| phinch1 said: PS1 had 2mb of ram PS2 had 32mb of ram, a power increase of x16 PS3 had 256mb of ram + 256mb of vram (512mb) again an increase of x16 PS4 8gb of ram plus 256mb for background tasks, again an increase of x16 the power
Ps5 131gb????
I can see next gen being very little difference between Ps5 and pc |
There's a flaw in your thinking.
Let's look at the read speeds of two of them...
PS2 had a memory bus bandwidth of 3.2 GB/s. PS4 has a memory bus bandwidth of 176 GB/s. That means that, in two generations, it's increased by x55, or a typical increase per generation of x7.5 (keep in mind, these are theoretical peak bandwidths, so it's not accounting for time required to search for data). So memory read speeds are limiting the benefit of the growth...
... but the bigger issue lies in the reading of data INTO memory. For discs... PS2 could read data in at a rate of 5.3 MB/s, PS3 could read data in at a rate of 9 MB/s, and PS4 can read data in at a rate of 27 MB/s. So it would take about 5 minutes to read PS4 blu-ray data into the PS4's memory to fill it up (compared with about a minute for PS3 and 6 seconds for PS2). Following this trend, even with a strong improvement in read speeds on the next disc format, it would end up taking of the order of 20 minutes to load it in.
Of course, a solution to this is to use hard disk data... but that only delays the issue. Even the fastest regular HDDs will only manage of the order of 100 MB/s, and while an SSD would improve on this, SSDs have an issue with frequent access. And even if we suppose the fastest of the SSDs were used, we're still talking less than 1 GB/s. So for 128 GB of RAM (not sure why you said 131 GB), it would still take 2 minutes to fill the RAM from the SSD.
The point I'm making, here, is that it's unlikely that so much memory would be used. Far more likely is that they'll save costs along the way by choosing a more manageable amount of RAM - maybe 16 or 32 GB, and instead spend the extra money on the cache, which is more likely to be beneficial following the trends at hand, or perhaps dedicated sections of RAM for special tasks that can be optimised in some way.
Keep in mind, this is likely to also apply to PCs, except the very top-end ones (as in, ones being used for professional/academic purposes). So it's possible that the RAM in the PS5 will be similar to that found in typical gaming PCs of the time. But it's unlikely that they'll go as high as 128 GB so soon, barring some significant breakthrough in technology (that would need to go beyond the trends given by Moore's Law, since that doesn't predict doubling every two years for speeds).
Also keep in mind that much of what gets loaded into RAM needs to essentially be hand-crafted at some point - we're talking meshes/textures/etc. There's already a problem with the cost of graphical designers in videogames, and increasing it by a factor of 16 again would basically kill most video game companies. Since very few would ever try to use that much, putting it in there would be a waste. It's likely that new techniques will be developed to try to cut the costs in this department... but such techniques would likely also cut amount of memory required (by being able to generate the textures/meshes/etc on the fly, avoiding the need to store them in memory).
To summarise, while the "optimal" amount of RAM that could be put into a system for a specific cost should roughly double every 2 years, other factors mean that RAM costs should instead begin to decrease, rather than keeping it constant with rapidly-increasing RAM size, as Moore's Law doesn't apply the same to every facet of technology.
