I/O Throughput Data Speeds

[Pemalite]

JRPGfan said:

Pemalite said:

Think of Ram as a water tank... And the CPU and GPU as the taps, by constantly running the taps, you lower the tanks water level... So in order to top of up the water tank you need to fill up with more water from the storage drive... An SSD will add water faster than a mechanical HDD.

In short for every second of screen time, the Playstation 4 is able to render up-to 5.5GB of data, because that is how large it's Ram pool is... But over 30 seconds the mechanical hard drive can extend that by an additional 3Gb of data by streaming in assets.

Hope that has cleared it up.

Everyone clinging to certain advertising buzzwords is how the confusion has set in... But it's nice to know every generation we have a group of people who become CPU/Ram/SSD experts over-night, so that's great.

I liked this way of explaining it.
Water tanks and water taps :)

"In short for every second of screen time, the Playstation 4 is able to render up-to 5.5GB of data, because that is how large it's Ram pool is... But over 30 seconds the mechanical hard drive can extend that by an additional *1,5-3Gb of data by streaming in assets."

So around 8,5 GB over 30 sec (if drive is not anywhere near full or fragmented, and is running near its optimal 100 mb/s).

Playstation 5 could have all that in its memory pool alone.
Or load it in less than 1sec.

Tried to keep it as simple as possible.

In the Playstation 5's case... If we were to assume the OS/Background stayed static at 2.5GB of memory consumption it could have 13.5Gb of data and being topped up with 165GB of data over 30 seconds... This is why it's such a huge selling point over mechanical drives... Because right now games are "guessing" data 30 seconds ahead of time out of need.

It's all about the water in/water out.

KratosLives said:

Do you know what the maximum data of game textures in gb that can be rendered on screen per second? Just curious as to how much more detail can be shown per frame compared to ps4.

Difficult to quantify as other assets need to be loaded into memory like game code, meshes, sound bytes and so forth... And then you have various compression algorithms with different degrees of compression ratios. (Plus lossless and lossy compression algorithms)

And of course... Asset streaming.

My suggestion for you is to recognize the ram pool, it's bandwidth and the I/O speed as separate constructs that can feed into each other rather than try and formulate a number to apply a label to it all as different developers and different games will all use them differently.

kirby007 said:

2 things :

how does this relate towards different framerates so a game running 30 fps or 60 fps or even 120 fps?

Does loading 16GB/s into the ram leave room for other tasks?

It doesn't affect total framerates. The limiting factor to framerates is not I/O.

It will ensure more consistent frame-rates however as there won't be "hitches" as data/cells gets loaded into memory.

Yes, loading 16GB/s into ram would leave room for other tasks.

kirby007 said:

higher FPS with the same quality assets has a higher strain on system memory doesn't it?

Sometimes. Games aren't a binary load... Depends on where the bottleneck lays.

Essentially a higher framerate will demand/require more bandwidth due to things like bandwidth heavy alpha effects rather than an increase in memory consumption itself.