News:

Willkommen im Notebookcheck.com Forum! Hier können sie über alle unsere Artikel und allgemein über Notebook relevante Dinge disuktieren. Viel Spass!

Main Menu

Micron announces first PCIe 6.0 SSD with over 26 GB/s transfer speeds

Started by Redaktion, August 06, 2024, 17:45:10

Previous topic - Next topic

Redaktion

The world's first PCIe 6.0 SSD developed by Micron will initially be available for data centers and edge applications as part of a portfolio of memory and storage products to support the broad demand for AI. More details to be revealed later this week at the FMS 2024 event.

https://www.notebookcheck.net/Micron-announces-first-PCIe-6-0-SSD-with-over-26-GB-s-transfer-speeds.871866.0.html

Pascal76

sequential performance is not what professionals will prefer to know.
What about random io ? :)


Real NikoB (name bl)

What's the point of discussing SSD for servers here on a site about laptops, editors? To the shame of their manufacturers, laptops don't even use PCI-E 5.0 for SSDs, while AMD's HX series has at least 16 5.0 lanes hanging in the air. Even NVidia's video cards, to the shame of NVidia, still use PCI-E 4.0, which has been morally obsolete for several years. Moreover, in x8 mode in most laptops...

Moreover, with the shameful bandwidth of the memory controllers of the latest x86 chips from AMD at 90-100 GB/s, 25 GB/s is exactly 1/4, and this is too much load on the data bus. It should not exceed 10% of the memory controller bandwidth per device, which means that the bandwidth of x86 RAM should have been higher than 250 GB/s long ago...

Goberman

Quote from: Pascal76 on August 06, 2024, 18:33:09sequential performance is not what professionals will prefer to know.
What about random io ? :)
"Random access" tests in server environment are bad data, they simulate an application doing 4K reads/writes sequentially to/from random blocks one after the other. Reality is SSDs are a mass-parallel system that has only one task - parallelized block-random-access. If your system does many random 4K block read/writes simultaneously (not sequentially) - speed will be high and will only be related to block size, 4K will be slow, 128K much faster, etc.
So in server environment, when there's millions of read/write requests coming in simultaneously, "4K random read" is a meaningless test. Also yeah, 4K is smaller than the SSD block size so why even do these writes.

A

Quote from: Real NikoB (name bl) on August 07, 2024, 14:34:17To the shame of their manufacturers, laptops don't even use PCI-E 5.0 for SSDs

Because in real world, even PCI-E 4.0 for ssds is borderline useless. Very few go faster for sequential reads (there are some), let alone more realistic random reads/writes. It has mostly use in servers

Real NikoB(bl)

Quote from: A on August 07, 2024, 21:07:44Because in real world, even PCI-E 4.0 for ssds is borderline useless.
Complete technical nonsense. They provide acceleration, but only where the system is balanced in terms of performance. Those. just in the latest Zen5, where the memory controller is capable of delivering 90-100GB/s - pci-e 4.0 x4 times are very profitable and just no more than 1/10 of the system memory bandwidth (and the lower the % of bandwidth, the better).

The problem is that there are STILL no energy-efficient mass-produced 4.0 SSDs on 3D TLC with a low-consumption DRAM buffer, and what's even worse is that in most laptop models, 2-sided SSDs will be used (if they even fit into the M.2 slot) in extremely unpleasant temperature conditions for NAND chips, especially taking into account the monstrous heating of current generation SSD controllers.

In fact, only Samsung on the market sells 2-4TB single-sided SSDs, but even they consume a monstrous 7-10W, which is unacceptable for laptops.

SSD consumption in a laptop should not exceed 2-3W at PEAK load.

In fact, today there are NO such controllers for mass production series 2-8TB with 3D TLC and DRAM buffer. And bufferless ones behave extremely poorly (especially the terrible models with QLC) as system disks, especially in a heavy mixed load of simultaneous read/write, which is typical for a system disk, which is also busy with other work besides the system load)

In fact, manufacturers of controllers and NAND chips are only now approaching energy efficiency that is acceptable for laptops in the pcie 4.0 x4 class with a dram buffer and on 3d tlc without the requirements of a radiator - the reason for this is the gradual slowdown of the "exhaust" from modern technical processes. The performance curve per 1W of consumption is becoming flatter and the cost of factories is rising exponentially due to the obvious failure in fundamental research around the world.

Quick Reply

Warning: this topic has not been posted in for at least 120 days.
Unless you're sure you want to reply, please consider starting a new topic.

Name:
Email:
Verification:
Please leave this box empty:

Shortcuts: ALT+S post or ALT+P preview