Just because they're rated the same...

[tlmiller]

So, recently been seeing REALLY good prices on PNY's OLDER 2TB NVMe SSD, the CS2130. This was replaced by the XLR8 CS3030 that I have. However, they're both available in 2TB size, and, interestingly, the specs from PNY show IDENTICAL ratings. Well, you know I can never pass up a cheap (and supposedly good performance) drive, so I bought a CS2130.
The pics will show everything, but just because they're rated identically, they really DON'T perform that way. The CS3030 is FAR faster throughout a variety of tests, and only with large files and high queue depths does the CS2130 keep up with it.
That said...for the price these CS2130's are going for ($125 for 2TB), they're not exactly what I'd consider a bad value. Far from it. If you need a high capacity PCIe 3 drive, and these are available for this price, I'd definitely suggest jumping on it. Just interesting at how massive difference there is in some fo the benchmarks for 2 drives that are rated identicall in terms of both read/write spead, and iops.

[crosscourt]

Ill check that out as I will need a 2tb drive of that type down the road. I also go for decent performance if the price is lower as high end drives are still too expensive. I still use standard hard drives because my wife needs more storage and I can get 2,3,4tb drives really cheap.

[tlmiller]

Yeah, I've seen 4TB spinning rust for like $60. Huge difference from my talking about how 2TB for $125 is an amazing deal on SSD's.

[crosscourt]

I just picked up a used 4tb drive for $38 2021 and a 2tb drive for $25 used also 2021. When I tested them they had very low usage and performance was excellent. One is a Seagate Barracuda and the other is a WD Blue.

[tlmiller]

I got bored the other day and actually opened the ticket for my T14G1A, tentatively scheduled to have repair done Wednesday (I buy onsite warranty w/ ADP...it's worth the extra money to have repairs done at my house)

[tlmiller]

So, interesting thing I noticed. I decided to throw a drive in the T14G1 with windows on for when they arrive to replace the Mobo. Put the Helix-L in there (it's one of the only 1TB single sided I have in an enclosure, most of them got put in enclosures because I couldn't use them in Lenovo's due to being dual sided (EX920 & SP34A80).
Anyway, since I installed Windows, I downloaded Crystal Diskmark and ran through the same file size benchmarks as I have with Kdiskmark. @16MB, linux reads were a little faster, windows writes were NEARLY DOUBLE the speed of Linux (800 Mbps Linux, 1500 Mbps Windows).
@128MB, linux reads were moderately faster, writes were basically tied.
@1GB, linux reads were quite significantly faster, writes were basically tied.

Benchmarks were probably not captured in the same system, but they'd both be PCIe 3x4 controllers in modern(ish) minimum systems, and if anything the T14G1 should be faster since it's much newer. Not sure what machine I had the Helix in for Kdiskmark benchmarks, but still. Not really drawing any conclusions, but I found it quite interesting that the same drive with 2 different OS's could get such drastically different results.

[wove]

I find that very interesting. I think it touches on the difficulty inherent in benchmarking. Manufacturers tend to rate their products based on theoretical maximums of the devices throughput, leaving it to upstream designers and engineers as to how to obtain those theoretical levels. The hardware controller on the systems plays a part in that. The kernel will determine how often the OS pays attention to the whole subsystem and of course the efficiency of the benchmarking software will play as big roll as well. To obtain the maximum will require a good deal of synergy between all these components.

I think benchmarking is most useful in looking for choke points on a system and sorting out cost effective upgrades. Benchmarking also provides insights into the priorities of OS designers and help developers find ways to compensate for weaker components.