Intel shows off Alder Lake system with a PCIe 5.0 SSD hitting almost 14GB/s

midian182

Posts: 9,741   +121
Staff member
What just happened? Just as more people are buying PCIe 4.0 SSDs, next-generation PCIe 5.0 storage is making headlines. The latest reveal comes from Intel’s Chief Performance Strategist and former tech journalist Ryan Shrout, who showed off a PCIe Gen 5 Samsung SSD running on an Alder Lake system and hitting some impressive speeds.

Shrout tweeted that the PCIe 5.0/Alder Lake demo was supposed to be part of Intel’s CES 2022 lineup, but with Chipzilla joining so many others in announcing that it will be severally limiting its physical presence at the show and going mostly virtual, we got to see the SSD in action a week early.

Shrout’s system comprises a 12th-gen (Alder Lake) Intel Core i9-12900K, an Asus Z690 motherboard, and an EVGA RTX 3080 graphics card. It’s a pretty standard high-end rig, apart from the Samsung PM1743 PCIe 5.0 SSD. The Gen 5 SSD is an enterprise-class drive, meaning the final consumer model will look different.

The demo begins by confirming the speed of a top-end WD Black PCIe 4.0 SSD, which reaches close to 7GB/s. That’s fast by anyone’s standards, but the Gen 5 Samsung SSD, which is connected via a PCIe 5.0 interposer card, is almost double the speed of the Gen 4 drive, reaching around 13.8GB/s.

In addition to those speeds, Samsung says that the PM1743 has around 30% better power efficiency (608MB/s per watt) than PCIe 4.0 drives. This enterprise version will arrive in 2022 with capacities ranging from 1.92TB to a massive 15.36TB and in 2.5-inch and 3-inch form factors.

In a separate tweet, Shrout also showed two PM1743 drives running simultaneously, giving a combined speed of over 28GB/s, though the limited number of PCIe 5.0 lanes meant the graphics card had to be removed.

Earlier this week, we saw this active cooler designed to keep PCIe 4.0 and PCIe 5.0 NVMe M.2 SSDs temperatures down and stop them from thermal throttling. These robust solutions may be necessary for Gen 5 drives such as the PM1743 and Adata’s ‘Project Nighthawk.’

Permalink to story.

 
These are cool and all, but PCI-E 4.0 SSDs already don't really show much real world benefit over 3.0. I just don't think sequential reads are the bottleneck anymore. Heck, most people would be hard pressed to tell any difference beyond SATA SSDs.
 
These are cool and all, but PCI-E 4.0 SSDs already don't really show much real world benefit over 3.0. I just don't think sequential reads are the bottleneck anymore. Heck, most people would be hard pressed to tell any difference beyond SATA SSDs.
It depends entirely on the application used. Think outside of games and facebooking.

The tested drive was an enterprise class drive, so there is your use case right there for starters.
 
It depends entirely on the application used. Think outside of games and facebooking.

The tested drive was an enterprise class drive, so there is your use case right there for starters.
Frankly, until we can solve the problem of drives only being capable of limited amount of writes there aren't many applications where it makes sense. Linus from Linus Tech Tips benchmarked one of the fastest PCI-e 4.0 drives and managed to use up ~10% of the SSDs writes in under a day.
 
Something tells me PC hardware will keep moving fast but our anchor or weakest link will be good performing GPU's that don't cost a fortune and easily attainable.

What's the point of having crazy fast hardware when you have either a CPU or GPU dragging you behind. Good example is in the 1990's RAM was slow and expensive. Sorry for being negative. I tend to think too much.
 
These are cool and all, but PCI-E 4.0 SSDs already don't really show much real world benefit over 3.0. I just don't think sequential reads are the bottleneck anymore. Heck, most people would be hard pressed to tell any difference beyond SATA SSDs.
Good thing "sequential reads" aren't the only thing that improved, but so did sequential writes, random reads, and random writes.
 
Well I would like them to improve random 4k read/writes. Also improve the speed drop that happens when u copy or move large files. Writes have low limit like 200GB on a 1TB drive, before the speed drops to 1/5 of sequential. Just boosting max sequential won't change anything. 7000MBps is already a lot for 99% of applications and games etc. What's the point of 14k sequential when it drops to 700mbps when u copy large files?
 
Frankly, until we can solve the problem of drives only being capable of limited amount of writes there aren't many applications where it makes sense. Linus from Linus Tech Tips benchmarked one of the fastest PCI-e 4.0 drives and managed to use up ~10% of the SSDs writes in under a day.
Which video and which drive did he use? If he tested a TLC (or god forbid, a QLC) drive, then yeah, no duh he did. These drives have much lower TBW ratings then MLC drives did.

Of course, using up 10% of the TBW rating on a consumer drive doesnt mean much. SSD can easily last multiple times the TBW rating, and that's jsut writes. You can read as much as you want, and for applications like large databases, particularly SQL servers, SSDs are a godsend, and the more speed and IOPS the better.

There's also the fact that enterprise quality drives usually use MLC NAND, alongside SLC cache. There's a reason they are rediculously expensive, and that type of NAND has much higher endurance, and dedicated servers for storage can also use system memory as a cache before writing data, minimizing drive wear.
 
So, basically, sIntel is still resorting to desperation to market its products. :facepalm:
If they were showing this on a server system, I might think differently. :rolleyes:
 
These are cool and all, but PCI-E 4.0 SSDs already don't really show much real world benefit over 3.0. I just don't think sequential reads are the bottleneck anymore. Heck, most people would be hard pressed to tell any difference beyond SATA SSDs.
This. I'm all for fast storage, but who really needs 14GB of bandwidth?
 
Last edited:
Back