Predict: The Next Generation Console Tech

Status
Not open for further replies.
Huh? 7-25% of a console cost (for a $399 machine) should be considered just for RAM (not even the BOM for them either), not to mention slow RAM that can be anywhere from 16-48 chips and be utilizing a DIMM format rather than lower latency tracing on the motherboard for the next 10 years? :oops:

To be fair, we're talking about a large global company buying in bulk, so I think they could get a better price.

I've been thinking about your comment. Maybe it IS worth it to have gobs of RAM be a major portion of your costs. It's a common complaint this gen. Just kind of kicking it around in my head.

The more I think about it, the more I think SOMEBODY will try 256 bit bus or at least give it serious thought til late in the process.
 
If ram is going to get them the best performance than that's where they should spend it. There's no point in spending the money on higher perf CPU/GPU if they are going to be bandwidth starved. I remember for years and years assembling PCs and ram being a significant portion of the build cost (at least 10-20%) it's only in the last few years that ram has really gotten cheap.

I don't know where that is going to be 2 years or so, but unless there's some major improvements in ram clocking I just don't see how they are going to meet bandwidth requirements without either gobs of edram or a 256bit bus. They've certainly proven this generation they can hold the price of consoles at a higher level.
 
I've been thinking about your comment. Maybe it IS worth it to have gobs of RAM be a major portion of your costs. It's a common complaint this gen. Just kind of kicking it around in my head.

Well, it's going to be the cost of the number of chips, not the capacity per se, and there's clearly the implications to the physical mainboard as well.

It's a little too easy to say that there should be 8-16GB of RAM just because one can do it on PC...

There's no point in spending the money on higher perf CPU/GPU if they are going to be bandwidth starved.

I'd say that's debatable where focus will be on ALU ops going forward. :) Don't forget that texture bandwidth reqs can be alleviated to a degree with larger caches on the GPU, for instance.
 
You're absolutely right,its 14 chips....and picture shows EE + GS and a south bridge give us possibility PS4 with the same dimensions ps3 could be more than enough to fit 16 DRAMs(30nm ?) XDR2/8GB and 1TB/sec bandwidth.

Ive already mentioned this before ... in the latest revision of MB they are using 256MB chips for both GDDR3/XDR

file.php


If sony will use 256MB chips in PS4 we will see 12x256MB or 14x256MB which is

3072MB or 3584MB of totall RAM ...



I dont know if 512MB chips exists of XDR2 or GDDR4/5
 
Well, it's going to be the cost of the number of chips, not the capacity per se, and there's clearly the implications to the physical mainboard as well.

It's a little too easy to say that there should be 8-16GB of RAM just because one can do it on PC...
Indeed. The consoles effectively use only the VRAM component of the PC. If a PC can be bought with 8Gbs for VRAM, then there's an argument. At the moment, 2GB GDDR5 graphics cards cost hundreds of pounds. And they are driving resolutions twice as high as 1080p, possibly even multiple monitors. 4GBs of GDDR5 will be ample storage in console. The idea of a pool of DDR3 as streaming cache makes some sense, but given the complexities of the memory architecture, perhaps it doesn't present a great economical argument? Thus we're back to a common theme of this thread - Fast ram for the system, and multiple GBs flash RAM to aid loading and streaming.
 
This discussion has one major flaw that consistently pops up.

People are trying to build 2014/2015 consoles with 2011 tech and prices.

When people with PC hardware background come here, they assume correctly that since prices of computer components halve every 18 months, this reduction should be taken into account when talking about the 2014/2015 timeframe (that would be a 4x reduction by 2014 or a 8x reduction in 2015,5 compared to current prices).

Like I said, even 8GB of GDDR5 is not off the table by 2014 (this winter brings first 4GB GDDR5 graphics cards), when talking about what is technologically possible as well as monetarily feasible in 2014/2015.
 
People are taking into account falling costs when talking about 128-bit busses, edram, using 8 memory chips etc. The cost of processors will fall in a similar fashion to the cost of memory (possibly more due to greater reductions in power hardware and cooling) so bear in mind that when you're speccing PC levels of memory you're inevitably taking from other areas, such as the processing side.

If you'd done this for the PS3 or the 360 you'd have ended up with much weaker systems.
 
The idea is not to take something away from the existing hardware. It is to add extra RAM and eat the initial moderate costs of maybe $15 extra per box on the first 2-5 million units in the knowledge that RAM prices come down very quickly, maybe even by the end of the first year on market, after which you'd have a well-specced box which no longer loses money on account of RAM.

It plainly makes sense to put into your box more of the component which will have the greatest price reduction over time.

CPU/GPU are much harder to manufacture and much harder to shrink to bring about cost reductions. In the console world especially, because you pick one architecture and you're stuck on it, meaning that only production process node reductions can help you bring down the price significantly (90nm > 65nm > 45nm > 32nm > 28nm > 22nm etc). On the PC, you can do architectural redesigns to squeeze out more power or take up less area even if the process node stays the same.

Not all of these process node reductions will work out in expected timeframes and deliver acceptable yields *and* be delivered on time, therefore fixed console CPU/GPU price reduction is not as certain as the reduction in RAM prices. To illustrate the unpredictability of the process node advances:

90 nm — 2002
65 nm — 2006
45 nm — 2008
32 nm — 2010
28 nm — 2011
22 nm — 2011
 
90 nm — 2002
65 nm — 2006
45 nm — 2008
32 nm — 2010
28 nm — 2011
22 nm — 2011

By this chart the consoles today are on a 2008 process and will be into 2012. Consider that.

Intel is years ahead of everybody else and imo process nodes are slowing down, not speeding up as this table erroneously suggests at a glance. TSMC had a huge fight getting 45 nm out and 28nm will likely be as bad or worse, every step down is more like pulling teeth.
 
Ive already mentioned this before ... in the latest revision of MB they are using 256MB chips for both GDDR3/XDR

http://www.edepot.com/forums/download/file.php?id=260&sid=3f667b87c33aa3f7e95233e894706c51[/IMG]

If sony will use 256MB chips in PS4 we will see 12x256MB or 14x256MB which is

3072MB or 3584MB of totall RAM ...



I dont know if 512MB chips exists of XDR2 or GDDR4/5

Perhaps in 2013/2014 the mainstrean is 4gbit modules(16 modules theory...enough to 8GB RAM) or even more,today we have currently only 2gbit modules at GDDR5*(at 45nm only) and probably Rambus could reach same,but if sony (going with new XDR2/3) contract and needing millions modules maybe reduce size and reach 4gbit or more(28/32nm?).

Until i see MB ps3 launch i think its only possible come console with 8 modules but im change...see my post about DDR4 possibilities too:

http://forum.beyond3d.com/showpost.php?p=1575596&postcount=6987


kaigai-09.jpg




*
http://en.wikipedia.org/wiki/GDDR5
 
Last edited by a moderator:
People are trying to build 2014/2015 consoles with 2011 tech and prices.
That's untrue. People have been speculating by extrapolating trends. Also bear in mind this thread has had a moving release schedule. We don't know when the next boxes might launch. There's even suggestion of a surprise appearance by the next XBox by 2013. If someone were to estimate XB3's specs based on 2015 tech and it launches in 2013, they'll be as off the mark as the other way round. ;)

When people with PC hardware background come here, they assume correctly that since prices of computer components halve every 18 months, this reduction should be taken into account when talking about the 2014/2015 timeframe (that would be a 4x reduction by 2014 or a 8x reduction in 2015,5 compared to current prices).
Okay. So Googling, a £366 Graphics with 4GBs RAM now will lead to a £183 graphics card with 4 GBs VRAM in 2013 leading to £90 for a 4GBs VRAM in 2014/2015 with a 2011 GPU. Add all the other gubbins of a console like CPU, up-to-date GPU, motherboard, controllers, devices, and you have a 4GB console for £300, no?

Like I said, even 8GB of GDDR5 is not off the table by 2014 (this winter brings first 4GB GDDR5 graphics cards), when talking about what is technologically possible as well as monetarily feasible in 2014/2015.
And we've said 8GB isn't off the table for next-gen consoles.

The idea is not to take something away from the existing hardware. It is to add extra RAM and eat the initial moderate costs of maybe $15 extra per box on the first 2-5 million units in the knowledge that RAM prices come down very quickly, maybe even by the end of the first year on market, after which you'd have a well-specced box which no longer loses money on account of RAM.
Except that doesn't factor in the other aspects. And lets not forget the financial situation of the consoles. MS and Sony have lost how many billions on these boxes? An extra $15 (a very conservative pricing I think) for 10 million units (first year sales before the RAM price drops to an extra $7 overhead) would cost an extra $150 million. The next ten million units would lose another $70 million. Realisitically, it'll be way more than that. A quick Google shows 4 GBs VRAM costs a £60 premium over 2 GBs RAM. Using your estimation method, and taking £60 retail price difference to be of the order of $50 added cost, the difference between 4 GB and 8 GB of GDDR5 in 2014/15 will be around $25.

Now what does Sony or MS gain for that? If their box has more RAM, at added losses to themselves, but devs don't make use of it because they are creating cross-platform titles that'll sell to a box with less RAM and adding improved assets to the more-RAM console won't net them more sales on that console than sticking with the low-RAM assets, then it's loss for no purpose. You only need see previous generations to see RAM advantages rarely lead to significant gains.

I'd say quite contrary to PC users having added insight, they have a limited perception. In the PC space, specifications are everything. People buy a 4 GB laptop over a 2 GB laptop because it's 'twice as good' for £10 more, even if they'll never even hit the limit of that 2 GB machine. As the numbers are the only differentiating factor in the PC space, there's no point in not pursuing better numbers for your money when the economies are so good. I'd buy a £65 1TB HDD over a £60 500GB HDD even though I'll never fill it because it's way better value (actaully a false economy, because it'll cost me £5 more for no actual gain, but we human's like bigger numbers! ;)). Consoles are instead CE devices, and they are all about the experience. One machine with $50 more RAM can readily lose out to a rival with that same expense invested into services or marketing. Or pulling a Wii and flipping the specs on their head to offer the experience. And this is kinda happennig with computing. £400 will net you far more laptop for your money than a tablet PC, but people are choosing the functionality of tablets despite less RAM and processing power. Smaller numbers don't matter compared to the overall experience.

What seems so straightforward on paper - an extra $15 per unit BOM to double the RAM - is not at all such a straightforward choice. Every dollar spent on hardware is a dollar less to spend on the rest of the chain. A dollar towards a vaulable exclusive deal (Gears has done wonders for MS this gen). A dollar towards a slick marketing campaign (Wii wouldn't be doing as well as it is if not for the superb lifestyle happy-families marketing campaign that has accompanied it). A dollar towards securing a valuable content partnership such as Netflix to gain added-value to your system. Then there's also the consideration of what it costs to use added RAM, and to design a system that can load and work with 8GBs without being slow versus a 4 GB system. Or picking 4GBs of faster RAM, clocked higher with a more expensive and effective cooling system, over 8 GBs of slower RAM. Or investing in devtools so they can actually use your machine's hardware. Many, many choices to be made, with none able to be viewed in isolation. We can't just add only $5 more flash there, only $15 more on RAM, and only $7 more on IO ports, thinking they are small numbers that won't impact things. The only way to do it is set yourself a cost and design stringently to that, building a system to a strict budget carefully balanced out to give the best bang-per-buck.
 
Well, it might not be as straight forward, but if you factor in that a small change like this can mean that all games will look better, it might be worth it. It's the same change as what MS did with the 360. Imagine this generation if 360 only had 256MB of RAM and compare that to a PS3 with 512MB RAM. ALL multiplatform games would've looked a LOT worse on 360, although it has a better GPU and eDRAM. In the end, it probably would've killed the 360 in the long run.
 
Shifty, listening to you one might get the idea that (V)RAM is not important at all :)

I can tell you that in view of a 10 year life cycle, it's a game-changer if you have twice the RAM of your competitor's box. Yes, it will not all be used immediately, but Cell and Xenos weren't either.

2-3 years after launch though, that's when these "small numbers" begin to mean the difference between the DreamCast and the PS2.

8GB could potentially mean twice as crisp textures, twice the rendering resolution (assuming no EDRAM), or who knows how many times the level size compared to 4GB.

And talk of the billions that MS/Sony "lost" (this is known as an investment, not loss) is inappropriate in the face of the billions they have made and will still make. ;)
 
Disingenious?

Ok, so what evidence do you have that RAM is a gamechanger between consoles? Now you're just making up excuses for why Xbox's double RAM didn't push its sales forward. Clearly, having double the RAM didn't make a huge difference them. ergo, it's not a game changer.


2-3 years after launch though, that's when these "small numbers" begin to mean the difference between the DreamCast and the PS2.
You do realize that RAM was not the instigator for DC's eventual failure.
 
Well, it might not be as straight forward, but if you factor in that a small change like this can mean that all games will look better, it might be worth it. It's the same change as what MS did with the 360. Imagine this generation if 360 only had 256MB of RAM and compare that to a PS3 with 512MB RAM. ALL multiplatform games would've looked a LOT worse on 360, although it has a better GPU and eDRAM. In the end, it probably would've killed the 360 in the long run.
That's true, but if PS3 had launched with 2 GBs RAM, it wouldn't have been any better off and would have lost Sony loads more money. It's a balancing act. 512MB was just enough this gen. Less would have been disasterous. More would have been good for gamers, but lost MS and Sony hundreds of millions on top of what they've already lost, so it was probably the right business decision even if not the best choice for the best value hardware.

Shifty, listening to you one might get the idea that (V)RAM is not important at all :)
Not at all. I'm just explaining why 4GB is a very realistic target.

I can tell you that in view of a 10 year life cycle, it's a game-changer if you have twice the RAM of your competitor's box.
XB had twice the RAM of PS2. Which one lasted ten years?

8GB could potentially mean twice as crisp textures, twice the rendering resolution (assuming no EDRAM), or who knows how many times the level size compared to 4GB.
It'll mean zero times the level size for cross-platform titles. It'll possibly mean some better texture fidelity, but the difference between 4GB of assets at 1080p and 8GB of the same assets probably won't be hgely perceptible to Joe Gamer. More RAM will give exclusives like whatever ND produce an edge, but will that be enough swing the market against all the other forces in play?

I don't disagree with any of your points about the benefits of more RAM. If I had the choice, I'd pick 8GB for my next console over 4GBs. However, it's not down to gamers to design these machines, but businessmen and engineers in collaboration with their software partners. There are lots of reasons to argue against 8 GBs, such that it'd be a lower-probability result IMO.

I can't find any talk predicting this gen during the early 2000's. The internet was fairly young then, and gamer talk wasn't as huge. I'd be very interested to see what people were reasoning for the next console RAM amounts.

And talk of the billions that MS/Sony "lost" (this is known as an investment, not loss) is inappropriate in the face of the billions thay have made and will still make. ;)
MS hasn't made any money off their console business yet. Factoring in the billions lost with XB, it'll be a while before they become profitable. Sony seems to have squandered the profits of their last two runaway successes on this generation alone! It extremely hard to make big bucks from consoles!
 
I do, I meant only graphics.

Well of course more ram is going to have an effect on graphics, but now you're trying to justify a significant cost when there is no evidence to show that graphics are some mythical gamechanger for sales.

Would Sony have bothered with 512MB total if MS didn't? The second company to launch (in the "true" next-gen) will want to match the first company in some aspects to the hardware due to multiplatform considerations.

Consumers won't care about texture resolutions as long as they're in the same ballpark. Will they even care about resolution differences? Only the technically informed will care, and that's unlikely to be a majority.

Although RAM is necessary for resolution on multiple fronts (textures, render targets, shadows etc), you're also asking for a lot more from the GPU. Doubling the RAM does not allow you to double the render target dimensions without sacrificing per pixel work or just raw pixel throughput. Then you have to consider the extra load on the TMUs... then the basic jump in bandwidth requirements...
 
It'll mean zero times the level size for cross-platform titles. It'll possibly mean some better texture fidelity, but the difference between 4GB of assets at 1080p and 8GB of the same assets probably won't be hgely perceptible to Joe Gamer. More RAM will give exclusives like whatever ND produce an edge, but will that be enough swing the market against all the other forces in play?

Try 2-4x times bigger levels on for size ;)

The number one reason why X360/PS3 game levels currently look like tunnel instead of a maze is too little RAM (and I'm saying it should've been 1GB).
 
Status
Not open for further replies.
Back
Top