Discussion in 'Video Cards & Monitors' started by Sphinx2000, Sep 1, 2020.
I think it is to do with the IHS as opposed to the actual chip itself.
i have always used liquid metal when delidding my intel cpu's, between the die and ihs.
Have never used it on a GPU before, never seen anyone else use it either. only in cpu delid's
Most use Thermal Grizzly Kryonought Thermal paste for GPU, and between IHS and cooler.
Usually on professional/factory uses of liquid metal, they would apply some kind of "diffusion barrier" to copper as well as an actual physical barrier around the sides to prevent "leakage" such as the foam carrier on the PS5 APU, I expect that the PS5 would have a patch of gold or nickel plating on the heatsink.
Using liquid metal on bare copper kind of sucks, requiring preparation/multiple applications. Liquid metal was used on some 2008 Mac Pro's (Mac Pro 8X), IIRC they used gold as a diffusion barrier on the processor/heatsink and a grease as a "leakage" barrier (prob just dielectric grease).
One tip I've seen people use is a small amount of regular thermal paste in a square around where the liquid metal goes.
I think you're far better off using a varnish (clear nail polish works) around the die to prevent any overspill creating conductivity where it shouldn't. I did this on my 6700k just to be sure, so far so good. Also my temps dropped like crazy, like 30%.
Stop being so bloody level headed, and get into the drama of it all!
ZOMG you're right! NVidia cancelling unannounced rumoured SKUs can only mean one thing.... 40Gb 200CU Big Navi confirmed!
Mind, I always thought that about the '20gb' 3080.
Zero sense quotient in RAM numbers so close to the 3090.
16gb 3080 - now that might make sense... big enough increase to appeal to gamers & creators, and also compete with AMD, yet still leaving the 3090 a healthy EP margin.
Also might be somewhat more reasoned in price, but then again - this is Nvidia, so...
THERE you are
Just did a benchmark with rtx on vs off. Did an average of 77 with RTX On and 106 with RTX Off. The minimum was 55 with it on vs 84 with it off. I couldn't tell the difference in the graphics quality tbh.
Probably better article. And wouldn't say a nothing story, Nvidia is clearly having issues - perhaps and hopefully they will make these cards on TSMC 7nm node -> 3080TI
would have to fit in between a 3080 and 3090 performance yeah?
small gap to hit.
perhaps a 7nm, non-X memory 16gb 3080 Super card might fit the bill?
for 5-10% faster then a 8nm X mem 10GB 3080, knowing nvidia, that card would be $2k-2.5K
really wouldnt make much sense.
What would make more sense, is a 3090 Ti, that is 10-15% faster then a 3090, and drop 8nm 3090 to 2.2k, and sell the 3090Ti for 2.8-3.1K
hey bud how’s your 3080 going? Did you end up getting one from the US? For dirt cheap?
my 1200 card is showing up on Monday can’t wait
Simple doubling of the memory chips (either number or volume), keeping the bus bandwidth the same. AFIK, to do a 16Gb 3080 would require a repeat of the 3.5gb+500mb 970 kerfuffle.
Not necessarily. Remember the second gen 12Gb Titan X? (as in the first Pascal cored Titan). Pipped by the 11Gb 1080ti, then stomped by the Titan X Pascal. So a 24Gb 3090 beating 20Gb 3080ti has form. Quickly followed by the proper Ampere Titan.
Oh yeah, forgot you bought the delid kit off me. Glad to hear you got the results you wanted.
Well I won't be getting mine today....thanks Auspost.....it just took a week to go from Redbank QLD to Melb, /le sigh....
Having visions of it being on the back of an electric postie bike, having to stop every 100 kms to recharge on the way down lol.