Lexar is showing off a PCIe 4.0 SSD with 7GB/s read sequential speeds

Copy thousand RAW files or 20k RAW files. Makes no difference, process of processing is the same. I just gave it as an example of being extremely slow storage and extremely fast storage. You can do 20k files from a large HD or from very large SD card.

Here is another way to prove it that SSD does nothing for workflow. If you have a desktop that has LED light connected so it shows SSD access do your workflow and watch the light. You will not see it blinking because read and writes are so fast that SSD work is minuscule compared to the rest of the work CPU has to do. I am not arguing that SSD is slow but it's speed adds nothing to the workflow. I have 3 desktops and 4 laptops some with HD and some with SSDs so I do know what I am talking about. The fastest workflow in on 3930K liquid cooled desktop running all 6 cores at 4.5Ghz. And it has only small SSD for OS but large HD for data. Second computer has CPU running at 4Ghz and 2 SSDs one for OS and one for data but it is no where close to the first one because it is not overclocked. That right there proves that SSD does not add any speed doing photographic work.

SushiEater. I now understand your earlier comments with your latest clarification.
Bottom line is you're correct. The faster SSD speed plays a very small part here. I also understand now what you mean by 'cache' although in this case, the caching benefit probably mostly comes from the photo app, not Windows i.e. the photo app is smart enough to know how best to read/write data.
On your comment about just copying large number (e.g. 20K) of files. The bottleneck is similar because lots of system overheads from dealing with each file (open/access/close ...) and the time needed to copy the (photo) data is relatively small compared to system overheads. Depending on the hardware config, a whole hi-res raw file could be read in just one single operation so that's why the SSD LED hardly comes on

From my experience with Adobe products you need huge amount of RAM and a lot of time and patience to complete some photo-related work.
But faster R/W speed is very useful if you need to work with RAW/High bit rate video.
And cataloging is not a must if you are not using LR but PS/Bridge.
Even with classic 7200rpm HDD it is easy to get 80-150MB/s sequential read speed. Which is common with big files processed in batch. This is something like 6x24MP pix or 2x 100MP pix per second. If we say this are 24Mpix files for 20K we will get around 1 hour time to copy/read them. Not really sure how this will relate to LR cataloging speed. And how faster will be if you R/W with 50x speed.
In addition from my experience faster flash drives are very useful for very large files like high quality/RAW video and/or for high I/O operations. Which is really nothing related to photo workflow.
But this is just my experience.

another guy who slightly relate to photography and never work on projects with tens of thousands of RAW files. you see, bro, when you load dozen of RAW files there is no noticeable difference. when you working with hundreds of files there is still not a really big difference. difference starts with >10K of photos / project

imagine you have to load fast (and as much faster is as much better) catalog with tens of thousands of RAW files. common practice in industry.

yeah, for sure, few files per second is good. wait. if you have 10K of RAW files (just another intense photo session for casual professional in many sectors of industry) and you want to have them loading at least in few seconds entirely?

here comes the bottleneck. people who are halfway in photography would never have a clue how is this is problem. while with common intense photo shoot or project consisting from few or more sessions you have tens of thousands of RAW files and you need to catalog and edit them fast enough

cause than faster you cataloging them than faster you can edit all of them. and what if you need deep colorgrading and create solid concept of outcome look? even got a clue what about i'm talking here? i really doubt it

and this is common practice for many people in the industry

as AcerK completely right said above there is many different workflows. this worklow is common and typical for many.

and thus way speed of reading/writing of data storage can boost your workflow not a few times more but 10x and more. got it dude? it's so simple please don't bother me anymore. i'm tired of explaining obvious things

MrSolo, thanks for a link. i just have not so much time and stamina to eternally arguing about things that i really don't need to argue about

@iozx did you checked the tables with results from provided link from MrSolo above?
Please do not throw specs digits here, I have asked for your experience as you said multiple times how many times you have worked with projects big as 10-30K RAW files. If I need to compare specs I can do it muuuch better than you, so don't worry about specs. As I said I have extensive experience with SYSTEM support. Not with LR projects files support!!!
We are talking about real experience with catalogs and LR here.
And yes you can get some better responsiveness when move your system drive from HDD to SSD. But between SSD and NVMe and LR you will gain almost nothing. Does not matter the specs!

Of course you will not participate ........
https://www.pugetsystems.com/labs/articles/Adobe-Lightroom-2015-8-Storage-Performance-Analysis-875/
-
https://www.pugetsystems.com/recommended/Recommended-Systems-for-Adobe-Lightroom-Classic-141/Hardware-Recommendations
-
Primary Drive - ...........An SSD is highly recommended as it will greatly improve how fast the OS and programs startup, but there is usually not much of a performance benefit to upgrade to a faster NVMe drive.
Secondary Drive - .............. For most users even a platter drive should be more than fast enough, although a SSD tends to be snappier and will often smooth out your workflow.
-
Maybe we need to advise them to replicate all tests with minimum of 10K RAW files. But you say nothing about resolution. Do we need to use 100Mpix or maybe 150Mpix files to replicate your unknown results. Because you provided 0 ( zero ) digits, just feelings and definitions about our "unprofessionalism".

Quote from above links comments section:
"Matt Bach Mod mclaren777 • 2 years ago
Lightroom Classic has greatly improved performance in a lot of areas, so while we saw no difference between a standard SSD and the Samsung 960 Pro NVMe drive in this article, there is a small chance that NVMe drives in general might give you a small performance bump now. I don't think it will be much (if anything), and if there is an improvement, I don't think the 900P is really going to be better than something like the Samsung 960 Pro. The 900P's strength is in it's endurance and lower latency, but for Lightroom I would imagine that the higher sequential read/write of the 960 Pro would be more important. Again, however, I don't think we're talking about more than a percent or two performance improvement using either of those drives over a standard SSD, so I wouldn't really worry about it too much"

Based on above tests for LR experience NVMe is no must! And almost all of the Adobe products are same or worse. No good hardware utilization.

iozx- the 10x difference in sequential transfer isn't particularly relevant to sqlite performance. That's the weakest part of your argument. Latency is key, not bulk transfer rate, as the data objects are small.

Seems many latest comments have been removed by DPR?
Anyway Gentlemen, no need to disagree too strongly with each other. As I've said earlier this performance eval thing was very complicated and carries plenty of caveats.
Everybody here have his/her own valid observation IMO and those links mentioned here should be taken with plenty skepticism too.
I've done perf benchmarks with top 5 hardware and software vendors over the years, including Intel beta enterprise NVMe in 2016. Intel sent us 'disks' and controllers and we had direct access to Intel NVMe engineers to optimize their hardware speed. On the other end we also had help from the developers who wrote the code for the app being tested to tune their code for whatever numbers they want to show.

The vendors would do anything to help with the testing so the best numbers are in their favor. For example one time they fedex us a controller taken straight from their own lab and another time I was sitting next to the lead app developer in a big lab evaluating test results and making code changes on the fly.
So even with Adobe+Lexar intimately involved in a perf eval, the results should be parsed carefully as the test cases might not reflect common real life cases, hardware setup and almost always attacked by competitors. It's always with lots of disclaimers and with the label YMMV attached.
Bests

@MrSolo - Exactly, point here is that no one show better performance results with LR and ultra fast disk drives. Does not really matter the HW specs. This is why I asked about some real experience but went like ...... whatever. :)

SushiEater
You had the correct general idea about cache. In practice there are many schemes/levels of caching. I'm not sure how things work with LR. You need Windows+LR engineers who actually know what the underlying program code does to get the correct picture
In theory for performance reason LR could be fed just part of the raw file from the 1st read so it can get started right away w/o waiting for the whole thing while the rest is being read into a cache then made available when LR wants it.
As for the copying of large number of files, Windows/OS overheads is one major problem since the raw files themselves, even MF raw data, are relatively small (<2GB). There's a way to handle it in the enterprise world where large number of files typically means >1M

Missing a sentence.
"Imagine Windows has to repeatedly look up a file from a list of 20K names"

I have had only experience with a lot of RAW files (about 4000) few times. About 6 hours of shooting twice per year. In one of my laptops I have two SSDs, both M.2. 1TB for OS and data and another Intel 660p NVMe with 1650Mb speed all dedicated to Photos. I shoot for multiple companies so I need to generate enough good photos to send them all. Few times I had to use 1TB drive to process photos from because I was running out of the room on the larger drive. I see no difference in processing speed. I usually break 4000 photos into 6 sub-folders. If I leave all 4000 in one then it takes forever to generate thumbnails. Also, I need to write metadata to each photo and it would take also a long time. But with small amount of photos in each folder it takes just a few seconds for metadata and maybe a minute or less for thumbnails. So obviously amount of photos in one folder has something to do with the workflow.

“Wasted”?

Troll much?

Not being able to upgrade the sSD and RAM without voiding the warranty or having to go to a Genius Bar is the reason I don't buy Macs anymore. I can deal with the Windows 10 hassles but at least MIcrosoft isn't preventing me from doing what I want to my system's hardware. It does seem, though, on many laptops (ultrabooks especially) that mfrs have resorted to soldering things onboard so you can't upgrade them even if you wanted to void the warranty by breaking any tamper proof tape.

What are you saying you can change the RAM of Macs...

I'm saying on some of them, you can't (it's soldered on board). Same for the SSD in some cases. It's Apple's way of "making" you to buy what you think you need. You go for the 8GB version and think you need 16GB so you get the 16GB version instead. This tactic is not limited to Apple. Some of the Lenovo ultrabooks have soldered on storage and memory.

Some models (like the Mac desktops) you probably can upgrade the memory but on many of the newer laptops it's getting harder to do so.

PhotoDog2:

Which Leica has/had a tiny internal SSD?

Cameras from a variety of manufacturers have had internal flash storage over the years.

Yeah. A camera with really fast and large internal storage would just be better if it has USB3.x interface and you can backup to an external hard drive or computer. And of course backup with fast w-lan.

Maybe it's not worth the cost of hardware and development, but it would definitely be better. Maybe we'll see ilc cameras based on smartphone hardware that can more easily do this.

Mnemon,
" I don't see the negative - price aside."
Increased power consumption...

Red has SSD's for their cameras, if you are willing to pay their ridiculous over-pricing

https://www.red.com/RED-MINI-MAG-PARENT

Video-vs-Photo:

Flash storage can, and does, fail.

Not having removable cards (or SSDs) means one has to cable the camera to a computer to remove the files from the camera for processing. No, wireless isn't fast enough for say 300 raws, or 2 hours of "just" 1080p MP4s.

@HowaboutRAW - every type of storage can and WILL fail at some point!
Buuut flash chips used for SSD/NVMe are much better than the one used for all kind of cards. And because no external contacts with readers this will add to reliability.
Current Wi-Fi is fast enough but if you like faster speeds it is ok. But then what is the problem with USB 3/3.xx cable? Wire camera for 5-10 minutes and then start working. Seems like you prefer fiddling with cards and readers instead of USB cable.

I have a 1st Gen Corsair PCIe-4 NVMe that is 2-degrees warmer than my Samsung PCIe 3 NVMe when fans are on quiet. When the fans are higher it is 2-degrees cooler. It is likely caused by there position or maybe the heatsink doing a better job dissipating heat with more airflow.

Regardless, I've read that the reason for the slightly higher temps on 1st generation PCIe-4 cards is that that they are using older controllers at max capacity. But that future generation PCIe -4 cards will have newer more efficient controllers, is this one, I don't know. But it likely judging by the improved speeds. Overall, I am very happy with my purchase since it is cheaper than current PCIe-3 Samsung and the temps are comparable.

Same here, Ryzen build done recently with 3900X. I have already 500GB Samsung 970 EVO Plus PCIe 3.0 (x4), NVMe SSD, 3500MB/s Read,3200MB/s Write. Synthetic test shows the very same results but in reality it's not faster than my regular 2.5" SSD Samsung 860 Evo. Also locations of m.2 slots are very bad. Right under CPU and GPU. These new Lexar pieces would only add extra heat and could actually degrade performance IMO.

@milkod2001 - this poor performance experience is not because bad slot location or heat. It is just because we continue to receive poor and bad made software! For well made apps this NVMe drives does make difference. But for everyday desktop apps they are overkill. Same as more than 4 cores, triple or quad channel memory and so on.

@panther, It's not a question of absolute performance but what is required to sustain the workload without slowing it down. The ratio of instructions to memory determines that and for many workloads these NVMes are sufficient. And I was speaking to primarily read-bound workloads.

@panther, Optane is fine but a bit more expensive. There are competitors to Optane as well that beat it on some metrics.

Linus tech tips on YouTube made a fun video about replacing their server. Even with slower drives you can almost max out read and write speeds of an epyc cpu with raid. I think they clocked around 27 gigabytes a second read from memory, with write not too far behind