NVIDIA 4000 Series

[Murphy]

I wonder... right now the 80 cards give 70-class performance, the 70-class cards give 60-class performance, and leaks indicate that the 60-class cards give 50-class performance, so could Nvidia perhaps have avoided a lot of the opprobium if they'd straight-out said something along the lines of, "We know GPUs are expensive so we're changing our tiering to create additional room at the low end."?

Problem with that is what would an even lower-end card actually offer apart from price, Nvidia themselves are pitching the 4060 @ 1080p gaming without max graphics settings so would a 4040 be targeted at people who play minesweeper.

How much more performance would such a low-end card offer over Intel and AMD iGPU's in some of their CPUs.

 

[CAT-THE-FIFTH]

When is the next generation of consoles due? The lifespan of a console seems to be 6-7 years, so that puts us at 2026-27. I expect the next consoles to support 4k gaming with full RT. So I expect we've one or two more generations (24,26) of high-priced video cards before competition from consoles bites.

There will be a refresh maybe in the next year or so? Remember the Xbox One and PS4 had refreshes around 3~4 years after release. The current consoles will be 3 years old in a few months. If there isn't a refresh,the next consoles will be arriving before 2026/2027 IMHO.

The rumours say the next Nvidia generation won't be releasing until 2025:

NVIDIA 3nm GPUs Not Launching Till 2025, Expected To Utilize TSMC For Next-Gen Chips

NVIDIA's next-gen GPUs based on TSMC's 3nm process node might not be coming any time soon as reported by DigiTimes.

wccftech.com

 

[Phixsator]

It does make you wonder, if they have done this with 4000 series what will they do with 5000 series and the generations after that. I don't feel like they will give up on trying to make these high prices the norm, if anything they will try and squeeze more out in profit by further reducing the cost of each tier whatever way they can even if it feels like to us they are taking the p***. They've proved this gen that they absolutely do not care how consumers feel about them because they know people will still buy their products regardless of how anti consumer they become.

Simple solution is to stop giving them money. A concept that is apparently lost on many people.

 

[Jay85]

When is the next generation of consoles due? The lifespan of a console seems to be 6-7 years, so that puts us at 2026-27. I expect the next consoles to support 4k gaming with full RT. So I expect we've one or two more generations (24,26) of high-priced video cards before competition from consoles bites.

Thats a tough one to swallow. Me and the 3090 might be together for awhile longer then.

 

[Jay85]

There will be a refresh maybe in the next year or so? Remember the Xbox One and PS4 had refreshes around 3~4 years after release. The current consoles will be 3 years old in a few months. If there isn't a refresh,the next consoles will be arriving before 2026/2027 IMHO.

The rumours say the next Nvidia generation won't be releasing until 2025:

NVIDIA 3nm GPUs Not Launching Till 2025, Expected To Utilize TSMC For Next-Gen Chips

NVIDIA's next-gen GPUs based on TSMC's 3nm process node might not be coming any time soon as reported by DigiTimes.

wccftech.com

crazy to think its been 3 years already, it didn't seem that long ago at all!

 

[Jay85]

Simple solution is to stop giving them money. A concept that is apparently lost on many people.

Fomo usually eats them up

 

[robgmun]

Fomo usually eats them up

But what are they missing out of? Just DDLS3 and that's something still in its infancy

 

[Jay85]

But what are they missing out of? Just DDLS3 and that's something still in its infancy

Fancy new tech, new toy, something new to show off, something to say i got one on a public forum .

 

[Bencher]

Thats a tough one to swallow. Me and the 3090 might be together for awhile longer then.

You criticizing people buying nvidia, while you have a 3090. A 3090...probably one of the worst value propositions of any GPU ever in the history of GPUs. Ohkay I guess.

EG1. I had a 3090 as well, but at least I realize I make a mistake, ,im not criticizing other people for buying a 4070ti for half the money while matching the 3090s performance.

 

[mrk]

Just watched this, not sure if it was posted ages ago but at 15mins or so in he talks and shows the power target vs fps range and power draw, very good insight, basically the most efficient power target peak is 50% for a 4090:

 

[RavenXXX2]

 

[Neil79]

 

[Jay85]

You criticizing people buying nvidia, while you have a 3090. A 3090...probably one of the worst value propositions of any GPU ever in the history of GPUs. Ohkay I guess.

EG1. I had a 3090 as well, but at least I realize I make a mistake, ,im not criticizing other people for buying a 4070ti for half the money while matching the 3090s performance.

Bought during the pandemic, what can i say i had no other gpu at the time . Would i do it now? Not without careful consideration.

 

[mrk]

Alright we've all bought an overpriced GPU or two, time to relax!

 

[no_1_dave]

Alright we've all bought an overpriced GPU or two, time to relax!

I've been very lucky and always managed to grab cards at the right time which weren't over priced

9800pro
6800 GT bios modded it to Ultra
8800 GT
GTX 470
GTX 680
980 TI
1080 TI just after launch from the MM with a sizable discount
3080 for MSRP at launch

Although it's very tempting to go against this trend because the 4090 is just sitting there and all this talk around vram is making me get that upgrade itch. Patience has never been my thing

 

[TNA]

My most expensive card was my 3080. Even my 295x2 was cheaper (though that was used)

 

[mrk]

Interesting about prices, just read a comment where someone compiled a list of technologies afforded to being an Nvidia user, ok some are trivial since AMD has FSR, but DLSS is superior anyway so.... yeah

• Shader Execution Reordering
• Micro Opacity Maps
• Frame Generation
• DLSS
• Cuda and Cuda Deep Neural Network Library (cudNN)
• NVIDIA Hairworks, Gameworks, and Physx
• NVIDIA Omniverse
• NVIDIA developed Text-Image AI model/NN coming to Blender
• DLSS 3 integration into professional applications such as D5 Renderer which means Blender and other popular 3D applications will get DLSS 3 into their software improving performance and decreasing render times
• NVIDIA Optix
• Real Time Ray Tracing in consumers GPU’s
• Fastest Ray Tracing performance in consumer GPU’s and professional GPU’s
• Better In-Game Modding Capabilities and Compatibility
• Better support in UE5 and Unity or other popular game engines
• PHYSX
• NULL
• NVIDIA Reflex + Boost
• Neural Radiance Caching
• NVIDIA's Path Tracing SDK (currently version 1.0) that is currently in Cyberpunk and the not yet released Path Tracing SDK 2.0 that contains the rest of the SDK and features with the SDK
• DLDSR - this is not DLSS
• Better Professional Application Integration
• 3D Rendering being much faster and reliable on NVIDIA
• Proper Professional drivers that are updated more than a quarter per year (AMD only updates their Pro Drivers every three months or barely so in comparison to NVIDIA's pro drivers)
• Consistent game driver updates while AMD can go months without updating their drivers or patching important fixes or security flaws, NVIDIA fixes them within 24 hours (CPU bug)
• Better efficiency from using a custom TSMC Node 4N; instead of the regular 5 nm nodes produced by TSMC this node was designed by NVIDIA with TSMC producing the wafers
• NVENC and better H264/265 capabilities and support in OBS and Video Editing software
• Shadow Play
• G-Sync
• A new AI Denoiser integrated into DLSS 3 (Not yet released) this supposedly similar to how Optix works in Blender
• Simulating 100,000 hair strands in real time, seems to be Hairworks 2.0
• Neural Network Compression to reduce VRAM usage by 27% on 40 series GPU’s
• Better Open GL and Open CL compatible APIs in the driver
• Better 3rd Party tools and App support for NVIDIA for Overclocking, Hardware Analysis, and OBS/Encoder integration (AMD took a much, much longer time to integrate their AV1 Encoder into OBS)
• NVIDIA Eye Contact (Still in Beta)
• NVIDIA Broadcast
• NVIDIA ReShade

The list is still missing some of the technologies or features NVIDIA developed.

Also, ask Radeon where their RND money is and what it is utilized for. More Cash = More RND Money. The technologies are only possible with NVIDIA demanding a premium to pay their RND and Software Teams.

I am not defending NVIDIA’s pricing but that is how they fund these research papers and new technologies.

The Hairworks thing is really quite interesting, game changer for simulating hair:

 

[CAT-THE-FIFTH]

Interesting about prices, just read a comment where someone compiled a list of technologies afforded to being an Nvidia user, ok some are trivial since AMD has FSR, but DLSS is superior anyway so.... yeah

Sounds like a lot of marketing speech and a number of those were not invented by Nvidia. A bit like with Apple saying why they are better than Google(Android) or Microsoft(Windows). In the end all that matters is really the generational uplift per pricing tier and price/performance. It also seems the worse the generational price/performance uplift the more the marketing slogans get wheeled out.

What is quite clear,neither Nvidia or AMD seem to be doing much in this regard this generation.

 

[Grim5]

Nvidia found it's new crypto boom

The AI boom is real and demand is currently so high that ordering H100 GPUs now come with a 6 month waiting period and then cards are sold 40% over MSRP

NVIDIA AI GPU Demand Blows Up, Chip Prices Increase By 40% & Stock Shortages Expected Till December

NVIDIA GPU business is booming thanks to AI but may not be able to keep up with demand despite increasing production at TSMC.

wccftech.com

 

[Nexus18]

Interesting about prices, just read a comment where someone compiled a list of technologies afforded to being an Nvidia user, ok some are trivial since AMD has FSR, but DLSS is superior anyway so.... yeah



The Hairworks thing is really quite interesting, game changer for simulating hair:

But software features don't count towards the price of a gpu!!! Companies shouldn't be charging for these and instead should be offering them for free....

There really is no comparison though when you look at the vast majority of features that nvidia have but most importantly, also the quality and consistently better experience (not perfect but a damn sight better than their competition) hence why they can charge somewhat of a premium for these alone especially since they are first to the market by months/years...... Like I have always said before, I got no time/patience to sit around waiting for several months/years to get a sub par experience, this is why I have no sympathy for amd at times, it's not like they are some startup company with little investment.....

DLDSR has been a real game changer for me recently especially when paired up with upscaling.

This is something which many overlook too and quite frankly shows how little "insider" knowledge some really have on these things:

The technologies are only possible with NVIDIA demanding a premium to pay their RND and Software Teams.

I am not defending NVIDIA’s pricing but that is how they fund these research papers and new technologies.

In every single development project I've been on, the most expensive thing by far is the staff/employees on projects.


That video Move aside tressfx and hairworks

My most expensive card was my 3080. Even my 295x2 was cheaper (though that was used)

Same! More than I was willing to spend too but looking back, probably the best bargain for a long time and probably won't be another gpu to offer the same bang per buck for some time, well, maybe Jenson will surprise us with the 5080.....