, but I cannot argue with your vision and talent. Enviable and inspiring.

Thanks, I know fine well that people mainly just look at the pictures and will read the text if it interests them further, I'm guilty too. Pictures and 100 words and all that!

I totally understand what you say about the separate cooling; a small PC should just be that and I've seen some incredible builds where they achieve that. This PC was all contained at first, mainly to test the components but also to see if it was actually worth the effort of external cooling. The system worked but by god it was hot and noisy! Now with most of the heat being redirected to the radiator which sits on a window sill (to extract the heat out the room rather than bring in cool air), the PC case itself is so quiet. It's still cheating though !