I have a 'problem' with one of my two Asus directcu ii OC r9 290 cards. I got those cards about two weeks ago and the first thing i did, was checking the temperatures with Unigine Valley. Apperently those cards should run 20 procent cooler than the reference r9 290, but for one card this isn't the case at all. I started testing those cards in a crossfire setup first. These were my results: Top card: 90-94 C Bottom card: 70 - 75 I wasn't happy with those results at all, so i swapped the cards around to see if i would make any difference: Top card: 82C Bottom cards: 84C The answer: yes, it did make a huge difference. To confirm my results, i tested the cards individually. One card only reached 71C, while the other reached 84C. There is obviously something wrong with the card the runs hotter. I know that the card isn't overheating, but there shouldn't be a difference of 13 degrees celcius between those two cards. After my tests i took a closer look at the coolers of the two cards and the first thing i noticed was that the heatsink of the card that runs hotter is slightly bent (the heatsink of the cooler card is perfectly leveled). I thought that i could fix this issue by revoming the heatsink and put it back on again. The temperatures did decrease only a little (2 degrees at best), but probably only because Asus used a whole tube of thermal paste instead of a tiny amount. As far as i can tell, the heatsink seems to make good contact with the die. I reapplied thermal paste (Artic Cooling MX 4) a couple off times, but it didn't have any effect on the temperatures (i used the rice method and it did seem to spread nicely over the die). I also put my fingers on the end of the heatsink while running Unigine Valley and i noticed that the heatsink of the cooler card gets hotter than the heatsink of the hotter card. The hotter card also cools down a little faster when there isn't any load on the GPU anymore. A couple of days ago i tried to level the heatsink by bending it a little. Strangly enough, this lowered the temperature by 2 degrees. What could be the problem here? I really don't have a clue. I know that it is probably hard to tell, but any suggestions are welcome. * I'm not a native speaker and i'm sure my english isn't always correct. Sorry for that
I'm just guessing here, but maybe the card with the bent heat spreader makes better contact with the die now that it's bent. Or possibly it's making it easier for air to be pushed along the fins. But tbh, I think it could also be that one chip is better than the other, or a simple thing as slightly different soldering... I wouldn't be surprised to see that this could be totally normal as it's not dangerous to your cards, but somebody with experience in CFX / SLI could probably give better feedback on this.
I did try to level the bended heatsink by bending it back, but i didn't seem to work. Only thing i noticed was a difference of 2 degrees celcius. So something changed, but i don't know what. I also thought about differences between the chips. i don't know if one chip can use a bit more power than the other chip. By the way, i noticed a difference in the RPM of the fans. The RPM of the hotter card is 5 - 10 procent lower than the RPM of the cooler card. At 100 procent fanspeed the difference is 150 - 200 RPM.
It sounds odd that those two fans on the same card (I take it their basically the same card, as in the same model / vendor) would have different RPM settings, and at 100% speed they should actually run the same RPM. Maybe something's wrong with the fan, and not the chip / heatsink.
i did use a wattage meter to measure the power consumption during furmark (+50 procent power limit). System idle (incl monitor and speakerset): 181 watt One card: 472 watt two cards 761 watt Both cards seem to use the same amount of power, about 290 watts.
Yeah, the two cards are exactly the same. Like i said before, i have two asus directcu ii OC r9 290 (R9290-DC2OC4-GD5).
I now know for sure why the hotter card gets hotter. It's definitely the fact the the heatsink isn't completly leveled. I used a wire to attached the PCB from the hotter (bottom) card to the cooler (top) card and this seems to have stabilized the heatsink. The heatsink of the hotter card is now leveled with the PCB. The result? A decrease of 7 degrees celcius. The card now runs at 77C instead of 84C. As soon as i remove the wire, the card destabilizes again and i get higher temperatures. The card still gets about 6 degrees hotter than the other card and it's stupid that i have to use a wire to get better temperatures, but i'm fine with these results. I still wonder though. Why does leveling the heatsink have such a huge impact on the temperatures?
A wire? That indeed sounds strange... so will you be able to leave the wire in there? I still have a hard time imagining how this has to look tbh.
I was talking about those wires that are used for tying together mouse and keyboard cables for example. I don't know if 'wire' is the right word for it, but i hope you get the idea now. Thanks to the 'wire', the top card is lifting the PCB of the bottom card up a little. The result is that the heatsink of the hotter card is now leveled with the PCB. Yes, i think that i'm able to leave the 'wire' in there. But the heatsink of the hotter card is still bent and like i said, without that 'wire' the temperatures are up again.
I’m afraid that I voided the normal warranty, because I removed the cooler from the card. One of the screws had a seal on it. There is however another form of warranty left. The warranty that is giving to me by the Dutch law. It’s a bit hard to explain, but according to the Dutch law you still have warranty for a certain amount of time. The length of this warranty is determined by how long someone can expect a product to last. I would say that the bare minimum for a GPU would be at least 1 year. So based on the Dutch law I still have one year of warranty left (maybe a bit longer, but it’s all a matter of interpretation). The question is however, if the cooler is not working as intended. If you compare it to the other card, then the answer of course is very simply. But if you don't compare it. How should someone know how good this cooler is supposed to cool? Is there a standard for how good this card should cool? In fact, if I didn’t have the cooler card, I probably wouldn’t have noticed that there was something wrong here. The answers to the question are very important, because I might not get a new card if I cannot prove that the cooler should cooler better.
Just checked my cooler again and it turns out that one of the 5 heatpipes isn't working correctly. The heatpipe of the cooler card is very hot when i touch it, the heatpipe of the hotter card is actually luke warm. The heatsink around the heatpipe is also alot cooler on the card that runs hotter. Only 3 of the 5 heatpipes are in direct contact with the dei. Luckily the heatpipe that doesn't work correctly is one of the 3 heatpipes that make direct contact with the die. That's why the cooling performance is much worse and as a result the card runs 10 degrees celcuis hotter than it actually should. I don't know how easy it is to get a replacement. Is the support of ASUS any good?
I just wanted to make sure that the heat pipe was defective, so i removed the cooler and heated it up with a hear dryer. Turns out that the heat pipe actually works and it did get pretty hot fast. I removed the cooler a couple of times and it did not have any effect on the temperatures. Now it is working as it should. While mounting the cooler i did push the cooler as far to the left as i could, so that the 'defective' heat pipe is covering the dei as much as possible and apperently that did the trick. I always tested the temperatures with unigine valley. I checked the temperatures after one run. Before: first card: 80 second card: 84 Now: First card: 74 second card: 72 I'm happy for now. Hope it stays this way.