setsunatenshi said:
1- on the speculation itself, we have to assume a more likely situation, there is only so much cost the console manufacturers are willing to take. My point still stands, keeping the esram is a mistake and the sooner they ditch it, the better for them. Judging by the rumors that they are going to have a 5.5 TFLOPS console, they can't have their cake and eat it too.
|
Disagree. If Microsoft moves the eSRAM off chip and builds it at 28nm, it could actually be very cheap to manufacture, there is going to be tons of free capacity on the 28nm node across the planet and an idle fab is money being lost, not all of them will be retooled either, so fabs will often make special deals to lower the price just so they are being used and bringing in revenue.
Again, there are benefits in doing this, lower power consumption, more performance, that sort of thing, Microsoft could also forgo excessive amount of L2 and L3 cache in it's successor CPU's and maintain the eSRAM/eDRAM as a chunky L4, which does close the transister gap considerably.
5.5 Teraflops isn't high-end or expensive either in the PC landscape, nor is it an accurate denominator for gauging performance.
Polaris is managing it, but that also includes Gigabytes of GDDR5 just for itself, PCB with tons of traces and layers, probably an expensive copper cooler, likely with a vapor champer design with a blower, over-engineered power delivery system, etc'.
Microsoft will be buying just the chip, in bulk with other chips likely consolidated into the same package/silicon.
setsunatenshi said:
2- it's irrelevant for my point if the best hardware is the one being targeted first, I agree that it won't actually since the news are that the games will run both on the vanilla and Neo PS4s. it stands to reason that the game must be playable well on the vanilla system first and foremost. my hypothesis is that they have an opening to mantain the same architecture for a future PS5 (keeping R&D costs low, getting cheaper parts, no exotic hardware in it) and then be able to give a reason for people to upgrade, knowing they will have 100% backwards compatibility with their older titles.
|
I think the consoles will just use the extra power to turn on the extra effects that are typically only turned on in their PC releases, so likely won't be much of any extra dev time.
setsunatenshi said:
3- i'm well aware of the differences between the console and PC ecosystems. what i'm saying is that starting with this generation there is very little seperating the 2, other than the OS the machines are running. the reason why games need to be patched and tweaked around sometimes to work is mostly because of the OS itself. If I try to play half life 1, probably what I would need to do would be to use compatibility mode for win xp and the game will work just fine. Notice that when the game came there was no such thing as a PCI-e port, SSDs or GDDR5 graphics cards, and still the game will work through brute force.
|
There are a few big differences between PC's and Consoles.
For one, games on PC's target high-level abstraction layers, which can conceal differences in hardware to an extroadinary degree.
On a console, you have low-level abstraction layers where games target the various nuances in the hardware more intimitely, which means that things get "broken" if there is a change in hardware that it relies on.
Now if we look at the difference in technology between the Xbox 360 and Xbox One, the first obvious change is the change in how Tessellation is done, the Xbox One doesn't use N-Patches in order to perform tessellation, meaning that the Xbox 360's and Xbox One's Tessellation is actually incompatible from a hardware perspective, even though they are made by the same company.
What this means for games that used the Xbox 360's approach to Tessellation (Aka. Truform) is simple, they don't have any Tessellation on newer PC's.
GPU's tend to have a myriad of fixed function hardware, even today, that games rely on, over time it gets consolidated into other parts of the chip to introduce flexibility... An example of this being TnL, it's no longer fixed function anymore.
There is also software and games that relied on specific SIMD instructions found in MMX and 3D Now! Which are also no longer present in newer processors, those same pieces of software and games are no longer functional.
Of course, there are ways around that, you can use yet another API to interface and abstract things farther or you could take the Xbox One's approach and virtualize the old software environment and emulate the individual pieces of hardware, which does come with some caveats of course.
setsunatenshi said:
The same think should be valid of a future PS5 in which keeping the same architecture as the PS4 but with much better performing hardware, all previous games should work with little effort.
In the end I guess we can agree there's several variables at play here and we don't have the full information yet. Perhaps in a few months it will become more clear what paths both companies will take. I doubt that specs will be given next week at E3, but who knows, maybe a mid-long term vision could be discussed by both Sony and MS, and we'll have some more data to speculate on.
|
Yep. We do need more information, then we can talk about it in more detail then and there. :)
