![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
I do not think you can share the water pool has a reservoir (modern water block with ultra-thin fin system need pretty near perfect water and some would worry about the many pool product used), you just past tube in it to make some heat exchange
yeah its laid under the pool bottoms cement and it radiates heat into the pool, like heated floors.
If they're planning on charging 700 for the 4080 12 GB this generation is already a massive disappointment. We'll see, but I'm not hopeful. If this is their plan it'll allow AMD to do basically the same thing.
It is possible, $700 in 2022 dollars will be around $590 2018 dollars when Turing released, the 2070 FE edition was $600 on launch, 2080 FE was $800 (or $950 now). The cheapest 3080 MSRP was $800 in 2022 dollars and impossible to buy at that price tag, I am not sure why the 16 gig 4080 would not be at least a bit more than the same price (i.e. $800 USD).
It's the cut down AD 103 chip, my guess it'll be marginally faster than the 3080 12 gb we already have. These chips were typically reserved for 70 series parts in the past. At the very least, they should name it a 3070 Ti to avoid confusion.
The 4070 rumored specs were very similar but at 160 bit bus instead of 192 bits and with the 160 bits the rumours were around a bit over a 3090 TI in games, it is a bit strange.
Well now, they've finally taken GPUs to 11; we have finally caught up with spinal tap amp!
Seriously?
Looks like that stupid back fan was added 'last minute'. Notice how it overlaps the X in 'RTX'. Was probably originally meant to have just holes there.
Fermi 2.0, Hot and Loud Edition, four fans...
If the rumored TGP is true, then the 4090 will dump up to 660W into the cabinet. I think there will be plenty of offers on watercooling for those cards then, so don´t worry.
Hope nvidia is making the most of this new node because AMD is leap frogging.
Where is the info? All I see is a twitter post with a pic of Lisa Su...
Its going to be an interesting launch this time. Nvidia with monolithic die on 4n and AMD with chiplet design on 5n.
Isn't the chiplet only separating memory controller and IO task from the main chip, still a single monolitic main compute die seem to be the latest rumors, will see how significant it will be to have a memory die and a graphic one. Maybe it will let them to have only one on 5nm and the other on a cheaper node, most of the gain on that generation could be on the cost side (not only maximising node cost but less waste of the expensive graphic unit die if you do not have to throw away those that do not work because of the memory controller part of the chip).
There have been rumored more then a single gpu chiplet and an I/O chiplet (7950 XT). Always take it with a huge bucket of salt though. In the Zen space, you can have 2 (consumer) to 8 (pro/threadripper) CPU cores combined with I/O and infinity fabric cache. A bit harder with GPU and parallell tasking, but who knows ...
The chiplet design might allow for more voltage and higher clocks too. Its a bit exciting nevertheless. 
It provide the thing we actually care about: AMD RDNA™ 3 is on track to deliver an estimated >50 percent better performance per watt than AMD RDNA™ 2 architectureOfficial AMD communication is promoting power efficiency but not higher clocks
https://community.amd.com/t5/gaming...gAfCsJiyJfp00InSABRvAJsXcMXTtc6BDEN6PPSU2vBD4
It has slightly larger fans, nothing too drastic.
Oh look, the 4080 proper costs more or equal to the 3090ti (current price). Still laughing?
According to their site 3.0 is exclusive to the 40 series.The claim is with DLSS 3 the 4090 is 2-4x faster than the 3090ti With DLSS 2... But DLSS 3 will be exclusive to RTX 4xxx? No mention there
Of course it will. I doubt there isn't even a technical limitations on the 30 series cards that would prevent it from take advantage of 3.0.
I think they were showing an extra chip that RTX3 contains which sequences the ray tracing for better performance aka DLSS 3
