Next Gen GPU's will be even more expensive

[Krit]

Is the 5nm-4nm the end of the road forever?

We need new competition back Samsung or Global Foundries.

 

[evernessince]

Expensive is the right word..... TSMC is overpricing it's technology because of monopoly.

Also due to yield issues the node has. Those yield issues are magnified by large GPUs dies as you loose a greater proportion of usable wafer per defect. One can hope Samsung catches up but they've had even more issues on 3nm than TSMC.

 

[eidairaman1]

We need new competition back Samsung or Global Foundries.

Tsmc is in a way a monopoly for chip fab

 

[Sir Beregond]

We need new competition back Samsung or Global Foundries.

God help us. The most advanced node GloFo is working with is 12LP+. They are not even close to any of the advanced stuff TSMC is putting out.

As for Samsung, their 3nm and 2nm yields are abysmal.

 

[3valatzy]

DOA generation.
RTX 5080 only 50% of RTX 5090 is a disaster. The worst is yet to come.

ASUS GeForce RTX 5080 and Radeon RX 9070 XT custom GPU names leaked, 16GB memory confirmed - VideoCardz.com

ASUS GeForce RTX 5080 and Radeon RX 9070 XT ASUS is set to launch new graphics cards in ROG (Republic of Gamers) series. It looks like gamers will have a new option in the ASUS graphics card lineup. A product called GeForce RTX 5080 ROG ASTRAL has been listed by several retailers and...

videocardz.com

 

[Zach_01]

TSMC is N°10 on top companies of the planet...

Companies ranked by Market Cap - CompaniesMarketCap.com

Ranking the world's top companies by market cap, market value, revenue and many more metrics

companiesmarketcap.com

 

[evernessince]

@evernessince
except:
groceries = need
gpu = want

dont have to buy the latter.

Just wanted to add a bit my last reply (shown below for clarity):

GPUs are used in medical, engineering, scientific, AI, streaming, and Computer rendering fields.

For a large number of professions they are in fact a need and their pricing will impact you in one way or another as GPUs are increasingly an necessary element of the modern world. That's whether you use one personally or use a service that uses one (like a medical imaging device).

Mind you, whether a product is a want or a need isn't a rebuttable as to the pricing increase of a product. It's a regressive argument that doesn't address the topic, only avoids it. You could argue you don't need anything but food, water, and sleep. Doesn't change arguments for or against pricing of any product.

In addition to what I said above; if you were to utilize the logic you are using here, in that any pricing is fine so long as you don't physically need it to live, the result is being permissible to nearly every monopoly and predatory pricing maneuver that has transpired over the markets. Do you physically need telephone service? Nope, guess that makes the Bell Systems monopoly ok. What about steel? Railroads? Monitors or system memory? This list could go on forever on monopolies that would have been fine under your standard.

The definition of "needs" is a tricky one, everyone has a different bar but just because something isn't a need for you doesn't mean it shouldn't be fairly priced. The more something is used or integrated across industries and the populace, the more it's in the best interest of the people to ensure reasonable prices. It is not a need sure but goes to advance society. If we accepted the telephone monopoly as it was we'd be paying astronomical amounts for cell service today that many consider an essential service. I believe that GPUs with their increasing uses should come under scrutiny as such.

 

[freeagent]

Groceries increased some 28%, not the 400% we've seen in GPUs. A 970 cost $330 and would get you 78% of flagship performance. Rumored 5080 is likely to be around 70% flagship performance for $1,370. That's before you consider that the performance gap between the two (like the 4080 vs 4090) will only widen over time due to the 5080 only having half the die size, 16GB of VRAM, and a tiny bus for such an expensive GPU.

This isn't the cause of inflation, it's the result of a complete monopoly with Nvidia exceeding even Bell System's marketshare at the height of their power.

Of course it is, but at the same time, what is the cost of the grocery list? That isn't even the expensive part, that would be all the machine time.

Edit:

No matter what you have made for you, usually the machine time, and engineering time are what you pay for, the materials are the cheapest part.

Edit again:

Then you have to have some profit.

 

[Krit]

Interesting perspective that typical nvidia user should listen to. Especially now when new expensive RTX 50 series drops.

I stumbled on this video quite accidentally and agree a lot.

 

[Hecate91]

It isn't just some profit, Nvidia is profiting on about 60% margins, BOM cost has to be pretty cheap too, at least on the x70 and below sku's.
Also gaming products are just money on the side for Nvidia, probably some defective dies that would otherwise be extremely expensive datacenter hardware.

 

[3valatzy]

The new 5080 is a 5060 Ti
5070 is 5050 Ti
5060 is 5030.

Groceries increased some 28%, not the 400% we've seen in GPUs. A 970 cost $330 and would get you 78% of flagship performance. Rumored 5080 is likely to be around 70% flagship performance for $1,370. That's before you consider that the performance gap between the two (like the 4080 vs 4090) will only widen over time due to the 5080 only having half the die size, 16GB of VRAM, and a tiny bus for such an expensive GPU.

This isn't the cause of inflation, it's the result of a complete monopoly with Nvidia exceeding even Bell System's marketshare at the height of their power.

Half the shaders, actually, which is the most important performance hit. The card will be heavily handicapped, and the lowest value.

But, this product line is a very high risk for Nvidia. If AMD manages to release the RX 9070, RX 9060 XT and RX 9060 very competitively in the price range 149-249, Nvidia is game over.

Nvidia is game over because its new RTX 5060 will be slower than the new integrated GPUs made by AMD.

 

[igormp]

It doesn't exactly work like that on Linux - there is no executable that you just click to install. I'm not exactly experienced with package managers, and with AMD, I don't even need to be.

Nvidia does provide an executable that you can just double click from their website (although that's not the ideal way), but as I said, I understand your point: not having to do anything at all is way easier, no matter what you were used to.

 

[agent_x007]

If you want margins/profits, 1600$ USD for 5080 seems like good compromise (from company money maker perspective).
NV may be simply increasing price to squeeze more from early buyers (before tariffs hit in second half of January). Market should adjust price based on demand (as always), it will just take a quarter or two of 2025.

But, this product line is a very high risk for Nvidia. If AMD manages to release the RX 9070, RX 9060 XT and RX 9060 very competitively in the price range 149-249, Nvidia is game over.

Nvidia is game over because its new RTX 5060 will be slower than the new integrated GPUs made by AMD.

Assuming performance is where they claim it will be, what game over are you talking about ?
AMD cards can only as fast 4080 Super/7900 XTX.
If you want/need something faster you WILL buy Nvidia card.
There is no Crossfire/SLI to help cheaper cards to get higher performance (this is 2x1 =/= 2 moment).
If you are fine with AMD level of performance, NV does get you options (xx70/xx60), but you may not take it (depending on actual pricing). NV however won't care though, because margins on top end are simply ridiculous (on top of using cut dies for this vs. full ones that probably will go AI stuff with even bigger margins).

 

[3valatzy]

Assuming performance is where they claim it will be, what game over are you talking about ?
AMD cards can only as fast 4080 Super/7900 XTX.
If you want/need something faster you WILL buy NVidia card.

No, you don't buy Nvidia. AMD's cards could do RTX 4080 Super, but the price will be 40% of that RTX 4080 Super.
Look at the prices, not the performance.

RTX 5090 ~ $3000
RTX 5080 which is in fact RTX 5060 Ti - $1600
RTX 5070 Ti which is in fact RTX 5050 Ti - $1200
RTX 5070 which is in fact RTX 5050 - $900
RTX 5060 which is in fact RTX 5030 - $600

If you are fine with AMD level of performance

This is games native, all Nvidia shenanigans and seemingly "higher performance" is because they have cheating technologies enabled in the driver as a default setting - DLSS.

 

[agent_x007]

No, you don't buy Nvidia.... Look at the prices, not the performance.

Actually, I personally buy both AMD and NV (I also got A770 recently too), but that's beside the point

"Gamers" first check performance, after that price - it's why 4090 is thing that we have "pleasure" of buying.

As for chip vs. name - not sure why it's relevant if 5030 chip is called 5090 in grand scheme of things ?
You pay for performance you get. Don't like it - don't buy it.
Telling me it's "bad for consumers" or that "NV is holding back" - doesn't mean anything (insert old man yelling at cloud meme). You can't change what a company like NV or AMD does (as I doubt you are a share holder in any of them).

PS. As for "cheats" you conveniently forgot about FSR which works with AMD cards (but not only them).
Everyone can "cheat", isn't democracy great ?
Note : Some testers/reviewers use native no DLSS/FSR to get results, which makes your point moot (since you can check what is "non cheat" performance of everything).

 

[sn2x]

The new 5080 is a 5060 Ti
5070 is 5050 Ti
5060 is 5030.

If that's the case, the upcoming RX 9070 XT could be competing against what you call 5050-tier.

 

[Hecate91]

DOA generation.
RTX 5080 only 50% of RTX 5090 is a disaster. The worst is yet to come.

View attachment 377506

ASUS GeForce RTX 5080 and Radeon RX 9070 XT custom GPU names leaked, 16GB memory confirmed - VideoCardz.com

ASUS GeForce RTX 5080 and Radeon RX 9070 XT ASUS is set to launch new graphics cards in ROG (Republic of Gamers) series. It looks like gamers will have a new option in the ASUS graphics card lineup. A product called GeForce RTX 5080 ROG ASTRAL has been listed by several retailers and...

videocardz.com

It might be DOA for some buying a 5080 or 5090 for gaming, as when even enthusiasts start to question the high pricing thats just how broken the market is.
Although, as Vex points out in their latest video, according to someone at Nvidia the 5080 and 5090 weren't intended for gamers. Businesses will probably be buying up 5090's for AI as its still cheaper than buying dedicated compute cards.

No, you don't buy Nvidia. AMD's cards could do RTX 4080 Super, but the price will be 40% of that RTX 4080 Super.
Look at the prices, not the performance.

RTX 5090 ~ $3000
RTX 5080 which is in fact RTX 5060 Ti - $1600
RTX 5070 Ti which is in fact RTX 5050 Ti - $1200
RTX 5070 which is in fact RTX 5050 - $900
RTX 5060 which is in fact RTX 5030 - $600



This is games native, all Nvidia shenanigans and seemingly "higher performance" is because they have cheating technologies enabled in the driver as a default setting - DLSS.

I'm not so sure the 5080 would really be a 5060Ti, it probably should be a 5070Ti since the 5080 has half the shaders of a 5090, and the 5070Ti is heavily cut down from the 5080.
Also I'm not defending how Nvidia has totally shifted around the product stack, it was really upsetting how much Nvidia nerfed the 40 series on the xx70Ti and below, like the 4070Ti should've been the 4070 and the 4070 should've been a 4060. If the 5060 comes with only 8GB of VRAM then it shouldn't exist at all, selling an 8GB card is a complete joke.
And I have to wonder how the AI DLSS is going to work, but if the frames are totally made up then it should be considered cheating, at least in multiplayer games.

 

[DemonicRyzen666]

If you want margins/profits, 1600$ USD for 5080 seems like good compromise (from company money maker perspective).
NV may be simply increasing price to squeeze more from early buyers (before tariffs hit in second half of January). Market should adjust price based on demand (as always), it will just take a quarter or two of 2025.


Assuming performance is where they claim it will be, what game over are you talking about ?
AMD cards can only as fast 4080 Super/7900 XTX.
If you want/need something faster you WILL buy Nvidia card.
There is no Crossfire/SLI to help cheaper cards to get higher performance (this is 2x1 =/= 2 moment).
If you are fine with AMD level of performance, NV does get you options (xx70/xx60), but you may not take it (depending on actual pricing). NV however won't care though, because margins on top end are simply ridiculous (on top of using cut dies for this vs. full ones that probably will go AI stuff with even bigger margins).

Not entirely true ALL AMD based RDNA cards support mGPU. everything from RX 5,000 to RX 7,000 support it & with two cards install you just lose "Rebar" when enabling it.
Nvidia is the one is lacking on the support for mGPU. I have seen RTX 4090 in mGPU in a game & at 4K no need for any D.L.S.S & you can use better AA than crap TAA.
The other lack is of mGPU games makes no sense. Since that most games are still going to be GPU bound anyways. Even the RTX 5090 will be GPU bound by UE5 being such a terrible built & designed game engine.
Also Intel ARC cards are suppose to support mGPU too, no one has tested it.

If anyone would like a list of games to test mGPU on with dual gpu's Just PM the list is only about 10 games but they play a lot better than Crossfire & S.L.I games did.

 

[igormp]

If anyone would like a list of games to test mGPU on with dual gpu's Just PM the list is only about 10 games but they play a lot better than Crossfire & S.L.I games did.

Now I'm curious, do any of those run on linux? If so, I could give it a try with my pair of 3090s just for the fun of it.

 

[DemonicRyzen666]

Now I'm curious, do any of those run on linux? If so, I could give it a try with my pair of 3090s just for the fun of it.

Seems like only two support Linux, one just does uses DX11 to OpenGL draw calls for it, most the mGPu support is for windows.

 

[tfdsaf]

In 8 years of RT "revolution" we barely have several good'ish games and even in them RT adds significant artefacts and/or blur and still tanks performance.

Second even in CP 2077 with the latest expansion and patches, RT only looks different in most scenarios. Its not really clear which image is clearly more appealing and better, it usually comes down to individual preference. Though I'd give it to you that in very high RT games every surface looks like a mirror or a wet floor, so there is quite a bit of bling and neons and colors. If you are into that UNREALISTIC look you'd love RT, if you are into realistic or developer artistry you'd hate RT.

 

[Sir Beregond]

RTX 5080 which is in fact RTX 5060 Ti - $1600
RTX 5070 Ti which is in fact RTX 5050 Ti - $1200
RTX 5070 which is in fact RTX 5050 - $900
RTX 5060 which is in fact RTX 5030 - $600

That ship long since sailed away. You need to go back to 2012 and ask why all the review publications gave the GTX 680 a good review. Many noted it was the smaller silicon typically used for previous gen 60-class cards, but gave it a good review anyway...including TPU. Since that time, with the exception of the 780 and 3080, the base 80-class product (non-Ti) has been on the step down die from the flagship in every single generation since the 600-series. In that sense, the 4080 and 5080 are absolutely no different. The 70-class cards on the other hand are definitely a step down from what they were, because go back to Kepler, Maxwell, Pascal, and even Turing Super refresh, and the 70-card was a cut-down of the 80-card on the same silicon. In that sense, the 4070 Ti Super, and rumored 5070 Ti are kind of keeping with that, but note they are suddenly the "Ti" and not just the base 70.

And with all that, what is different, is the price.

You went from the GTX 580 on GF110 silicon (520mm2) to the GTX 680 on GK104 silicon (294mm2). By comparison, the GTX 560 Ti was on the Fermi generation's equivalent GF114 silicon (332mm2). The difference? Oh the GTX 560 Ti was $249. The GTX 680? $499. And absolutely everyone gave it a glowing review and let Nvidia get away with it.

And thus it all started.

 

[oxrufiioxo]

That ship long since sailed away. You need to go back to 2012 and ask why all the review publications gave the GTX 680 a good review. Many noted it was the smaller silicon typically used for previous gen 60-class cards, but gave it a good review anyway...including TPU. Since that time, with the exception of the 780 and 3080, the 80-class has been on the step down die from the flagship in every single generation since the 600-series. In that sense, the 4080 and 5080 are absolutely no different.

What is different, is the price.

That was definitely the beginning of the end. It was also the generation we got the Titan which Nvidia used to test what gamers were willing to pay even though they split it in 2 with the 600/700 being Kepler based cards other than some low end options being Maxwell.

Don't get me wrong I'm part of the crowd that made it possible but after the last couple generations even I'm fatigued by it.

 

[Neo_Morpheus]

I know that you directed that to me, since every time I post something, you reply with such nonsensical answers.

But I will byte this time and feel free to delete my post.

I don't understand all the dislike for RT either.

Its ok, the world doesnt revolve around you, some people like it, some dont.
In my case, I simply said as many others had, that we dont see the big deal of RT to justify the performance hit.
That doesnt mean that others need to stop using it, we simply dont see the reason.

Read it and keep going.

ooks great. Same with all the talk of DLSS..

Again, as said before, many prefer native.
And, again in my case, I hate ANY tech that exists to keep locked into one vendor.
Again, thats my prerogative and I am not imposing it on anyone.
The white knights are the one that always come out all offended and rowdy.

you don't have to use it..

Even if I wanted to use it, I cant because its not open to anyone. Even among the loyal Ngreedia customers.

its just there if you want.

See above.

Like how I can use the AMD version.. but I don't.

Your choice and its ok if you dont want to.

People will buy what they buy, on both sides. Does it really matter?

Yet here you are, all bothered because little ol'me dont want to give Ngreedia any money.

It doesn't bother me that people use AMD GPUs, but it bothers some people that I use an Nvidia GPU?

Yet, you created the post based on my comment about not liking Ngreedia.
What did you said once? Oh thats, right, leave feelings out it?

I don't even have a leather jacket.

It will eventually come with more displays of loyalty.

I dont call anyone out, I dont care whatever anyone is buying with their money.

They are not my problem, my problem is the current situation that we all are experiencing because of the Ngreedia monopoly.

I am simply expressing my opinion and its up to everyone else to analyze if its valid or not and use it in a constructive way..or not.

 

[nguyen]

That was definitely the beginning of the end. It was also the generation we got the Titan which Nvidia used to test what gamers were willing to pay even though they split it in 2 with the 600/700 being Kepler based cards other than some low end options being Maxwell.

Don't get me wrong I'm part of the crowd that made it possible but after the last couple generations even I'm fatigued by it.

I'm quite sure you are not part of the crowd that enabled Nvidia dominating the market at all, you bought like what, 1 Nvidia GPU in the past 2.5 years? And also an Radeon too.

No point getting too work up over some toys anyways, like buying LEGO and then listen to people complain about why it's bad for LEGO smaller competitors