The Team Group MP34 512GB SSD Review: Refining Phison's NVMe Workhorse
by Billy Tallis on May 15, 2019 9:00 AM EST- Posted in
- SSDs
- Storage
- Toshiba
- Phison
- M.2
- NVMe
- BiCS
- 3D TLC
- Team Group
- PS5012-E12
AnandTech Storage Bench - The Destroyer
The Destroyer is an extremely long test replicating the access patterns of very IO-intensive desktop usage. A detailed breakdown can be found in this article. Like real-world usage, the drives do get the occasional break that allows for some background garbage collection and flushing caches, but those idle times are limited to 25ms so that it doesn't take all week to run the test. These AnandTech Storage Bench (ATSB) tests do not involve running the actual applications that generated the workloads, so the scores are relatively insensitive to changes in CPU performance and RAM from our new testbed, but the jump to a newer version of Windows and the newer storage drivers can have an impact.
We quantify performance on this test by reporting the drive's average data throughput, the average latency of the I/O operations, and the total energy used by the drive over the course of the test.
The overall performance of the Team Group MP34 on The Destroyer is the highest we've seen from a 512GB-class drive and a clear improvement over earlier Phison E12 firmware. We haven't tested the latest 500GB TLC drives from Samsung or WD, but even if Phison's latest firmware can't match them it is still competitive. The 1TB Phison E12 drives with older firmware are only about 12% faster overall than the 512GB MP34, so they clearly need to get updated.
The average and 99th percentile latencies on The Destroyer have both improved slightly for the MP34 compared to the Gigabyte drive with older firmware, but the 1TB class drives still have a larger advantage here than shown on the overall average data rate.
The new Phison E12 12.2 firmware seems to have a bigger impact on average read latencies than average write latencies. The Team MP34 is virtually tied with the Samsung 970 EVO for read latency, and Phison's existing lead in average write latency is barely improved.
The 99th percentile read and write latency scores show roughly the same effects as the average read and write latencies: the new 12.2 firmware is a real improvement to read latency, but has minimal effect on write latency—not that Phison needed to work on the latter.
The Team MP34 uses less energy over the course of The Destroyer than any other 512GB-class high-end NVMe drive we've tested, but SATA drives and particularly efficient entry-level NVMe drives are still more efficient overall despite being significantly slower to complete the test.
17 Comments
View All Comments
ssd-user - Sunday, May 19, 2019 - link
I see that you are still in denial about how it was you who couldn't read diagrams. I'd also like to point out that I'm actually trying to be the change I want to see exactly by asking for the sorting to be fixed.Because the sorting clearly is wrong. I pointed out a very stark example of when a much worse drive sorts above the better ones.
Also, your lack of reading comprehension is showing in how you think this is only about TRIM. As I said, this is about disk full situations. And even with TRIM, the disk may simply be close to full. Not everybody buys an SSD that is twice as big as it needs to be.
I was also pointing out that even if your drive isn't full, it may well show the full behavior in reality.
Sorry for not being your ideal party buddy.
peevee - Monday, May 20, 2019 - link
Who uses their SSDs full to the brim and in sustained write mode? Honestly, that scenario is not even realistic for properly managed DB servers, let alone in client systems where the only wait time which actually happens is during system boot/application launch/data load on up to 80% full (in Anandtech-speak "empty" system).Client writes are all cached first and the write itself happens in background, the user does not have to wait anything.
AT does not even test this scenario properly, even their "Light" test is WAY too write-heavy for that.
A synthetic which would reflect that is something like "64kb random read" (runs are 16 clusters=64k on NTFS, and most DLLs are close to that size).
MDD1963 - Wednesday, May 15, 2019 - link
660P from Intel is $109 for 1 TB....; even though it is 'only' 2x PCI-e lanes capable, it is still more than 'snappy' for that sort of cost/capacity ratio....peevee - Monday, May 20, 2019 - link
Why do they even use x4 PCIe when they cannot even saturate x2? Really, peak read of 1.4GB/s is pathetic.DyneCorp - Monday, June 17, 2019 - link
"The write endurance ratings are still competitive with high-end drives that offer five year warranties"The MP34 has over twice the endurance of any SSD utilizing the SM2262 with Micron NAND. I apologize, but I'm not understanding what you mean by "still competitive". Seems as if Phison is outclassing the competition in certain regards. A small sacrifice in performance for exceptionally more endurance.
DyneCorp - Monday, June 17, 2019 - link
Metrics*, not regards ha.crimson117 - Friday, April 24, 2020 - link
Looks like the new MP34's offer a 5-year warranty:256GB - TM8FP4256G0C101
512GB - TM8FP4512G0C101
1TB - TM8FP4001T0C101
https://www.teamgroupinc.com/en/product/mp34
https://www.teamgroupinc.com/en/catalog/act.php?ac...