* RTX 3080, 699$, 2x faster than RTX 2080
* RTX 3090 (Titan), 1500$, 1.5x faster than RTX Titan, 8k resolution @ 60 FPS with RTX on Control.
---
I hope that if somebody bought an RTX 2080 or similar in the last 2-4 weeks, that they bought it over Amazon, and can return it.
8N manufacturing process is presumably Samsung, which will probably be beat by TSMC 7nm.
I'm holding out for RDNA2.
> To that end, our engineers designed a much smaller PCB, shrank the NVLink and power connectors, and still managed to pack in 18-phases for improved power delivery. Don’t worry, we included an adapter that allows Founders Edition cards to work with users’ existing power supplies.
[1] https://www.nvidia.com/en-us/geforce/news/introducing-rtx-30...
Why would you keep a 1200$ RTX 2080 Ti or a 2500$ Titan when you can get at least the same perf for 50-75% of the price with the new products, and much better RTX perf ?
This assumes that with RTX off the new gen won't be slower than the old one, but I think that's a fair assumption, and if it isn't, you can always return the 30xx card in 2 weeks, and buy an used 2080 Ti or Titan on ebay for pennies, once the market overflows from people upgrading for the 30xx series.
People were still asking for 900$ for a used 2080 Ti on ebay this morning, and the 3080 700$ price just destroys that offer. Many owners are going to try and dump these cards for as much as they can get in the next two weeks. I wouldn't buy a used 2080 Ti for more than 250$ today. In one month, these cards are going to sell used for 100-200$. If AMD can only match the 2080 Ti perf, they are going to have a very hard time pricing against the used 2080 Ti market.
> And let's not forget the huge power draw, requiring a new power connector and a 3 slot cooler.
That's only for the 3090 IIUC. All other cards were announced as being actually much smaller than the 20xx series ones.
Do note that the 3080 is still 2 weeks away (17 sept), 3090 24 sept, 3070 in October.
I would consider getting a non-founders edition though, in the past other brands have had better/more silent cooling for pretty much identical pricing.
Edit: While the 3090 (350W) is the only one requiring a 3 slot cooler, all 3 use the new 12 pin power connector. The 3080 founder edition power draw is still 320W, vs 220W for the 2080.
That's not that difficult considering the RTX cards had terrible price/perf.
Certainly the market for them will take a beating compared to new prices now. But I can't imagine it collapsing like that purely based on the supply. How many people are really going to drop their 2080 just because a new thing is out there?
I'd love to be wrong, as an original 2070 owner ;)
Pretty much everyone who paid for a 2080 Ti on launch and always need to have the latests bestest thing.
Well for one, keeping a card you already bought is free, buying a new one costs money.
Why are you all getting so excited about the prices given in a paper launch? Wait until you can actually buy one at that price, which I'm guessing isn't going to be until well into 2021.
Also if find a 2080ti for 200 before the year 2022 let me know I'll buy you a beer.
They explicitly noted that it was Samsung.
[1] https://www.nvidia.com/en-us/geforce/news/introducing-rtx-30...
I still don't understand the raytracing craze when it comes to games. Years ago when I wrote my first raytracer I might have got excited about it. But we have advanced so much with rasterization that I don't really understand why this is something we need.
Raytracing complexity is tricky and it is likely to prove challenging to do in a game with consistent frame rates. Soft shadows are expensive even for hardware.
I would be more excited about some other global illumination improvements, like photon mapping.
"Need" is a strong word, but I'm sure someone will make a game where the shadows _matter_ for telling the story, or solving puzzles, or even just making the game world look more realistic and believable.
Or selling GFX cards..
RTX cards don't render everything with ray tracing. They use it for light calculations, for example it can be used exactly for global illumination.
> why this is something we need.
The answer is simple. Because there is a limit to how realistic we can make the lighting and shadowing using the old models. We are almost at that limit. Using ray tracing removes that limit.
> challenging to do in a game with consistent frame rates
Try Quake II RTX. It is very consistent right off the bat, and if you turn on the resolution scaling, you can get completely consistent framerates (within 2-3 fps continuously). While it might not be a game with the most advanced graphics, it does have variety of scenes, and there is no issue with the framerates being "inconsistent".
- Ray tracing only is not 200% but 195% in real world tests.
- In modern games like Tomb Raider the fps increase is between 165 and 180%, averaging 175%.
This is comparing the 2080 to 3080 playing real world games.
It looks pretty impressive.
Regarding the new power connector: All the partner cards showcased by Gamers Nexus have the regular two or three 8-pin connectors and the FE cards seem to include an adapter. Nvidia states that a 750W power supply is required, so depending on the CPU even a 650W should be fine.
Anyway, thank you AMD for the competition.
https://www.newegg.com/p/pl?d=GTX1070
https://www.newegg.com/p/pl?d=RTX2070
People are going to get a massive jump in performance for just $500. I don't think we've ever seen this before.
It looks pretty impressive.
If this pricing holds up though, I need to get a large NVDA position, because they'll sell at TON of 3090. I'll buy 8.
You likely meant something else; Titan can't be faster than Titan.
There is also a previous gen HPC "Titan V" with HBM memory, but AFAICT no Ampere Titan card with HBM2 was announced.
You can see this because the chip name is GA102, the 2 at the end indicates that this is a cut-down chip.
With one exception: the Titan V card that I mentioned which comes with the GV100 chip. But as I mentioned, that one targets a very different market segment with HBM memory, which the "RTX 3090" obviously doesn't target, since otherwise it would come with HBM2 memory like all the GA100 products.
>With one exception:
This is simply wrong. The majority of Titans have been full chips: Titan, Titan Z and Titan Black as well as the Titan V and Titan X.
The other Titans were about as fast as manufacturer-overclocked 80Ti models.
The naming is super confusing. The 100 versions are essentially HPC chips on a PCI-express board, while 10x are completely different products.