The standard rule in hardware is, if you want it to go faster, you have to work on the bottleneck. There’s no point in using the latest, fastest GPU if you’re going to put it next to a CPU from 10 years ago.
When it comes to NVMe drives, you have to ask the question, where is the bottleneck today?
On the PCIe bus? Nope, PCIe 3.0 is more than enough. The only reason you would want to go to 5.0 is if you have a server and you have 100Gb switches and you want 12 NVMe drives, you need PCIe 5.0.
sequential speeds? Well, they’ve had enough already.
at constant speeds? There is a margin to that, especially with consumer drives that run with cache or where the speed drops when the drive is nearly full.
in response times? Bingo, this is where the big bottleneck lies and that’s the most important thing to look at. This means that you have to look at one specific performance number, random 4K read/write with a depth of 1.
And guess what, not much progress has been made with consumer drives because those consumers aren’t looking at it. It’s a completely different story for enterprises.
Another issue is if you actually lower those response times, is how do you address that drive without adding response time to it again? DirectStorage targets gamers, NVMe-oF NAS targets, RoCe targets networking, RDMA targets Windows OS/NICs, Intel VMD raid The name says it all.
“Professional web ninja. Certified gamer. Avid zombie geek. Hipster-friendly baconaholic.”
Apple TV gets a four-player picture-in-picture sports broadcast mode with iOS 16.5 – Picture and sound – News
Dolphin Emulator for Nintendo Wii and GameCube Games Coming to Steam – Games – News
Nintendo Shows New Gameplay The Legend of Zelda: Tears of the Kingdom – Games – News