TheRaidenPT
Member
If they do go for 3D cache with the PS6 I think it will be a beefy upgrade.
ars technica "Chips aren’t improving like they used to, and it’s killing game console price cuts"
- Thread starter ArtHands
- Start date
REDRZA MWS
Member
If chips aren't advancing that much, they shouldn't cost more, so neither should consoles.
PeteBull
Member
Chips aren’t improving like they used to, and it’s killing game console price cuts
Op-ed: Slowed manufacturing advancements are upending the way tech progresses.arstechnica.com
twilo99
Member
At current prices the Mac mini m4 makes even more sense as a gaming machine but studios just don't care to make games for that gorgeous ARM based hardware…
Apple are missing a golden opportunity here because console manufacturers are able to convince their customers that buying a console based on some archaic x86 architecture from 5 years ago at a premium by sticking "pro" on it or simply raising their prices is actually a good deal…
It blows my mind that people are actually buying zen2 architecture CPUs in the middle of 2025 for gaming purposes
poppabk
Cheeks Spread for Digital Only Future
Thats what I said?
Console - same performance target for 5+ years.
Reduced chip advancments = no price reduction
PC - performance target increases roughly every two years
Reduced chip advancements = smaller and smaller jumps in performance
ZehDon
Member
Due to inflation, if nothing changes, they'll just cost more through organic market forces over time. Add the current economic turmoil caused by tariff ideocracy, and it's pretty easy to explain why the hardware hasn't gone down in price. Now, going up in price is something else - Sony in particular has been pretty open that the PS5 underwent multiple revisions since launch to achieve hardware profitability. They could make less profit - still a profit - but have elected not too, passing the cost on to the consumers. The same is true of Nintendo. The one outlier is Microsoft; their hardware is still sold at a loss - so if anyone can justify raising their prices, it's the green team. But Microsoft are a trillion dollar company and aren't selling much hardware anyway, so, frankly, they could eat the costs if they wanted to.
Rosoboy19
Member
At this point backwards compatibility is expected.
If PS6 has a 3D cache CPU and UDNA graphics it will be a nice step forward. Not a leap, but a nice step.
Last edited:
U are technically right, but those 3nm chips apple is using in their phones are mobile low powerdraw ones console gonna need around 300mm² with 250W powerdraw.
Those need different node compared to small mobile chips apple is using.
AMD Playstation 5 Pro GPU Specs
AMD Viola, 2350 MHz, 3840 Cores, 240 TMUs, 64 ROPs, 16384 MB GDDR6, 2250 MHz, 256 bitwww.techpowerup.com
https://en.wikichip.org/wiki/3_nm_lithography_process#google_vignette here some deep down info about it.
What prevents anyone from making a large die or high power chip on the current 3nm nodes? No doubt the HP-optimized libs (N3P/N3X) will be better but Intel ships Arrow Lake with 250W TDP on N3B and Apple ships M4 Max with >100B transistors which is like ~500sqmm on N3E right now.
There's still innovations to be found, ai will also speed up new discovery. This one the other week is pretty nuts incase people missed it:
www.livescience.com
China's '2D' chip could soon be used to make silicon-free chips
Advances in materials and architecture could lead to silicon-free chip manufacturing thanks to a new type of transistor.
ArtHands
Thinks buying more servers can fix a bad patch
The rational is that the advance is costlier than before.
diffusionx
Gold Member
PS5 has already gone up in price in every region but the US. It happened a few years ago. It seems Sony intentionally tried to keep the US away from this and instead decided to just put us through the pain with the Pro.
Sony is raising the price of the PS5
Price increase affects almost all territories, but not the US
www.polygon.com
Price and manufacturing capacity. Whenever a new node comes in, Apple swoops in with their gazillions of dollars to make sure their chips have it first.
I am not sure the juice is worth the squeeze on PC parts. It makes a lot of sense on a cell phone where every little bit matters but Intel was able to, as you point out, just use a less efficient node and run more power through the chip. Now, we might be reaching the end of the road on that strategy too.
spookyfish
Member
I do like the improvement called "all dressed" chips.
HerjansEagleFeeder
Member
PeteBull
Member
3nm node isnt mature enough yet, even nvidia's premium rtx 5090 that has streetprice of 3,5k+ usd is made on tsmc 4n finfet process which is 5nm(large die size -750 mm²). Thats over 4x base ps5 raw performance btw.
NVIDIA GeForce RTX 5090 Specs
NVIDIA GB202, 2407 MHz, 21760 Cores, 680 TMUs, 176 ROPs, 32768 MB GDDR7, 1750 MHz, 512 bit
Schmendrick
Banned
We´ve been in a "what`s next? Gosh I hope there is a next" - Limbo for almost half a decade now with the fabrication nodes getting dangerously close to what is physically possible (to our current understanding) at least.
Now we`ve arrived at the squeezing point where any and all progress gets prohibitively expensive and honestly, I have no clue what the way forward will turn out to be at this point.
We obviously have the roadmaps exchanging nm for "angstrom" from the big foundries, but in the end most of that is just PR material and alternative technologies are nowhere to be seen.
Short term we`ll just get more and more aggressive shortcutting via ML/AI ..no idea where the limits for that will turn out to be.
Right now is pretty much the most
- moment ever in the semiconductor industry since it was established.
Now we`ve arrived at the squeezing point where any and all progress gets prohibitively expensive and honestly, I have no clue what the way forward will turn out to be at this point.
We obviously have the roadmaps exchanging nm for "angstrom" from the big foundries, but in the end most of that is just PR material and alternative technologies are nowhere to be seen.
Short term we`ll just get more and more aggressive shortcutting via ML/AI ..no idea where the limits for that will turn out to be.
Right now is pretty much the most
Last edited:
TheThreadsThatBindUs
Member
This so hard!
We have hardware features in the current gen APUs that was present in the previous gen and is still currently underutilized.
And computing hardware tech aside, there are also a plethora of peripheral hardware features that have been largely ignored by devs, that can be utilized to create innovative, whole new gaming experiences that would be more impactful than just being able to bump framerate, resolution or texture res.
If we had longer gens, with no mid-gen refreshes, devs could start focusing on optimization as well as utilizing some of these other features.
cireza
Member
Except that costs to extract the ore, to manufacture the chips, to apply to regulations etc... are rising on a constant basis.
Last edited:
Alebrije
Member
Not all electrónics need a last gen chip , most of them uses chips that were created 20 years ago, we find them from a washing machine to a blender.
Those basic chips have huge demand and factories are busy producing them instead of developing state of the art chips for niche /smaller markets.
This is is also a reason why new consoles price behave different from previous generations
Those basic chips have huge demand and factories are busy producing them instead of developing state of the art chips for niche /smaller markets.
This is is also a reason why new consoles price behave different from previous generations
REDRZA MWS
Member
Exacrly, and the jump isn't worth it, gens will last longer. They already are. Pro consoles etc.
RCX
Member
Thats just a gussied-up cheese & onion
Guardian Monkey
Member
Last twenty years of halo GPUs by architecture by Nvidia with an attempt of an apples to apples comparison of computing power increases.
Now I know people are going to throw the current prices into the mix on the halo cards so bumping down to the xx70 line here are some rough estimates on baseline MSRP cards.
Roughy estimates above. Are we really seeing a decline on price vs performance?
Year | GPU | Architecture | FP32 Teraflops | Delta from Prior Gen (TFLOPS) |
| 2004 | GeForce 6800 Ultra | Curie | ~0.040 | — |
| 2006 | GeForce 8800 GTX | Tesla | ~0.345 | +0.305 |
| 2008 | GTX 280 | Tesla (2nd gen) | ~0.933 | +0.588 |
| 2010 | GTX 480 | Fermi | ~1.35 | +0.417 |
| 2012 | GTX 680 | Kepler | ~3.09 | +1.74 |
| 2013 | GTX 780 Ti | Kepler Refresh | ~4.50 | +1.41 |
| 2015 | GTX 980 Ti | Maxwell | ~5.63 | +1.13 |
| 2017 | GTX 1080 Ti | Pascal | ~11.34 | +5.71 |
| 2018 | RTX 2080 Ti | Turing | ~13.45 | +2.11 |
| 2020 | RTX 3090 | Ampere | ~35.58 | +22.13 |
| 2022 | RTX 4090 | Ada Lovelace | ~82.58 | +46.99 |
Now I know people are going to throw the current prices into the mix on the halo cards so bumping down to the xx70 line here are some rough estimates on baseline MSRP cards.
Year | GPU | Architecture | FP32 TFLOPS | Launch Price (USD) | TFLOPS per $ |
| 2004 | GeForce 6800 GT | Curie | ~0.040 | $399 | 0.0001 |
| 2006 | GeForce 7900 GT | Curie | ~0.145 | $299 | 0.0005 |
| 2008 | GeForce 9800 GTX | Tesla | ~0.504 | $349 | 0.0014 |
| 2010 | GeForce GTX 470 | Fermi | ~1.21 | $349 | 0.0035 |
| 2012 | GeForce GTX 670 | Kepler | ~2.46 | $399 | 0.0062 |
| 2014 | GeForce GTX 970 | Maxwell | ~3.49 | $329 | 0.0106 |
| 2016 | GeForce GTX 1070 | Pascal | ~6.46 | $379 | 0.0170 |
| 2018 | GeForce RTX 2070 | Turing | ~7.46 | $499 | 0.0149 |
| 2020 | GeForce RTX 3070 | Ampere | ~20.31 | $499 | 0.0407 |
| 2022 | GeForce RTX 4070 | Ada Lovelace | ~29.15 | $599 | 0.0487 |
| 2025 | GeForce RTX 5070 | Blackwell | ~42.00* | $549 | 0.0765 |
Roughy estimates above. Are we really seeing a decline on price vs performance?
Last edited:
Audiophile
Member
The thing is the average consumer doesn't care about the cost of process nodes going up or their being major engineering challenges, they expect something for their money. If PS6 is highly iterative and just something like a 3nm monolithic chip with bump in CPU/GPU architectures and no more than 60% more raw power over pro; then paired with another excessive cross-gen period it'll end up as little more than a PS5 Pro+. I think a significant number of people will just hang around on PS5, it already has the drastically improved load times and general responsiveness / quality of life improvements.
I think they need to find some way to pull a miracle out of their arse as there needs to be some self-evident leaps on screen to make it worth it on this go around and not someone saying "if you look closely, you can see the ray tracing on this surface here, it's really impressive what we're doing here because the calculation is really complicated and it's being accelerated by special units, it's running at sub-1080p and the particles look like tv static, but it's really impressive cause of the calculations, honest...".
I feel like they got away with it this gen, but I wonder if they get away with it again. I just don't see significant chunk of people being inclined to switch again to what'll probably be a $600 console (before a disc drive) so they can play what are fundamentally the same games with some extra bells and whistles. I think they really have to find a way to wow people or to differentiate the experience somehow. I know totally exotic setups are out of the window but I feel like they need to find a way to lean parts of the system towards that again to punch above their weight and at the same time make it incredibly easy to utilise, they also need to drop a few experiences that simply cannot be had on PS5 or even PS4.
Many people I know are frustrated with things as they are and I've explained to them as clearly as possible why things have slowed and how process nodes effect that; and to paraphrase their response it's basically "ok, I get it, but that's not really my problem, I expect a certain step up for my money and they need to find a way to do it, if not then I'll wait and buy a PC down the line, just not bother anymore or stick with my current console for most of the generation".
They [console makers] have to find a way.
I think they need to find some way to pull a miracle out of their arse as there needs to be some self-evident leaps on screen to make it worth it on this go around and not someone saying "if you look closely, you can see the ray tracing on this surface here, it's really impressive what we're doing here because the calculation is really complicated and it's being accelerated by special units, it's running at sub-1080p and the particles look like tv static, but it's really impressive cause of the calculations, honest...".
I feel like they got away with it this gen, but I wonder if they get away with it again. I just don't see significant chunk of people being inclined to switch again to what'll probably be a $600 console (before a disc drive) so they can play what are fundamentally the same games with some extra bells and whistles. I think they really have to find a way to wow people or to differentiate the experience somehow. I know totally exotic setups are out of the window but I feel like they need to find a way to lean parts of the system towards that again to punch above their weight and at the same time make it incredibly easy to utilise, they also need to drop a few experiences that simply cannot be had on PS5 or even PS4.
Many people I know are frustrated with things as they are and I've explained to them as clearly as possible why things have slowed and how process nodes effect that; and to paraphrase their response it's basically "ok, I get it, but that's not really my problem, I expect a certain step up for my money and they need to find a way to do it, if not then I'll wait and buy a PC down the line, just not bother anymore or stick with my current console for most of the generation".
They [console makers] have to find a way.
Last edited:
We've just hit diminishing returns in general. So even when we see things that are "big upgrades" by the numbers, they don't translate into "big upgrades" for your eyes and they certainly don't translate to "big upgrades" for games overall.
For me, this would be completely fine. It should be so easy for developers to take advantage of the hardware that makes it EASY for them to have games look and perform competently and now just focus on making compelling games. Instead a lot of these devs sit around and spend their time trying to get marginal graphical upgrades that provide very little to the actual experience.
Even Switch 2, which will be the weakest console on the market when it launches next month, is more than enough power to make a great game that looks plenty fine. If you can't accomplish that, you're just not a good dev.
For me, this would be completely fine. It should be so easy for developers to take advantage of the hardware that makes it EASY for them to have games look and perform competently and now just focus on making compelling games. Instead a lot of these devs sit around and spend their time trying to get marginal graphical upgrades that provide very little to the actual experience.
Even Switch 2, which will be the weakest console on the market when it launches next month, is more than enough power to make a great game that looks plenty fine. If you can't accomplish that, you're just not a good dev.
Tams
Member
They can't just find a way. They are pushing the edge of known science to produce these.
While those people will get some benefit from switching to PC, it will force console makers to make more games exclusive.
I'm fine with that, but I doubt those people will be.
kiphalfton
Member
Calling it now...
RTX 5090 > RTX 6090 = 1% increase (but more expensive)
RTX 5090 > RTX 6090 = 1% increase (but more expensive)
ReBurn
Gold Member
ARM architecture is a 40 year old derivative of RISC, so somehow implying that it's the new hotness and x86 is old and busted is kind of weird. Especially since x86 is only 7 years older than ARM. ARM and 32-bit x86 were introduced in the same year.
The thing that makes Apple silicon machines as efficient as they are isn't just the CPU architecture. It's just as much the integrated nature of the entire system. The M-series system design with unified memory and integrated storage allows for Apple to fine tune things like fast-swapping processes between RAM and SSD, which lets it make much better use of memory than a similar spec Windows machine. They know exactly how fast memory and storage are as well as the bandwidth between all of the parts. Windows still has to support nearly infinite spec combinations where Apple gets to target and tune to a finite number of configurations. It's one of the reasons why a MacBook Air can run really well with 8 GB RAM and a Windows laptop with 8 GB RAM runs like crap.
An M4 Mac mini is architected more like a console than it an x86 PC, so I agree that it could potentially be a great machine for gaming. Getting developers to support it is the challenge.
Last edited:
Wasabi Ginger lays, limited time only around 2015. I lament their discontinuation.
Radical_3d
Member
Melon Husk
Member
Chips are improving, but teracorporations have cornered the chip market.
I still believe in Moore's law. The production side isn't focused on consumer side right now. Unless you're willing to drop 2k on a GPU you get the scraps. The demand is in the ultra-high-end, for multiple reasons.
Last edited:
StreetsofBeige
Member
Whats also killing console prices is I dont get a sense MS/Sony are willing to subsidize their consoles losing a couple hundred dollars for years. And when they'd be dumping systems in old gens for $100-150 were they making money off it?
They're in money making mode. Keep console prices high, gamepads cost as much as a game (elite pads more), sub plans, and whatever other rev streams they got.
Only gen ever console makers raised hardware prices.
They're in money making mode. Keep console prices high, gamepads cost as much as a game (elite pads more), sub plans, and whatever other rev streams they got.
Only gen ever console makers raised hardware prices.
Last edited:
UltimaKilo
Gold Member
Someone who gets it. We are in a strange period that will last until someone develops a new tech, probably sometime in the next 6-8 years.
However, that does not mean we are not seeing some incredible things come out in the next year or two:
TSMC's SoIC-mH (System-on-Integrated-Chips Molding Horizontal), a 2.5D packaging technology. This enables 3D chip stacking, improving thermal management and reducing electrical leakage probably significantly.
Wafer-Level Multi-Chip Module (WMCM) Packaging: by 2026 we may see adoption of WMCM packaging, integrating multiple components (CPU, GPU, Neural Engine, DRAM as mentioned by others posts) into a single package. This enhances performance by improving interconnect speed and efficiency, potentially enabling more complex chip designs for high-end tasks.
Combine with TSMC's 2nm tech and the new Gate-All-Around (GAA) transistors that will come with it, we're going to see things over the next 2-5 years that blow everyone's minds.
The near future is very, very bright and as soon as 2029, you'll be blown away by how far we have come from May 2025. Heck, we have some big things coming by late 2027!
DeepEnigma
Gold Member
Engagement farming.
MacReady13
Member
It's because the VAST majority of us gamers aren't fucking nerds who look at that shit too deeply and just want to play good games. If we looked too hard at things like that and actually cared about that bullshit, the Switch would've been a massive failure. Thank FUCK gamers don't care about bullshit like this to be honest.
Physiognomonics
Member
Bullshit. Since Covid they all increased their margins because they saw they could do it so it inevitably led to serious price increase accross the board.
There was only 40 years between the invention of salt and vinegar crisps and the iPhone. if you were born before 1987 you were born closer to the crisps than the smart phone revolutions
Insane Metal
Member
We need those Bismuth chips right now. Or the light ones.
Just imagine what intel could do with a radioactive material....
It's a low rate of decay but has been rejected before as even one atom decaying every couple of years is pretty bad in a CPU.
Dorfdad
Gold Member
Turned out well for everyone the S is their best selling console with the worst games, Microsoft has lost all credibility as a serious hardware manufacturer going forward, and they are now slowly ruining major developers with gamepass fillers.. yeah team!!
Too old for it.
Neo Member
AI is going to change the game I think.
diffusionx
Gold Member
yes, it's Apple's control of the entire stack - hardware, OS, API, drivers - that gives them their primary edge. They can also put specific things in their chips to help their hardware and their software that Intel or AMD or Samsung or Qualcomm can't do. They're also going to be getting rid of Qualcomm's modems so that advantage will only grow.
Apple already runs the biggest game store in the world, and devs can put their games on iOS and Mac with little trouble, I am not sure what you want them to do exactly?
In the past, process shrinks would have allowed console manufacturers to increase their margins while keeping the price the same.
PaintTinJr
Member
I would disagree about the "splitting will not help much" part.
With where we are in terms of trying to render natively as low as possible (540p) and use ML to deliver final pixels, two revised PS5 Pro APUs for a PS6 would be enough for a gen leap over PS5 and likely hit the target power and price given the age of the PS5 Pro silicon and its use of older RDNA parts.
I could see an asymmetric approach with a high CPU clock in APU1 along with a reduced GPU clock for the low resolution ML AI raster input - and for doing low priority ML/RT - and a low CPU clock and high GPU clock on APU2 for the ML AI and the lighting simulation - with a one frame latency - so as to keep both APU's power draws below 100Watts to stay below 250watt system limit.
Sony showed in the PS3 gen prototype GT5 rendering features (to DF) splitting across multiple PS3s to get a 240fps mode, and had another mode doing 4K60fps. Crossfire/SLI styled software solutions aren't ideal but far easier to achieve in a closed console environment IMO.
Schmendrick
Banned
the engineering nightmares some people fantasize about.....
all that´s missing here is duck tape.
Last edited:
PaintTinJr
Member
What? First time? That is child's play of a design compared to the PS2 and PS3 architectures and wouldn't workout much different to a PC with a CPU and discrete GPU
64gigabyteram
Reverse groomer.
this is exactly why i believe the whole ARM hype started over the M1 macs 5 years ago was kinda unfounded and didn't do much. Apple machines are apple machines, they are designed bespoke from the ground up. We've not really seen equivalent performance on ARM from most other machines/manufacturers, because it isn't ARM doing the heavy lifting there.
Swapping architectures will not suddenly yield massive gains in performance. No CPU architecture holds the insane amount of speed and power X86 does at the moment.
The only way out is to basically change what our CPUs are made out of and nobody has any leads on that. I don't think our current computing landscape will drastically change for the next 15 years at least
Last edited:
Denial is the most predictable of human responses
PaintTinJr
Member
Okay
S
Schmendrick
what's your expected solution?
They'll have 250watts a their disposal, they'll need double the TOPs and RT of the PS5 Pro and at least the same or better raster capabilities for B/C. The Pro also shows their hand that an NPU is out of the question. They need to parallelize ML AI in the system or incur latency or get a modest improvement over the PS5 Pro.
They still have the option to delid and use liquid metal and expensive cooling to hit high frequencies/thermals at launch again, but the power consumption will be a surprising design choice if it isn't locked below 250watts. Even taking BoM out of the equation, other than my asymmetric twin APU suggestion how would they even double PS5 Pro performance at 250watt?
They'll have 250watts a their disposal, they'll need double the TOPs and RT of the PS5 Pro and at least the same or better raster capabilities for B/C. The Pro also shows their hand that an NPU is out of the question. They need to parallelize ML AI in the system or incur latency or get a modest improvement over the PS5 Pro.
They still have the option to delid and use liquid metal and expensive cooling to hit high frequencies/thermals at launch again, but the power consumption will be a surprising design choice if it isn't locked below 250watts. Even taking BoM out of the equation, other than my asymmetric twin APU suggestion how would they even double PS5 Pro performance at 250watt?
Last edited:
diffusionx
Gold Member
the M1 laptops got so much hype because they delivered better performance AND battery life than the Intel equivalents, which nobody expected.
The issue here is that speed and power isn't what it used to be. Most consumers don't give a shit. And a lot of intense applications that theoretically could benefit from the speed and power are even better on GPU. This is leaving Intel on the outside looking in because they don't have GPUs for those applications and they can't compete on the stuff people want out of their laptops which is good battery life and efficiency.
ReBurn
Gold Member
I agree. Simply swapping to ARM isn't likely to yield performance gains over x86. While ARM processors can be very powerful there isn't some magic that just makes them better than other architectures. Windows ARM laptops with the Snapdragon processors have great battery life and are good for lighter workloads, but they can struggle in use cases that require higher performance.
For video game consoles, much like gaming PC's, x86 is probably the better choice where the expectation is higher overall performance with power consumption being a lesser concern. That's not to say that one day ARM won't be the right choice. I just don't think today is that day.
64gigabyteram
Reverse groomer.
That doesn't say much about ARM, just Apple. Which was what I meant by my post.
We've yet to really see any non-Apple ARM chips compete with x86 performance. The closest is Switch 2 and that's moreso just Nvidia's GPU solution doing the carrying than anything substantial. This just reflects more on the benefits of Apple's engineering and approach to computing as opposed to the architecture itself.
Haven't they tried to make their own break in with Arc? They're taking steps to try and enter the dGPU market and their tech is pretty decent.