Once in properly and expediently please!
What I immediately replaced were the abysmally hard pads, which put way too much pressure on the memory and are not particularly valuable conductors thermally. I replaced the fun with 0.5mm pads from Fujipoly, you could have used 0.75mm for safety. I replaced the completely cured paste, which had zero adhesion between GPU and heatsink, with a matching blob of Alphacool Apex. Fits. Now let’s move on to the first modification. To illustrate, I took the modified card apart again to show you the imprints and contact areas as well:
I applied paste between the cooled vapor chamber and the frame above the memory modules. Thus, the chamber not only lies loosely, but is also thermally connected to the frame. We’ll see in a moment that this all makes sense.
Let’s quickly recap what I did to the card: Better thermal paste, performant and slightly softer or thinner pads for the memory (and thus more contact pressure on the GPU) as well as the thermal connection of the vapor chamber to the frame. I also added two more pads for better support and stabilization. All this was done in less than 15 minutes and it took longer to write it down for you afterwards.
Measurement of the rebuilt card and a conclusion
Why AMD lets the quality control at the contract manufacturer slip so much is a complete mystery to me. The trouble with the Vapor Chamber of the RTX 7900XTX should have been a final warning shot. Apart from this Radeon Pro W7600 with the factory blackout, there is certainly more going on. For example, the tensioning cross of the vapor chamber was only loosely screwed on my sample of the Radeon Pro W7900, so there were still a few turns to be done on two screws.
The missing screw on the slot bracket of the W7900 above the Mini-DP socket is due to a Mini-DP adapter, which is locked by a screw. I had originally noted this as an error. However, the old adapters and native cables no longer fit (as I have been using them for years), but a new connector with screw is used, which also only fits on such a special threaded solution. On the left the old solution and on the right the connection with the screw:
All of this is absolutely unacceptable in the workstation sector, and you can stand by NVIDIA however you want – I haven’t seen any cards that are so screwed up in the last 15 years, and there were quite a few cards in the tests. Well, nobody is error-free, but the error of the Radeon Pro W7600 would have actually been noticed in the QC stress test, if they had really taken everything seriously. For the screw, even a simple visual inspection is actually enough. The fact that neither would be done is almost frightening. But let’s get back to the test result AFTER the conversion.
I have packed this once in a before-after slider. The card doesn’t heat up any more after the conversion than before, the temperatures are still the same after two hours as they were after 6 minutes. So everything runs as it should and should have run. I can help myself of course, but a customer would now have a loss of productive work time and an RMA case.
I am actually tired of constantly making or fixing what could have been expected from the factory. It costs me unnecessary time and resources that are then sadly lacking on other tests and articles. Besides, such a correcting is always unattractive. I have communicated the problems to AMD, but unfortunately have not received any feedback yet. It was no different with the Vapor Chamber at the beginning. I would wish that AMD really puts more resources into quality management and finally takes the contract manufacturer to task. Something like here with the Radeon Pro W7600 is actually an impossibility. But unfortunately it’s true, too bad.
91 Antworten
Kommentar
Lade neue Kommentare
Veteran
Mitglied
Mitglied
Urgestein
Urgestein
Urgestein
1
Urgestein
1
1
Urgestein
Veteran
1
Veteran
1
Urgestein
Urgestein
Urgestein
Urgestein
Alle Kommentare lesen unter igor´sLAB Community →