nalyDylan

Yah, I hadn't considered the PCI-E bandwidth limits. So it would ultimately have to be on the GPU. I'm hesitant to compare it to something like H264/265 or especially the PNG decompression just because of the scale, H265 is usually <100mbit, and we're talking about 10+ Gbit/s stream decompression, that's huge and is going to take a fairly beefy processor and probably HBM2 memory for the bandwidth to be able to do this while also rendering graphics, I know the PS5 uses GDDR6 but it's the console advantage to be able to optimize for specific hardware that will make the difference probably.

Great video! Mad respect Linus, it takes to serious juevos to show that kind of humility with that many people watching. For a PC equivalent to this though, I think it's a hard shrug for all the reasons you pointed out. Workstation GPUs have DirectGMA(GPUOpen) and the new version of GPUDirect(coming soon, CUDA 11.0) to talk directly to a PCI-E SSD, but neither address the issue of the stream decompression that could potentially give Sony's SSDs an edge in throughput, but may not be a huge deal since like Linus pointed out originally, PC already has some pretty fast SSDs If compression is a big deal though, it would fall on GPU manufacturers to implement stream decompression using either block level decompression that can be parallelized on the CUDA cores/Stream processors, or going the Sony route and putting that compression hardware on the board somewhere. I suspect the stream decompression is out of scope for an M.2 SSD to have this kind of stream processor on it due to power/heat constraints in the form factor though, so it would have to end up on the GPU somewhere or as an entire new PCI-E device, which could be cool but expensive, and another thing eating PCI-E lanes.

Unfortunately it hasn't worked for me, I'm waiting on ASUS to try again. It seems it's 50/50 with people who had issues on previous 14xx builds if 1405 fixes it though, so they are at least attempting to unbreak things. This feels like it was rushed to get the new AGESA version out before the holidays so when people were buying parts they'd run and get the one that was supposed to boot 20 seconds faster or whatever.

I was contacting ASUS about this issue and several others I've had with this update and just got jerked around between support people who never read my questions completely to begin attempting an appropriate answer. 1) I have an issue where sometimes after a CMOS clear the only way for me to get the computer to do anything (Del or F2 doesn't work, none of the 3 methods described in the manual for accessing BIOS work after clearing CMOS) is to unplug one of my hard drives and break my RAID array so it falls back to an old school AMI BIOS screen where I press F1 to acknowledge the error and enter setup. Other people on the internet seem to be less fortunate and not be in RAID mode for this and have to RMA their boards at this point. 2) Saving User profiles in the new build seems to sometimes randomly insert fClk of 666Mhz, even on profiles where previously it was left on Auto 3) All these overclocking issues. For other people who encounter a bad BIOS flash (Hello future people on Google): That is, ASUS CrashFree Bios 3, or if you get to, "Warning! BIOS Recovery Mode has been detected." All USB ports on my X570-Plus TUF Gaming are created equal, but, the BIOS Recovery mode I was only able to make work with very specifically a GPT formatted FAT32 USB 3.0 drive. (I rotated between 5 different thumb drives and configurations with the fine people of the LTT Discord for almost 2 hours to scrape this one out)