Benchmark: MS IDE vs. nForce IDE (Updated)

[Unbreakable]

Hi!

I've done a few benchmarks with HD Tach to see how the most recent nforce IDE drivers perform against the generic MS drivers. Scroll further down for regularly updated benchmark results.

Here's my system setup:
AMD Athlon 64 3500+ (Venice E6)
1GB DDR400 RAM
ASUS A8N-E (nForce4 Ultra, revision 2.01G, BIOS 1013)
Old system setup:
AMD Athlon XP 2600+ (333MHz FSB)
1GB RAM (DDR400 but running in sync with FSB)
MSI K7N2 Delta-L (nForce2 Ultra 400, BIOS v5.9)

HD drive:
Samsung HD080HJ (7200rpm, 8MB cache, SATA 3.0Gb/s, NCQ, AAM disabled, master on prim. SATA controller)
Western Digital WD2000BB-22GUA0 (7200rpm, 2MB cache, UDMA5, master on prim. IDE port, using ATA133-cable)
pre-5.18 benchmarks were done with:
Quantum Fireball lct20 40GB, 128kB buffer, UDMA5=ATA-100, single drive on primary channel (THE slowest 40GB drive on earth...)
Optical drives:
Samsung TS-H352A (f/w: TS05, UDMA2, master on sec. IDE port, using ATA133-cable)
Samsung SH-W163A (f/w: TS01, UDMA2, master on sec. SATA controller)
old DVD burner:
LG GSA-4163B (f/w: A106, UDMA2, slave on sec. IDE port)
BIOS settings (IDE related):
- Access Mode = Auto
- IDE Prefetch = Enabled
- IDE DMA transfer access = Enabled
- IDE HDD block mode = Enabled
- IDE controller running in native mode
- Ports that are not being used are disabled in BIOS.
- APIC is and should be enabled in BIOS. nForce IDE drivers starting with UDP 5.10 rely on it else installation fails (after restart Windows hangs during boot).
OS:
Windows XP Pro SP2 (defragged, no background proggies, rebooted after each test, waited 1 minute after bootup so that no OS disk I/Os interfere with the benchmark)
Drivers:
Audio driver 4.65 (WHQL)
Ethernet driver 50.25 (WHQL)
IDE driver 5.21, 5.52, 6.66 (WHQL)
SMBus driver 4.57 (WHQL)
for benchmarking the MS IDE driver: pciide.sys, pciidex.sys, and atapi.sys (build 2180, 2714, and 2939)
Benchmark apps:
HD Tach 3.0
PCMark05 1.1.0
HD Tune 2.52
Nero CD-DVD Speed
Benchmark updates:
Updated: HD Tach results with nForce IDE 5.07 driver (read caching on vs. off)
Updated: PCMark05 HDD scores with nForce IDE 5.35
Updated: PCMark05 HDD scores with nForce IDE 5.52
Updated: ATTO v2.0.2 with nForce IDE 5.52 (read caching on vs. off)
Updated: PCMark05 with nForce IDE 6.22 vs. build 2180 and 2714 of atapi.sys
Updated: PCMark05 and HD Tach on nForce4 SLI X16 with Western Digital WD2500KS
Updated: Nero CD-DVD Speed 4.06 test
Updated: Samsung HD080HJ benchmarks
Updated: NCQ - good or bad?

HD Tach results of the short test with nforce IDE 4.79:


HD Tach results of the long test with nforce IDE 4.79:


"Facts":
- Surprisingly (at least for me), the MS IDE has a lower CPU utilization than the nforce driver.
- Again surprisingly, the MS IDE driver has a higher burst rate than the nforce IDE.
- The nforce IDE driver has a lower random access time.

It really surprises me to see that the MS driver beat the nforce driver in its supposed "domain" that is lower CPU load and higher burst rate. And I even recommended the nforce driver to people on this forum because of that... *me hides*

[Update: Due to further testing (see updates above) it has become evident that read caching is hindering the performance of the nForce IDE driver. Once read caching is disabled, the nForce IDE driver shows better results in CPU utilization and burst rate than the generic MS driver (which has read caching disabled by default).]

Recommendation:
One can't certainly compare drivers with one benchmark (and one single drive!) and come to the conclusion that this or that one is better. It would be interesting to see how the nforce drivers "behave" with a faster drive, a different IDE controller or even a different benchmark! So anybody who wants to put his/her results in here (or another thread) should feel free to do so! :)

Here's a link to Imageshack for free uploading of images: http://www.imageshack.us/

[PnP]

Is it possible to reinstall a MS IDE driver? If yes how? thanks...

[Unbreakable]

Yeah sure. Just uninstall or roll-back the nforce driver in the Device Manager. Alternatively you can go to "Add or Remove Programs" (in control panel) and then start the uninstall process of the "Nvidia Drivers". There you will see an option "Remove only the following". Select this option and place a checkmark next to "NVIDIA nforce IDE Driver" or similar...

[PnP]

Yeah sure. Just uninstall or roll-back the nforce driver in the Device Manager. Alternatively you can go to "Add or Remove Programs" (in control panel) and then start the uninstall process of the "Nvidia Drivers". There you will see an option "Remove only the following". Select this option and place a checkmark next to "NVIDIA nforce IDE Driver" or similar...

Ok but when find new device shall install "standard dual channel PCI ide controller" or other?

[Unbreakable]

Yes, that's it!

[PnP]

Ok... and shall I set some of options or leave default? thanks for help and info...

[Unbreakable]

You don't have to. The controller auto-detects all the drives and sets the appropriate speed. But to be sure you can check if the drivers are operating in Ultra DMA mode. To do so go to Device Manager and open the properties of the Primary IDE Channel (the same for the secondary). Now go to the Advanced Properties tab and make sure that the "Current Transfer Mode" equals to the setting your hard-, CD- or DVD-drive is capable of.

[PnP]

Ok thanks you so much now all are set up correctly and work very good...

[Monarch]

I still use nForce IDE...there has to be something else being better about them.

[EoRaptor]

My system performs better with most nForce IDE drivers (some are just screwy, though). HD Tach has benchmarking issues, check out www.storagereview.com's take on it for why it's not a terribly good benchmark.

Most tests won't show whjat the NV-IDE driver does best, and that's when multiple independent requestsa are made of differentr hard drives. MS-IDE bogs down badly copying two files from or to the same HD, and while the NV-IDE does slow down, it's way better at amnaging such conflicting requests.