Velocity Architecture vs PS5 SSD

Yup, this. There are also many tasks that could be done using ML that don’t need to be doing huge computations every frame the way SS does. Any compute spent doing it though is coming from a budget that could be spent for other stuff. In the Spiderman implementation so far it looks like traps and biceps are the only muscles using it?

Getting a 1Tb expansion right off the bat was a solid decision on my part.

It is good to know that games on the expansion drive will operate at the same speed as internally.

For my PS5, I don’t know what the heck is going on with the expansion…

Hey, but Cerny said…!

1 Like

Why? They can just up the RAM requirements on PC which will negate the reason for having an SSD.

1 Like

True, good point.

As simple as this. Special SSD on console are not creating a new paradigm. They are simply closing the gap of DRAM.

A Gaming system having 3060 (12 GB) won’t have any problem running PS5’s SSD games. That system will generally have a total of 28GB total RAM (12 GB vram and 16 GB RAM). Consoles are still capped on 16GB total memory.

Having more RAM is fine if you know what you’re going to load in advance. If you don’t then RAM is not going to be a substitute for SSD.

1 Like

This is a great counterpoint actually too.

It’s just 5-6 months in and PS5 is barely faster to load on most 3rd party games by 2-3 seconds.

This isn’t even using VA yet.

The SSD’s on both consoles are architected to make up for lack of Ram. Each are using their SSD’s to stream assets, we know the xbox has 100Gb reserved for this and I assume the PS5 would have something similar. If you watch Xbox’s hotchips presentation on the Series X they had expected ram prices to fall to low levels but this didn’t happen. Originally I think Microsoft had planed the Series X to have 20Gb but with Ram costing more than expected we got a mix of 2Gb and 1Gb chips

Xbox allows for more than that, the 100 GB was used strictly as a mere marketing point. It could be 200 GB if the game was that large.

Is there any real difference in the hardware decompression tech between the two? From memory, the PS5 has two processors, while the XSX only has the one?

Would the Direct storage API have any effect on the real world speeds? Does it reduce latency, or make the seek times quicker in anyway?

Simply understand the pipeline first. Decompression will occur on the data recieved.

Assume a case where bandwidth is 100% utilised all the time for both the consoles.

Now, decompressor’s power should be enough to support the RAW bandwidth.

Let’s take for example - both the consoles’s decompressers have same power. Now because the RAW bandwidth is more on PS5 then Series X, decompression becomes a bottleneck on PS5.

So in theory - PS5’s decompression blocks should atleast be twice as fast as series X’s decompressorers.

The hardware decompression in them handles different formats. Sx has zlib compression, Ps5 uses kraken which is newer/more effective. But in the other hand SX supports a format aimed at textures that in order to get the same compression rates you must use cpu/gpu on Ps5.

The IO apis on both are said to allow just that, less overhead when dealing with api calls. And SSDs on their own have lower seek times than a HDD, though they both claimed some customizations (likely in the SSD controller) and a better filesystem to reduce that even further.

1 Like

So do you think the PS5 has its own Direct Storage API equivalent? Microsoft said that the decompression block saved the equivalent of 3 Zen 2 cores and then the Direct Storage API saved an additional 2 Zen 2 cores. That’s a massive saving.

Of course Sony has their own API for accessing the SSD. How else would this work?

Sony definitely has their own apis, that most likely are very similar to direct storage as they all have the same goals