Samsung's 970 pro designed to combat Optane on performance?

[The Elder Smurf]

Samsung released the 970 pro (and evo) and it performs quite admirably compared to its 3D-Xpoint counterpart. Remember when Optane was wiping the floor with random io and latency, it looks like samsung closed the gap by quite a bit.

https://arstechnica.com/gadgets/2018/04/samsung-announces-970-pro-and-evo-nvme-ssds/

https://news.samsung.com/us/samsung-970-pro-970-evo-nvme-ssds/

https://hothardware.com/reviews/samsung-ssd-970-evo-and-pro-review?page=1

[The Benjamins]

I wish Samsung would focus more of price and density. I want bigger drives for less.

[Mira Yurizaki]

The question I'm wondering is... does that mean anything in terms of single-client practical performance? When I started using an NVMe drive, the system didn't really feel faster.

[davrosG5]

Isn't Samsung developing Z-NAND to compete with Optane?

The 970 is just an evolution of their 3D NAND going from 48 layer to 64 layer IIRC as MLC (970 Pro) or TLC (970 EVO).

 

7 minutes ago, M.Yurizaki said:

The question I'm wondering is... does that mean anything in terms of single-client practical performance? When I started using an NVMe drive, the system didn't really feel faster.

I wouldn't have thought so, not noticeably anyway.

[GDRRiley]

8 minutes ago, M.Yurizaki said:

The question I'm wondering is... does that mean anything in terms of single-client practical performance? When I started using an NVMe drive, the system didn't really feel faster.

no not for most use cases. scratch disk and highly random environments its a big change but for most no not much.

[DuckDodgers]

12 minutes ago, M.Yurizaki said:

The question I'm wondering is... does that mean anything in terms of single-client practical performance? When I started using an NVMe drive, the system didn't really feel faster.

NAND-based NVMe drives don't contribute much (if anything) to single-threaded random access performance that's most critical for the consumer workloads.

NVMe allows for much faster sequential and multi-threaded transfers and this is where Optane adds to the table its top speed random access performace.

[GoodBytes]

 

- Moved to Storage until fix -

[Mira Yurizaki]

On 4/27/2018 at 9:13 AM, DuckDodgers said:

NAND-based NVMe drives don't contribute much (if anything) to single-threaded random access performance that's most critical for the consumer workloads.

NVMe allows for much faster sequential and multi-threaded transfers and this is where Optane adds to the table its top speed random access performace.

This is making me think that future storage devices are only useful in server scenarios. Particularly databases.

[TechyBen]

In the future... everything is the cloud [storage].

[blueskull]

On 4/27/2018 at 12:00 PM, M.Yurizaki said:

The question I'm wondering is... does that mean anything in terms of single-client practical performance? When I started using an NVMe drive, the system didn't really feel faster.

I do embedded system design, and I compile large projects often, mostly Yocto (a collection of Linux kernel and apps).

With a 22-core CPU, the SSD chokes the system.

And I'm not saying I have a SATA or cheap entry level 2D TLC NVMe. I have a 1.2TB 750, and it still sucks.

Compiling is a really stressful operation on SSDs -- many threads, shallow (mostly 1) QD, tons of small files and mixed R/W.

 

I'm waiting for the new 970 Pro. 900P was great, but my chassis can't house it. It only supports up to 12mm 2.5" drives, while 900P is 15mm thick.

Ans my mobo has only one PCIe, so AIC cards are also not possible. Yes, silly me, buying an ITX X99.

[Mira Yurizaki]

1 hour ago, blueskull said:

I do embedded system design, and I compile large projects often, mostly Yocto (a collection of Linux kernel and apps).

I started in embedded systems and I was on the opposite end of the spectrum where there was no Linux. It was all bare metal programming because "megabytes" wasn't a thing on the target.

 

But I get where you're coming from.