• Hey Guest. Check out your NeoGAF Wrapped 2025 results here!

Why does the RX VEGA 64 not perform as well as Nvidia's GPUs?

Why the VEGA 64 12.6 Tflops has the same performance as the basic GTX 1080 9Tflops.
Can this change with the arrival of the NAVI architecture?
 
Last edited:
That measurement isn't exactly relevant.

Nobody knows exactly how Navi is gonna be.
Speculation would be pointless at this point, AMD is still playing catch up to Nvidia.
 
I'm no expert, but I know that they both have different architectures, so that's why a Nvidia card with less Tflops can be faster than an AMD card with more. Tflops aren't everything when measuring GPU performance anyway.
 
Nvidia has a lot more money to develop technology and invest in development along with R&D. Nvidia will always be 2 steps ahead of AMD. And Nvidia is always 1 gen ahead with their technology as well. I will never buy an AMD GPU or CPU. But AMD needs to be around to keep NVIDIA developing their tech. Having 1 GPU manufacturer alone in the market will basically leave Nvidia doing what they want and not moving technology further because NVIDIA won't have a reason to

My apologies if it doesn't make sense or is confusing but English is not my 1st language
 
Last edited:
The rated flops are only theoretical. Realistically you're restricted from ever reaching 100% utilization on the graphics cores because you'll be limited by other things at some point. Number of ROPS, TMUs, hardware scheduling, software etc.
 
Why the VEGA 64 12.6 Tflops has the same performance as the basic GTX 1080 9Tflops.
Can this change with the arrival of the NAVI architecture?

What are you talking about?

The Vega 64 does better in higher resolutions. That's the entire point of the card, it for people who prioritize 4K so at higher resolutions is has better/more stable fps. Which is something you claim to care about.

That's the entire point of the card is that it has better 4K and high resolution performance.
 
Nvidia has a lot more money to develop technology and invest in development along with R&D. Nvidia will always be 2 steps ahead of AMD. And Nvidia is always 1 gen ahead with their technology as well. I will never buy an AMD GPU or CPU. But AMD needs to be around to keep NVIDIA developing their tech. Having 1 GPU manufacturer alone in the market will basically leave Nvidia doing what they want and not moving technology further because NVIDIA won't have a reason to

My apologies if it doesn't make sense or is confusing but English is not my 1st language
Bullshit. NVIDIA wasn't always 2 steps ahead of AMD. AMD just overslept their chances some time ago. Nevertheless, they really had great CPUs and GPUs some years ago like Radeon 9700Pro/9500 or Athlon 64 or XP CPUs. Back then they were ruling the market. NVIDIA had problems with implementing Shader Model into GeForce 5700/5800/5900 cards which resulted in poor performance.

AMD is different than NVIDIA because it has to pay extra money for R&D for both CPU and GPU departments. NVIDIA has some mobile processors like Tegra but they aren't X86. More like ARM and for mobile devices rather than desktop solutions.

AMD lost the market some time ago not only for having slower CPUs but as we all know Intel paid biggest corporations for exclusivity which resulted in additional fees. But still, Intel doesn't care about using illegal actions.
 
This makes me think that 12 tf on a console will be faster than 12 tf on a pc because everyone is using the same hardware and games will be even more optimized. How many teraflops are needed to display any possible moving images at 60fps 1080p? And for 4k?
 
This makes me think that 12 tf on a console will be faster than 12 tf on a pc because everyone is using the same hardware and games will be even more optimized. How many teraflops are needed to display any possible moving images at 60fps 1080p? And for 4k?

Also I think in a console 12Tflops can perform much better than on a PC. Especially if the PS5 is an APU (Zen 2 or Zen 3). We do not know how the unified components will behave. And maybe there are APIs that are perfectly compatible with these hardware until then. We can have as example the Vulkan API. So we can wait for a significant leap, yes. The PS4 can do incredible things with CPU so lagging. This time the PS5 will have a powerful CPU. I think this will change the way developers will program the games. I think they will program the games based on the CPU and secondarily on the GPU. The opposite of what was done in the current generation in which the developers programmed games based on the GPU and secondarily on the CPU.

Anyway ... Correct me if I'm wrong! I like to read your comments!
 
Why the VEGA 64 12.6 Tflops has the same performance as the basic GTX 1080 9Tflops.
Can this change with the arrival of the NAVI architecture?

It doesn
Maybe the way the game is programmed, directx, and driver support favor nvidia in new computer games.

No the way the games are programmed as well as the review favor Nvidia. The new Vega cards as specifically designed to take advantage of new APIs and directx 12. We're seeing 30%-50% increases just from switching between the directx 11 and 12/Vulcan versions of the same game.

The problem industry wide is that delaying console refreshes hurts the entire gaming industry. Let's say it's 6 years between consoles... at the end of that 6 years you're still running on hardware that's 6 years old using tools and APIs that may be close to 10 years old at that point. So your games are really always close to half a decade behind even at launch.

Having reliable consistent hardware refreshes at the console level means your baseline target refreshes often and make be able to take advantage of the latest tools at the silicon level. Ultimately making your games faster, better, more efficient. Everyone wins. The current industry development process is toxic to everyone.
 
it is all up to optimization. Nvidia has most times optimized their architectures to run optimal with current tools/games. But they fast loos performance with newer titles that use newer features. With amd cards you get supoptimal performance at lauch but most times better in a few years. But it might be that this is just my way to see the last 15 years ;)
 
The title is terrible.
You are asking "why isn't Vega 64 beating all nVidia GPUs with lower TFLOPs ranking".

And the answer is easy: GPU is about more things, than raw tflops.

nVidia and AMD cram different stuff into the same die area. E.g. :
AMD RX 580 - 6.17 Tflops, 232 mm^2
1060 6GB - 4 Tflops, 214 mm^2
 
The title is terrible.
You are asking "why isn't Vega 64 beating all nVidia GPUs with lower TFLOPs ranking".

And the answer is easy: GPU is about more things, than raw tflops.

nVidia and AMD cram different stuff into the same die area. E.g. :
AMD RX 580 - 6.17 Tflops, 232 mm^2
1060 6GB - 4 Tflops, 214 mm^2

What do you mean they cram different stuff into same die area, they are probably using different manufacturing processes and at different manufacturing sizes etc.

Teraflop is simply one floating point instruction per second, so one teraflop on an AMD card should equal one teraflop on a nvidia card. Maybe someone else pointed out earlier that there are co-processors and other chips on the card that are operating differently so the cards might be a little bit faster or slower.
 
Last edited:
What do you mean they cram different stuff into same die area, they are probably using different manufacturing processes and at different manufacturing sizes etc.

RX580 is only slightly bigger, but has 1.5 more TFLOPs than 1060 (6Gb, very nice trick with 3GB, NV)
 
Also I think in a console 12Tflops can perform much better than on a PC. Especially if the PS5 is an APU (Zen 2 or Zen 3). We do not know how the unified components will behave. And maybe there are APIs that are perfectly compatible with these hardware until then. We can have as example the Vulkan API. So we can wait for a significant leap, yes. The PS4 can do incredible things with CPU so lagging. This time the PS5 will have a powerful CPU. I think this will change the way developers will program the games. I think they will program the games based on the CPU and secondarily on the GPU. The opposite of what was done in the current generation in which the developers programmed games based on the GPU and secondarily on the CPU.

Anyway ... Correct me if I'm wrong! I like to read your comments!
I wouldn't compare PC vs console architecture. OK, since PS4 it's a x86 architecture but still it operates quite differently. GPU/CPU Low-level access available in console was also presented to PC when DirectX 12 showed up. And show me games supporting DirectX 12?
 
The rated flops are only theoretical. Realistically you're restricted from ever reaching 100% utilization on the graphics cores because you'll be limited by other things at some point. Number of ROPS, TMUs, hardware scheduling, software etc.
This doesn't make sense. If the entire card is rated at 10 teraflops, then it doesn't matter what the other chips on the card are doing. Maybe they are using API's differently or directx differently.
 
But AMD needs to be around to keep NVIDIA developing their tech. Having 1 GPU manufacturer alone in the market will basically leave Nvidia doing what they want and not moving technology further because NVIDIA won't have a reason to

If Nvidia released a GPU each year that had a 5-10% gain because AMD didn't exist and they own the market, no one would buy it. They would wait 3/4 years minimum until the gains were half decent, otherwise its throwing money away.
 
It doesn


No the way the games are programmed as well as the review favor Nvidia. The new Vega cards as specifically designed to take advantage of new APIs and directx 12. We're seeing 30%-50% increases just from switching between the directx 11 and 12/Vulcan versions of the same game.

The problem industry wide is that delaying console refreshes hurts the entire gaming industry. Let's say it's 6 years between consoles... at the end of that 6 years you're still running on hardware that's 6 years old using tools and APIs that may be close to 10 years old at that point. So your games are really always close to half a decade behind even at launch.

API's and software tools evolve in the console ecosystem and get more and more out of the hardware thanks to the fact that hardware is still for a non trivial amount of time and there is greater ROI for every bit of effort you pour in maximising the hardware utilisation (and there are more and more consumers that are positively affected too). Consoles had very low overhead low level API's before they were a thing on PC, before Vulkan or DirectX 13 were a thing.

Console HW always used to be designed to have forward looking features before they were mainstream on PC or that would be underused on PC where there is little incentive to maximise every card you run on or where there are other bottlenecks that could lower their effective potential.

Having reliable consistent hardware refreshes at the console level means your baseline target refreshes often and make be able to take advantage of the latest tools at the silicon level. Ultimately making your games faster, better, more efficient. Everyone wins. The current industry development process is toxic to everyone.

Sorry, but I will file it under the same category as "Digital downloads... cheaper costs... cheaper games... better for everyone", "games cannot be made without DLC", "games cannot be made without loot boxes... loot boxes and micro transactions do not affect gameplay design... better for everyone", etc...

Besides, we already have the definition of reliable hardware refreshes: often enough not to create a stale frustrating environment, but not too frequent to make investing time in games and tools that exploit it efficiently a waste of resources (over-engineering and poor returns... look at what happened with the console refreshes). Every five years or so :).
 
This doesn't make sense. If the entire card is rated at 10 teraflops, then it doesn't matter what the other chips on the card are doing. Maybe they are using API's differently or directx differently.

"The entire card" isn't rated at 10 TFLOPS. The CUDA cores or Steam processors are. One way to see that your statement is just dumb is to take your graphics card and clock the memory at 1/10th the speed it normally runs. The theoretical TFLOPS is still the same but the performance sure won't be.
 
Why the VEGA 64 12.6 Tflops has the same performance as the basic GTX 1080 9Tflops.
Can this change with the arrival of the NAVI architecture?

Because nVidia has had better performance per flop compared to AMD for a very long time. Could it change with Navi? Yes it could change, but I don't expect it to, they'd have to improve their IPC by a lot and nVidia would have to stay stilll on the area for AMD to catch up. Polaris and Vega did virtually nothing to improve in this area...
 
API's and software tools evolve in the console ecosystem and get more and more out of the hardware thanks to the fact that hardware is still for a non trivial amount of time and there is greater ROI for every bit of effort you pour in maximising the hardware utilisation (and there are more and more consumers that are positively affected too). Consoles had very low overhead low level API's before they were a thing on PC, before Vulkan or DirectX 13 were a thing.

Console HW always used to be designed to have forward looking features before they were mainstream on PC or that would be underused on PC where there is little incentive to maximise every card you run on or where there are other bottlenecks that could lower their effective potential.



Sorry, but I will file it under the same category as "Digital downloads... cheaper costs... cheaper games... better for everyone", "games cannot be made without DLC", "games cannot be made without loot boxes... loot boxes and micro transactions do not affect gameplay design... better for everyone", etc...

Besides, we already have the definition of reliable hardware refreshes: often enough not to create a stale frustrating environment, but not too frequent to make investing time in games and tools that exploit it efficiently a waste of resources (over-engineering and poor returns... look at what happened with the console refreshes). Every five years or so :).

But this is where I feel like you're going to have a difference of opinion, if you're an old school enthusiast the idea of coding to the metal is appealing, however as were talking about 4K art assets, 5 year development cycles and hardware optimization's at a high level you can any a process that is going to be hardware agnostic and generate the most performance in the shortest period of time as well as the hardware being designed with the most current tools and APIs in mind.

The end result would be something more akin to getting 80-90% out of all compatible hardware than spending twice time and effort to close that last percent to close the gap on performance. The idea of trying to eek out the last bits of performance become something for first parties to worry about as a showcase while cross platform developers can actually worry about what they need to do which is making games.

Yeah you can get 10-20% better performance spending 2x as long on a specific hardware configuration but the that's what these API solution are trading off, it's saying the time and expense put into that small amount of optimization could be better used creating a whole new game. It's a different mindset of game development for the entire industry.

Frequent consistent hardware updates supporting the latest software tools that significantly reduce development time across the board which ultimately creates more games in shorter time periods and curbs the cost of development. In a macro sense looking at a single game you can argue that it's not it's absolute best, but the overall picture for the industry is much healthier and better long term.

We aren't going to see that until consoles reached parity with PCs which will may finally see happen in the coming generation. Remember Microsoft is working on their "code once" philosophy which they are trying to extend to the Xbox as well.
 
But this is where I feel like you're going to have a difference of opinion, if you're an old school enthusiast the idea of coding to the metal is appealing, however as were talking about 4K art assets, 5 year development cycles and hardware optimization's at a high level you can any a process that is going to be hardware agnostic and generate the most performance in the shortest period of time as well as the hardware being designed with the most current tools and APIs in mind.

The end result would be something more akin to getting 80-90% out of all compatible hardware than spending twice time and effort to close that last percent to close the gap on performance. The idea of trying to eek out the last bits of performance become something for first parties to worry about as a showcase while cross platform developers can actually worry about what they need to do which is making games.

Yeah you can get 10-20% better performance spending 2x as long on a specific hardware configuration but the that's what these API solution are trading off, it's saying the time and expense put into that small amount of optimization could be better used creating a whole new game. It's a different mindset of game development for the entire industry.

Frequent consistent hardware updates supporting the latest software tools that significantly reduce development time across the board which ultimately creates more games in shorter time periods and curbs the cost of development. In a macro sense looking at a single game you can argue that it's not it's absolute best, but the overall picture for the industry is much healthier and better long term.

We aren't going to see that until consoles reached parity with PCs which will may finally see happen in the coming generation. Remember Microsoft is working on their "code once" philosophy which they are trying to extend to the Xbox as well.

I do not think you NEED constant frequent hardware updates to support better tools and API's, consoles had better tools and API's (what programmers were asking Microsoft and the Khronos consortium) and you do not need constant frequent hardware updates to support better tools and API's... as compute power on GPU is getting to be more and more flexible the more features move out of hardware support that must be embedded in an API anyways.

Also, in an era where performance and feature are slowing down in terms of time it takes from one performance jump to the next (HW cycles are getting longer and not shorter), increasing the rate of HW refreshes is counterintuitive at best and an attempt to copy Apple and hoping that such strategy brings a lot higher profits per unit sold (Apple is the one that gets the actual benefits of yearly product releases ;)) and allows some big AAA publishers to consolidate in even bigger ones.

Frequent consistent hardware updates supporting the latest software tools that significantly reduce development time across the board which ultimately creates more games in shorter time periods and curbs the cost of development.

Again, until I see other reasons supporting this and not data that actually works against it (as Inwas mentioning earlier in this post), it will have to sit with the same kind of tak around "going to all digital distribution being best for everyone and bringing all this many many benefits... and this panacea of goodness for gamers and the gaming industry" or "loot boxes and F2P promises". I do not see the logic link between the two statements because I do not agree with the logic suppprting the first statement.

Btw, thanks for posting as you do take your time in your replies. Glad to see that :).
 
Last edited:
"The entire card" isn't rated at 10 TFLOPS. The CUDA cores or Steam processors are. One way to see that your statement is just dumb is to take your graphics card and clock the memory at 1/10th the speed it normally runs. The theoretical TFLOPS is still the same but the performance sure won't be.

No, but, if you just give the default card, it is rated at 10 tf, so then it outputs 10 tf at default settings. So the card is rated at 10 tf. So, if APIs are the same then the performance should be the same.

Maybe the API is compiled differently for different architecture so it does not run as fast.
 
Last edited:
There are more reasons why these cards perform the way they perform.

1) Games get shipped utterly broken and it's up do Nvidia / AMD to address this with there drivers even while that shouldn't be there job to start with. It's pure laziness of those developers.

A EX nvidia dev basically said the following:

"Nearly every game ships broken. We're talking major AAA titles from vendors who are everyday names in the industry. In some cases, we're talking about blatant violations of API rules – one D3D9 game never even called BeginFrame/EndFrame. Some are mistakes or oversights – one shipped bad shaders that heavily impacted performance on NV drivers. These things were day to day occurrences that went into a bug tracker. Then somebody would go in, find out what the game screwed up, and patch the driver to deal with it. There are lots of optional patches already in the driver that are simply toggled on or off as per-game settings, and then hacks that are more specific to games – up to and including total replacement of the shipping shaders with custom versions by the driver team. Ever wondered why nearly every major game release is accompanied by a matching driver release from AMD and/or NVIDIA? There you go."

It's obvious Nvidia does a hell lot better effort on this department then AMD but i believe they have a bigger department after AMD sacked most of there team at some point in history for reasons.

2) Bribing.

It's well known Nvidia and both AMD play dirty tricks in games or benchmarks to up each other. Crysis 2 where Nvidia pushes loads of tessellation objects of no meaning just to tank AMD performance, AMD also had it's fair share of shady practices towards Nvidia.

This means dirty tricks get applied in games to make the competitor tank performance.

3) Support in games from the GPU vendor.

Nvidia: Hey we want to work with you to make your game look even more great then it already is. Lets ram it full with physx ( tanks on amd ) and high tesselation hairworks. ( tanks again on AMD cards ). Look it's amazing right.

Dev team: Sure why not.

AMD: ( witcher 3 ) Oh crap, can you witcher 3 devs implant our stuff in the last month?

Dev team: Lol no thanks, to late now.

Gameworks implemented = AMD tanks performance once again.

4) Optimization.

More TFLOPS specially on PC doesn't automatically mean more performance.

It's because developers just dump there games on PC and leave it there and up the specifications because why spend money when PC gamers can just brute force it. Skyrim was really obnoxious in this department. But instead of spending time to optimize it they just drop it and let modders deal with it. Up the specification and there you go.

This is also a reason why i am not a big fan of having huge leaps on hardware updates specially on the CPU department. It will just makes older hardware faster outdated even while that shouldn't really be needed even remotely. With only 5% minimal increases each year, devs will have to spend more time and effort for better performance which eventually you as consumer will benefit from.

It's a whole range of reasons why things happen the way they do.












3)

As nvidia has a hell of a
 
honestly, I don't trust teraflops numbers because people know that this is something gamers pay attention to they have a motive to fudge the numbers. just "buyer beware" I guess, the actual performance (FPS) is all that really matters anyway. performance seemed playable at least, it's at least a decent port. one thing to remember is that this will ship on a "game card" with limited storage capacity and if there is a difference between platforms that would be the first place I would hold accountable. Nintendo charges more to the developer if their game is over a certain size. the game cards come in 1GB, 2GB, 4GB, 8GB, 16GB and 32GB capacities. they have a financial incentive to make their game as small as possible.
 
If Vulkan is any indication, and it definitely is, it boils down to software. Radeon chips are often substantially beefier then GeForce and yet often fall short in the end.

Now, I am not a fanboy but I am also not blind to nvidia efforts to undermine competition any way they can. Overblown tesselation, hairworks, anything related to TWIMTBP, you name it. That's why AMD created Mantle, then gave it up to the Khronos group, who made it into a standard and renamed it to Vulkan. And do we see when I title is using Mantle? That Radeon chips are actually really fast but are often severely limited by software, especially DirectX. Hm, I wonder why that is.

Just look at Doom. That's how game engines should be made. Looks amazing and runs 100 FPS faster than anything.
 
honestly, I don't trust teraflops numbers because people know that this is something gamers pay attention to they have a motive to fudge the numbers. just "buyer beware" I guess, the actual performance (FPS) is all that really matters anyway. performance seemed playable at least, it's at least a decent port. one thing to remember is that this will ship on a "game card" with limited storage capacity and if there is a difference between platforms that would be the first place I would hold accountable. Nintendo charges more to the developer if their game is over a certain size. the game cards come in 1GB, 2GB, 4GB, 8GB, 16GB and 32GB capacities. they have a financial incentive to make their game as small as possible.

Your comment reminding people of financial incentives for making games smaller on Switch reminded me of the pre reveal discussions where the hypothesis of using carts was heralded as positive for everyone as Nintendo would have subsidised the carts costs for developers sharing with them enormous shipment costs savings (Nintendo obviously went with much bigger boxes than those people were expecting because... having tiny boxes makes them disappear at retail in the sea of attention grabbing items on display)...
 
From my lazy understanding, AMD's drivers are not as optimized as Nvidia's drivers for their hardware, that's why AMD has to use more powerful hardware to compete with Nvidia's less powerful GPUs. Basically brute-forcing it. Whenever Vulkan/DX12/low-level APIs become the norm for PC games, this should become a non-issue.
 
The simplest answer is that most games/technology's are not optimised for AMD architecture. The biggest reason is Nvidia have the higher market share (3-1 ration or there abouts) so if a dev needs to optimise for a specific chip/architecture then its going to be for the highest market share.

The Vega64 has shown it can match performance with Nvidias cards, and it has a lot of raw performance power, its just there are no games out there to take advantage of it.

http://gpu.userbenchmark.com/Compar...AMD-RX-Vega-64-LC-Liquid-Cooled/3603vsm353890

Also the cost has gone through the roof, because the Vega cards are apparently amazing at mining due the "Raw Performance" so bang for the buck Nvidia win it.
 
There are more reasons why these cards perform the way they perform.

1) Games get shipped utterly broken and it's up do Nvidia / AMD to address this with there drivers even while that shouldn't be there job to start with. It's pure laziness of those developers.

A EX nvidia dev basically said the following:



It's obvious Nvidia does a hell lot better effort on this department then AMD but i believe they have a bigger department after AMD sacked most of there team at some point in history for reasons.

2) Bribing.

It's well known Nvidia and both AMD play dirty tricks in games or benchmarks to up each other. Crysis 2 where Nvidia pushes loads of tessellation objects of no meaning just to tank AMD performance, AMD also had it's fair share of shady practices towards Nvidia.

This means dirty tricks get applied in games to make the competitor tank performance.

3) Support in games from the GPU vendor.

Nvidia: Hey we want to work with you to make your game look even more great then it already is. Lets ram it full with physx ( tanks on amd ) and high tesselation hairworks. ( tanks again on AMD cards ). Look it's amazing right.

Dev team: Sure why not.

AMD: ( witcher 3 ) Oh crap, can you witcher 3 devs implant our stuff in the last month?

Dev team: Lol no thanks, to late now.

Gameworks implemented = AMD tanks performance once again.

4) Optimization.

More TFLOPS specially on PC doesn't automatically mean more performance.

It's because developers just dump there games on PC and leave it there and up the specifications because why spend money when PC gamers can just brute force it. Skyrim was really obnoxious in this department. But instead of spending time to optimize it they just drop it and let modders deal with it. Up the specification and there you go.

This is also a reason why i am not a big fan of having huge leaps on hardware updates specially on the CPU department. It will just makes older hardware faster outdated even while that shouldn't really be needed even remotely. With only 5% minimal increases each year, devs will have to spend more time and effort for better performance which eventually you as consumer will benefit from.

It's a whole range of reasons why things happen the way they do.












3)

As nvidia has a hell of a

So would it be safe to say that if the PS5 receives 12Tflops AMD architecture(RXVEGA 64 equivalent), will be more powerful than a PC with the same components? Why? Can consoles better leverage AMD architectures with new APIs and good optimizations? I realize that Xbox One X is doing almost unbelievable things with an optimized RX 580. It's amazing how he's been running games in native 4k and 30fps without compromises. Not only has it increased the resolution of the games but it is receiving countless enhancements in each of the games. I particularly think if the PS5 or Xbox Two have twice the power of the Xbox One X GPU, they will be powerful machines. What do you think?
 
How come PC vega is not optimized if game developers are making games for console vega? Xbox One X is not selling well enough so not enough games that do this?
 
How come PC vega is not optimized if game developers are making games for console vega? Xbox One X is not selling well enough so not enough games that do this?
The changes that are part of Polaris and some of the Vega ones are only in the stop gap console updates and neither of them has reached a critical mass of users yet to make the ROI from those change significant enough. Also do remember that there are strong console maker mandates about excellent support for the original consoles by titles that are "enhanced" too which further limits how far some developers are willing to go on exclusive mid generation console updates changes.
 
How come PC vega is not optimized if game developers are making games for console vega? Xbox One X is not selling well enough so not enough games that do this?

Vega on PC is performing about where it should, it's basically a Fury X with higher clock-speed with little to no IPC gain over previous gen AMD GPUs.

Xbox One X is based on Polaris and has a similar Tflop rating as RX 580. Xbox One X GPU and RX 580 perform very similarly...
 
I do not think you NEED constant frequent hardware updates to support better tools and API's, consoles had better tools and API's (what programmers were asking Microsoft and the Khronos consortium) and you do not need constant frequent hardware updates to support better tools and API's... as compute power on GPU is getting to be more and more flexible the more features move out of hardware support that must be embedded in an API anyways.

Also, in an era where performance and feature are slowing down in terms of time it takes from one performance jump to the next (HW cycles are getting longer and not shorter), increasing the rate of HW refreshes is counterintuitive at best and an attempt to copy Apple and hoping that such strategy brings a lot higher profits per unit sold (Apple is the one that gets the actual benefits of yearly product releases ;)) and allows some big AAA publishers to consolidate in even bigger ones.

Again, until I see other reasons supporting this and not data that actually works against it (as Inwas mentioning earlier in this post), it will have to sit with the same kind of tak around "going to all digital distribution being best for everyone and bringing all this many many benefits... and this panacea of goodness for gamers and the gaming industry" or "loot boxes and F2P promises". I do not see the logic link between the two statements because I do not agree with the logic suppprting the first statement.

Btw, thanks for posting as you do take your time in your replies. Glad to see that :).

I have to agree that hardware updates are not a necessity strictly speaking, but I believe it is one for a healthy game industry. The fact that we are seeing less and less specialized silicon more programmable silicon is the exact reason why I think the frequent updates are neccessary.

We are reaching a critical mass very quickly where the minimum amount of power to run photo-realistic graphics can be incorporated as part of the APU.

Tim Sweeney feels that 40 Tflops will net us photo realism realtime and I'm inclined to agree.
https://www.digitaltrends.com/computing/40-teraflops-tim-sweeney-epic-photo-realism/

In terms of hardware this target will be hit easily in the next 5 years, and in the next 10 years it will be incorporated as part of an APU.
At this point hardware become trivial because it becomes the tools in use and very few will expect to core to the metal of hardware.
The processing power to produce games will be an implied expectation in the same way we no longer have discrete sound cards and we do not code to the metal for things like a Roku or a Fire TV.

I separate these things from lookt boxes and f2p because they are business models which don't have much issue with.

What we are talking about here is the inevitable diminishing returns which will occur within 1 or 2 generations. My personal belief is that the industry and consumers should get use to the idea that the experiences across platforms will largely be the same in terms of the quality of the content delivered and I think frequent hardware refreshed which prompt the developers to use the latest tools available due to the large install bases is what will push use to the best games faster.

Personally I see 2025 as the turning point for gaming. That will be the last generation of consoles in my opinion as the point of real-time photo-realism will be achieved on a visually technical level.
 
I don't think AMD sees the value in optimizing a card that no one's using for gaming. The launch was so weird and the timing with the mining craze so bad that even hardcore AMD fans like myself knew we were better served just buying a 1080 than paying the hundreds of dollars premium still attached to Vega cards after all this time. Their efforts are better spent looking forward to their next lineup and the new console(s), I'm sure they'd like to turn the page after parting ways with Raja.
 
I have to agree that hardware updates are not a necessity strictly speaking, but I believe it is one for a healthy game industry. The fact that we are seeing less and less specialized silicon more programmable silicon is the exact reason why I think the frequent updates are neccessary.

We are reaching a critical mass very quickly where the minimum amount of power to run photo-realistic graphics can be incorporated as part of the APU.

Tim Sweeney feels that 40 Tflops will net us photo realism realtime and I'm inclined to agree.
https://www.digitaltrends.com/computing/40-teraflops-tim-sweeney-epic-photo-realism/

In terms of hardware this target will be hit easily in the next 5 years, and in the next 10 years it will be incorporated as part of an APU.
At this point hardware become trivial because it becomes the tools in use and very few will expect to core to the metal of hardware.
The processing power to produce games will be an implied expectation in the same way we no longer have discrete sound cards and we do not code to the metal for things like a Roku or a Fire TV.

I separate these things from lookt boxes and f2p because they are business models which don't have much issue with.

What we are talking about here is the inevitable diminishing returns which will occur within 1 or 2 generations. My personal belief is that the industry and consumers should get use to the idea that the experiences across platforms will largely be the same in terms of the quality of the content delivered and I think frequent hardware refreshed which prompt the developers to use the latest tools available due to the large install bases is what will push use to the best games faster.

Personally I see 2025 as the turning point for gaming. That will be the last generation of consoles in my opinion as the point of real-time photo-realism will be achieved on a visually technical level.

Going once more on the performance angle you mention here again... if the time between major performance improvements is getting longer and longer (and it is), why would releasing HW updates more frequently be helpful (to anyone but the console makers that blindly hope they could replicate Apple's success by doing Apple style frequent iterations)?

To reiterate in a different way. As GPU's are actually at or near the point where the focus is on software techniques harnessing the raw power (and few bits, fewer and fewer ones too, of fixed function HW) of programmable processing clusters, given that the number of such processing clusters is physically limited by the evolution of manufacturing nodes and wafer sizes, and given that the cost of each new manufacturing node/silicon wafer size update is increasing exponentially and taking longer and longer to bring to market... how would releasing HW more frequently help?
 
Last edited:
Going once more on the performance angle you mention here again... if the time between major performance improvements is getting longer and longer (and it is), why would releasing HW updates more frequently be helpful (to anyone but the console makers that blindly hope they could replicate Apple's success by doing Apple style frequent iterations)?

To reiterate in a different way. As GPU's are actually at or near the point where the focus is on software techniques harnessing the raw power (and few bits, fewer and fewer ones too, of fixed function HW) of programmable processing clusters, given that the number of such processing clusters is physically limited by the evolution of manufacturing nodes and wafer sizes, and given that the cost of each new manufacturing node/silicon wafer size update is increasing exponentially and taking longer and longer to bring to market... how would releasing HW more frequently help?

But that's the point. Some software technique will simply not be possible without the raw power of new hardware. Because those techniques are impossible to use on the older hardware, it prevents that software from gaining widespread use and utilization. It artificially holds the industry back because while the power to utilize the software techniques exist in newer systems, in older to accommodate aging hardware software which would utilize the newer hardware isn't put into use.

"and given that the cost of each new manufacturing node/silicon wafer size update is increasing exponentially and taking longer and longer to bring to market... how would releasing HW more frequently help? "

But this isn't a given, while the R&D going into the process is definitely more costly and takes longer, we are seeing shorter and cheaper manufacturing relative to the performance gains and their previous costs.
In the long run we are getting more power for less memory and frequent consumer demand will only increase this cycle in the same way that cellphone processors continued to increase in power and going down in cost.

So my argument is this:

To use the newest software techniques which benefit the industry frequent and consistent hardware refreshes will encourage the latest software techniques because the power to run them is there.
This helps the hardware market in the long run as the constant demand helps accelerate the hardware development process from a business perspective until we reach a point of diminishing returns.
In the long run it gets us to the end goal where the market stabilizes because software tools become king and the hardware becomes almost irrelevant, but the process happens faster and cheaper on the business side because constant demands pushed it in that direction.
 
Cellphones started from a much lower starting point which is why their performance increase pattern cannot be just extrapolated in the console industry and just because it worked for Apple, and a bit Samsung too, to fan / help manufacture this demand, it does not mean that the console makers can replicate it and reap the same products. Still, the phone market is a good example of who the main beneficiaries are: Apple and Samsung instead of consumers (many third party apps that use a fraction of what the phone can do for years... cross generation forever...).

In the consoles and the PC space, but you are seeing it on phones too (look at the performance increase between X and iPhone 7 Plus, between iPhone 7 Plus and iPhone 6S Plus, etc... keep going for a bit), we are seeing the pattern of industry cycles slowing down, new manufacturing advancements getting much much costlier, etc... look at Intel even: first they were doing big impactful changes often-ish, then the Tick - Tock (optimise architecture on new process - then release new architecture on existing now stabilised process), to a three phase process where the same node is now used for yet an additional stabilisation phase.

Just because there is demand does not mean that magic happens... look at battery life/power consumption/etc... on phones: we still do not have magic technology powered batteries no matter the quite serious payoff for anyone able to deliver it...

Releasing HW more often means more and more R&D pours in marketing, manufacturing the new device, presenting it, and shipping it to stores (new HW launches are not free) and less in actual HW improvements.

Power is there already and a more consistent user base all able to make use of it does actually a lot of good in terms of ensuring that more and more people have the same predictably performing HW and have not slowed down but helped the progress of the game industry in the past.

Latest software techniques can be used right now and keep being developed in the future, the industry is providing nice meaningful performance jumps and then giving time to developers to iterate on what they do best: software.

I can understand why people that love to always have a shiny new product think they would want consoles to come out every year, why some mega Uber AAAA publisher would want forward compatibility and yearly hardware releases, and why some HW maker may want people to desire and beg for this transition to a iterative model to start and thus help said HW maker increase its profit margins year after year... not sure what is in it for the overall gaming industry or for gamers.

I still see the kind of hand waving used by the industry over initial distribution, over micro transactions, over DRM, over license/not sold vs ownership and consumer resale rights, against used games, over loot boxes, etc... in which there is a hypothetical future full of goodies for everyone if we just trust the industry to put itself into a potential position where they could keep all the profits and give consumers nothing back (but they just wouldn't because that is just how nice and selfless they are ;)).
 
Last edited:
Cellphones started from a much lower starting point which is why their performance increase pattern cannot be just extrapolated in the console industry and just because it worked for Apple, and a bit Samsung too, to fan / help manufacture this demand, it does not mean that the console makers can replicate it and reap the same products. Still, the phone market is a good example of who the main beneficiaries are: Apple and Samsung instead of consumers (many third party apps that use a fraction of what the phone can do for years... cross generation forever...).

In the consoles and the PC space, but you are seeing it on phones too (look at the performance increase between X and iPhone 7 Plus, between iPhone 7 Plus and iPhone 6S Plus, etc... keep going for a bit), we are seeing the pattern of industry cycles slowing down, new manufacturing advancements getting much much costlier, etc... look at Intel even: first they were doing big impactful changes often-ish, then the Tick - Tock (optimise architecture on new process - then release new architecture on existing now stabilised process), to a three phase process where the same node is now used for yet an additional stabilisation phase.

Just because there is demand does not mean that magic happens... look at battery life/power consumption/etc... on phones: we still do not have magic technology powered batteries no matter the quite serious payoff for anyone able to deliver it...

Releasing HW more often means more and more R&D pours in marketing, manufacturing the new device, presenting it, and shipping it to stores (new HW launches are not free) and less in actual HW improvements.

Power is there already and a more consistent user base all able to make use of it does actually a lot of good in terms of ensuring that more and more people have the same predictably performing HW and have not slowed down but helped the progress of the game industry in the past.

Latest software techniques can be used right now and keep being developed in the future, the industry is providing nice meaningful performance jumps and then giving time to developers to iterate on what they do best: software.

I can understand why people that love to always have a shiny new product think they would want consoles to come out every year, why some mega Uber AAAA publisher would want forward compatibility and yearly hardware releases, and why some HW maker may want people to desire and beg for this transition to a iterative model to start and thus help said HW maker increase its profit margins year after year... not sure what is in it for the overall gaming industry or for gamers.

I still see the kind of hand waving used by the industry over initial distribution, over micro transactions, over DRM, over license/not sold vs ownership and consumer resale rights, against used games, over loot boxes, etc... in which there is a hypothetical future full of goodies for everyone if we just trust the industry to put itself into a potential position where they could keep all the profits and give consumers nothing back (but they just wouldn't because that is just how nice and selfless they are ;)).

But the cellphone and consoles mirror each other.

-They both started as highly specialized hardware for very specific applications. They migrated from very specific applications to more general purpose applications.

-As part of that progress they adopted computer systems which are closer to general computing devices. In fact we know have arm OSs that can run on the same chips as our cellphones.

-As part of that progress to more programmable and general purpose hardware the phones features, applications, and product have gotten better.

-Consoles are set to follow the same track, Sony & MS are platforms, game developers make games. When their resources aren't spend pouring time and effort and R&D into the hardware they can focus more on delivering what it is we want as an industry and what their primary task should be: Games. The Playstation & Xbox platforms should first and foremost be able the games. This should be the primary objective and making the best possible games.
This is why like cellphones, converging the hardware will only produce better applications at cheaper costs of the hardware. It's been our frequent hardware cycle that has allowed for the creation of sub $100 smartphones, and it will be the frequent hardware refreshes that allow for sub $100 consoles which can play photo-realistic games in 20 years.

Yes when viewed in the lens of traditional console making:
-highly specialized- custom built production
-R&D and market
-a product that still isn't comparable to a PC

Then yes, it would not make sense. But the industry doesn't look like that, or at least won't look like that in 5-10 years.
Consoles and PCs are converging. Its unavoidable. Clinging to the model of the past with products of the future will no improve the products.

We talk about the power of the consoles and how time has helped them, but those are the consoles of the past and that time has only helped because of the model.
-specialized hardware
-not as powerful as PC
-Thus programmers time to create new and different techniques to get the most of our exotic and under powered hardware when compared to the latest PC running the latest tools.
The time wasn't a benefit, it was a necessity due to the model.

We talk about the latest software techniques but there's no chance the PS5 will be able to fully take advantage of the advanced real-time ray tracing techniques.
The hardware just doesn't exist in this form yet. But we can still get some performance gains in a PS5 Pro, developers can begin developing the ray tracing techniques for the PS6 and so-on.
The see a benefit to the roi of implementation because they have a wider base which can use these techniques now instead of figuring it out 3 years into a new console's life that releases every 6 years.

To put it another way, We wouldn't have to wait nearly 10 years to see the improvements in gaming. It would drastically cut the time new techniques, especially those which are hardware dependent, make it into our games.

And yes, this will have publishers and hardware manufacturers salivating, but the fact is for all it's flaws it will benefit the industry more than it will harm it just like the cellphone industry.

The argument you make is that digital distribution is inherently a negative, however that's not shared by the general public. While understandably things like micro-transactions and loot boxes are and inevitable issue with it, as well as the DRM. It doesn't take away from the fact that people find value in things like Netflix, or Pandora, or iTunes, any other media as a service product. The digital distribution model has already been proven in every industry, even gaming where PC has almost completely switched to a digital distribution model. Claiming that it doesn't work or does not benefit the market would be disingenuous when the market decided that's what they wanted. If they didn't digital distribution wouldn't exist because the market wouldn't have bought those products.

The fact is when we look at something like Steam, we see tremendous value to consumers. We see humble bundles and people with hundreds or even thousands of games. On PSN the flash sells generate a ton of revenue to the original game developers. You're complaint about resell and ownership is ironic given how much we know about resell hurting the industry. In a digital distributed console world, developers can drop their prices in the marketplace to whatever they want without worrying about the resell market cutting into their costs.

I've said it before and I whole-heartedly believe this. The current console model, is dead/dying. And It's a good thing. It's toxic to the entire gaming industry at every level:

-Developers are forced to code on 5+ year old underpowered hardware because that's the largest install base
-Adapting current techniques to this older hardware costs money because new API or techniques may not be supported or run well on aging hardware
-These concessions reduce the quality of the games on more powerful hardware such as PCs, the same way cross gen games do
-Publishers demand safe games because they need pre-orders and day one sales to make up their costs
-They lose money in the secondary market because of game resells
-Consumers end up with games designed for 5+ year old plus hardware that basically looks like every other game before it and its sequels because taking risks in the current gaming environment is too costly.

How this is solved:
The convergence of console and PC hardware gives developers consistent and frequent hardware refreshes.
-Low-Level APIs for PC and console help developers get 80-90% of the performance of the hardware right out of the gate before optimizations. Saving time and money
-Designing for better hardware because you have a larger install base no longer limits the games as much and results in more games doing more with the hardware
-Because game development should be faster, cost less, and use more recent tools, publishers should be able to take more risks.
-If digital distribution becomes the norm, publishers and devs can also recoup costs after launch by dropping their prices and could actual have less reliance on things such as loot boxes.
-Gamers get newer hardware, sooner, possibly cheaper, games shouldn't be as held back by previous tools and hardware and a digital distribution model should make console gaming more similar to the steam environment.

These are merely the first steps into a badly needed overhaul for the gaming industry as a whole.
 
Top Bottom