I’d like to have a game with those graphics. That felt like a long QuickTime event.
Interesting news about Nvidia Ampere here.
https://www.anandtech.com/show/15801/nvidia-announces-ampere-architecture-and-a100-products
If these numbers are correct, Nvidia somehow managed to fit 59% more transistors per square millimeter than AMD, even though both are on a 7nm process node. This is an absolutely monstrous advantage for Nvidia that will make the 30-series GPUs much cheaper to manufacture and much, much faster.
It’s not entirely clear how this is possible. Process improvements often happen within a given size, but 59% is unheard-of.
Oh hooray, now we don’t have to satisfy our GPU news cravings with WCCF bollocks.
When I see graphics like that I am reminded of this video.
Does that info suggest the rumored specs we saw a couple of weeks ago seem any more probable?
It’s certainly more probable with 59% greater transistor density, yeah. If Nvidia pulls that off they’re going to flush AMD down the toilet.
@stusser Intel or AMD cpu to go with that RTX3080? :)
vyshka
4439
My plan is AMD Ryzen™ 9 3950X
I’d say Zen 3, since they will be available at around the same time. My planned April build has officially become an October/November build
Yeah when the days shorten and the economy is in a full on death spiral, that’s when I too shall slay the piggy bank and build a monster rig for the ages. And that machine shall go on to serve until it’s about as powerful, relatively speaking, as a 486 is now ;-)
There must be some typo with the transistor count. It more than doubled on the same die size. The compute increase does not match the doubling of the transistor count.
So I think there is a print error regarding the transistor count
I would certainly go with AMD, but I’m still not seeing any really compelling reason to upgrade from my 5 year old overclocked i7-6700K.
The transistor counts are correct, but it seems the A100 has an enormous 40MB cache on the die, and SRAM transistors are extremely small. So Nvidia may have around a 15-17% density advantage by using a 7nm+ process, but not 59%. That’s what people are saying, anyway, nobody really knows for sure.
What you said does not make sense with the counts. 54B vs 21B transistors.
Even if it dedicate a portion to cache. It still cannot explain the massive shrinkage on non-cache portions of the CPU. They’ll have to shrink the 21B existing transistor types also on the same die size.
It’s inconceivable that they can squeeze an extra 33B transistors onto 11mm2 increase in die size (826mm2 from 815mm2).
Edit: The only way we can explain this that someone counted 2x transistors as in they have to use a pair of chips but forgot to add the die areas of 2 chips. Or magically managed to etch both sides of a die…
Edit2: Oh wait, I didnt see that the process shrunk from 12nm to 7nm. Ok that clears it up.
RTX 3080 pictures leaked. Actually looks rather innovative, pulling in air directly over the GPU at the front bottom of the card and exhausting it out the back top of the card, right where your case CPU-aligned exhaust fan should be. If it’s fake, someone put a ton of effort into this!
If this is real it means the 3080 is well into production and will be ready for release soon. Probably August.
Both fans look like intakes, spinning in opposite directions.
That’s a user render so it could be wrong. I don’t see how that would actually work.
They’ d have to drill holes in the card :)
If you look at the last pic, the actual PCB only goes like 2/3 the length of the card, with the rear capped off with just a heatsink and the top fan.
The sides aren’t solid, like in the render. I think the air gets pushed out from the sides.
Edit: or just pushed out the opposite side through all the fins?
I suppose if the whole thing is sealed and it works like a double-sided blower, it could exhaust out the rear? Doubt that would make it much more effective than a standard blower. It could be quieter though.
