New PCIe Gen 5 SSDs

[Shmee]

So I got a promo email from Newegg, looks like Western Digital is now selling the Black SN8100 NVMe SSDs to compete with Samsung's 9100 Pros, but they seem to be even more expensive right now.

https://www.newegg.com/western-digital-2tb-nvme/p/N82E16820250279?Item=N82E16820250279&ignorebbr=1&cm_mmc=EMC-ignefl051325-_-EMC-051325-Index-_-Internal%20SSDs-_-N82E16820250279-_-USA_CP_20250513_01&ignorebbr=1&cvtc=29303885

Anyway, good to see newer, faster and less power hungry PCIe gen 5 drives coming out, and hopefully more competition soon. Obviously, many previous PCIe gen 5 SSDs were crazy hot and required active cooling, which made them a bit more of a gimmick.

 

[Hail The Brain Slug]

The Hynix P51 Platinum is also now available in the US in 1TB and 2TB sizes. They just kind of silently came available on Amazon, so there are no formal reviews yet - computerbase says they are in the process of testing one now.

I ordered an SN8100 4TB as soon as it went live, but its not going to ship for weeks so I have time to decide if I actually want it. I've been waiting for this wave of gen 5 drives, but I'm looking for the absolute best of the bunch. The SN8100 seems like it might be it.

 

[CakeMonster]

Yep, I'm gonna need to read several reviews with proper testing of temperatures and throttling behavior before I pull the trigger.

 

[Hail The Brain Slug]

Yep, I'm gonna need to read several reviews with proper testing of temperatures and throttling behavior before I pull the trigger.

Well, they're definitely out there. Sandisk did an excellent job of sampling it to numerous outlets and giving them time to have a review ready to post on launch.

A bit sad that Anandtech and their Destroyer test will never be run again. Always fantastic information out of that review.

 

[thigobr]

The P51 seems more expensive than the WD SN8100 currently and its NAND is slower on paper. Need to wait for reviews to be sure. There's some chance Phison E28 drives later this year will be even faster than the current wave of SMI2508. But it's good to finally have alternatives and be able to forget about older and hot Phison E26 based drives.

I am playing the waiting game this year as I don't have any PCIE5.0 enabled devices yet.

 

[Hail The Brain Slug]

The P51 seems more expensive than the WD SN8100 currently and its NAND is slower on paper. Need to wait for reviews to be sure. There's some chance Phison E28 drives later this year will be even faster than the current wave of SMI2508. But it's good to finally have alternatives and be able to forget about older and hot Phison E26 based drives.

I am playing the waiting game this year as I don't have any PCIE5.0 enabled devices yet.

I'm really not that hopeful the P51 can match or best the SN8100.

I expect the E28 to be highly competitive with the standard SM2508 drives. The SN8100 seems to be quite a bit faster due to the NAND and proprietary hardware/firmware changes Sandisk has implemented, so I'm not so confident the E28 will be super competitive with the SN8100 immediately out of the gate.

 

[gdansk]

SanDisk stopped developing their own consumer SSD controllers?

 

[Hail The Brain Slug]

SanDisk stopped developing their own consumer SSD controllers?

They've used off the shelf controllers before, but the SN850 and SN850X was an in house design.

They said the SN8100 controller is based on an SM2508 but sandisk has made hardware changes to include "proprietary IP"

I am a little disappointed, but whatever they've done seems to work given the performance figures.

 

[Hail The Brain Slug]

P51 Platinum has been discounted to the same price as the 9100 Pro - $270 for 2TB. If I had my test bench setup up and working I'd be tempted since it would seem there are still no reviews at all despite being available for over two weeks in the US.

Sad that nobody will review a product without getting sent a free sample these days. Eventually reviews will crop up a few weeks after Computex, I assume.

 

[thigobr]

There are some number from StorageReview and it seems to be competitive with Samsung's 9100Pro but slower than the SN8100.

 

[Makaveli]

Sandisk WD Black SN8100 2TB SSD Review: The fastest overall consumer SSD ever made

An incredible, if late, top-tier SSD from Sandisk

www.tomshardware.com

 

[fastandfurious6]

can someone describe to me what makes an SSD actually deliver faster system responsiveness?

system responsiveness arguably ranks as #1 in benefits of ssds

but what I find incredibly striking:

16 years ago, I first got an 80GB Intel X25-M.... arguably the first popular release of good client ssd, MLC

diff in system responsiveness was day and night... everything opened instantly... it was crazy

but as years went by and SSDs became norm and m2 slot etc... responsiveness significantly dropped... even with top end models not much diff. nothing comes close to the snappiness experienced back then with that intel proto-ssd

so what gives?


also the x25-m, iconic

 

[igor_kavinski]

16 years ago, I first got an 80GB Intel X25-M.... arguably the first popular release of good client ssd, MLC

diff in system responsiveness was day and night... everything opened instantly... it was crazy

but as years went by and SSDs became norm and m2 slot etc... responsiveness significantly dropped... even with top end models not much diff. nothing comes close to the snappiness experienced back then with that intel proto-ssd

so what gives?

I had the same too. I wouldn't say it was insanely fast. But it helped a great deal. I don't find today's SSDs that much wanting in speed and responsiveness. I use RAM drives regularly so I know what maximum responsiveness feels like and no, you can't feel much difference between a RAM drive and a modern SSD, unless your brain stored some data about a several seconds slow operation and you later do the same on RAM drive and it completes in a second or two.

Mostly, it has to do with how our brain works and experiences delay. When we used Win98 or Windows XP, those systems seemed pretty snappy to us on good hardware. But now try to use those same systems like an Athlon or a Core 2 Duo with Windows XP and you'll be like, KILL ME ALREADY! Because the reference data regarding delays in computing operations has been updated in your brain and the new point of reference makes the older hardware seem really, really slow. I'm betting if you were able to use the X-25M today, it wouldn't feel that fast to you.

One thing that may be a factor is RAM's CAS latency. That one has only increased since DDR2 and may be responsible for modern systems not feeling as fast.

 

[fastandfurious6]

what stuff you put on ramdrive?

on OS, what is the real extra computation and real extra benefits win11 offers over winXP for day-to-day stuff?

and why is linux still crap and not 5 times faster if winXP -> win11 adds more and more clutter compute?

 

[mikeymikec]

can someone describe to me what makes an SSD actually deliver faster system responsiveness?

system responsiveness arguably ranks as #1 in benefits of ssds

but what I find incredibly striking:

16 years ago, I first got an 80GB Intel X25-M.... arguably the first popular release of good client ssd, MLC

diff in system responsiveness was day and night... everything opened instantly... it was crazy

but as years went by and SSDs became norm and m2 slot etc... responsiveness significantly dropped... even with top end models not much diff. nothing comes close to the snappiness experienced back then with that intel proto-ssd

so what gives?

Windows sucks, is what the problem is. I have a Win7 VM running in Linux which is way more responsive than my native Win11 install. I have never seen Win11 Explorer run as snappily as Win7's does, and the effect is even more pronounced when you disable visual effects on Win7 (doing the same makes no discernable difference on Win11 in my experience). My Win7 VM has 36 processes running when simply sitting idle at the desktop, my Win10 VM has about 115, and my Win11 install has 160. Having said that, my Linux install has 397 and yet runs rings around Win11 while using much less RAM

From a performance perspective, HDDs are basically suped-up record players with platters that spin at extremely high speed, but because the drive has to wait for the platter to turn to a point where the drive head can read the data from a particular sector, there is a noticeable data access latency (my 4TB 7200RPM HDD benchmarks at ~15ms in this respect). SSDs have no moving parts in, so the data access latency is orders of magnitude lower (IIRC something like 0.03ms). Win10 and Win11 like to give the boot drive five jobs to do at the same time which is why a HDD booting system takes so long to start up and settle down (on average in my experience: 1.5min to get to the desktop, another 10 minutes before it performs optimally... for a HDD system).

A SSD *can* deliver faster system responsiveness up to a point, and trust me: if you tried Win10/11 on a system booting from a HDD, you would agree that a SSD does deliver faster system responsiveness, but the fact of the matter is that most of the time (ie. if we ignore the bottom-of-the-barrel SSDs), comparing one SSD to another is like comparing one F1 car to another, largely because the data access latencies are so brief that as far as we're concerned they barely happened at all.

 

[igor_kavinski]

what stuff you put on ramdrive?

Mostly when working on files like large Excel sheets (they get saved much quicker). Running portable apps on the RAM drive etc.

on OS, what is the real extra computation and real extra benefits win11 offers over winXP for day-to-day stuff?

No idea to be honest. Ask Microsoft

and why is linux still crap and not 5 times faster if winXP -> win11 adds more and more clutter compute?

I think Linux does feel faster from a UI responsiveness perspective but I haven't used it extensively or longer than several minutes at a time, mostly in VMs or booting as Live desktop. By the way, forget Linux. I have an M1 Macbook Air and it doesn't feel blazing fast. It is impressively fast for something that doesn't need active cooling but in normal operation, it won't feel like you are running a modern Ryzen or Alder Lake or Arrow Lake system. But again, if your brain has stored data for some operations, like mine knows how long the overclock.net Excel benchmark takes to run, seeing that benchmark run on the M1 and complete in just 9 to 11 seconds feels blazing fast.

 

[igor_kavinski]

A SSD *can* deliver faster system responsiveness up to a point, and trust me: if you tried Win10/11 on a system booting from a HDD, you would agree that a SSD does deliver faster system responsiveness

Yup. A guy's Core i3 Ivy Bridge laptop with a Toshiba mobile HDD was taking about 10 minutes to load Win11 crap. Cloned the HDD to a Teamgroup SSD and in less than a minute, Task Manager showed the SSD to be idle.

 

[coercitiv]

can someone describe to me what makes an SSD actually deliver faster system responsiveness?

diff in system responsiveness was day and night... everything opened instantly... it was crazy

but as years went by and SSDs became norm and m2 slot etc... responsiveness significantly dropped... even with top end models not much diff. nothing comes close to the snappiness experienced back then with that intel proto-ssd

When it came to the speeds that powered systems responsiveness (mostly random 4k read/write) the gains you had with first gen of SSDs were over an order of magnitude. We went from random operation performance like 0.5 Read (MB/s) and 1.5 Write (MB/s) to something like 15-20 Read and 20-60 Write. In order to experience a similar jump, today we would need 200+ Reads and 600+ Writes for Q1T1 tests. One of the fastest commercial SSDs available today, the 9100 Pro, only does ~90 Reads and ~250-300 Writes.

The only disks that would indeed make you feel the SSD revolution again would be the extinct Optanes:



AFAIK there are two reasons for these limited gains on mainstream SSDs:

• limits in NAND flash technology
• reliance on ever faster CPU cores to power these higher random I/O numbers

This is what I experienced when I upgraded my system from 8700 non-K to 12700K, with the same SSD and OS version:

A fun little fact considering the same SSD was used on both systems - i7 8700 & i7 12700K:

View attachment 54243View attachment 54244

The sequential speeds are of less importance and far less relevance considering the move from PCIe 3.0 to 4.0 for the SSD.

That jump in Q1T1 performance was quite noticeable for me, though obviously not nearly as impressive as the one I got from my first SSD.

One last observation with regards to modern computing and system responsiveness, I would advise everyone to aim for more RAM than they think they need for a new build. If you "need" 32GB of RAM then see if you can squeeze 48-64GB instead, even if that means buying DDR5 6000 instead of 6400 (or higher) or not buying the most high-end SSD on the market.

On my last build I ended up with 64GB due to a curious set of circumstances, and I kinda love it. The OS is currently using ~32GB for caching, that's more than many people have for system RAM and it obviously improves system responsiveness for certain apps. (modern phones use this trick too if you think about it)

 

[DavidC1]

I use RAM drives regularly so I know what maximum responsiveness feels like and no, you can't feel much difference between a RAM drive and a modern SSD,

RAM disks can't boost responsiveness in general Windows operations because the whole point of BOOT in computers is moving the data from storage drive to RAM, thus most is already in RAM anyways, which is much faster than what RAM Drive with software overhead can ever deliver.

An Optane NVMe drive is about same responsiveness as a RAM drive for this reason.

We went from random operation performance like 0.5 Read (MB/s) and 1.5 Write (MB/s) to something like 15-20 Read and 20-60 Write. In order to experience a similar jump, today we would need 200+ Reads and 600+ Writes for Q1T1 tests.

Even then it still won't feel as fast because you start being more limited by other parts of the system such as the CPU. Reading about Optane experiences, it sped up more "niche" parts of computing experiences, such as deleting large amounts of data, or even more niche such as running out of RAM and going on pagefile.

The real revolution would have been if they offered Optane DDR sticks on the consumer side and worked with Microsoft for "bootless" systems, and pushing software vendors towards loading directly from it.
Mostly when working on files like large Excel sheets (they get saved much quicker). Running portable apps on the RAM drive etc.

I think Linux does feel faster from a UI responsiveness perspective...

Linux is infinitely better in terms of security, bloat and how it handles updates, and even in the HDD days the structure is such you didn't need defragmentation.

Microsoft couldn't reliably update until about Windows XP. Windows 95, 98, and even XP a lot of times struggled to simply get updates completed. And it took hours upon hours and many restarts.

Linux? After giving it the command it finishes updates in no more than a minute. Doesn't ever need to restart.

Technically Linux is far superior. No wonder Android is a distant cousin of it. Microsoft's idea of a "mobile OS" with Win 8/10/11 is a skin on top of it, while Android's is a seamless one.

@mikeymikec Windows is basically officially approved spyware. It costs to have something that spies and collects data on users.

 

[igor_kavinski]

RAM disks can't boost responsiveness in general Windows operations because the whole point of BOOT in computers is moving the data from storage drive to RAM, thus most is already in RAM anyways, which is much faster than what RAM Drive with software overhead can ever deliver.

Not everything is loaded into RAM. Take a portable browser for example. It still needs to write to its various SQLite DB files and store the temp files. Putting such a browser on the RAM drive makes that bottleneck go away, even with the slight software overhead.

Running a SQL Server DB on RAM drive is usually not easy but I managed to trick it by creating the DB, detaching it, moving to RAM drive and re-attaching it. HammerDB scores jumped 10x over the SSD scores. Using the RAM drive guarantees that RAM is used, bypassing the weaknesses of OS data caching strategies.

The drawback is obviously volatility so need to ensure to take regular backups. It's an annoyance but very much worth it in my opinion.

 

[DavidC1]

Not everything is loaded into RAM. Take a portable browser for example. It still needs to write to its various SQLite DB files and store the temp files. Putting such a browser on the RAM drive makes that bottleneck go away, even with the slight software overhead.

Yes but that's the problem. You have to move everything to RAMDrive to truly benefit.

Intel truly missed an opportunity to revolutionize compute with Optane. But hey, they were in a company-ending execution trouble, so who can blame them?

Intel's strategy: Optane Memory is the next attempt at caches for HDDs, and aimed at super value systems. In servers, Memory Drive is a software that allowed RAM extension, at a cost of $700 per Memory Drive software. Another example of why I say Intel is a finance company with engineers working for them.

My idea: Optane Memory would have took the Memory Drive software(free), and it would have aimed at enthusiasts to boost memory capacity and for pagefile storage. The latency is essentially low as RAMDrives anyway.

Future Optane Memory would have been the DRAM versions, and considering the server version was $600 for 128GB, client version could have been $200-250. It would be bundled with patched Windows which allowed demonstration systems where it can have a "BOOTless" mode and it would go from zero power use to desktop nearly immediately. Call it "Zero Power Sleep" or something. Then you would bring in software ecosystem to allow you to directly install to Optane Memory and advertise "bootless and loadless computing".

 

[coercitiv]

Even then it still won't feel as fast because you start being more limited by other parts of the system such as the CPU.

I intentionally left that out of the post as it was getting bloated already. Even if we somehow solve the CPU bottleneck, there's still the issue with diminishing returns in terms of perception: for a human being the jump from 5 seconds to 0.5 seconds has a bigger impact than the jump from 0.5s to 0.1s

Intel truly missed an opportunity to revolutionize compute with Optane.

I tried many times to convince myself to buy the consumer oriented modules, especially after they started getting big discounts. Every time the math did not work out, they were too small for an OS partition or too slow for caching/pagefile. The small size I could undertstand considering their cost, but gimping their speed marked them for commercial death. I never found out if this was a price issue or Intel just segmented these products into oblivion.

 

[mikeymikec]

Linux? After giving it the command it finishes updates in no more than a minute. Doesn't ever need to restart.

Pretty sure Linux Mint asks me to restart after say kernel updates. Sure, it doesn't happen anywhere as often as Windows asking for a reboot after a Windows update, but it does happen.

 

[fastandfurious6]

why optane extinct? intel bad decisions.... big opportunity for rival to step in

does it make sense to move any parts of OS (if viable) and browser to Ramdrive? would that quicken anything?

my new 9955hx3d behemoth is 96gb ram so at least 30-50gb is free ramdrive lol

 

[igor_kavinski]

does it make sense to move any parts of OS (if viable) and browser to Ramdrive? would that quicken anything?

Probably not unless you are doing I/O intensive stuff like saving large documents or editing uncompressed audio or image files etc.

If you are doing something you know involves a lot of random 4K I/O, doing it on the ramdrive will feel faster since that's one thing where SSDs haven't improved that much.