AMD Ryzen 7000 Series Processor Runs Phison PCIe 5.0 SSD with Micron 232-Layer NAND Flash

[AleksandarK]

During this year's Flash Memory Summit, Phison, a company known for SSD controllers and now flash drives, demonstrated a system running AMD Ryzen 7000 series processors based on Zen 4 architecture. What is interesting about the shown specification is that the system was running an engineering sample of an upcoming Zen 4-based CPU with the latest storage technologies at impressive speeds. Using a Phison PS5026-E26 SSD controller, also called E26, the PCIe 5.0 SSD is powered by Micron's latest 232-layer TLC NAND flash. This new NAND technology will also bring greater densities to the market by promising higher endurance, higher read/write speeds, and better efficiency.

With AMD's upcoming AM5 platform, support for PCIe 5.0 SSDs is a welcome addition. And we today have some preliminary tests that show just how fast these SSDs can run. In CrystalDiskMark 8.0.4, it achieved over 10 GB/s in both read and write. We know that the E26 controller is capable of 12 GB/s speeds, so more fine-tuning is needed. Being an early sample, we expect final specifications to be better. The system is powered by an engineering sample of a six-core, twelve-threaded Zen 4 CPU running at unknown clocks, codenamed 100-000000593-20_Y. We can expect to see more of this technology once AMD's AM5 platform lands and Phison-powered SSDs hit the shelves in September.


View at TechPowerUp Main Site | Source

 

[Kirederf]

I am curious what performance to expect on Alder Lake or Raptor Lake.

 

[ZoneDymo]

I am curious what performance to expect on Alder Lake or Raptor Lake.

well yeah....everyone is

 

[Jimmy_]

Ouuuuuuuuhooo! numbers are looking pretty awesome on AMD.
will waiting for intel platforms results as well

 

[jesdals]

That motherboard though - wonder if its a BioStar or something else broken

 

[Durvelle27]

That motherboard though - wonder if its a BioStar or something else broken

Its a OEM board

They've been released into the channels already

 

[windwhirl]

I love how they ran only the first sequentials test and nothing else /s

 

[AnthonyC]

Its a OEM board

They've been released into the channels already

That the wierdiest Motherboard i ever saw.
the shipset fan
The shipset location next to the io
The amounts of led and alike.
what is the pci-e x8 card?
btw, the pci-e x8 mechanical socket

nothing look normal

 

[Bwaze]

Any real world consequence of these results? Faster boot? Faster game load? I suppose no...

 

[chrcoluk]

Any real world consequence of these results? Faster boot? Faster game load? I suppose no...

Probably no benefit real world.

 

[windwhirl]

Any real world consequence of these results? Faster boot? Faster game load? I suppose no...

The only thing that actually matters at this point is the random test and they did not run that.
Once you cross the 1 or 2 GB/s speed mark for sequential speed the diminishing returns hit very hard unless you're running something that really requires extreme sequential speeds.

 

[Chomiq]

Any real world consequence of these results? Faster boot? Faster game load? I suppose no...

In the end it's all gimped by Windows IO.

 

[Bwaze]

There was some speculation DirectStorage will change that, and the fast PCIe 4 (and 5) drives will benefit greatly. That was quite a long time ago.

 

[Chrispy_]

We're well past the point where bandwidth matters. What isn't show is the RND4K Q1T1 score. That's latency and IOPS, which is the single biggest storage bottleneck in almost any PC built within the last decade.

If you're actually working with 8K RED footage all day every day or something like that, then scrubbing performance might improve slightly because of the extra bandwidth - though even in that best-case scenario for a fully-local, 100% sequential workload, there's a fair bit of CPU/GPU bottlenecking in decoding from each keyframe to the selected point. I've personally witnessed someone at work upgrade from an old Samsung 950 Pro for use exclusively as a scratch disk for editing 4K survey footage. He jumped from ~2.5GB/s to 5.5GB/s and nothing changed. He initially thought it was actually slower but in jumping back to a colleague's PC that still had a 950 Pro, he decided he was just subconsciously disappointed that nothing had measurably changed.

I genuinely can't think of any other scenario currently that actually benefits from faster sequential streaming. DirectStorage isn't here yet, and even then there are other decompression and processing overheads on the GPU that will be a bigger bottleneck to performance than the SSD bandwidth.

 

[windwhirl]

There was some speculation DirectStorage will change that, and the fast PCIe 4 (and 5) drives will benefit greatly. That was quite a long time ago.

Developers have to implement DirectStorage in their software. I can't recall it being implemented anywhere tho.

 

[phanbuey]

That is awesome.

 

[Chrispy_]

That is awesome.

You know what 10,098MB/s is, right?

 

[TheLostSwede]

That the wierdiest Motherboard i ever saw.
the shipset fan
The shipset location next to the io
The amounts of led and alike.
what is the pci-e x8 card?
btw, the pci-e x8 mechanical socket

nothing look normal

It's because it's what AMD uses to develop their hardware with, not something you will ever see outside of a product development department at AMD or possibly a motherboard maker or someone large OEM like HP, Dell or Lenovo.

 

[Tomorrow]

I love how they ran only the first sequentials test and nothing else /s

Hiding their embarrassingly low random r/w numbers no doubt. Also downgraded sequentials. First it was reported 14GB/s. Now 11-12GB/s and the screenshot only shows 10GB/s.

It's because it's what AMD uses to develop their hardware with, not something you will ever see outside of a product development department at AMD or possibly a motherboard maker or someone large OEM like HP, Dell or Lenovo.

AMD Myrtle.

 

[Minus Infinity]

I don't think we'll see meaningful random performance improvements until the new 232/238 layer SSD's ship. PCI-E 5 will just bring more heat, more throttling, more cost, more power use. The massive sequential performance will make bugger all difference to most people.

I will only buy a PCI-E 5 if it brings tangible random performance increase and or much more affordable 4TB TLC drives, otherwise I'll get a much better value PCI-E 4 drive.

 

[Bwaze]

Why would the new high layer chips bring random performance increase? And even if they do, we need an order of magnitude better results to be noticeable in real world.

I think we're kind of stuck in this regard, norhing seems to be even in development to compete with now years old and discontinued Intel Optane.

 

[Tomorrow]

Why would the new high layer chips bring random performance increase? And even if they do, we need an order of magnitude better results to be noticeable in real world.

I think we're kind of stuck in this regard, norhing seems to be even in development to compete with now years old and discontinued Intel Optane.

More layers should mean better parralelism/performance and smaller/cheaper chips.
SLC NAND still exists. For example Kioxia's XL-FLASH tho second gen could be MLC there too.
Tho it's also enterprise only with a price tag and poor availability.

But yes aside from SLC i dont see anytthing else in development that could mirror Optane. Not unless someone else will develop Phase Change Memory from ground up. Also it seems no one was willing to buy the Optane business from Intel. Micron said no and exited. Hynix only bought the regular NAND business.