TR Forums

I was just watching a YouTube video showing God of War on the PS4 and realized how good it is, not just graphically. Makes me wonder what makes consoles so much more efficient than personal computers. The PS4 is using only AMD Jaguar cores. You can't have that sort of performance from those Jags in a PC even if you slap in a nice GPU. Same with all the other past video game consoles. Look at the PS1 for example. 33MHz, little RAM. You'd probably need a Pentium MMX or something at 150MHz to get that sort of performance in a PC. NES was something like 1.7MHz. When did we get side scrolling games in our PCs?

Coding to a single bare metal design means you can tweak everything to be *perfect* on that one platform and you can take advantage of every low level access point into the hardware that is available. Time consuming, but actually possible on one platform.

Yup, it's all about the hardware optimization. Also, console OSes don't have nearly as much of the cruft and crap as desktop OSes, and Windows in particular.

Xbox os is basically windows.

But it's not peasant-tier desktop Windows, it's a heavily modified derivative like what you'd get out of Windows Embedded. Totally different animal, like comparing a Yocto Linux image to Debian/CentOS/whatever.

ronch wrote:

I was just watching a YouTube video showing God of War on the PS4 and realized how good it is, not just graphically. Makes me wonder what makes consoles so much more efficient than personal computers. The PS4 is using only AMD Jaguar cores. You can't have that sort of performance from those Jags in a PC even if you slap in a nice GPU. Same with all the other past video game consoles. Look at the PS1 for example. 33MHz, little RAM. You'd probably need a Pentium MMX or something at 150MHz to get that sort of performance in a PC. NES was something like 1.7MHz. When did we get side scrolling games in our PCs?

As far as newer generations are concerned, it's optimisation. Do keep in mind the gap is not nearly as huge as it once was, though.

In older generations you had custom graphics hardware on that 1,7mhz nes, or that 3,68mhz super famicom (and the PS1), while the PC only had a general purpose CPU. Think of it sort of like a software renderer competing with a system that has a GPU. You bet you're going to need a much beefier general purpose CPU just to scroll the damn screen.

As far as my home computer memories go , going back way before 3D cards, only the Amiga had competitive -for its time- dedicated gaming oriented custom graphics chips (the blitter, was it?).

Nowadays, we have specialized gaming accelerators on our PCs, so it is not even close to what it used to be, but console optimization still makes a difference.

ronch wrote:

When did we get side scrolling games in our PCs?

I think the first console-level SMOOTH scrolling PC platformer was Commander Keen, about 1990 or 1991. Don't remember the minimum requirements, but it did scroll buttery-smooth on my 12Mhz 286, and the graphics were quite good. Most other PC platformers up to then (and even some time later) scrolled like crap. Compare, for example, the scrolling on PC versions of Titus the Fox with the original Amiga version (you can find youtube playthroughs)... it made me want to cry.

Think about how close-to-metal APIs like Vulkan and DX12 can perform better than more abstracted APIs like previous OpenGL versions and DX11, DX10, etc. Consoles are the same idea.

The emphasis on "can" is because it all depends on how optimized the game is with the close-to-metal API. This approach moves the onus for performance improvements out of the hands of the driver developers and into the hands of the game developers. Some are better at it than others.

The other consequence is that it makes cross platform development more challenging. A game moving from PS4 to DX12 will need more re-optimization by the game developer to run well on PC than would be required in further-from-metal approaches. Exhibit A: Batman Arkham Knight.

Today's consoles are actually alot more like a PC, in that they always have an OS with many services running in the background eating up RAM and CPU cycles. Yesterday's DOS PCs were likewise more console-like in that they were close to metal and gave the game/program almost complete exclusivity.

Of course being able to optimize for the same hardware is a big benefit, but I always felt that the anti-piracy measures and everyone having an identical game controller was a bigger boon for consoles. On a PC you have to assume that the player doesn't have a controller, and back in the DOS days, even having a mouse wasn't a given. This discouraged entire genres from taking hold.

As others have noted, it is optimization. Not only is it possible to craft a code path which is optimally targeted at a specific set of hardware with minimal compromises, the devs also know that they can't just tell people "if it is running slow, upgrade to a better GPU". They have no choice but to provide a smooth experience on the existing console hardware.

Efficient? The PS4 was drawing 264W at launch, and even then it was upscaling from 900p a lot of the time. Since improved with die shrinks, but I don't think they are especially electrical power, GPU power, or CPU power efficient. I do think a very small minority of first party developers can and to push their software efficiency to the limits, but even they have to use a lot of tricks to hide or fake details to get there, it's not just about code or shader or API efficiency. And spending the time and effort to get there is not always "efficient" from a manpower or project perspective, because that is manpower that could be spent on other things (not that I'm excusing the many horribly optimized games that come out on PC *and* console all the time).

Maybe calling the consoles efficient is the wrong way around. Perhaps a better way of putting it is why are PCs so hideously slow given multiple gigahertz cores that are capable of performing billions of operations a second? What on earth are they doing for all those seconds?

Absolutely nothing.

I'm not sure what else you'd call it other than being efficient though. Higher performance means higher power consumption. There's only so much we can do to get around that. If someone wants to make a case that building good software on a large scale without driving all your developers off the deep end is still somewhat of an open problem, then I'd agree with that, but that's not really the same question.

I hope this also helps shed some light on why the whole "games are bad because developers suck" trope is so tiresome. Consoles can do what they do because developers don't suck.

NovusBogus wrote:

But it's not peasant-tier desktop Windows, it's a heavily modified derivative like what you'd get out of Windows Embedded. Totally different animal, like comparing a Yocto Linux image to Debian/CentOS/whatever.

This is false. Xbox One OS is built on top of Windows 10 Core - the same standard one on top of which the desktop skus are built on. It's not Windows Embedded or anything like it at all. It's nowwhere close to the comparison of Yocto to Debian.

In fact, Xbox One OS is a fully featured OS SKU, with full support for UWP apps, just like the desktop OS. Games and apps on top of Hyper-V. The divergence between desktop OS and Xbox OS is primarily in the driver and UI layers. Xbox OS naturally includes its own first party drivers and UWP UI that's specific to Xbox. Under the hood, it's all standard WinNT APIs.

Source: Former softie

Waco wrote:

Coding to a single bare metal design means you can tweak everything to be *perfect* on that one platform and you can take advantage of every low level access point into the hardware that is available. Time consuming, but actually possible on one platform.

This.

A good friend of mine is a coder who specializes in writing the code that allows game engines to work over multiple platforms - basically he writes the code that tells your GPU exactly how to render the stuff that the game engine wants it to draw. Actually does the math and whatnot. There is actually a ton of processing that goes into that - it has to take into account not only what card you're using, but your drivers, memory available, and a load of other stuff... it can actually be much more resource heavy than the game engine itself.

When you can write the game to work on one specific platform, you can avoid all that and can make things work remarkably fast.

The 8th-gen consoles have two whole cores set aside for OS stuff. PC games actually come much closer than console games to using the whole CPU in that regard.

The gap isn't all that big in the first place. CPU-side, if you take for any given game [console_framerate * 3.8 / 1.6] and compare to an i5-8400, you're probably a bit on the low side. GPU-side, everything matches up really well with equivalent PC graphics cards, PC gamers just tend to have much higher pixel-per-second targets and/or be running much higher settings.

Jaguar's IPC should still be dragging down consoles more than it is in a direct comparison, but the specific-hardware-target thing helps a lot there, and consoles have some more small advantages like not having to go over PCIe to communicate with the GPU.

synthtel2 wrote:

Jaguar's IPC should still be dragging down consoles more than it is in a direct comparison, but the specific-hardware-target thing helps a lot there, and consoles have some more small advantages like not having to go over PCIe to communicate with the GPU.

eDRAM on the xbox helps that out too.

NES was something like 1.7MHz. When did we get side scrolling games in our PCs?

The NES and all the 8/16bit consoles had custom ASICs that handled game-oriented 2D graphics (sprites, backgrounds, smooth scrolling) in hardware with minimal CPU intervention. The NES, for example, could do sprite collision detection in hardware while drawing the frame. I don't know much about graphics hardware on contemporaneous PCs, but I'd hazard a guess the common hardware you could rely on being present was less game-oriented, and I suspect in the worst cases the CPU would have to bang pixels directly into some framebuffer.

As for modern consoles, I think the gap isn't nearly as wide. Most of these games are running at pretty conservative visual quality settings and frame rates compared to what you'd typically go for on a PC.

Usacomp2k3 wrote:

synthtel2 wrote:

Jaguar's IPC should still be dragging down consoles more than it is in a direct comparison, but the specific-hardware-target thing helps a lot there, and consoles have some more small advantages like not having to go over PCIe to communicate with the GPU.

eDRAM on the xbox helps that out too.

In the Xbox One S yes, but in the Xbox One X that's been removed in favor of GDDR5 memory on a wider, 384-bit bus. Source: https://www.anandtech.com/show/11992/th ... x-review/3

ronch wrote:

I was just watching a YouTube video showing God of War on the PS4 and realized how good it is, not just graphically. Makes me wonder what makes consoles so much more efficient than personal computers. The PS4 is using only AMD Jaguar cores. You can't have that sort of performance from those Jags in a PC even if you slap in a nice GPU. Same with all the other past video game consoles. Look at the PS1 for example. 33MHz, little RAM. You'd probably need a Pentium MMX or something at 150MHz to get that sort of performance in a PC. NES was something like 1.7MHz. When did we get side scrolling games in our PCs?

The performance isn't as good as a PC.

God Of War on the PS4 only runs at 30fps (or less), which the console citizens have just learned to accept. While you an get an inexpensive 6 core (i5 8400) that will just run circles around the Jaguar.

For a direct comparison, in the latest BF5, you get 135fps average from the Core i5 8400 in 64-player match, Ultra terrain. The original PS4 sits at around 40fps average, and that's with way less detail in your terrain. Same number of cores as the game uses on the PS4.

They make up the motion difference by using the oldest tricks in the book: motion blur. I find that a travesty when PC console ports insist on using it.

In addition to everything that's been said, Santa Monica Studios (as well as Naughty Dog with TLOU/Uncharted) are downright wizards, man. I wonder what they could do with those same effectively unlimited budgets and talent if it was turned towards maxing out a modern PC.

Voldenuit wrote:

Efficient? The PS4 was drawing 264W at launch, and even then it was upscaling from 900p a lot of the time.

The PS4 most often ran things at 1080p at launch, and even now with demanding titles targeting the Pro more, 1080p is still more common than 900p, even if more of the latter are coming in. The XBO was and is the 900P box, sometimes down to 720p.

Its arguably best looking title and the topic of the OP happens to also be native 1080p even on the base 2013 box.

tipoo wrote:

The PS4 most often ran things at 1080p at launch, and even now with demanding titles targeting the Pro more, 1080p is still more common than 900p, even if more of the latter are coming in. The XBO was and is the 900P box, sometimes down to 720p.

Its arguably best looking title and the topic of the OP happens to also be native 1080p even on the base 2013 box.

There's running at 1080p and running at 1080p with visual quality settings you wouldn't feed to a dog. The original PS4/XBO is pretty outdated but the XBO X is probably around the level of a GTX970/980 which is not bad. If you've got a current high-end PC you're still the king.

Out of curiosity, I throttled my 8700k to 800mhz (using Windows 10's power plan settings, though curiously having the "Game Mode" Windows 10 option on seems to override it... always wondered what that did) which puts it roughly in the ballpark of the PS4's 1.6ghz Jaguar cores, going by the Geekbench scores for the Athlon 5150. With my GTX 1080, I'm averaging about 30-40fps in PUBG and 25-50 Monster Hunter World (the only current-gen games I have on my computer right now).

Doesn't seem like it would take any magical efficiency gains to get the PS4 CPU to run those games at 30fps.

blargh4 wrote:

Out of curiosity, I throttled my 8700k to 800mhz (using Windows 10's power plan settings, though curiously having the "Game Mode" Windows 10 option on seems to override it... always wondered what that did) which puts it roughly in the ballpark of the PS4's 1.6ghz Jaguar cores, going by the Geekbench scores for the Athlon 5150. With my GTX 1080, I'm averaging about 30-40fps in PUBG and 25-50 Monster Hunter World (the only current-gen games I have on my computer right now).

Doesn't seem like it would take any magical efficiency gains to get the PS4 CPU to run those games at 30fps.

And there's six of the Jaguar cores available for games (two reserved for OS), so that's between 25 to 50% better performance in online games. So yeah, you get 40-50fps in most online shooters, and 30fps in more demanding single-player games like God of War.

I'll give console game makers credit for finally scaling to engines with 6+ major threads, but this multi-core revolution was GOING TO HAPPEN. Between 8-cor e Jaguar consoles and DX12 on PCs, the multi-core future was a given.

But there is nothing inherently more efficient about consoles. One example: the scaling from 4 to 6 Jaguar cores is painful, because the communication goes over the local bus between the two separate dies. The dual quad-cores are hacked together, because it was easier to design it that way to release 6 months after Jaguar was officially released.

Consoles (aside from PS3) have always been about prioritizing GPU power, ram bandwidth and disc storage, while making the rest of the system as cheap to build and develop for as possible. That is why you get compromises like the Dreamcast's Hitachi SH-4 CPU, 360s in-order triple core PPC the Wii-U's triple -core PPC 750, (the same core they used on the Gamecube and Wii) or the PS4's dual-quad core Jaguar.

defaultluser wrote:

Consoles (aside from PS3) have always been about prioritizing GPU power, ram bandwidth and disc storage, while making the rest of the system as cheap to build and develop for as possible. That is why you get compromises like the Dreamcast's Hitachi SH-4 CPU, 360s in-order triple core PPC the Wii-U's triple -core PPC 750, (the same core they used on the Gamecube and Wii) or the PS4's dual-quad core Jaguar.

Not sure why you think the SH-4 was a compromise. It was a beautiful, beautiful RISC architecture with gobs of floating point grunt. PPCs - well, not exactly beautiful - but still much more elegant than x86. And then we entered the era where x86 ate everything...

defaultluser wrote:

Consoles (aside from PS3) have always been about prioritizing GPU power, ram bandwidth and disc storage, while making the rest of the system as cheap to build and develop for as possible.

What? PS3's design was not about prioritising GPU power, it was all about Cell, doing the whole thing in-house, and developers be damned 'build it and they will come'.

Cell was both a CPU that was way beyond most developers ability to harness and fully utilise for a good long time after getting access, and was originally planned to be used for graphics too, as a 2S config.

IIRC reading at the time, and in 2013 or so when people (Mark Cerny) started talking more about it, the NVIDIA GPU was not in the hardware spec until very late in development, it was bolted on in maybe late 2004 or early 2005, and doing so delayed the launch by a year, let Microsoft have that year's sales, and cost Sony its console market lead and maybe 20-100 million unit sales.

Usacomp2k3 wrote:

synthtel2 wrote:

Jaguar's IPC should still be dragging down consoles more than it is in a direct comparison, but the specific-hardware-target thing helps a lot there, and consoles have some more small advantages like not having to go over PCIe to communicate with the GPU.

eDRAM on the xbox helps that out too.

It's eSRAM on the One (eDRAM was the 360), and it isn't a general-purpose cache. It makes the lack of main memory bandwidth less crippling for the GPU, but the whole setup can't be classified as a clear advantage in any way aside from possibly bill of materials.

Topinio wrote:

IIRC reading at the time, and in 2013 or so when people (Mark Cerny) started talking more about it, the NVIDIA GPU was not in the hardware spec until very late in development, it was bolted on in maybe late 2004 or early 2005, and doing so delayed the launch by a year, let Microsoft have that year's sales, and cost Sony its console market lead and maybe 20-100 million unit sales.

The way I heard it described (in The Race For A New Game Machine), Sony were not planning on a 2S Cell. They were developing an in-house GPU design, much like how the PS2 had an in-house design with the Graphics Synthesizer. But something went wrong with development (I'm a bit fuzzy on the details), hence they killed it and went with NVidia's GPU instead.

(It's been several years since I read the book, so forgive me if I'm not remembering 100% accurately. I can dig it out and look it up if you want more detail!)

Pancake wrote:

defaultluser wrote:

Consoles (aside from PS3) have always been about prioritizing GPU power, ram bandwidth and disc storage, while making the rest of the system as cheap to build and develop for as possible. That is why you get compromises like the Dreamcast's Hitachi SH-4 CPU, 360s in-order triple core PPC the Wii-U's triple -core PPC 750, (the same core they used on the Gamecube and Wii) or the PS4's dual-quad core Jaguar.

Not sure why you think the SH-4 was a compromise. It was a beautiful, beautiful RISC architecture with gobs of floating point grunt. PPCs - well, not exactly beautiful - but still much more elegant than x86. And then we entered the era where x86 ate everything...

The SH-4 was fine if all you were doing was game engine control.

But the GPU on the Dreamcast had no T&L unit, so doing complex lighting quickly brought the CPU to it's knees. That's why most Dreamcast games use software lighting tricks for most effects (lightmaps), and couldn't keep up with with the rest of the consoles of that generation (with dedicated T&L units). It was becoming clear that the Dreamcast was all tapped out by year 2000, while all the other consoles had room to grow.

The PS2, Gamecube and Xbox all had hardware-accelerated lighting (even if developers had trouble harnessing at-first). If they wanted to stay relevant, they should have gone dual SH4, or a more powerful single core. It wasn't the ONLY reason for the Dreamcast's failure, but it was a major contributor to third parties dropping support so quickly.

Guys, I`ve got a question. PC upgrade or new console? What to spend money on in the fall? Buy yourself a new RDNA2 or Nvidia Ampere graphics card? Or buy a new console? The humor is that the choice isn't really that easy. Whether to take a console or prefer a PC - some pitfalls are not obvious at first glance. And another question, is it possible to play in a casino https://onlinecasinoscyprus.com/ using a console? Thanks

I'll give console game makers credit for finally scaling to engines with 6+ major threads, but this multi-core revolution was GOING TO HAPPEN. Between 8-cor e Jaguar consoles and DX12 on PCs, the multi-core future was a given.