ada lovelace NVIDIA Project Beyond GTC Keynote with CEO Jensen Huang: RTX 4090 + RTX 4080 Revealed

[BiG StroOnZ]

Summary

Today is finally the day that NVIDIA will announce its first Ada GPU architecture graphics cards for the consumer market. NVIDIA CEO Jensen Huang is expected to take center-stage to launch these new RTX cards.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Quotes

Quote

NVIDIA just kicked off the GTC Autumn 2022 Keynote address that culminates in Project Beyond, the company's launch vehicle for its next-generation GeForce RTX 40-series graphics cards based on the "Ada" architecture. These are expected to nearly double the performance over the present generation, ushering in a new era of photo-real graphics.

 

The RTX 40 Series GPUs feature a range of new technological innovations, including:

 

• Streaming multiprocessors with up to 83 teraflops of shader power — 2x over the previous generation.
• Third-generation RT Cores with up to 191 effective ray-tracing teraflops — 2.8x over the previous generation.
• Fourth-generation Tensor Cores with up to 1.32 Tensor petaflops — 5x over the previous generation using FP8 acceleration.
• Shader Execution Reordering (SER) that improves execution efficiency by rescheduling shading workloads on the fly to better utilize the GPU’s resources. As significant an innovation as out-of-order execution was for CPUs, SER improves ray-tracing performance up to 3x and in-game frame rates by up to 25%.
• Ada Optical Flow Accelerator with 2x faster performance allows DLSS 3 to predict movement in a scene, enabling the neural network to boost frame rates while maintaining image quality.
• Architectural improvements tightly coupled with custom TSMC 4N process technology results in an up to 2x leap in power efficiency.
• Dual NVIDIA Encoders (NVENC) cut export times by up to half and feature AV1 support. The NVENC AV1 encode is being adopted by OBS, Blackmagic Design DaVinci Resolve, Discord and more.

 

My thoughts

It appears that many of the rumors regarding the RTX 4090 and RTX 4080 were true, which is really unexpected. I'm mostly surprised by the segmented 16GB and 12GB RTX 4080. Pricing for all three cards is pretty crazy too; RTX 4090 is $1599, RTX 4080 16GB is $1199, and RTX 4080 12GB is $899. They must be really trying to get rid of old excess Ampere stock with these prices. I'm also pretty amazed by the fact that they reached 3GHz OC in their labs. That means there will definitely be AIB/AIC variants of Ada that can hit 3GHz out of the box.

 

Sources

https://www.techpowerup.com/299056/nvidia-delivers-quantum-leap-in-performance-introduces-new-era-of-neural-rendering-with-geforce-rtx-40-series

https://www.hardwaretimes.com/nvidia-rtx-4090-to-launch-on-the-12th-of-oct-for-1499-rtx-4080-16gb-for-1199-claims-4x-perf-gain-over-rtx-3090-3080/

https://www.hardwaretimes.com/nvidia-rtx-40-series-announced-with-dlss-3-3x-perf-boost-in-ray-traced-games-but-exclusive-to-rtx-40-series-gpus/

https://www.techpowerup.com/299058/nvidia-introduces-dlss-3-with-breakthrough-ai-powered-frame-generation-for-up-to-4x-performance

https://www.anandtech.com/show/17583/the-nvidia-geforce-project-beyond-and-gtc-fall-2022-keynote-live-blog-starts-at-8am-pt1500-utc

https://www.kitguru.net/components/graphic-cards/matthew-wilson/nvidia-launches-rtx-4090-for-1679-rtx-4080-starts-at-949/

https://videocardz.com/newz/nvidia-introduces-geforce-rtx-4090-4080-series-rtx-4090-launches-october-12th-for-1599-usd

https://www.overclock3d.net/news/gpu_displays/nvidia_officially_reveals_their_rtx_4090_and_rtx_4080_graphics_cards/1

https://arstechnica.com/gadgets/2022/09/nvidias-ada-lovelace-gpu-generation-1599-for-rtx-4090-899-and-up-for-4080/

https://www.kitguru.net/components/graphic-cards/matthew-wilson/nvidia-is-releasing-an-rtx-remaster-of-valves-portal/

https://www.tweaktown.com/news/88521/nvidias-new-geforce-rtx-4090-2x-faster-than-3090-ti-for-1599/index.html

https://www.techpowerup.com/299055/nvidia-project-beyond-gtc-keynote-address-expect-the-expected-rtx-4090

https://videocardz.com/press-release/nvidia-introduces-dlss-3-upscaling-technology-35-games-on-a-list

https://www.guru3d.com/articles-pages/geforce-series-4000-announcements-(preview),1.html

https://www.nvidia.com/en-us/geforce/news/rtx-40-series-graphics-cards-announcements/

https://www.tomshardware.com/news/live/nvidia-geforce-rtx-4090-reveal-live-blog

https://www.engadget.com/portal-rtx-ray-tracing-nvidia-155126135.html

https://www.engadget.com/nvidia-rtx-4090-announced-152529456.html

https://www.engadget.com/nvidia-dlss-3-ada-lovelace-160404110.html

https://www.pcgamer.com/nvidia-geforce-rtx-4090-revealed/

[schwellmo92]

RTX 4080 12GB should really be the RTX 4070 but is called the 4080 so they can say the 4080 starts at $899 and make it look like prices haven’t gone up much. Change my mind.

[igormp]

TLDR; Portal RTX, DLSS3 fast boi, 4090 $1.6k, 4080 16gb 1.2k, 4080 12gb $900.

[starsmine]

RTX remix to generically inject RTX mods into older games I think is the biggest news to me.

[Senzelian]

I had to laugh so hard when they said the Deutsche Bahn is using Omniverse for more efficient train routes so they wouldn't have to build more tracks, which is exactly what the Deutsche Bahn would do, meaning: Spend millions on some bullshit software so they wouldn't have to do actual work and improve the terrible network xD

 

Also, $1200 for a 4080  

$900 for 12GB 

[porina]

Random thoughts:

 

With the usual disclaimer of waiting for actual game tests, for those scared about power consumption, it looks like you can reduce the power limit a LOT and still be far ahead of Ampere.

 

Is out of order processing a new thing on GPUs? It's been in CPUs since forever.

 

DLSS3 looks like adding temporal frame generation or am I misunderstanding it? 

 

Looks like the move to TSMC N4 helps a lot. Jensen used the term optimised, I wonder if that's in reference to N4 being a refinement of N5, or if it is N4 but further optimised? Reason for asking is that AMD using optimised N5 for Zen 4 isn't just N5 but something "better". Presume same for RDNA3?

[igormp]

Just now, porina said:

Is out of order processing a new thing on GPUs? It's been in CPUs since forever.

GPUs are really simple and dummy when compared to CPUs.

[BuckGup]

Just now, porina said:

Random thoughts:

 

With the usual disclaimer of waiting for actual game tests, for those scared about power consumption, it looks like you can reduce the power limit a LOT and still be far ahead of Ampere.

I'm sorry but if you are paying $1600 MSRP you shouldn't have to modify your product on the companies behalf because they didn't

[shermantanker]

looks like I'm upgrading to a 4090. Need a tax write off lol.

[DeerDK]

Anything about power usage? I still want a quiet lean build

[Blademaster91]

The 4080 at $899 is too expensive, no TDP numbers either, and the 4080 12GB is more of a 4070Ti to justify the higher cost of the 4080 16GB.

Also no wonder EVGA got out of the market, as cards get more expensive and larger coolers cost more, it doesn't make sense to have a 5% margin or even taking a loss.

[jasonvp]

2 minutes ago, BuckGup said:

I'm sorry but if you are paying $1600 MSRP you shouldn't have to modify your product on the companies behalf because they didn't

I'm pretty sure he wasn't talking about modifying anything.  MSI Afterburner is a thing, and let's you dial up or down the power usage of the card.

[porina]

2 minutes ago, BuckGup said:

I'm sorry but if you are paying $1600 MSRP you shouldn't have to modify your product on the companies behalf because they didn't

It is an option for those who look to reduce power in they choose to do so. Right now I have a 2070 and 3070 running a compute challenge elsewhere. Both are set to minimum power limit, since the impacts to their throughput are much smaller than the reduction to power. However for gaming, I take off those limits and go back to stock.

[cj09beira]

6 minutes ago, porina said:

Random thoughts:

 

With the usual disclaimer of waiting for actual game tests, for those scared about power consumption, it looks like you can reduce the power limit a LOT and still be far ahead of Ampere.

 

Is out of order processing a new thing on GPUs? It's been in CPUs since forever.

 

DLSS3 looks like adding temporal frame generation or am I misunderstanding it? 

 

Looks like the move to TSMC N4 helps a lot. Jensen used the term optimised, I wonder if that's in reference to N4 being a refinement of N5, or if it is N4 but further optimised? Reason for asking is that AMD using optimised N5 for Zen 4 isn't just N5 but something "better". Presume same for RDNA3?

not exactly sure, but nvidia was doing all their scheduling on the cpu side which likely has some limitations which could lead to more pipeline stalls than if the scheduler was in the gpu itself

[Quackers101]

As I thought new DLSS or update to it. might be card specific though.

Also look USD else the price increase likely not going to be fully price to performance and Nvidia going to try and push for max profit.

To that Nvidia used some of the features that intel showed for their ARC GPUs, I guess its nice to see.

[starsmine]

8 minutes ago, DeerDK said:

Anything about power usage? I still want a quiet lean build

Graph up there shows 450W 

[Quackers101]

3 minutes ago, starsmine said:

Graph up there shows 450W 

PC's being 600W space heater. But yes it seems to be around 450W.

[DeerDK]

2 minutes ago, starsmine said:

Graph up there shows 450W 

Oh yea. Missed that. Thanks

I hope they make a 150w model. 

Otherwise ill just have to wait and see what AMD brings to the table

[melete]

I generally buy xx80 series cards, but $1199 for the true RTX 4080 is nuts. 

[starsmine]

Just so people are aware in terms of GPU costs
TSMC N7 is 9k a wafer
Samsung 8nm is well under that
TSMC N4 is closer to 18-20k per wafer

For theses 500mm+ sized chips, THATS HUGE 
With AIBs already unable to make money AT MSRP with 30 series, that price increase has to exist

 

[cj09beira]

no card below 900 dollars, to help sell the 3000 series cards no doubt. 

[Rauten]

2-4X times the performance? Excuse me good sire, what do you think we're smoking?

Luckily, the first cards are available October 12th, so we don't have to wait too long to see how much bullshit Jensen's jacket is trying to feed us.

 

Edit: I wouldn't mind playing RTX Portal, though.

[Quackers101]

oh god the gameplay sections for DLSS looked awful, either something is wrong or such bad resolution.

[Dracarris]

2 minutes ago, starsmine said:

For theses 500mm+ sized chips, THATS HUGE 

how many (functioning) GPU dies do you roughly get out of one N4 wafer?

[LAwLz]

The decode engine in the 40 series is the same, but they have updated their encoding engine to support AV1 encoding! 

It seems like they have also decided to include two encoding engines, so we might see a doubling in encoding performance in addition to the new format support. 

 

Edit: It's confirmed that the dual encoders can work on the same encoding task, thus doubling performance. 

They are also working with multiple software vendors such as OBS, Discord and DaVinci to add support for hardware accelerated AV1 encoding. They are also making improvements to the older generation of GPUs in OBS, reducing encoding overhead by around 35%. 

 

 

Saw someone ask about TDP. 

RTX 3090 - 450W

RTX 4080 16GB - 320W

RTX 4080 12GB - 285W