New Intel RST drivers hit performance scaling ceiling

I was quite excited when Intel released its latest storage controller drivers last week. We knew these drivers were coming, and I’d been waiting on them to begin the next wave of storage testing with our new suite and test systems. And so a new benchmarking binge began.

The hangover didn’t hit until this weekend when, as I was importing some results into Excel, I noticed performance scaling issues with Intel’s new Rapid Storage Technology drivers. The problem manifests itself in IOMeter, where after hitting a load of 32 concurrent I/O requests, the new RST drivers effectively cap transaction rate scaling. Intel’s old 8.9.0.1023 storage drivers continue to ramp transaction rates all the way up to 256 concurrent IOs, as do the Microsoft AHCI drivers embedded in Windows 7. Below are a few graphics I whipped up showing the difference in performance between those drivers on a Western Digital Caviar Black 2TB hard drive.

32 IOs just happens to be the queue depth of Native Command Queuing, which makes me suspect that a change in the drivers’ NCQ implementation is responsible for the performance discrepancies above. I’ve been in touch with Intel, which is looking into the matter.

Interestingly, this issue doesn’t only afflict the new RST drivers. We’ve seen AMD’s SB850 south bridge exhibit hit a similar scaling ceiling. Marvell’s 6Gbps SATA controller appears to be afflicted, as well.

While the average desktop PC probably won’t encounter too many situations where the number of simultaneous IO requests exceeds 32, it’s under those demanding conditions that one would want a hard drive to perform as well as it can. Given the fact that previous Intel drivers haven’t capped performance in this way, I’m hopeful that Intel will be able to address the issue with a driver update. Until then, though, we’ll be doing storage testing with the Microsoft AHCI drivers available in Windows 7.

Comments closed
    • Raskaran
    • 10 years ago

    From from the two other links you provided I’m judging the new drivers just disable software requests reordering and use only the drives native queue, just look at the cpu utilization, its far smaller and is what i would except from hardware only approach. Atleast this is my understanding of how NCQ works.

    • Spotpuff
    • 10 years ago

    This is just for RAID? Or for any P55 chipset board?

    • ChrisDTC
    • 10 years ago

    I cant prove it yet, but Im fairly certain the new drivers can also cause system instability in certain cases. I have a Windows 7 x64 machine that has the OS on a RAID 0 array and not long after I upgraded to the RST drivers the machine would just lock up after varying periods of use. The only way to recover it was to cut the power. I would always check the Windows Logs but they never gave any indication of any problems.

    I rolled back to v8.9 and so far so good.

      • LoneWolf15
      • 10 years ago

      Not sure what to say –they work fine with my RAID-5 array (three 500GB Seagate 7200 drives) on an ICH10R setup (Gigabyte EP45-UD3P board). I haven’t had any stability issues related to them so far, and they’re the first set I put in to upgrade from the former Matrix Storage kit.

        • Ryu Connor
        • 10 years ago

        I’ve seen BSOD problems from 9.6 on both ICH7 and 8 systems. ICH10 systems seem to be happy with them by comparison.

      • maxxcool
      • 10 years ago

      Interesting. i suspect my achi raid to be my issue as well. my system sill lock up, by having the screen go “white”.. and then dead as a door nail. right before the lockup all apps will freeze, but the mouse will work for about 3 seconds before the whiteout occurs.

      maybe its time to install again and leave the driver pack off the box. (fyi this is server 2008)

    • Usacomp2k3
    • 10 years ago

    Does this affect regular users? How does something like PC-Mark change?

    • eitje
    • 10 years ago

    Hey! I think these are the drivers that support TRIM over RAID!

      • derFunkenstein
      • 10 years ago

      as long as the drives aren’t in the RAID array themselves, that is.

    • eternalmatt
    • 10 years ago

    lolers double post

    • eternalmatt
    • 10 years ago

    Does only allowing 32 requests benefit some other area? What was the point of this driver update if it only harms performance?

      • Veerappan
      • 10 years ago

      I believe this might be the new version of the intel drivers that allows pass-through of TRIM commands for SSDs when an SSD is hung off the same controller as a disk-based RAID array.

    • Ryu Connor
    • 10 years ago

    I’ve seen some BSOD behavior on a fresh install of Vista using the new IRST 9.6 drivers on ICH7 hardware. The 975x Bad Axe 2 Intel motherboard more precisely. Rolling back to 8.9.1023 solved the issue.

    • shank15217
    • 10 years ago

    Looks like they are trying to limit cpu utilization. You guys might want to see performance with a raid configuration as well.

      • Dissonance
      • 10 years ago

      CPU utilization isn’t an issue here–it’s well under 1%.

        • UberGerbil
        • 10 years ago

        Even with Atom?

          • eternalmatt
          • 10 years ago

          Who is using RAID with an Atom

            • Prototyped
            • 10 years ago

            There are several SOHO NAS devices with driver-based RAID and Atom 330 and D510 processors.

            Since all disk I/O is DMA anyway, I doubt you’d see substantial CPU utilization even with Atom.

            • imtheunknown176
            • 10 years ago

            Even if you did see cpu usage in an Atom in the 20-30% range why would it matter? That’s what the storage server is for.

            • derFunkenstein
            • 10 years ago

            a storage server using RAID is a great use for an Atom, assuming you can find an Atom board with enough SATA ports. The only one I’ve seen lately is a Gigabyte board with a D510 CPU and 4 ports.

    • _Sigma
    • 10 years ago

    I have a p35-ds3l board from gigabyte, so I can’t install these drivers. I remember hearing about some tricks that allow you to get around the intel installer checks. Do these still work? Is it worthwhile doing?

    • Vaughn
    • 10 years ago

    interesting find…

    I’m sticking with the new Intel drivers because when my machine wakes up from sleep I can see all my hard drives. With the MS drivers one hard drive is always missing. I have to go to device manager and scan for new hardware to pickup the drive.

    Since this is my home pc I don’t have to worry about going over 32 IO request, and the missing drives from sleep is a far more annoying issue for me.

    • Prototyped
    • 10 years ago

    So now, the only vendor AHCI drivers that don’t suck are nVidia’s.

    LOL.

    Just goes to show what I’ve been saying for ages, IHVs couldn’t write decent drivers to save their lives. (Operating system makers like Microsoft can, however.)

    q[

    • cygnus1
    • 10 years ago

    How stable are those MS drivers? I had a home server setup (2008 R2) on an AMD/ATI motherboard and it bluescreened fairly regularly until i loaded the real storage drivers. I had wanted to stay with the MS ones because of all the good things i’ve read about their AHCI drivers.

    • continuum
    • 10 years ago

    Uhoh, that’s not good… wonder if it’s just a mistake, or if there’s some deliberate reason why?

Pin It on Pinterest

Share This