Quick Review: PNY XLR8 CS3030 2TB PCIe Gen3x4 NVMe SSD

Having recently started using my Lenovo Legion 5 (17ITH6) laptop, one thing I felt was also lacking was storage. The included Western Digital SN730 1TB SSD wasn’t bad by any measure, but 1TB was just not quite big enough for dealing with large datasets (like SDR I/Q recordings), so I decided that I needed to add additional storage to the available M.2 slot.

I decided that I would be looking for a 2TB-class NVMe SSD, with preference to units that do not have a heatsink (as it may not fit the laptop), containing DRAM cache for better performance, have reasonable endurance and was available for immediate purchase at a price below AU$300. While PCIe-4.0 based SSDs are all-the-rage because of PS5 users, I decided this was not necessary as the laptop couldn’t run the interface at that speed, and because such SSDs may run hotter and generally cost more (outside of a pricing error).

Trawling the pages of Amazon, I eventually shortlisted my selections down to just a pair of models. I settled for the PNY XLR8 CS3030, having read a few reviews that indicated its performance to be in the mainstream-performance category. As I had made the silly mistake of buying this before the “Prime day” sales event caused a bit of a drop in prices, I probably overpaid a bit for this.

It was only after my purchase did I realise that I had likely been duped – PNY have, like some other vendors, changed the product without changing the model number. Worst of all, they have cut the claimed TBW endurance significantly from 3115TBW to 660TBW, without clarifying the reason (likely a change to QLC alongside a change of controller). This was very unfortunate and definitely an irritating outcome, but I decided to see what would turn up as I think most reviews probably reflect the results for the older model.

Unboxing

The unit arrived in a matte-finish colour-print retail box. It seems they’ve taken the red colour scheme to indicate speed, proudly proclaiming “up to” 3500MB/s read, 3000MB/s write and a 5-year warranty (although, subject to TBW limits). This unit clearly indicates it is a PCIe Gen 3×4 SSD, a fact which is confirmed by the claimed speeds.

The rear features the key specifications in multiple languages and some clarifications about warranty details in different regions. This particular unit has a model number of M280CS3030-2TB-RB with a barcode number 4712847098510.

There really isn’t much provided in it – just a plastic tray with the SSD itself and that’s it! Note how the PCB appears different from the review units and the ones on the website – the DRAM cache IC is visible nearest the connector, suggesting this is not one of the original units, but a newer “downgraded” unit.

The rear side of the SSD is also covered with a label. Unfortunately, this means no peeking at the markings on the NAND chips and controller, or risk voiding the warranty on this brand-new drive.

No surprise – for a 2TB SSD, there are chips on both sides.

Firmware Update

To start with, I decided to get the PNY PCIe Toolbox software and check for firmware updates. The drive apparently did not have any updates available at the time of testing.

SMART Data

On the initial boot, the SMART data was very clean … nothing to report.

Testing took several boots of the machine (due to family emergency, I had to shut down the machine and test in several sessions), but otherwise a total of about 10TB was written and the media wearout indication still reads 0%. This is perhaps a good sign, as the expected value would have been 1.5% based on the 660TBW figure.

Performance Tests

Testing was performed in the second (non-heatsink) slot inside the Lenovo Legion 5 17″ Intel Edition (17ITH6) running Windows 11 from the OEM Western Digital SN730 SSD in the main slot. Because of this, the lack of cooling may have limited the performance of the drive due to thermal throttling behaviour.

HDTune Pro

Sequential reads on a fresh drive averaged 2956.8MB/s.

Once data is written to the drive, a minor slowdown is observed to 2830.8MB/s. Under thermally-constrained conditions (or fresh after filling), it seems to throttle back to just under 2000MB/s.

Sequential writes showed a relatively steady speed of about 2956MB/s right up to 600GB when it started to taper slightly, before dropping down at about 720GB. The write speed then maintained about 1500-1750MB/s with a fresh SSD.

With the SSD full, a follow-up write saw full speed only for about 75GB, then the speed reduced. In a dirty/thermally-constrained scenario, write speeds were in the 950MB/s-1000MB/s range.

Random Access tests were performed, but I’m not sure how reliable they are as it could be constrained by the program and its specific I/O patterns.

Extra Tests were also performed even though they are more relevant to hard drives.

A full surface random fill erase and verify was undertaken, completing in just over one hour, suggesting an average transfer rate around 1.1GB/s under realistic circumstances.

CrystalDiskMark

The default settings in CrystalDiskMark 8 yielded the above results, again not quite reaching the 3500MB/s and 3000MB/s on the box.

Choosing to run Peak Performance + Mix profile, the values are now closer for read and exceeding the claimed value for writes. Oddly, the mixed scenario sequential results were even better than pure read/write.

The Real World + Mix profile, however, shows a bit more of a reduction in performance, as expected. However, not having any firm numbers of non-boot PCIe NVMe drives to compare to, it is hard to judge,

AS-SSD Benchmark

AS-SSD gave the drive a score of 6646, with numbers for writes being a bit more positive than CrystalDiskMark.

The copy benchmark completed pretty quickly, although the program scenario seemed to take quite a bit longer than expected.

As with most modern SSDs, no compression effects seem to be measurable.

ATTO Disk Benchmark

ATTO results suggest the drive has practically reached top throughput for writes by 128KB accesses and has peaked in reads by 256KB.

Anvil Storage Utilities

Anvil Storage Utilities awarded the drive a combined score of 15,937.05.

H2testW

Testing with H2testw did not find any errors. Speeds may have been limited because of the program, dirty state of the drive or thermal throttling, however, the time taken is quite similar to the HDTune Pro random fill erase which may suggest the latter.

Conclusion

Unfortunately, it seems that I’ve been duped by PNY, with the CS3030 provided differing notably from the review units by a visual inspection, with a quiet change to the datasheets reducing the endurance TBW ratings significantly also meaning that the warranty is voided earlier with fewer writes. While the unit seems to operate happily without a heatsink, it will throttle operation when it reaches a measured 60 degrees C, as is common.

Regardless, it seems the performance of the PNY XLR8 CS3030 is quite solid overall, with HDTune Sequential figures peaking at 2956MB/s, and throttled writes hitting around 980MB/s. CrystalDiskMark mixed real-world load was less impressive, clocking in 2064MB/s, but that’s still miles ahead of any SATA-based drive. The endurance data from SMART doesn’t seem to be in-line with the datasheet reduction in endurance (expected 1.5%, observed <1%), so perhaps there are some inconsistencies here.

While this latter discovery may be encouraging, without voiding my warranty, I cannot confirm the memory and controller type, nor can I know just how long it will last. Had I paid this amount for a QLC drive (which I suspect I may well have), I will feel particularly shortchanged as I was specifically on the lookout for a scratch drive with well over 300-cycles endurance. Before anyone asks – yes, I do have SSDs with over 300 cycles on them still serving me on a daily basis … my scratch-disk for working on blog assets is one of them.

As a result, while the SSD does seem to post some respectable mainstream-performance scores, the fact that it has a TLC price-tag and may be QLC material makes it difficult to recommend. The behaviour of the company in quietly changing out or reformulating the product is one that has caused many to lose trust – for example, Patriot, Crucial, Kingston, ADATA, WD and Samsung have been caught doing this in the past. Worse still, quietly changing the datasheet to reduce the TBW endurance figure which has a direct impact on warranties is practically unforgiveable. This has been perhaps a regrettable purchase, but as it is technically functioning just fine, is not one I will return.

I do not recommend you take the gamble …