Guldo 2024年2月20日 14時25分
3
THe RTX 5000 series is gonna be a flop and heres why.
If the leaks are true nvidia is coming out with a 16 pin power connector wich is devestating news for anyone that prefers nvidia over amd because that means these cards are going to be too big to fit in anyones case and over heating will also become an issue.

Now i can understand why amd is making no high end cards next gen because why even bother.

This definitly has to do with diminishing returns of shrinking the node as they arent getting the same uplift they used to so they have to make up the slack by making the cards bigger.
投稿主: PopinFRESH:
You are absolutely clueless. Smaller process notes doesn’t just equate to “more power”. Generally smaller process nodes result in less loss / power waste for a number of reasons. GPU power changes are a design choice balancing power:performance:die size. Hence why the same die on a smaller process node (e.g. a die shrink) generally use less power.

The actual graphics card for the 30 and 40 series is significantly smaller than earlier cards going back to the GTX 200 generation. My liquid cooled 3090 is tiny in comparison to my 2080, 1070, 970, 780 Ti etc. There are still a plethora of options across the product segments that don’t have large heatsinks on them.

“Is there a way to ditch the big heatsinks…” yeah, it is called a custom water cooled loop.
< >
136-150 / 234 のコメントを表示
_I_ の投稿を引用:
SuKI の投稿を引用:
All of those problems are non existent with an ATX 3 power supply and standard cables
no, its still the same plug on the gpu thats is the problem
No
Guldo 2024年2月23日 14時34分 
SuKI の投稿を引用:
Dutchgamer1982 の投稿を引用:

they would :

2200 euro = price 4090.
*380 euro = Vat
*90 euro = profitmargin retailer
*450 euro = cost to make a 4090
gross profit = 1280 euro.

thats an obsee profit margin

now if that 4090 chip gets undervolted to 250w to create a 4080ti that performs between a 4080S and 4090.. it likely can be made even cheaper.

if 700 :
120 vat
30 retail margin
450 priduction cost
100 gross profit

which would be a far more normal profitmargin.
Hahaahahahahahah!

4080 ti does not cost $450 to manufacture! It costs a lot more!
I thought it was called the 4080 super lol. i wasnt aware a 4080ti existed.
4080 Ti doesn't exist, and the 4080 SUPER wouldn't be it because it's basically just a 4080 with 5% more CUDA cores and performs within 5% faster as a result.

If they even add a 4080 Ti, would just be built off of bad 4090 bins.
Guldo 2024年2月23日 14時49分 
r.linder の投稿を引用:
4080 Ti doesn't exist, and the 4080 SUPER wouldn't be it because it's basically just a 4080 with 5% more CUDA cores and performs within 5% faster as a result.

If they even add a 4080 Ti, would just be built off of bad 4090 bins.
with how greedy nvidia is they might come out with a 4080ti.

im skeptical that they arent charging a huge premium for these cards because they have brand loyalty. how much do these cards really cost vs how much money they want for them.

wouldnt be surprised at all if the 4090 only costs like between 800 or 1000 dollars to produce but they want 1600 for it.
Kenshiro の投稿を引用:
r.linder の投稿を引用:
4080 Ti doesn't exist, and the 4080 SUPER wouldn't be it because it's basically just a 4080 with 5% more CUDA cores and performs within 5% faster as a result.

If they even add a 4080 Ti, would just be built off of bad 4090 bins.
with how greedy nvidia is they might come out with a 4080ti.

im skeptical that they arent charging a huge premium for these cards because they have brand loyalty. how much do these cards really cost vs how much money they want for them.

wouldnt be surprised at all if the 4090 only costs like between 800 or 1000 dollars to produce but they want 1600 for it.
The thing is, gaming GPUs stopped being their primary market quite awhile ago. They're making a lot more off of machine learning, because it reportedly only took them around $2000 to produce an H100 but they sell for $20000 per GPU, they make almost 1000% profit when it comes to AI aka machine learning. NVIDIA is worth trillions and Jensen himself is now the 21st most wealthy billionaire on the planet, at least in the official rankings that they keep public.

Which is exactly why they don't give a crap about the low end anymore, they can afford to let AMD and Intel fight over it, their main concern is the high end, which is why the 4090 was so beefed up from the 4080.

Not like AMD would be much better if the tables were turned. Corporations are in the game for money, that's literally their purpose. Growth of wealth, jobs, industry. Welcome to Capitalism.
最近の変更はr.linderが行いました; 2024年2月23日 14時58分
Kenshiro の投稿を引用:
SuKI の投稿を引用:
Hahaahahahahahah!

4080 ti does not cost $450 to manufacture! It costs a lot more!
I thought it was called the 4080 super lol. i wasnt aware a 4080ti existed.
4080 Super doesn't cost $450 to manufacture, either. It costs a lot more.
SuKI の投稿を引用:
Kenshiro の投稿を引用:
I thought it was called the 4080 super lol. i wasnt aware a 4080ti existed.
4080 Super doesn't cost $450 to manufacture, either. It costs a lot more.
For the material cost, it can easily be that low, but there are other factors in the total price, i.e. R&D, marketing, etc. And there have been rumours of the material cost by itself being in the ballpark of 300~400 USD based on the value of the process and VRAM, so unless you actually have proof of the contrary, I would just stop parroting in NVIDIA's defense.

The cost of manufacturing can be low, doesn't mean the price has to be too, because again, it's a business. The whole purpose of them producing and selling their product is to make profit, and Jensen has ambitions for NVIDIA, many others would be doing similar things in his shoes.
最近の変更はr.linderが行いました; 2024年2月23日 15時08分
Guldo 2024年2月23日 15時08分 
r.linder の投稿を引用:
Kenshiro の投稿を引用:
with how greedy nvidia is they might come out with a 4080ti.

im skeptical that they arent charging a huge premium for these cards because they have brand loyalty. how much do these cards really cost vs how much money they want for them.

wouldnt be surprised at all if the 4090 only costs like between 800 or 1000 dollars to produce but they want 1600 for it.
The thing is, gaming GPUs stopped being their primary market quite awhile ago. They're making a lot more off of machine learning, because it reportedly only took them around $2000 to produce an H100 but they sell for $20000 per GPU, they make almost 1000% profit when it comes to AI aka machine learning. NVIDIA is worth trillions and Jensen himself is now the 21st most wealthy billionaire on the planet, at least in the official rankings that they keep public.

Which is exactly why they don't give a crap about the low end anymore, they can afford to let AMD and Intel fight over it, their main concern is the high end, which is why the 4090 was so beefed up from the 4080.

Not like AMD would be much better if the tables were turned. Corporations are in the game for money, that's literally their purpose. Growth of wealth, jobs, industry. Welcome to Capitalism.
oh yea i agree and thats why i think brand loyalty is stupid. Thats true that they dont care about the mid range because they are making so much money off of AI.

what baffles me is people are still buying nvidia in the mid range even though they are getting a worse product than last gen for more money.

price to performance with nvidia in the midrange is absolutily terrible right now and the 4070 super doesnt have enough vram to be future proof.

12 gigs isnt gonna last the way things are going. If you dont care about ray tracing and dlss then the 7800xt is the superior option but the moment you say this nvidia fanboys get mad and say well the 4070super is faster wich might be true but 8% faster is nothing and you are sacrificing 4 extra gigs of vram for a tiny insiginicant performance uplift.
Kenshiro の投稿を引用:
r.linder の投稿を引用:
The thing is, gaming GPUs stopped being their primary market quite awhile ago. They're making a lot more off of machine learning, because it reportedly only took them around $2000 to produce an H100 but they sell for $20000 per GPU, they make almost 1000% profit when it comes to AI aka machine learning. NVIDIA is worth trillions and Jensen himself is now the 21st most wealthy billionaire on the planet, at least in the official rankings that they keep public.

Which is exactly why they don't give a crap about the low end anymore, they can afford to let AMD and Intel fight over it, their main concern is the high end, which is why the 4090 was so beefed up from the 4080.

Not like AMD would be much better if the tables were turned. Corporations are in the game for money, that's literally their purpose. Growth of wealth, jobs, industry. Welcome to Capitalism.
oh yea i agree and thats why i think brand loyalty is stupid. Thats true that they dont care about the mid range because they are making so much money off of AI.

what baffles me is people are still buying nvidia in the mid range even though they are getting a worse product than last gen for more money.

price to performance with nvidia in the midrange is absolutily terrible right now and the 4070 super doesnt have enough vram to be future proof.

12 gigs isnt gonna last the way things are going. If you dont care about ray tracing and dlss then the 7800xt is the superior option but the moment you say this nvidia fanboys get mad and say well the 4070super is faster wich might be true but 8% faster is nothing and you are sacrificing 4 extra gigs of vram for a tiny insiginicant performance uplift.
Their choice is valid, more valid than a mere 4GB of VRAM that they're not even going to be using in games any time soon, and by time we actually need that much VRAM consistently, neither the 4070 SUPER or 7800-XT will be particularly fast enough anyway, we're talking a few more generations down the line because 8GB is still enough in most cases, with 12GB at worst being needed 99% of the time.

RAM and VRAM are just capacity, unused capacity is effectively wasted capacity, you only need what you're actually going to use throughout the theoretical lifespan of that GPU in your system.

NVIDIA has the market share domination because they earned it by providing better product for years, AMD was playing catch-up.
最近の変更はr.linderが行いました; 2024年2月23日 15時12分
r.linder の投稿を引用:
SuKI の投稿を引用:
4080 Super doesn't cost $450 to manufacture, either. It costs a lot more.
For the material cost, it can easily be that low, but there are other factors in the total price, i.e. R&D, marketing, etc. And there have been rumours of the material cost by itself being in the ballpark of 300~400 USD based on the value of the process and VRAM, so unless you actually have proof of the contrary, I would just stop parroting in NVIDIA's defense.

The cost of manufacturing can be low, doesn't mean the price has to be too, because again, it's a business. The whole purpose of them producing and selling their product is to make profit, and Jensen has ambitions for NVIDIA, many others would be doing similar things in his shoes.
YOU NEVER PROVIDED A SINGLE BIT OF PROOF.
Guldo 2024年2月23日 15時12分 
r.linder の投稿を引用:
Kenshiro の投稿を引用:
oh yea i agree and thats why i think brand loyalty is stupid. Thats true that they dont care about the mid range because they are making so much money off of AI.

what baffles me is people are still buying nvidia in the mid range even though they are getting a worse product than last gen for more money.

price to performance with nvidia in the midrange is absolutily terrible right now and the 4070 super doesnt have enough vram to be future proof.

12 gigs isnt gonna last the way things are going. If you dont care about ray tracing and dlss then the 7800xt is the superior option but the moment you say this nvidia fanboys get mad and say well the 4070super is faster wich might be true but 8% faster is nothing and you are sacrificing 4 extra gigs of vram for a tiny insiginicant performance uplift.
Their choice is valid, more valid than a mere 4GB of VRAM that they're not even going to be using in games any time soon, and by time we actually need that much VRAM consistently, neither the 4070 SUPER or 7800-XT will be particularly fast enough anyway, we're talking a few more generations down the line because 8GB is still enough in most cases, with 12GB at worst being needed 99% of the time.

RAM and VRAM are just capacity, unused capacity is effectively wasted capacity, you only need what you're actually going to use throughout the theoretical lifespan of that GPU in your system.
i have played plenty of games on my 3070 where 8 isnt enough. games are becoming vram hogs these days but it depends on what you are playing.
SuKI の投稿を引用:
r.linder の投稿を引用:
For the material cost, it can easily be that low, but there are other factors in the total price, i.e. R&D, marketing, etc. And there have been rumours of the material cost by itself being in the ballpark of 300~400 USD based on the value of the process and VRAM, so unless you actually have proof of the contrary, I would just stop parroting in NVIDIA's defense.

The cost of manufacturing can be low, doesn't mean the price has to be too, because again, it's a business. The whole purpose of them producing and selling their product is to make profit, and Jensen has ambitions for NVIDIA, many others would be doing similar things in his shoes.
YOU NEVER PROVIDED A SINGLE BIT OF PROOF.
Neither did you, but I said there were rumours, a five second google search can pull that up. Rumours aren't proof, just speculation, and nobody outside of NVIDIA and others privy to that kind of data actually has that information. So again, unless you can actually prove that the cost to produce is much higher as you claim, I would just stop talking.
Guldo 2024年2月23日 15時14分 
Unfortunately, no. 8 GB of vRAM is no longer enough, even for 1080p gaming, let alone 1440p or 4K. Recently released AAA titles like The Last of Us, Hogwarts Legacy, Resident Evil 4, and Ratchet & Clank – Rift Apart can easily consume up to 12 GB of VRAM at 1080p on medium to high graphics settings with ultra-quality textures.

straight from bing.
r.linder の投稿を引用:
SuKI の投稿を引用:
YOU NEVER PROVIDED A SINGLE BIT OF PROOF.
Neither did you, but I said there were rumours, a five second google search can pull that up. Rumours aren't proof, just speculation, and nobody outside of NVIDIA and others privy to that kind of data actually has that information. So again, unless you can actually prove that the cost to produce is much higher as you claim, I would just stop talking.
You're telling me to stop talking when you offer nothing as proof.

That's just embarrassing, honestly.
Kenshiro の投稿を引用:
r.linder の投稿を引用:
Their choice is valid, more valid than a mere 4GB of VRAM that they're not even going to be using in games any time soon, and by time we actually need that much VRAM consistently, neither the 4070 SUPER or 7800-XT will be particularly fast enough anyway, we're talking a few more generations down the line because 8GB is still enough in most cases, with 12GB at worst being needed 99% of the time.

RAM and VRAM are just capacity, unused capacity is effectively wasted capacity, you only need what you're actually going to use throughout the theoretical lifespan of that GPU in your system.
i have played plenty of games on my 3070 where 8 isnt enough. games are becoming vram hogs these days but it depends on what you are playing.
I have a 3080 10G, which is just enough to run without any issues with VRAM capacity in the majority of games at ultra settings, nobody needs more than 12 for the foreseeable future. People who are always maxing out every single graphics setting in the latest AAA titles are going to naturally see higher VRAM usage, but the performance in those games with most cards isn't even great to begin with.
最近の変更はr.linderが行いました; 2024年2月23日 15時17分
< >
136-150 / 234 のコメントを表示
ページ毎: 1530 50

投稿日: 2024年2月20日 14時25分
投稿数: 234