What exactly will faster ram provide over a larger pool of slow ram?
in non-technical terms i think it will provide higher frame rates.. But this is just my educated guess.
http://semiaccurate.com/forums/showpost.php?p=171629&postcount=380 said:sirroman said:If they actually go with off the shelf parts, late 2013, but 2014 is far more likely as fatty2nd brought up.
It isn't off the shelf parts, not even close.
-Charlie
Don't you mean more everything? (provided the smaller pool of ram is not inhibitedly small). You can never have too much bandwidth
Charlie at SemiAccurate
Wow a one liner with MASSIVE implications!!!!!
I think the latest Durango rumor or leak confirms Jaguar CPU packages but it also has HDMI in and out which I think is going to be unique to a home Console connected to a TV (HDMI pass-thru with overlay) design to support XTV. The other parts of leaks confirming other AMD building blocks like 16 gig SSD flash and low power modes fit with the goals needed to make happy consumers (fast boot, fast app switching etc.) for any CE platform be it PC or tablet.
So not even close to off the shelf parts in performance?
Game console specific features?
Not current off the shelf building blocks but from future (2014+) designs?
What exactly will faster ram provide over a larger pool of slow ram?
What does 'heavily' mean?
I can't imagine they're doing too much stuff to the main chips. I think I remember one of the rumours talking about optimisation for tessellation and finer-grained compute, but I thought those things were just a by-product of GCN anyway.
Cramming a Pitcairn level GPU onto an APU could fairly be called a heavy modification. So could a 2.5D RAM stack. Neither would really cause headaches for developers, PS3 style, though. I don't expect them to be tinkering with the micro architecture, but whatever it has could look pretty different from AMD's consumer portfolio in terms of how it is packaged.
What does it matter what crammed into and onto what? They'll still be current off the shelf parts, even crammed into a one APU altogether. From a programming point of view, from what's possible on such system and from perfomance perspective it'll be mostly the same as a PC made of such parts right now.Cramming a Pitcairn level GPU onto an APU could fairly be called a heavy modification. So could a 2.5D RAM stack. Neither would really cause headaches for developers, PS3 style, though. I don't expect them to be tinkering with the micro architecture, but whatever it has could look pretty different from AMD's consumer portfolio in terms of how it is packaged.
What does it matter what crammed into and onto what? They'll still be current off the shelf parts, even crammed into a one APU altogether. From a programming point of view, from what's possible on such system and from perfomance perspective it'll be mostly the same as a PC made of such parts right now.
Sony should take some tablet chipset, just over-clock it, and call it a day. Would save a huge amount of money.
They could call it the Vitastation.
Sony should take some tablet chipset, just over-clock it, and call it a day. Would save a huge amount of money.
How about no?
For high-end gaming PCs these days you'd need too much eDRAM, at least until stacking becomes viable. Don't forget that for the high-end PC customers resolutions are higher and IQ is massively higher than on consoles. And for modern games it would also be a huge software challenge to figure out in the driver what to put into the eDRAM.probably(?) more cost efficient to go for lots of 'normal' ram and then a nice size of edram for bandwidth critical elements. At least enough for high quality deferred rendering at 1080p
actually surprised PC GPUs haven't tried that. Perhaps they can't rely on optimisation for that architecture so everyone is bound to an arms race of more and faster ram which costs a lot
You can back calculate how much ram (and how fast) you might need, if you use a few reference examples for different types of engines, target resolution/framerate/AA and your expected GPU specs.
For high-end gaming PCs these days you'd need too much eDRAM, at least until stacking becomes viable. Don't forget that for the high-end PC customers resolutions are higher and IQ is massively higher than on consoles. And for modern games it would also be a huge software challenge to figure out in the driver what to put into the eDRAM.
Really, at current graphics card prices it makes most sense (and is by far the most flexible) to just to put a lot of fast RAM on a wide bus on there.
Sure, I think it's a good idea. As long as you can make it large enough as not to seriously limit IQ.How about for a 1080p target (ie consoles)? feasible to use edram?
eDRAM can be a good design decision in a closed box regardless of how much external memory bandwidth you have. (Well, maybe not if you have some unrealistic number, but for every realistic one) It doesn't have to be in a low-end system just because that's how it is used in Wii U (see: 360 at launch).What's this talk about EDRAM? I thought that neither of the next gen boxes would be using it. Is the RAM going to be that slow to require additional support in the form of EDRAM? Please let it not be the 64-bit bullshit.
DDR4 hasn't even hit mass production has it? i think they'll stick with GDDR5.A 256Bit GDDR5 implementation would be faster than a 384Bit DDR4 solution - so with a realistic bus/cost estimation I don't see any way around eDRAM if Sony really goes with DDR4 to be "future" proof when it comes to shrinking and new processes. It seems GDDR5 will slowly die out in the next few years.
DDR4 hasn't even hit mass production has it? i think they'll stick with GDDR5.
I hear you but you have to understand that even some "ultra wide IO" in a stacked console die probably won't be anywhere near 256 bit DDR3 main memory bus (provided by Intel's socket 2011) plus a 256-384 bit GDDR5 busses in current GPUs. Thus it will somewhat help with bandwidth problem without resorting to wider external busses but ultimately it'll still be slower than current top end PC configurations. At best we may be looking at some sort of fast CPU-GPU interconnection that may allow such system to do some stuff much more efficiently than on todays PCs. But even that stuff won't drop anyones jaws I'm afraid.Early on this year an article I read had wide IO 3D stacked RAM as not a PC option for a few years because PC users like to be able to choose their own RAM Configurations with DIMM modules; So handhelds, tablets and CE platforms but not PCs. This means that it's likely that Game consoles will be some of the first CE equipment to have performance 3D stacked ultra wide IO DRAM rather than Low Power. This is why the Kabini (Tablet) design is more likely to have ultrawide IO DRAM than Kaveri the desktop design.
** The base design for next generation consoles should be Kabini (with performance 3D stacked RAM (DDR4)) not an A10. **
Both Intel and AMD fusion PC designs will be going embedded ultrawide DRAM at some point and the designs will reflect this. When is not known but already both have known bandwidth issues with AMDs design suffering more. Clocking AMD APU DRAM faster gives a higher increase in performance than overclocking DRAM connected to Intel Fusion designs. So a 1.5+ year future AMD Desktop PC design could be in next generation game consoles.
I hear you but you have to understand that even some "ultra wide IO" in a stacked console die probably won't be anywhere near 256 bit DDR3 main memory bus (provided by Intel's socket 2011) plus a 256-384 bit GDDR5 busses in current GPUs. Thus it will somewhat help with bandwidth problem without resorting to wider external busses but ultimately it'll still be slower than current top end PC configurations. At best we may be looking at some sort of fast CPU-GPU interconnection that may allow such system to do some stuff much more efficiently than on todays PCs. But even that stuff won't drop anyones jaws I'm afraid.
Effectively we're still looking at an off the shelf AMD x86 CPU and off the shelf AMD DX11 GPU, no matter whether they're on one die, or stacked, or ducked-taped, or whatever.
Building blocks off the shelf I think yes but the core that those building blocks are attached to, the Memory and registers, the glue logic, the Xbar switches will be custom with a possibility they are future high performance designs. Think about the HDMI pass-thru that is supposed to be in the Xbox 720 and I think in the PS4, that is not going to be off the shelf and will have to be interfaced with other AMD building blocks.http://semiaccurate.com/forums/showpost.php?p=171629&postcount=380 said:sirroman said:If they actually go with off the shelf parts, late 2013, but 2014 is far more likely as fatty2nd brought up.
It isn't off the shelf parts, not even close.
-Charlie
I hear you but you have to understand that even some "ultra wide IO" in a stacked console die probably won't be anywhere near 256 bit DDR3 main memory bus (provided by Intel's socket 2011) plus a 256-384 bit GDDR5 busses in current GPUs. Thus it will somewhat help with bandwidth problem without resorting to wider external busses but ultimately it'll still be slower than current top end PC configurations. At best we may be looking at some sort of fast CPU-GPU interconnection that may allow such system to do some stuff much more efficiently than on todays PCs. But even that stuff won't drop anyones jaws I'm afraid.
Effectively we're still looking at an off the shelf AMD x86 CPU and off the shelf AMD DX11 GPU, no matter whether they're on one die, or stacked, or ducked-taped, or whatever.
Sure, I think it's a good idea. As long as you can make it large enough as not to seriously limit IQ.
eDRAM can be a good design decision in a closed box regardless of how much external memory bandwidth you have. (Well, maybe not if you have some unrealistic number, but for every realistic one) It doesn't have to be in a low-end system just because that's how it is used in Wii U (see: 360 at launch).
If I had to chose a number, 64 MB.what would you consider a 'comfortable' amount of edram to support 1080p deferred rendering, assuming main ram is a little slower?
my friend was saying that uncharted 4 development is going to start soon for next gen
His couzin had done some outsourcing work for uncharted 3 and is currently working on few stuff on a new game from sledge hammer and uc4
dont know how true is this but he will get more info in dec
I would figure it would have been in development for a while by now. I don't think it'll be a launch title since ND likes to work with final hardware for a while and they said that after Uncharted 3 they're not going to lock themselves into a release date too long before launch but it'll almost definitely be out within the first year of the PS4.my friend was saying that uncharted 4 development is going to start soon for next gen
His couzin had done some outsourcing work for uncharted 3 and is currently working on few stuff on a new game from sledge hammer and uc4
dont know how true is this but he will get more info in dec
my friend was saying that uncharted 4 development is going to start soon for next gen
His couzin had done some outsourcing work for uncharted 3 and is currently working on few stuff on a new game from sledge hammer and uc4
dont know how true is this but he will get more info in dec
my friend was saying that uncharted 4 development is going to start soon for next gen
His couzin had done some outsourcing work for uncharted 3 and is currently working on few stuff on a new game from sledge hammer and uc4
dont know how true is this but he will get more info in dec
I mean outsourcing stuff has started that what i heard.
I mean outsourcing stuff has started that what i heard.
Unless they go completely overboard with specs it'll almost definitely be $399. You can make a pretty badass system for that price target and it won't be a $100 or more loss at launch like the PS3 was.If they can do this at 399$, they'll be in a good position. Specs sound about what I'm hoping for, so far.
Unless they go completely overboard with specs it'll almost definitely be $399. You can make a pretty badass system for that price target and it be a $100 loss at launch like the PS3 was.
Unless they go completely overboard with specs it'll almost definitely be $399. You can make a pretty badass system for that price target and it be a $100 loss at launch like the PS3 was.
The investors may take less of a liking to a loss that large per console (or any loss at all, frankly)
Typos on my phone messed up my post. Fixed it now.PS3 was a much larger loss. Hundreds of dollars per unit.
I don't expect Sony to take a huge hit on PS4. $50 at most.
Has anyone speculated on how the rumored PS4 compares/contrasts to the rumored Xbox World details on the Xbox 720 (...cant seem to find this thread on GAF for some reason)?
Also, Sony has overcome the Blu-ray hurdle and doesn't have to worry about the cost for that. I really hope they can hit $399. If this thing is not gonna have backwards comparability, they have to try to make it as cheap as possible to get the hardcore to buy it.
They'll have a gimped ps4 $399 (16gb with one controller). Then another proper one at $449 160gb with all the bells and whistles.
Hard drives make all the difference to the stupid consumer.