Hypothetically speaking, could the FDE in T239 be used to offset some of the speed/power related issues of SD Express? Wondering if there was any prior speculation on what the File Decompression Engine would be used for
There are three possible things that dedicated decompression hardware lets you do.
First, it can give you cheap CPU power. Just about every game will need to decompression at some point. Usually you need the CPU to do that. Having dedicated hardware frees that CPU up to do other things. Physically, the FDE is smaller on chip than a generic CPU core, so it's usually a pretty good trade off.
Second, it can support formats that might be slow on a generic CPU. You're not compressing "generic" data, you're compressing specific things, usually game textures, which you know a lot about. You've got custom hardware. You can take advantage of these two facts to make something that gets very good levels of compression (much smaller files) with very high speed. This is the place it might help with storage speeds. You've got smaller files on storage, so you need to read less, and the FDE makes decompressing fast enough for that to be worth it.
GDeflate is one example
Third, you can integrate the FDE with your storage controllers. This means that instead of reading the data into memory (one copy), then decompressing to a different chunk of memory (another copy) you can stream the data through the FDE before it gets to memory, decompressing as you go. This is really tricky to pull off,
but it has been done. A cartridge based console is kind of the best case scenario for this sort of tech, no idea if it will be used by Nvidia
It doesn't matter that the Playstation Q runs on a generic android os since were talking about a device that is exclusively made to stream games.
Exactly, but also, exactly. There is already a Remote Play app for Android, how much of a premium is Sony going to charge for a cheap Android table with a controller glued to it?
I wouldn't call it lazy, because I think that word gets thrown around too much in an industry loaded with abusive, low paying, crunch riddled studios. But Android is not a real time OS, and this is a device whose only operation is low latency streaming. This is ripe for a stripped back OS where the Remote Play software has direct access to the BT and WiFi stacks, as well as the framebuffer.
I mean, I'm waiting for the reviews, but if they can't get the latency down, the experience is gonna be real bad, and unlike the Wii U, they don't have the custom hardware in the Playstation itself, so it has to get solved on the Q itself.
Does Sony struggle with keeping secrets, or does Nintendo excel in maintaining secrecy?
We just heard about devkits for Switch 2. If you assume holiday 2024 for both, they're about at the same level of "leak."
But were also talking about a brand new console, which will have a slow drip of developer engagement, mostly focusing on launch titles, versus a revision, whose kits need to get into the hands of every major dev with an existing PS5 game. Put Sony and Nintendo aside, it's just a different ball game.
The one concrete thing there is, is the T239 production time. Would it make sense to tapeout 2 years before launch? 3 years? 4 years?
We have Nintendo's entire production schedule for the NX due to the gigaleak. The goal was to deliver Switch 18 months from chip sampling, but they had to delay to 2 years because the software wasn't ready.
There were rumors of Oberon samples as far back as late 2018, and there were definitely devkits with Oberon (the APU in the PS5) in May. So yeah, probably 2 years there as well.
) I don't think we'll reach April 2024 without a reveal because Nintendo will probably not want to go into their end of year financial meeting with investors saying: "We're predicting even lower Switch sales than before... have a good day."
