Showing results for 
Search instead for 
Did you mean: 

Intel 750 1.2Tb performance under Red Hat Linux

Okay, so I know there have been a lot of performance questions, but I'm going to toss another one out there because I'm wondering if there's a trick I'm missing.

My system:

ASUS X99 Sabertooth

Intel Xeon E5-2630 v3

32Gb Corsair DDR4-2400Mhz (which is useless with a Xeon unfortunately)

EVGA GeForce 950 < PCIe3x16 slot 1

Intel 750 SSD 1.2Tb < PCIe3x16 slot 2

Samsung 850 Pro 1Tb

2Tb WD Plate drive

Extra NIC

Blue Fans


Anyway, I have Red Hat 7 running. It installed right to the SSD with no problems. The device is there, I can partition it using fdisk, everything works great. I updated the system and the kernel to the latest version, and everything worked fine.

I also made sure it's in a PCIe3.0 slot and it's running at 4x. No worries there.

Then I start benchmarking it. I wanted to "condition" the drive as I've read that's important in order to receive Intel's performance metrics, so I did, writing a lot of data to a couple partitions over and over. I noticed that when I copied a directory tree with about 275Gb of data, it was copying at a rate of about 150Mb/s even after I had conditioned the drive. I turned to iozone to do some tests. I ultimately wound up doing the same exact benchmark on my Samsung that I did on my Intel.

What I found on my system was very surprising. For file sizes in the 256k-2M band, the Samsung seriously outperforms the Intel on all I/O tests. The most striking result was random write.. which for a 1MB file averaged across all record lengths at 386MB/s for Intel and 738MB/s for the Samsung. the Intel does better up until s 32MB file, at which point they basically converge on all IO tests.

So what am I doing wrong? Did I miss something with the kernel, like I was supposed to load a certain driver or something? Should I attempt to switch the cards, so that my VGA card is in slot 2 and Intel in 1? My BIOS is all updated to the latest version, no prob there. I've got a few more days to get this working or I'm returning the drive. At twice the price of the Samsung, it needs to outperform it.

0 Kudos
2 Replies
Honored Contributor II


The best performance possible is achieved when the correct combination of hardware and software is present on the PC. The hardware you have looks more than capable to run the max speed for this drive, but sometimes the software gets in the way. It could be that the software is the cause because of the results you have mentioned. Here is a link with some recommendations from and ASUS* about NVMe drives. ASUS Z97 & X99 Motherboards & Intel 750 series NVMe SSDs – All You Need To Know | ASUS PC DiY | If you want it done ri…

It is important to mention that An NVMe* drive running on an OS that is not correctly optimized for the NVMe* protocol is most likely going to perform below what is expected. Even below a previous standard like AHCI over SATA. It is exactly the same if the driver is not the Intel provided driver, the performance may be reduced. Please consider that the type of test is key to obtain proper results. We don't have a guide to evaluate performance in Linux*, however, you can use the document as reference in order to see what the recommended settings for each test are: Intel® Solid-State Drive 750 Series - Evaluation Guide


We understand, you are running RH7. Nevertheless here is a link to our NVMe* drivers: Download Intel® Solid-State Drive Data Center Family for NVMe Drivers This contains a driver for the drive and for the NVMe* controller that comes embedded on the drive.




Thank you. I did consult that first document prior to buying it.

Intel does not provide NVMe drivers for Linux. The drivers are contained within the kernel.

What isn't optimized for NVMe? Is it the kernel? Is it Red Hat? Certainly you have enterprise customers running some form of the 3.x kernel using your NVMe drives? If I were to walk into my CIOs office today, I'd need to tell her that Intel NVMe doesn't work very well with Linux. That's difficult for me to believe given everything I've seen with this technology.

Let me ask this: How can I verify 100% functionality of the drive? Is there some software you have for Linux that can inspect things and just verify? Or, can I safely assume that it is working perfectly because I can use it and boot from it?