• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

NVIDIA 4000 Series

Associate
Joined
12 Jun 2021
Posts
1,663
Location
Leeds
What would a good bump be compared to a 3080 performance wise?

The 4000 series has an excellent generational performance increase and with lower power consumption as well. The only problem is the price and how Nv has pushed each tier of product up a notch and is making the consumer pay top dollar for them.

https://www.techpowerup.com/gpu-specs/geforce-rtx-3080.c3621 - Key point Die Size is 628mmsquared on Samsung 8nm

https://www.techpowerup.com/gpu-specs/geforce-rtx-4080.c3888 - Die Size 379mmsquared on TSMC 5nm

Now the 4080 has a decent performance increase over the 3080, approx 50%, and the production costs should not be as wildly different as the two MSRPs suggests. Nvidia has massively pushed up the margins they want for their gpus and that is the problem. The 3080 was great value (if you could get your hands on one) whilst the 4080 and the other 4000 series cards are poor value in comparison.
 
Last edited:
Soldato
Joined
16 Sep 2018
Posts
12,727
If next gen is 2025 maybe will get a good bump in performance especially mid to lower end via a 4000 series refresh this year, hopefully giving gamers what the 4000 series should have been. I think NV will want to boost sales mid-life if it's gonna be dragged out to a three year life
Why wait so long, they could lower prices tomorrow if they wanted. The 4090 is already performant so a good bump in performance on that before 2025 seems a bit pointless, the problem is the prices not the performance.
 
Soldato
Joined
19 Oct 2008
Posts
5,954
It's been done to death I'm sure - the value of the rest of the range compared to the 4090 which itself has a lower core count % of the full fat chip than the 3090 had a GPU generation ago. The 4090 in that regard itself seems more like a 4085 (or 4080 Ti). To buy a 4080 you might as well just buy the 4090 as the specs are much better.
A 4080 Ti with the 4090 specs (core count at least) and then a 4090 Ti a notch up, both at same price levels as the current 4080 and 4090 would be an improvement. And of course do something with the rest of the range

The prices are the problem I agree, more accurately the price to performance vs the 4090. The 4090 is a great GPU already and seems like the only one worth buying which I think will change when/if they do a mid-life refresh.
 
Last edited:
Soldato
Joined
30 Jul 2012
Posts
2,774
Personally it is Microsoft who are sticking their noses in on FSR exclusivity for obvious reasons. You honestly think Bethesda would do this if they were still independent, nope.
 
Soldato
Joined
20 Aug 2019
Posts
3,035
Location
SW Florida
My most optimistic expectation (if sales stay as bad as they have been) is a 4080Ti with 20GB, 320bit bus, at $1200. (and lower the 4080 price below 1k)

But if Nvidia is content to just let their stuff collect dust, they could just maintain course and cut off their nose to spite their face.
 
Last edited:
Associate
Joined
16 Aug 2017
Posts
1,155
Location
London
To buy a 4080 you might as well just buy the 4090 as the specs are much better.
And this is exactly what NVIDIA communicates all the time - just buy their top GPU, give them the most monies. Their whole lineup is set up this generation to push people towards higher and higher SKUs. Very smart move for them, very bad for consumers. The more you buy the more you save. ;)
 
Man of Honour
Joined
13 Oct 2006
Posts
91,999
And this is exactly what NVIDIA communicates all the time - just buy their top GPU, give them the most monies. Their whole lineup is set up this generation to push people towards higher and higher SKUs. Very smart move for them, very bad for consumers. The more you buy the more you save. ;)

That is what I've thought - the whole line up seems intended to make people buy 1 up from their normal tier but using the stick instead of carrot - very cynical :( though I don't think it is the only reason for the messed up model names and prices.
 
Associate
Joined
16 Aug 2017
Posts
1,155
Location
London
My most optimistic expectation (if sales stay as bad as they have been) is a 4080Ti with 20GB, 320bit bus, at $1200. (and lower the 4080 price below 1k)

But if Nvidia is content to just let their stuff collect dust, they could just maintain course and cut off their nose to spite their face.
But their stuff do not collect dust anymore. They produce just enough 4090s to cover the market, the rest of the expensive silicone goes to Enterprise products (AI cards and the likes). They simply do not seem to care for gaming market anymore, as that's not where the real money is. And lower end products are in big part just failed silicon that couldn't be used for Enterprise grade products, so is being sold to consumers with still good margins instead of throwing it out. In other words, I doubt they will lower prices anytime soon - last I heard Jensen talking he was saying 4090 is way too cheap and this is the last time 90 series card will be that cheap. :p
 
Associate
Joined
12 Jun 2021
Posts
1,663
Location
Leeds
last I heard Jensen talking he was saying 4090 is way too cheap and this is the last time 90 series card will be that cheap. :p

Jensen is a great CEO but that sales carp only works for the very very best card. The best card can and does command a price premium and there are always enthusiasts who will buy the best no matter the cost and this was true , the 4090 sold very well.

You know what did not sell very well ? The 4080 the 4070ti the 4070 the 4060ti and now the 4060. The most sales come at a products launch but the yields and profit per sale can increase over time and all the consumer who wants to upgrade from their 970 , 1070 1060 1080 2060 is going to hear is how poor the value of current 4000 series cards are, not going to make them want to upgrade and many will just buy a console instead. If it is true and the next gen 5000 series cards are not out until 2025 then Nv has 18 months of TERRIBLE sales. I can see a 50% reduction in gpu sales over the next year.

They tried the same rubbish with the 2000 cards and they sold atrociously. Overpriced cards with features that were at that point in time worthless.
 
Associate
Joined
16 Aug 2017
Posts
1,155
Location
London
Jensen is a great CEO but that sales carp only works for the very very best card. The best card can and does command a price premium and there are always enthusiasts who will buy the best no matter the cost and this was true , the 4090 sold very well.

You know what did not sell very well ? The 4080 the 4070ti the 4070 the 4060ti and now the 4060. The most sales come at a products launch but the yields and profit per sale can increase over time and all the consumer who wants to upgrade from their 970 , 1070 1060 1080 2060 is going to hear is how poor the value of current 4000 series cards are, not going to make them want to upgrade and many will just buy a console instead. If it is true and the next gen 5000 series cards are not out until 2025 then Nv has 18 months of TERRIBLE sales. I can see a 50% reduction in gpu sales over the next year.

They tried the same rubbish with the 2000 cards and they sold atrociously. Overpriced cards with features that were at that point in time worthless.
The big difference is that now they simply do not care - AI/enterprise market is the main source of profit and potentially much more stable and long-term than crypto booms. Gamers pay pennies in comparison (yes, even for 4090), so it's just "by the way" market now, for NVIDIA.
 
Last edited:
Associate
Joined
12 Jun 2021
Posts
1,663
Location
Leeds
The big difference is that now they simply do not care - AI/enterprise market is the main source of profit and potentially much more stable and long-term than crypto booms. Gamers pay pennies in comparison (yes, even for 4090), so it's just "by the way" market now, for NVIDIA.

At the moment AI is the next big thing but Nv tried selling gpus to car makers for self driving cars and that soon died out.

The AI bubble will burst and Nv will come crawling back to their core market of selling gpus. A $30 billion Dollar yearly market is nothing to be thrown away lightly.

I may be wrong but I see this AI craze fizzling out quickly. I do not mean so called AI or the applications that it runs, I mean the profit that is currently on offer. What they spout as AI is actually just enhanced statistical analysis and you can easily build ASIC style chips to run those algorthyms once the software is ironed out. The profit margin for AI will go down very very quickly imho.
 
Associate
Joined
16 Aug 2017
Posts
1,155
Location
London
At the moment AI is the next big thing but Nv tried selling gpus to car makers for self driving cars and that soon died out.

The AI bubble will burst and Nv will come crawling back to their core market of selling gpus. A $30 billion Dollar yearly market is nothing to be thrown away lightly.

I may be wrong but I see this AI craze fizzling out quickly. I do not mean so called AI or the applications that it runs, I mean the profit that is currently on offer. What they spout as AI is actually just enhanced statistical analysis and you can easily build ASIC style chips to run those algorthyms once the software is ironed out. The profit margin for AI will go down very very quickly imho.
AI needs more and more computational power the better it gets - that's not going to change and we are far far away from hitting diminishing returns. I don't see it fizzling out anytime soon, unless governments step in and try to limit AI development as it is a dangerous tech. Not the ChatGPT kind, as that one is really stupid, but we're moving towards AGI, which will be a totally different beast. However, even the kind of ChatGPT requires loads of dedicated hardware and many companies want it running internally (to prevent data leaks to the outside world).

That said, NVIDIA has limited production capacity and most of their silicon seems to be redirected towards AI etc. these days. Left overs go to consumer's market and low sales are actually even better for NVIDIA in such case (less to store, not much to gather dust, higher margins still etc.). In ideal world, with unlimited production capacity they would care more, but with existing limits they definitely prefer to sell what they can to enterprise/AI than to gamers, as margins are hugely bigger in the former.
 
Last edited:
Associate
Joined
12 Jun 2021
Posts
1,663
Location
Leeds
I would say do not drink the cool aid they are trying to sell.

There is a difference between developing true AI using cubits and what they are calling AI. Most of the AI applications , I really hate calling it AI , is just using a computer to crunch numbers really really fast...erm that is what computers have always done.

What is now needing massive parallell computing will within a short period of time be reduced to a smart phone accesing the correct algorythms.IMO Jensen knows this and is trying to maximise profits whilst everyone is on the bandwagon, he is a business man and not a tech visionary.

True AI is a completely different animal and who knows if we will see it within our lifetime.

This is just my opinion so feel free to disagree and buy Nv stock if you like.
 
Soldato
Joined
26 Aug 2004
Posts
5,053
Location
South Wales
True AI is a completely different animal and who knows if we will see it within our lifetime.

This is just my opinion so feel free to disagree and buy Nv stock if you like.
John Carmack said a few years ago he believes an early AGI is possible within a decade, Ray Kurzweil has also been talking about it long enough.

But yeah everyone has their own opinion on it, I trust Carmack and Kurzweil's opinion myself.
 
Associate
Joined
12 Jun 2021
Posts
1,663
Location
Leeds
John Carmack said a few years ago he believes an early AGI is possible within a decade, Ray Kurzweil has also been talking about it long enough.

But yeah everyone has their own opinion on it, I trust Carmack and Kurzweil's opinion myself.

They said in the 80s we would have space lasers to shoot down nuclear missiles. I think we are still waiting for that particular Star Wars project to develop.

I am a pessimist and until I can see it , touch it and actually utilise it I will doubt its existence. Until then it is vaporware and meh.

Bit political here but Water privitisation happened in 87 and it took until 2023 for the majority to see it was a total con. My natural pessimision has been fed far more than my optimism over my lifetime so I tend to be a grumpy old git.
 
Associate
Joined
12 Jun 2021
Posts
1,663
Location
Leeds
Back
Top Bottom