But what actually happens here on the SUPRIM? When the pad has started to bleed out, the temperature of the memory will rise significantly. The thermal contact decreases significantly even after the silicone has run away and the RAM becomes continuously hotter.
Of course, this is a gradual process over weeks or perhaps even months, the speed of which also depends on user behavior. Unfortunately, the longer and more intensive the work, the faster the pads bleed. First the silicone collects (as you can see on the pictures) between the pad and the two contact surfaces and is still held up by the protruding and somewhat higher edges. Even now the temperatures of the GDDR6X are above those of the normal delivery state (approx. 10 degrees higher)
Later, however, it runs out at some point anyway and then the thermal stress on the memory also begins. The temperatures of the GDDR6X go up to 110 degrees and then thermal throttling already starts. This costs significant performance and can also be logged with suitable tools. In the picture below we can already see the second stage, where the silicone, which has excellent creeping properties, slowly spreads over the radiator. There have also been cases where the silicone ran downwards towards the PCIe slot when the card was mounted vertically (riser) and significantly impaired functionality. At the latest then you will notice and even see that something is wrong even without disassembly.
I have already spoken more intensively with MSI about this problem, which, moreover, is already known at the factory. However, most RMA cases so far (according to MSI) are likely to be the fault of miners who have even removed the bleeding pads and replaced them with supposedly better (and thus, unfortunately, harder) pads. But then exactly what I wrote above about the solder pills happens: the memory cracks away. The picture below shows MSI’s original pads on a normal customer card that was not used for mining. Exactly this kind of thing should not happen, however, and it constitutes a real RMA case.
Solution in sight?
MSI is actively swapping pads right now and looking for viable alternatives for mass production. Bridging the 1.5 to 2 mm with rather dry pads, which only perform 3 W/mK and as NVIDIA buys them e.g. from Laird, is not a solution either, because they are not suitable for machine mounting on custom coolers in the better performance classes. More expensive pads probably do, but again are harder. And so MSI is certainly not the only one looking for a solution. I will test there soon once what, see if it brings something.
As a buyer of such a card, do you now have to panic or can you sit it out comfortably? Neither, because you can also test yourself whether the cooling behavior may have deteriorated. Then you should go the way via the RMA, because the process of dissolution is irreversible. As long as everything runs as usual and bought, the world is still in order and paranoia is out of place. I wouldn’t just ignore it either.
71 Antworten
Kommentar
Lade neue Kommentare
Urgestein
Veteran
Urgestein
1
Veteran
Mitglied
Veteran
Mitglied
Mitglied
Urgestein
Mitglied
Mitglied
Veteran
Mitglied
Urgestein
Mitglied
Urgestein
Urgestein
Mitglied
Alle Kommentare lesen unter igor´sLAB Community →