News Sabrent Developing Ultra Fast PCIe 5 SSD That Could Hit 14 GBps

[Admin]

Sabrent is working on a new SSD that will outperform all of the best SSDs on the market today, with a targeted transfer rate of 14,000 MB/s on the final product.

Sabrent Developing Ultra Fast PCIe 5 SSD That Could Hit 14 GBps : Read more

 

[InvalidError]

If you cannot push higher clocks, you can still push extra NAND channels at lower individual channel speeds for the same or greater aggregate bandwidth at increased IO concurrency. Of course, then you need a controller with those extra channels.

 

[PlaneInTheSky]

We want fast random, who cares about sequential.

I'm not running a Petabyte cloud datacenter am I.

So far no PCIe 5.0 has shown any improvement in random reads over PCIe 4.0

Intel Optane is still the fastest SSD on the planet, nothing else even comes close.

 

[vern72]

I almost dread to see the size of that heat sink. It's probably the height of a PCI card.

 

[Alvar "Miles" Udell]

So we went from a Mustang in SATA to Porche to a Lamborghini to a Ferrari to a Veyron now to those 1800bhp area electric hypercars, with prices to match.

And this is great for the 1%, but for the other 99% of us, those who don't run hyperscale datacenters, we're waiting for the Ferrari with a Mustang price tag, we want these 4¢/GB drives with Samsung 980 Pro or SK Hynix P41 level performance. No doubt even the 1% want it as well...

 

[ezst036]

So we went from a Mustang in SATA to Porche to a Lamborghini to a Ferrari to a Veyron now to those 1800bhp area electric hypercars, with prices to match.

And this is great for the 1%, but for the other 99% of us, those who don't run hyperscale datacenters, we're waiting for the Ferrari with a Mustang price tag, we want these 4¢/GB drives with Samsung 980 Pro or SK Hynix P41 level performance. No doubt even the 1% want it as well...

I'm still using SATA for my daily driver. I have an NVME in another computer, can't even tell the difference. Not unless I have a specific workload and actively looking for it.

 

[TechieTwo]

The moto of the SSD industry appears to be: "Over-promise and Under-deliver".

A fast gen 4 M.2 SSD does make a noticeable difference in PC performance. These gen 5 SSDs -whenever they actually show, will not provide a major performance diff over gen 4 unless you have large files or many files being accessed concurrently such as in a large database or similar.

 

[DSzymborski]

Incoming thread of someone running an OS in RAID 0 with these, screwing it up, and blaming us somehow when they lose their data and for suggesting the whole thing was a bad idea.

 

[HideOut]

We want fast random, who cares about sequential.

I'm not running a Petabyte cloud datacenter am I.

So far no PCIe 5.0 has shown any improvement in random reads over PCIe 4.0

Intel Optane is still the fastest SSD on the planet, nothing else even comes close.

And sadly out of production

 

[HideOut]

Sabrent Developing Ultra Fast PCIe 5 SSD That Could Hit 14 GBps


These kind of filler "stories" crack me up. Reminds me of CNN's "BREAKING NEWS" for a dog pissing on a sidewalk. Like holy boy/girl who cried wolf. EVERY SSD OEM is doing this exact thing. Its not news, its the newest standard.

 

[usertests]

We want fast random, who cares about sequential.

I'm not running a Petabyte cloud datacenter am I.

So far no PCIe 5.0 has shown any improvement in random reads over PCIe 4.0

Intel Optane is still the fastest SSD on the planet, nothing else even comes close.

3D XPoint was a post-NAND technology that got to market ahead of its vaporware competitors. It wasn't fast, cheap, or dense enough to seriously replace DRAM or NAND, so it occupied a new niche tier in the middle that benefited some companies, but wasn't worth investing in. Intel cutting and running says less than Micron doing so first.

Maybe another company will take a stab at introducing some phase change/crossbar/memristor non-volatile memory aiming at NAND and/or DRAM. If you can kill both at the same time, you have universal memory and a new era of computing. I would be satisfied with something better than NAND in every way, even if it doesn't have the speed or endurance to replace DRAM. It would also be nice to see some 3D DRAM with higher capacity and lower cost-per-bit for consumers.

 

[peachpuff]

We need more storage capacity not faster speeds.

 

[InvalidError]

We need more storage capacity not faster speeds.

When more software gets written to leverage NVMe speed for faster load times, faster drives will become more relevant. Fir now, it is a case of hardware being 5-10 years ahead of most software developers.

 

[jackt]

will it require liquid cooling ?

 

[RichardtST]

So we went from a Mustang in SATA to Porche to a Lamborghini to a Ferrari to a Veyron now to those 1800bhp area electric hypercars, with prices to match.

And this is great for the 1%, but for the other 99% of us, those who don't run hyperscale datacenters, we're waiting for the Ferrari with a Mustang price tag, we want these 4¢/GB drives with Samsung 980 Pro or SK Hynix P41 level performance. No doubt even the 1% want it as well...

I'll pass on the Spamsung. I've lost faith in them. But the Sabrent is tempting, as would be an SK Hynix. Have you seen prices lately? pcie 4 drives (even the fast ones!) are like almost given away as door prizes. I'm already over-storaged or I'd go get a few more.

 

[RichardtST]

When more software gets written to leverage NVMe speed for faster load times

Lol! Good one! As if today's programmers can actually write efficient code!
Hah! Go ask one what a "thread" is and watch them look at their clothes...

 

[InvalidError]

Lol! Good one! As if today's programmers can actually write efficient code!
Hah! Go ask one what a "thread" is and watch them look at their clothes...

Efficiency and speed are only loosely related, even more loosely when you do multi-threaded stuff. The most heavily optimized single-threaded code will still be slow if it has too much stuff to process and breaking the thing in more manageable chunks tends to introduce significant overhead in both programming effort and management. Often not worth bothering with.

 

[sonofjesse]

I"m sure the prices on these won't be cheap.

GEN IV will be the sweet spot for a while.

4TB NVME is down to 216 now!!!!

Be nice as more larger sizes get affordable.

PCIE Gen 5 drives are super super slow to actually get to market.

 

[InvalidError]

I"m sure the prices on these won't be cheap.

I wouldn't recommend buying into the first couple of generations unless you absolutely need the bandwidth for specific reasons anyway. Future PCIe 5.0 controller implementations are pretty much certain to become more energy-efficient just like it did with 4.0. It took about four years for 4.0x4 SSDs to reach price parity with 3.0x4 ones, it'll probably be a similar story with 5.0x4.

 

[RichardtST]

Efficiency and speed are only loosely related, even more loosely when you do multi-threaded stuff. The most heavily optimized single-threaded code will still be slow if it has too much stuff to process and breaking the thing in more manageable chunks tends to introduce significant overhead in both programming effort and management. Often not worth bothering with.

Preaching to the choir bro. Been doing both efficiency and multi-threading for over 40 years.