NVMe P4600 speed partially lower than SATA SSD ?!

[JAnde19]

Hello,

i have one Intel P4600 DC with 4TB and one Samsung 850 EVO 500GB (normal SATA). Even with an old Intel SSD 520 Series the results in reading are not that different!

How is it possible that with small files the Samsung SATA SSD outperforms the Intel NVMe SSD (in read speed)?

The P4600 ist updated with the latest Firmware (with the current datacenter tool) and runs with the latest intel drivers (the windows drivers are even slower!)

Is this a normal behavior or a configuration/software problem?

Intel P4600 DC NVMe (on PCIe 3.0 x8 Port):

Samsung 850 Evo (SATA 3.0):

Intel 520 Series (SATA 3.0):

[idata]

Hello JemandAnderes,

Thanks for posting your question in the Intel® SSD community.Benchmarking for SSDs for comparison needs to be performed with specific procedures and with SSDs under the same conditions to make sure you are getting the expected results. For instance, is your OS installed in the P4600? If the OS is installed and running the benchmark on the same P4600, this would be the reason why the read lectures are lower than what it should. Since this SSD is running various tasks on top of the benchmarking tool.Intel's procedure to benchmark SSDs is the following: https://www.intel.com/content/dam/support/us/en/documents/ssdc/hpssd/ssd-750/sb/intel_ssd_750_series... https://www.intel.com/content/dam/support/us/en/documents/ssdc/hpssd/ssd-750/sb/intel_ssd_750_series...The recommended tools for benchmarking are the following: http://www.iometer.org/doc/downloads.html http://www.iometer.org/doc/downloads.htmlhttps://crystalmark.info/en/software/crystaldiskmark/ https://crystalmark.info/en/software/crystaldiskmark/Additionally after checking the only lectures that are higher are the initial Reads, after 512KB the P4600 runs away in speed and it is not even close to the other 2, in benchmarking the result comparison is done at the highest speed in the highest peak of usage and the P4600 is way faster than the other 2 SSDs. Since the PCIe port is faster than the SATA port this would be the normal behavior. How many times have you performed this tests? I will be waiting for your feedback. Hope this information helps to clarify your questions. Best Regards, Juan N.

[JAnde19]

Hello Juan,

thanks for your answer!

Benchmarking for SSDs for comparison needs to be performed with specific procedures and with SSDs under the same conditions to make sure you are getting the expected results. For instance, is your OS installed in the P4600?

-> No i did all the tests on the same system. All SSDs are completely empthy. OS is installed on another drive. So there are the same conditions for all tests i made.

Additionally after checking the only lectures that are higher are the initial Reads, after 512KB the P4600 runs away in speed

-> This is true, but why is this? I bought the P4600 for high speeds even with small data (Database/Access of small pictures), but for now even my small HDD RAID Array is FASTER than this P4600 with small data. I even tried to stack 4 of these P4600 together in a windows storage pool, but NOT even in raid 0 mode (i know softraid but still) they smash the hdd array. Thats what i find a bit disappointing, because i thought these NVMe SSDs would be way faster.

I'll try it again with IOMeter and CrystalDiskMark.

[idata]

Hello JemandAnderes,

I have been researching on your issue and I was able to find information for the reason why the ATTO tool for benchmarking shows the information this way. First I found the evaluation guide for the P4600: http://manuals.ts.fujitsu.com/file/12176/fujitsu_intel-ssd-dc-pcie-eg-en.pdfInitially, not all Benchmarking tools behave the same or treat data the same way at the time of running the tests. I have found the following thread that explains the difference on how ATTO and crystaldiskmark handle the information. https://linustechtips.com/main/topic/682597-why-do-crystaldiskmark-and-atto-benchmarks-report-differ...There are several reports of customers having the same issue as you with other tests, here is an example: https://www.overclock.net/forum/21-benchmarking-software-discussion/1060279-atto-vs-crystaldiskmark....Here is another article that you can check as a reference which covers the difference between some benchmarking tools in the market: http://www.thessdreview.com/our-reviews/runcore-t50-6gbs-msata-120gb-ssd-review-atto-crystal-disk-ma...Have you performed the tests with crystal diskmark?Hope all this information helps you. Let me know if you have any additional questions.Best Regards, Juan N.

[idata]

Hello JemandAnderes,

Were you able to perform the tests with the validated tools? Is there any additional feedback regarding this issue? Best Regards, Juan N.