The Crucial P1 1TB SSD Review: The Other Consumer QLC SSD
by Billy Tallis on November 8, 2018 9:00 AM ESTRandom Read Performance
Our first test of random read performance uses very short bursts of operations issued one at a time with no queuing. The drives are given enough idle time between bursts to yield an overall duty cycle of 20%, so thermal throttling is impossible. Each burst consists of a total of 32MB of 4kB random reads, from a 16GB span of the disk. The total data read is 1GB.
When the Crucial P1 has plenty of unused capacity and its SLC cache is large enough to contain the entire 16GB of test data, the burst random read performance is excellent. When the drive is full and the test data can no longer fit in the SLC cache, the performance falls behind the Crucial MX500 and most low-end NVMe SSDs.
Our sustained random read performance is similar to the random read test from our 2015 test suite: queue depths from 1 to 32 are tested, and the average performance and power efficiency across QD1, QD2 and QD4 are reported as the primary scores. Each queue depth is tested for one minute or 32GB of data transferred, whichever is shorter. After each queue depth is tested, the drive is given up to one minute to cool off so that the higher queue depths are unlikely to be affected by accumulated heat build-up. The individual read operations are again 4kB, and cover a 64GB span of the drive.
The sustained random read performance of the Crucial P1 at low queue depths is mediocre at best, falling behind most TLC-based NVMe SSDs and the Crucial MX500. By contrast, the Intel 660p manages to retain its high performance even on the sustained test, indicating that the Intel drive kept more of the test data in its SLC cache than the Crucial P1 does. When the test is run on a full drive, the P1 and the 660p have equivalent performance that is about 12% slower than the P1 with only the 64GB test data file.
 |
|||||||||
| Power Efficiency in MB/s/W | Average Power in W | ||||||||
The power efficiency of the Crucial P1 during the sustained random read test is less than half of what the Intel 660p offers, due almost entirely to the large performance difference. At just over 2W, the power consumption of the P1 is reasonable, but it doesn't provide the performance to match when the test data isn't in the SLC cache.
 |
|||||||||
The random read performance of the Crucial P1 increases modestly with higher queue depths, but it pales in comparison to what the Intel 660p delivers by serving most of the reads for this test out of its SLC cache. Even the Crucial MX500 develops a large lead over the P1 at the highest queue depths, while using less power.
Plotting the sustained random read performance and power consumption of the Crucial P1 against the rest of the drives that have run through our 2018 SSD test suite, it is clear that the drive doesn't measure up well against even most SATA SSDs, let alone NVMe drives that go beyond the SATA speed limit when given a sufficiently high queue depth. Thanks to its SLC cache being more suited to these test conditions, the Intel 660p is among those NVMe drives that beat the limits of SATA.
Random Write Performance
Our test of random write burst performance is structured similarly to the random read burst test, but each burst is only 4MB and the total test length is 128MB. The 4kB random write operations are distributed over a 16GB span of the drive, and the operations are issued one at a time with no queuing.
The burst random write performance of the Crucial P1 is good, but not quite on par with the top tier of NVMe SSDs. The Intel 660p is about 10% slower. Both drives clearly have enough free SLC cache to handle this test even when the drives are completely full.
As with the sustained random read test, our sustained 4kB random write test runs for up to one minute or 32GB per queue depth, covering a 64GB span of the drive and giving the drive up to 1 minute of idle time between queue depths to allow for write caches to be flushed and for the drive to cool down.
The longer sustained random write test involves enough data to show the effects of the variable SLC cache size on the Crucial P1: performance on a full drive is less than half of what the drive provides when it only contains the 64GB test data. As with the burst random write test, the P1 has a small but clear performance advantage over the Intel 660p.
 |
|||||||||
| Power Efficiency in MB/s/W | Average Power in W | ||||||||
When the sustained random write test is run on the Crucial P1 containing only the test data, it delivers excellent power efficiency. When the drive is full and the SLC cache is inadequate, power consumption increases slightly and efficiency is reduced by almost a factor of three.
 |
|||||||||
Even when the random write test is conducted on an otherwise empty Crucial P1, the SLC cache starts to fill up by the time the queue depth reaches 32. When the drive is full and the cache is at its minimum size, random write performance decreases with each phase of the test despite the increasing queue depth. By contrast, the Intel 660p shows signs of its SLC cache filling up after QD4 even when the drive is otherwise empty, but its full-drive performance is steadier.
Plotting the Crucial P1's sustained random write performance and power consumption against the rest of the drives that have completed our 2018 SSD test suite emphasizes the excellent combination of performance and power efficiency enabled by the very effective SLC write cache. The P1 requires more power than many SATA drives, but almost all NVMe drives require more power to deliver the same performance, and the very fastest drives aren't much faster than the peak write speed of the Crucial P1.
Facebook
Twitter
RSS
66 Comments
View All Comments
Lolimaster - Friday, November 9, 2018 - link
With worse of everything how is it going to be "faster", do any TLC SSD beat the Samsung MLC ones? No.Valantar - Thursday, November 8, 2018 - link
What's the point of increasing performance when current top-level performance is already so high as to be nigh unnoticeable? The real-world difference between a good mid-range NVMe drive and a high end one are barely measurable in actual real-world workloads, let alone noticeable. Sure, improving random perf would be worthwhile, but that's not happening with flash any time soon. Increasing capacity per dollar while maintaining satisfactory performance is clearly a worthy goal. The only issue is that this, as with most drives at launch, is overpriced. It'll come down, though.JoeyJoJo123 - Thursday, November 8, 2018 - link
^ This.For typical end users, even NVMe over SATA3 SSDs don't provide a noticeable difference in overall system performance. Moving to an SSD over an HDD for your OS install was a different story and a noticeable upgrade, but that kind of noticeable upgrade just isn't going to happen anymore.
Typical end users aren't writing/reading so much off the drive that QLC presents a noticeable downgrade over TLC, or even MLC storage. Yes, right now QLC isn't cheap enough compared to existing TLC products, but we've already done this dance when TLC first arrived on the scene and people were stalwart about sticking to MLC drives only. Today? We got high-end NVMe TLC drives with better read/write and random IOPS performance compared to the best MLC SATA3 drives back when MLC was the superior technology.
Yeah, it's going to take time for QLC to come down in price, the tech is newer and yields are lower, and companies are trying to fine tune the characteristics of their product stacks to make them both appealing in price and performance. Give it some time.
romrunning - Thursday, November 8, 2018 - link
Sure, we lost endurance and speed with the switch from MLC to TLC. But the change from TLC to QLC is much worse in terms of latency, endurance, and just overall performance. Frankly, the sad part is that the drive needs the pseudo-SLC area to just barely meet the lowered expectations for QLC. Some of those QLC drives barely beat good SATA drives.We now have a new tech (3D Xpoint/Optane) that is demonstrably better for latency, consistency, endurance, and performance. I'd rather Micron continue to put the $ into it to get higher yields for both increased density/capacity & lower costs. That's what I want on the NVMe side, not another race to the bottom.
JoeyJoJo123 - Thursday, November 8, 2018 - link
Sorry, you're not the end consumer that dictates how products get taped out, and honestly, if you were in charge of product management, you'd run the company into the ground focusing on making only premium priced storage drives in a market that's saturated with performance drives.The bulk of all SSD sales are for lower cost lower storage options. There is no "race to the bottom", it's just some jank you made up in your head to justify why companies are focusing on making products for the common man. Being able to move from an affordable 500GB SSD on TLC to an similarly priced 1TB SSD in a few years is a GOOD THING.
If you want preemium(tm) quality products, SSDs with only the HIGHEST of endurance ratings for the massive Read/Write workloads you perform on your personal desktop on a day-to-day basis, SSDs with only the LOWEST of latencies so that you can load into Forknight(tm) faster than the other childerm, then how about you go buy enterprise storage products instead of whining in the comments section of a free news article. The products you want with the technology you need are out there. They're expensive because it's a niche market catered towards enterprise workloads where they can justify the buckets of money.
You keep whining, I'll keep enjoying the larger storage capacities at cheaper prices so that I can eventually migrate my Home NAS to a completely solid state solution. Right now, getting even a cheap 1TB SSD for caching is super-slick.
romrunning - Friday, November 9, 2018 - link
"...how about you go buy enterprise storage products instead of whining in the comments section of a free news article."You are taking this way too personally.
I'm actually thinking more about the business side. I want 3D-Xpoint/Optane to get cheaper & get more capacity so that I can justify it for more than just some specific servers/use-cases. So I'd like Micron to focus more on developing that side than chasing the price train with QLC, which is inferior to what preceded it. With Micron buying out Intel's stake in IMFT for 3D-Xpoint, I just hope the product line diversification doesn't lessen the work to make 3D-Xpoint cheaper & even greater capacities.
JoeyJoJo123 - Friday, November 9, 2018 - link
>You are taking this way too personally.Talk about projecting. Micron is taping out dozens of products across different product segments for all kinds of users. They're working on 3D-Xpoint and QLC stuff simultaneously and independently from each other. What's happening here is that Micron is producing QLC NAND for this Crucial M.2 SSD, and you're here taking it personally (and therefore whining in a free news article comments section) that Micron isn't focusing enough on 3D-Xpoint and that supposedly their QLC is bad for some reason. Thing is, this news article isn't for you. This technology isn't for you. You decided your tech needs are above what this product is aimed for: affordable, large volume SSDs for lower prices.
Seriously, calm down. This wasn't an assault orchestrated by Micron against people that need/want higher performance storage options. More 3D-Xpoint stuff will come your way if that's the technology you're looking forward to. Again back to my main point, it's going to take some time for these newer technologies to roll out. Until then, don't whine in comments sections that X isn't the Y you were waiting for. If the article is about technology X, make a half-decent effort keep to the topic about technology X.
mathew7 - Tuesday, November 13, 2018 - link
"I'll keep enjoying the larger storage capacities at cheaper prices so that I can eventually migrate my Home NAS to a completely solid state solution."Wwwwwhhhhhhhaaaaaaaaaaattttt?? NEVER. You don't understand the SSD limits. I would not do that with SLC (assuming current quality at QLC price).
Enterprises with SSD NASes only use them for short-term performance storage with hourly/daily backup. Anyone who uses them differently is asking for a disaster.
Look for linuxconf Intel SSD. There is a presentation where they explain how reading a cell damages nearby cells and manufacturers need to monitor this a relocate the data that is only read.
I have 2 servers with only 1 SSD each for OS and 8-10TB HDDs for my actual long-term data.
All my desktops/laptops have SSDs (Intel 320, Samsung 830-860 evo+pro, Crucial BX100/MX300 etc). But anything important on SSDs will be backed-up to HDDs.
Oxford Guy - Thursday, November 8, 2018 - link
"That's what I want ... not another race to the bottom."That's what consumers want: value.
That's not what companies want. They want the opposite. Their wish is to sell the least for the most.
Mikewind Dale - Thursday, November 8, 2018 - link
"[Companies] want the opposite. Their wish is to sell the least for the most."Not true. Companies want to maximize net revenue, i.e. total revenue minus cost.
Depending on the elasticity of demand (i.e. price sensitivity), that might mean increasing quantity and decreasing price.
A reduction in quantity and an increase in price will increase net revenue only if demand is elastic.
But given the existence of HDDs, it makes sense that demand for SSDs is elastic, i.e. price-sensitive. These aren't captive consumers with zero choice.
Of course, nothing stops a company from catering to BOTH markets, i.e. high performance AND low cost markets.