The nForce PCI writes mystery

Since I purchased the MSI K7N420 Pro motherboard late last year, I’ve participated in one form or another in the MSI nForce forum. During that time, there has been a complaint or two about low write performance from hard drives, in particular with IDE RAID. The nForce boards’ peak write speeds on drive benchmarks using PCI disk controllers would never crest over 20MB/s, while read scores have hit as high as 113MB/s. Due to a relative lack of complaints and the fact that most users were using Windows XP, which has its own hard drive writes performance problems, the issue was largely ignored.

In light of some recent tests from SCSI and IDE RAID users from Windows 2000, it’s begining to seem that there is a problem with nForce, not just with the OS. In fact, Asus apparently has taken steps to address the problem. Among the fixes in Asus’ latest beta BIOS for its nForce board is a correction for poor PCI write performance.

Despite a benchmark or two from MSI users and some reports from Asus users, the overall picture remains spotty. The MSI nForce forum doesn’t have a large enough user base to prove that PCI writes are limited to 20MB/s. Nor does the forum have the resources to show whether or not this writes issue is limited to just MSI’s model or to the entire nForce chipset. So I need the community at large’s help to determine whether or not this issue actually exists and with which boards.

If you own one of the nForce boards from Abit, Asus, MSI, or Leadtek and you use a PCI IDE or SCSI controller, please take the time to download ATTO’s disk benchmark and run it. Go here and enter a name and e-mail address. Then choose the “Select” button to the right of “ATTO ExpressPCI HostAdapter Utilities” and select to download the “Windows SCSI Utilities Version 1.63” from the drop down menu at the bottom of the page. The file that begins to download should be EPT163.exe. Here is an example of a WDJB100B drive’s results using the appropriate test settings (I was using the south bridge chip’s controller here, so write performance is fine in the example).

Once you have installed the software and configured it to the above example like my sample picture, run the test. After the benchmark finishes, post a comment with your board brand, BIOS revision, controller model and driver, OS, and the scores themselves. For example:


  1. MSI K7N420 Pro, BIOS 2.4
  2. Promise 100TX2, 2.20.0.11, and 2.00.0.29
  3. Windows 2000 SP2
  4. 64k:Write:X Read:X
  5. 128:Write:X Read:X
  6. 256:Write:X Read:X
  7. 512:Write:X Read:X
  8. 1024:Write:X Read:X

…where the X’s are scores. The ATTO program does not create any Add/Remove Programs entries or registry keys, so you can just delete the folder and program group to remove it.

Remember, this problem will not occur with the nForce’s on-board IDE controller. Only drives using PCI IDE or SCSI controllers or the Highpoint controller integrated into the Abit NV7-133R will be affected by this issue. Please be courteous in the comments. The goal here is to figure out what’s up with nForce PCI writes. We’ll be following up with NVIDIA, as well. Thanks.

Comments closed
    • Anonymous
    • 17 years ago

    Using the sample picture on my pci controller card with a 20gb Western Digital i found the write speed never went over 16/17.

    here’s the results

    Write Read
    64 – 15572 22636
    128 – 15895 24732
    256 – 16676 24807
    512- 16659 24966
    1024- 16556 25069

    • Anonymous
    • 18 years ago

    will this affect me if I have a scsi card installed or only if i use the scsi as my main storage mine is currently attached to a zip drive and a cf reader but I do have a lot of issues with the ide bus and never ending ide controller errors in the event log.

    • Ryu Connor
    • 18 years ago

    Gerbil #126:

    The problem you are facing is not a issue with the board, but an inherent limitation within the IDE specification. Two drives on the same channel cannot use the channel at the same time. They must make alternate accesses. Copy your CD to the hard drive and then have the burner pull the image from the hard drive. That should resolve your problem.

    Telfordd #127:

    We’ve had several before and after updates on the Asus BIOS and have seen success from the beta in resolving this issue. I have to wonder if your problem may lay elsewhere. Bad cable, miss jumpered drive, defective card, BIOS bug, driver bug, etc. I’m not dismissing your claim, but the evidence is currently stacked against you.

    You can see the successful results (before and after) of other Asus A7N users in post #120.

    Runeve #128:

    Congratulations. It seems the Asus beta BIOS has resolved your issues.

    • Anonymous
    • 18 years ago

    Hello Runeve or runev again, cant remember my password, but i think this is great news!

    The lated beta bios from asus did somthing to my pc!
    Very very good! or is it good?

    Asus A7N266-C Bios 1003nc.003
    Promise 100TX2, 2.20.0.24 Driver: 2.00.0.25
    Windows 2000 SP2
    4x40GB Maxtor D740X Harddisk drives

    64k:Write: 40530 Read:97517
    128:Write:43668 Read: 97779
    256:Write:43905 Read:94659
    512:Write:43733 Read: 96032
    1024:Write:43835 Read: 96653

    Picture with 003 bios
    §[<http://home.no.net/rvenes/raid/atto2.JPG<]§

    • Anonymous
    • 18 years ago

    *[

    • Anonymous
    • 18 years ago

    Hi Guys,
    I have an ABIT NV7-133R motherboard and going thru some problems.
    Hope you can solve it.
    My configuration is as follows:
    AMD Athlon TBird 1400 C , 266FSB
    ABTI NV7-133R
    512MB PC-2100 DDR – Samsung
    Maxtor 60GB 7200RPM HDD on Primary Master
    Pioneer 16x DVD on Sec Master
    LG 32-10-40 CDRW on Sec Slave
    ATI All in Wonder 7500 64MB Video
    Win 2000 with all updates
    Easy CD Creator basic updated 5.3

    The system works perfect except for one problem.
    When you copy CD toCD , the buffer level goes down to 1% .
    CD Writes OK but very slow.
    I have changed three diff CDRW drives, changed diff configuration like putting drive on sec master instead of slave as well as tried cdrw on pri master. but to no avail.
    MB Bios is updated.

    Please help
    Thanks in advance
    Farooq

    • Ryu Connor
    • 18 years ago

    Looks like you have one of the Asus A7N266 boards. Apply the latest beta BIOS from Asus.

    • Anonymous
    • 18 years ago

    runeve…

    i forgot!

    win2k
    promise fasttrack tx2
    maxtor 7200 rpm 4x40gb raid 0 stripping

    • Anonymous
    • 18 years ago

    Hi

    is this thread dead?

    Runeve here!

    §[<http://home.no.net/rvenes/raid/atto1.JPG<]§ Looks fucked up to me? Right? <img src="http://home.no.net/rvenes/raid/atto1.JPG" alt="damn! loook!" />

    • Anonymous
    • 18 years ago

    MSI K7N420D Pro Bios 2.4
    512 MB (2 DIMM, 1,3)
    Maxtor DiamondMax Plus D740X (6L080L4) 80GB 7200 RPM
    Windows XP Professional
    Promise Ultra100 TX2 (1.43 Build 0603)

    64K Write: 19972 Read: 38508
    128 Write: 20020 Read: 35298
    256 Write: 19972 Read: 38036
    512 Write: 19972 Read: 37858
    1024 Write: 20016 Read: 35675

    I guess this is why PQDI takes so long to restore my system.

    • Doc Oc
    • 18 years ago

    [quote]MSI K7N420 Pro, BIOS 2.4
    Adaptec 19160: 10,000-RPM Atlas 10kIII x 1
    ?
    Max Writes: 17,953MB/s
    Max Reads: 108,464MB/s

    MSI K7N420 Pro, BIOS 2.4
    Adaptec 19160: 10,000-RPM Atlas10k x 1
    ?
    Max Writes: 17,435MB/s
    Max Reads: 43,733MB/s [/quote]
    Judging by the hardware mix, I’m sure these tests are AeroWB’s and mine. We both use Windows 2000.

    • Ryu Connor
    • 18 years ago

    Thanks to everyone who has participated thus far. Here is a summary of results:

    Abit NV7-133R, BIOS 84
    Built-in HPT372, 7,200-RPM 60GXP x 2
    Windows XP Professional
    Max Writes: 61,512MB/s
    Max Reads: 112,161MB/s
    Notes: System BIOS modified to include HPT372 BIOS 2.31

    Abit NV7-133R, BIOS ?
    Adaptec 29160: 10,000-RPM Atlas x 1
    Windows 2000 SP2
    Max Write: 34,779MB/s
    Max Read: 96,909MB/s

    Asus A7N266-C, BIOS 1003-02
    Promise 100TX2, 7,200-RPM 120GXP x 1(?)
    Windows XP Professional
    Max Writes: 29,155MB/s
    Max Reads: 95,650MB/s

    Asus A7N266-C, BIOS 1003-01
    Dawicontrol: 7,200-RPM 60GXP x 2
    ?
    Max Writes: 51,307MB/s
    Max Reads: 85,721MB/s

    Asus A7N266-C, BIOS 1001.E
    Promise 100TX2: 7,200-RPM 120GXP x 1(?)
    Windows XP Professional
    Max Writes: 3,300MB/s
    Max Reads: 95,490MB/s

    Asus A7N266-E, BIOS 1003-02
    Adaptec 19160: 15,000-RPM Cheetah x 1
    ?
    Max Write: 50,477MB/s
    Max Read: 50,855MB/s

    Asus A7N266-E, BIOS 1003-02
    Iwill Side RAID, 7,200-RPM D740X x 2
    Windows XP Professional
    Max Writes: 52,428MB/s
    Max Reads: 102,811MB/s
    Notes: 2.31 RAID driver

    Asus A7N266-E, BIOS 1003-02
    Promise 133TX2: 7,200-RPM D740X x 1
    Windows XP Home Edition
    Max Writes: 31,304MB/s
    Max Reads: 84,984MB/s

    Asus A7N266-E, BIOS 1001.D
    Promise 133TX2: 7,200-RPM D740X x 1
    Windows XP Home Edition
    Max Writes: 3,261MB/s
    Max Reads: 84,822MB/s

    Asus A7N266-E, BIOS 1001.B
    Iwill Side RAID, 7,200-RPM D740X x 2
    Windows XP Professional
    Max Writes: 3,223MB/s
    Max Reads: 102,515MB/s
    Notes: 2.31 RAID driver

    Leadtek K7N420DA, BIOS 4-19-2002
    Iwill Side RAID 100: 7,200-RPM D740X x 2
    Windows XP Professional
    Max Writes: 26,410MB/s
    Max Reads: 31,968MB/s
    Notes: RAID using stock XP driver

    Leadtek K7N420DA, BIOS 4-19-2002
    Iwill Side RAID 100: 7,200-RPM D740X x 2
    Windows XP Professional
    Max Writes: 58,428MB/s
    Max Reads: 101,743MB/s
    Notes: 2.31 RAID driver

    Leadtek K7N420DA, BIOS 5-30-2002
    Iwill Side RAID 100: 7,200-RPM D740X x 2
    Windows XP Professional
    Max Writes: 58,369MB/s
    Max Reads: 101,483MB/s
    Notes: 2.31 RAID driver

    Leadtek K7N420DA, BIOS 5-30-2002
    Iwill Side RAID 100: 7,200-RPM D740X x 2
    Windows 2000
    Max Writes: 57,608MB/s
    Max Reads: 101,458MB/s
    Notes: 2.31 RAID driver

    MSI K7N420 Pro, BIOS 2.4
    Adaptec 2100 RAID: 15,000-RPM Cheetah x 2
    Windows XP
    Max Write: 17,801MB/s
    Max Read: 116,624MB/s

    MSI K7N420 Pro, BIOS 2.4
    Adaptec 19160: 10,000-RPM Atlas III x 1
    ?
    Max Writes: 17,953MB/s
    Max Reads: 108,464MB/s

    MSI K7N420 Pro, BIOS 2.4
    Adaptec 19160: 10,000-RPM Atlas x 1
    ?
    Max Writes: 17,435MB/s
    Max Reads: 43,733MB/s

    MSI K7N420 Pro, BIOS 2.4
    Mylex AcceleRAID 170: 10,000-RPM Cheetah x 1
    Windows XP
    Max Write: 12,732MB/s
    Max Read: 22,783

    MSI K7N420 Pro, BIOS 2.4
    Promise 133TX2: Quantum 30GB x 1
    Windows 2000 SP2
    Max Write: 19,997MB/s
    Max Read: 64,267MB/s

    MSI K7N420 Pro, BIOS 2.4
    Promise 100TX2: Western Digital 60GB x 2
    Windows XP
    Max Write: 19,737
    Max Read: 41,840

    MSI K7N420 Pro, BIOS 2.4
    Promise 100TX2: 7,200-RPM 120GXP x 2
    Windows XP
    Max Writes: 20,219MB/s
    Max Reads: 111,254MB/s

    MSI K7N420 Pro, BIOS 2.4
    Promise 100TX2: 7,200-RPM 75GXP x 2
    ?
    Max Writes: 20,095MB/s
    Max Reads: 111,282MB/s

    MSI K7N420 Pro, BIOS 2.4
    Promise 100TX2: 7,200-RPM Barracuda IV x 2
    Windows XP Professional
    Max Write: 19,886MB/s
    Max Read: 63,123MB/s

    MSI K7N420 Pro, BIOS 2.4
    Iwill Side RAID: 7,200-RPM D740X x 2
    Windows XP Professional
    Max Writes: 1,729MB/s
    Max Reads: 100,964MB/s
    Notes: 2.31 RAID driver: 266/266MHz FSB

    MSI K7N420 Pro, BIOS 2.4
    Iwill Side RAID: 7,200-RPM D740X x 2
    Windows 2000
    Max Writes: 1,638MB/s
    Max Reads: 101,524MB/s
    Notes: 2.31 RAID driver: 266/266MHz FSB

    MSI K7N420 Pro, BIOS 2.4
    Iwill Side RAID: 7,200-RPM D740X x 2
    Windows XP Professional
    Max Writes: 17,476MB/s
    Max Reads: 100,162MB/s
    Notes: 2.31 RAID driver: 200/200MHz FSB

    MSI K7N420 Pro, BIOS 2.4
    Iwill Side RAID: 7,200-RPM D740X x 2
    Windows 2000
    Max Writes: 16,748MB/s
    Max Reads: 99,986MB/s
    Notes: 2.31 RAID driver: 200/200MHz FSB

    MSI K7N420 Pro, BIOS 2.3
    AMI MegaRAID 428: 10,000-RPM Atlas x 3
    Windows 2000 SP2
    Max Write: 4,042MB/s
    Max Read: 8,398MB/s
    Notes: Is this a typo?

    MSI K7N420 Pro, BIOS 2.3
    Iwill Side RAID: 7,200-RPM D740X x 2
    Windows XP Professional
    Max Writes: 1,565MB/s
    Max Reads: 32,480MB/s
    Notes: RAID using stock 2000 driver

    MSI K7N420 Pro, BIOS 2.3
    Iwill Side RAID: 7,200-RPM D740X x 2
    Windows XP Professional
    Max Writes: 1,694MB/s
    Max Reads: 101,021MB/s
    Notes: 2.31 RAID driver

    MSI K7N420 Pro, BIOS 2.3
    SIIG U-66: 5,400-RPM U80020 80GB x 1
    Windows ME
    Max Writes: 20,164MB/s
    Max Reads: 41,527MB/s

    • Anonymous
    • 18 years ago

    I have the oposite problem on the ASUS AZN266-VM (220D) system I built for my wife. On board IDE around 5 MB/s on channel 1, and 6 MB/s on channel 2. I put an old ATA66 card in and the Sandra marks went from 5000+ to 24000+ with a continuios read/write around 40 MB/s. Anyone have any ideas? Am I missing something?

    • Anonymous
    • 18 years ago

    *[

    • cass
    • 18 years ago

    LeadTek K7N420 DA 5/30/2002 bios
    PCB REV B
    assy Rev A
    256mb x 2 (512 total) Kingston value ram
    system drive: c: D740X 40 GB J2 series
    Iwill Side Raid 100 bios 2.31 pci slot 1 (beside agp)
    Drivers 2.31.0.0
    2x D740X 40GB J2 series Raid 0
    Clean install of windows 2000
    200MB page and file index off
    default cluster size
    apic enabled irq’s to 22
    133/267/66/33 fsb/mem/agp/pci verified with speedgear

    C: drive onboard IDE
    64 22896 31378
    128 25435 31405
    256 25381 31508
    512 25336 31406
    1024 25318 31384

    D: Raid on pci
    64 57147 91226
    128 57527 90214
    256 57360 101458
    512 57171 100964
    1024 57608 101268

    I turned file indexing on/off and set the timing to 100/200/66/33 and there was about 1% difference so I won’t waste space here with detailed results.

    • Anonymous
    • 18 years ago

    *[

    • Anonymous
    • 18 years ago

    Make sure you aren’t using Seagate Baracuda’s, otherwise the problem you think you have, is not the problem you actually have.

    And don’t forget to RMA your IDE Seagate Baracuda’s if you are using them in RAID-0 or make sure if you purchase them, that the retailer guarantees that they will work up to spec in IDE RAID config, otherwise you will wonder why your 7200rpm drives in RAID-0 are slower than a single 5400 rpm drive of a smaller capacity.

    • cass
    • 18 years ago

    MSI K7N420D PRO ver 1.0 bios 2.4
    256mb x 2 (512 total) Kingston value ram
    system drive: c: D740X 40 GB J2 series
    Iwill Side Raid 100 bios 2.31 pci slot 2
    drivers 2.31.0.0
    2x D740X 40GB J2 series Raid 0
    Windows 2000 (5.00.2195) page file 200MB
    clean install with drivers from MB disk with no updates
    default cluster size
    apic enabled irq’s to 22
    133/266 cpu/mem
    xp 1800+

    c: onboard ide
    64 25427 30962
    128 24793 31483
    256 29112 31508
    512 25336 31406
    1024 29127 31530

    D: raid on pci
    64 1631 90177
    128 1628 89429
    256 1615 101196
    512 1638 101483
    1024 1617 101524

    them are not typos

    timing changed to 100/200 cpu/mem

    c: onboard ide
    64 23502 20403
    128 26860 31483
    256 28970 31431
    512 26905 30738
    1024 28793 31359

    D: raid on pci
    64 16746 91226
    128 16664 90214
    256 15924 99913
    512 16748 99926
    1024 16676 99986

    • cass
    • 18 years ago

    MSI K7N420D PRO ver 1.0 bios 2.3
    256mb x 2 (512 total) Kingston value ram
    system drive: c: D740X 40 GB J2 series
    (this drive has been a j2 series the whole time
    I missread the label)
    Iwill Side Raid 100 bios 2.31 pci slot 2
    drivers default from M$
    2x D740X 40GB J2 series Raid 0
    windows XP pro with performance settings,system restore off, file indexing off, and page file off
    clean install with drivers from MB disk with no updates
    default cluster size
    apic enabled irq’s to 22
    133/266 cpu/mem
    xp 1800+

    c: onboard ide
    64 28621 31145
    128 25826 60029
    256 31021 31431
    512 31330 31482
    1024 31311 31384

    D: raid on pci
    64 1551 30626
    128 1553 30626
    256 1565 32480
    512 1562 32437
    1024 1554 31797

    them are not typos they mean 1551 etc.

    Raid drivers updated to 2.31.0.0

    c: onboard ide
    64 30702 36336
    128 28855 39970
    256 30916 31173
    512 30320 32068
    1024 28884 30029

    D: raid on pci
    64 1687 91521
    128 1694 91847
    256 1688 100694
    512 1657 100983
    1024 1676 101021

    Bios Updated to 2.4
    timing changed to 100/200 like always after flash

    c: onboard ide
    64 29328 29069
    128 25435 58740
    256 30690 31277
    512 29994 31482
    1024 29685 30654

    D: raid on pci
    64 17304 88906
    128 17391 89429
    256 17434 100162
    512 17476 99950
    1024 17438 99742

    Bios 2.4
    timing changed to 133/266/66 like always after flash

    c: onboard ide
    64 25695 29661
    128 25826 84712
    256 30765 31277
    512 31180 31078
    1024 30372 31384

    D: raid on pci
    64 1729 90737
    128 1684 91067
    256 1679 100964
    512 1678 100725
    1024 1681 100509

    This is my Nforce Xperience the pci write transfer is like this always when the timing is 133/266 when it is 100/200 I get about 20MB/s Surely I am not alone.

    I have w2k ready… more testing.

    • Anonymous
    • 18 years ago

    Gerbil #99 here again:

    Re-ran the tests with 256k, 64k, and .5k Total Lengths. Write speed never exceeded 62XXX

    MODDING the NV7-133R BIOS’s HPT372 BIOS ENTRY:

    Download CBROM V2.07 from Binary

    • Ryu Connor
    • 18 years ago

    [q]These things may be outside the scope of this article however, perhaps the Author can suggest another place to discuss them?[/q]

    If you would like to show some burst results using a even smaller file size than 1MB here, you may. For most people 1MB will be fine (as most modern drives have a 2MB buffer.) Please remember to denote any variance from the control set by the news piece.

    I also don’t have a problem with you sharing the modified HPT372 BIOS with him.

    If you aren’t comfortable doing that here, the “Processors, Chipsets, & Motherboards” or “Storage” section of our forums would be a good place.

    §[<http://www.tech-report.com/forums/<]§

    • Anonymous
    • 18 years ago

    *[

    • Anonymous
    • 18 years ago

    Anonymous Gerbil #98 and 99 here:

    drruhe, I will re-run tests with what you’ve requested this evening. I’ve got to run errands this afternoon. I modified the BIOS myself using CBROM. I’ll find and post a link later and also some NV7 & HPT particular instructions.

    These things may be outside the scope of this article however, perhaps the Author can suggest another place to discuss them?

    Be back around 7pm EST.

    Anonymous Gerbil 99

    I’ll register for a real nick tonight as well.

    • Ryu Connor
    • 18 years ago

    drruhe:

    I’m begining to think that there is some communication difficulties.

    I received your results, but you obtained them through two instances of ATTO on two different drives. I have to assume they both tested simultaneously, nonetheless I’m leary of those results because of various factors that could influence two seperate threads trying to operate at the same time. To speak nothing of the fact that the test were configured incorrectly. Nonetheless, Gerbil #98 proved what you had been trying to show. He provided the results I needed. Albeit I still need more results.

    I’m going to be blunt with you, because I’m at that point. You have tried to hijack this thread from the first moment you posted and it’s time for it to stop. You’ve complained about the method, the reasons for the method, and the results from square one. This has really begun to upset me, because you for one don’t fully grasp the technical details and for two you apparently have not read the news post that preceeds these comments.

    For example:

    [q]In any case it reports lower wrinting than reading performance on bot raid and buildin IDE ports.[/q]

    It does, does it?

    From the news post above:

    [q][i]. Here is an example of a WDJB100B drive’s results using the appropriate test settings (I was using the south bridge chip’s controller here, so write performance is fine in the example).[/q][/i]

    §[<http://www.tech-report.com/etc/2002q2/example.gif<]§ You'll notice that I referenced a picture with the test results I wanted and even showed you a drive on the on-board controller that scores higher writes than reads. How am I supposed to take you seriously when I see you behave in this fashion? The result you e-mailed to me were not only configured wrong; insinuating you never looked at the picture. On top of that you appear to insinuate that the benchmark is broke because it always has slower writes than reads. When in fact that behavior is determined by the drive, not the benchmark and I had already provided an example of a drive that could write faster than read. But what really put the nail in the coffin was this: [q]The previous poster, Anonymous Gerbil , reports 112MB/s reading performance. This is a clearly wrong results, because each IBM 60GXP=IC35L020AVER is only capably to read/write with 36MB/s at the beginning of the disk. So 72MB/s is the best result a raid array could achieve here. In this case, the writing performance seems to be OK for these disks - minus the ATTO penalty.[/q] Not only does it further confirm that you didn't fully read my news post and did not look at my configured example. It also shows that you do not understand that the kind of testing we needed done was burst testing. That we needed to push the PCI bus to the max. Something I tried to convey to you in post #88 when I asked for single burst results. I understand that you are anxious. I understand that you've invested in a board that you feel is fantastic and that I am presenting a story and results that might put a dent in that perfection. This thread is not the place for unrelenting dedication to the nForce platform or a particular nForce implementation. This thread is to find a bug and then for us to be able to fix that bug for you and for everyone else. It is not meant to be pretty or happy, it is simply the plain ugly truth. Contrary to popular belief, I know what I am doing. You have provided your results and your opinion. I appreciate the results and respect your opinion. It is now time for you to be a silent observer and let the TR staff and me deal with the issue. Thank you.

    • Anonymous
    • 18 years ago

    *[

    • cass
    • 18 years ago

    It would beg the question now that we have seemed to have Identifed some good Nforce PCI boards, what would the atto Write numbers look like on those same boards with W2K?

    Since the low write numbers might be a symptom of the XP SCSI problem.

    Is the Write cache an Atto problem or an XP Problem?

    I think sustained writes at 60MB/s are good for two IDE drives, and don’t know if the write not bursting over 60 is a problem or a measurement anomally.

    • Anonymous
    • 18 years ago

    *[

    • Anonymous
    • 18 years ago

    *[

    • Doc Oc
    • 18 years ago

    [quote]The previous poster, Anonymous Gerbil , reports 112MB/s reading performance. This is a clearly wrong results, because each IBM 60GXP=IC35L020AVER is only capably to read/write with 36MB/s at the beginning of the disk. [/quote]
    You’re overlooking the fact that the 1MB test pattern fits in the HDD’s cache memory. It’s the reason we chose 1MB and not a larger pattern. Remember that it’s not harddrives we’re testing, it’s the PCI bus.

    • Anonymous
    • 18 years ago

    *[

    • Anonymous
    • 18 years ago

    *[ruheejih@linux.zrz.tu-berlin.de. The performance monitor of Windows is indded of great use for measuring performance parameters

    Julian

    • Doc Oc
    • 18 years ago

    [quote]the Total Length setting of 1MB is probably within the drives physical cache size[/quote]
    That was intentional. We’re testing a suspected PCI bug, not a suspected hard drive bug. Choosing a pattern that fits the cache makes sure that PCI is the limiting factor, not the HD. 🙂

    • Anonymous
    • 18 years ago

    I’m the same guy as Gerbil #98

    I noticed in another thread of this type (the SCSI thread about dynamic vs. basic disks i[<(which BTW does not apply to IDE)<]i) that the b[

    • Anonymous
    • 18 years ago

    Abit NV7-133R, BIOS 84 i[

    • cass
    • 18 years ago

    Asus A7N-266e Rev 1.03 Bios 1001.b
    256mb x 2 (512 total) Kingston value ram
    system drive: c: D740X 40 GB J3 series
    Iwill Side Raid 100 bios 2.31 pci slot 4
    drivers 2.31.0.0 from High Point
    2x D740X 40GB J2 series Raid 0
    windows XP pro with performance settings,system restore off, file indexing off, and page file off
    (system drive from last test just plugged to the A7N266e and drivers for MBD and raid updated)
    default cluster size
    apic enabled irq’s to 21
    xp 1800+

    c: onboard ide
    64 25052 60993
    128 23330 43389
    256 24941 24732
    512 24834 27962
    1024 24658 31384

    D: raid on pci
    64 3222 93352
    128 3219 94189
    256 3223 102515
    512 3223 102016
    1024 3217 98975

    A7N266e bios updated to 1003n.002 beta
    c: onboard
    64 24718 44642
    128 23244 40749
    256 26150 31431
    512 21122 31482
    1024 24897 31384

    D: raid on pci
    64 52298 93875
    128 52298 94189
    256 52170 102776
    512 52428 102791
    1024 52177 102811

    If the board had worked like this to start with, there would be some more happy people in the world.

    Next up… MSI K7N420D

    • Doc Oc
    • 18 years ago

    [quote]What is PCI latency timer it is the bios… it is at 32 at my asus a7n266-c mb [/quote]
    PCI allows for [b]bus mastering[/b] devices to use it. This means that a device can take control of the PCI bus and read/write to memory, without the CPU getting involved. To make sure that a single PCI device does not suffocate other devices of their bandwidth, there’s a limit to how long a PCI device is allowed to write to the memory this way. This is the PCI latency timer.

    A lower setting makes for better multitasking, cause the PCI system will check more often whether anyone else uses the bus, while a high setting is good to establish higher transferrates for few devices.

    • cass
    • 18 years ago

    OK, got complete results on LeadTek k7n420 da now.
    K7N420 DA 4/19/2002 bios
    PCB REV B
    assy Rev A
    256mb x 2 (512 total) Kingston value ram
    system drive: c: D740X 40 GB J3 series
    Iwill Side Raid 100 bios 2.31 pci slot 1 (beside agp)
    2x D740X 40GB J2 series Raid 0
    Clean install of windows XP pro with performance settings
    default cluster size
    apic enabled irq’s to 22
    133/267/66/33 fsb/mem/agp/pci verified with speedgear

    I forgot and left the queue depth at 4 in atto.

    default windows raid driver, file indexing on, system restore on default page file.

    C: drive onboard IDE
    64 28881 40850
    128 31378 39289
    256 31378 92829
    512 31250 41268
    1024 31304 40964

    D: Raid on pci
    64 26410 30702
    128 26279 30702
    256 26343 30709
    512 26342 30734
    1024 26341 31968

    Raid drivers 2.31.0.0 installed, sysrestore off, page file off. file indexing still on c: drive

    C:
    64 25103 33387
    128 26794 37115
    256 25176 92337 = no typo
    512 25319 45122
    1024 25063 31457

    D: Raid on pci
    64 58312 91260
    128 58428 92829
    256 58428 94163
    512 58398 94424
    1024 58369 101743

    motherboard bios flashed to 5/30/2002, file indexing turned on C:Drive

    C:
    64 20555 30962
    128 27717 30182
    256 25124 31223
    512 24868 31431
    1024 26658 31277

    File indexing turned off
    c:
    64 28881 38700
    128 28240 38932
    256 31405 53599
    512 31173 56561
    1024 30213 39160

    D: raid on pci
    64 58312 93352
    128 58283 93613
    256 58283 93875
    512 58138 93929
    1024 58369 101483

    no bsod with clean install of XP. careful with mem timings here, because there is a Stupor stability on/off setting in bios. I jumpered the fsb to 133, then turned off super stability in bios.
    Left Xbios mem frequency on auto. post said 133/267/66/33, but when it booted i only had 133/200/66/33 using speed gear.
    I went back to Xbios and set the mem to 266 and then speed gear verified 267. System bios cached and video ram. did not try aggressive mem timing settings.

    • Anonymous
    • 18 years ago

    Anyone got one of these?

    §[<http://www.ioss.com.tw/web/English/RD2PCGeiger.html<]§ ? One of it's claimed functions is measuring PCI bandwidth? §[<http://www.scan.co.uk/products/overclock.html<]§ <-

    • Anonymous
    • 18 years ago

    [q]What is PCI latency timer it is the bios… it is at 32 at my asus a7n266-c mb

    anyone[/q]

    Based on a review of the nForce A7N266-E at LostCircuits.com I set mine to 64.

    • Anonymous
    • 18 years ago

    *[

    • Doc Oc
    • 18 years ago

    Oh wait. It had something to do with the way the linebreak came out on my browser display. Seeing it in my own quotation suddenly cleared it up.

    If you take a look at this page: §[<http://www.nvnews.net/previews/nforce/mcp_features.shtml<]§ (which is a sub-page of this one: §[<http://www.nvnews.net/previews/nforce/nforce.shtml<]§ ) you'll notice that the PCI controller is housed in the southbridge and not the northbridge. So it in fact [b]is[/b] the way you though it was not. Easy mistake of course, the majority of chipsets, the Intel 440 series in particular, has the PCI controller residing in the Northbridge.

    • Doc Oc
    • 18 years ago

    [quote] What has the Hyper Transport Link to do with all? I think data goes that way PCI<->Northbridge<->memory and not PCI<->southbridge<->HTL<->northbridge<->memory[/quote]
    I think it’s supposed to be [b]PCI <-> SthBr <-> HTL <-> NthBr <-> RAM[/b]. I don’t understand how you came up with the first four steps (left to right) of your suggestion.

    • Anonymous
    • 18 years ago

    *[

    • Ryu Connor
    • 18 years ago

    [q]Let me iterate: My Abit NV133-R is able to write with ~70MB/s to disk over the highpoint controller and that does not seem to be the limit.[/q]

    Do you have that result as a single burst? Can you provide ATTO results showing that 70MB/s burst?

    • Anonymous
    • 18 years ago

    *[

    • Ryu Connor
    • 18 years ago

    [q]I think some MSI nForce boards have an old revision of the southbride, or the BIOS is just buggy.[/q]

    If that was the case the Asus board wouldn’t be suffering as well. Nor does it completely explain number 70’s results. Any ways there’s a good chance this isn’t a southbridge bug.

    The PCI bus within the southbridge is able to read from memory (PCI Reads) right up against the real world wall of the PCI bus. Meanwhile data from memory written to the PCI bus is getting interrupted at 3MB/s, 20MB/s, and 30MB/s on the Asus, MSI, and Abit boards. That could mean the IGP or HyperTransport Link is at fault.

    In the end will still need more results from MSI, Asus, Abit, and Leadtek users. In particular more results from Abit and Leadtek. Each additional result make things that much clearer.

    • cass
    • 18 years ago

    LeadTek k7n420da (duh.. forgot to write down bios)
    Iwill Side Raid 100 bios v2.31 PCI slot 2
    2 x D740X 40GB Raid 0 64k cluster 64k stripe.
    D740X 40GB (J3) sys drive
    Win XP home Settings still on cutesy appearance and all default services running.

    I installed the Raid card straight in from another comp. leaving the data and stripes set up. Atto would not run on the Raid array. It could not create the test file. The array was created in Win XP pro and this comp has XP home.

    I could however copy files back and forth and access and play all my multimedia files on the raid using windows. I will format the array if I have to, but I will have to move my data first.

    I tried to update the XP home to XP pro, but got bad pooler error and then IRQ_less_than_equal on reboot.

    Any how, The onboard ide results to the system (c) drive from atto are:
    64 24492 37374
    128 27893 41630
    256 26860 41734
    512 29594 41630
    1024 29746 41638

    copied a 3,825,784 KB file from c: to the Raid Drive 98 seconds
    39,038.6 KB/s

    I like the looks of that, but I want to get Atto running so any suggestions welcome. Is there anyway to get XP home to do a 64k cluster with hdd format/partition?

    • Doc Oc
    • 18 years ago

    [quote](1) Has anybody taken a look into the taskmanager during a run of ATTO?! On my XP1800+ I get 100% CPU load!!! I dislike ATTO more and more… [/quote]
    That’s probably just a side effect of the way its programmed. WGENS (a sega emulator) does that too, but it definately doesn’t need 100% of an AthlonXP’s 1700+ CPU time to emulate a 10Mhz CPU. Any game does it too. So do SiSoft Sandra and very likely HDTach as well. You shouldn’t be worried about this.

    Besides, if my Celeron 433 scores writes three times the speed of my Athlon XP, using a slower harddrive, CPU power certainly isn’t the limiting factor.

    • kraquen
    • 18 years ago

    [q](1) Has anybody taken a look into the taskmanager during a run of ATTO?! On my XP1800+ I get 100% CPU load!!! I dislike ATTO more and more… [/q]
    atto maxes out my xp 2100+ as well when running the test.. That, logically, does not seem to be good, however, if the processor [i]was[/i] limiting the benchmark result, my athlon 2100+ would perform better than a lesser equiped computer.. but such is not the case, many different processor speeds are reporting numbers all very close to 20mb..

    • Anonymous
    • 18 years ago

    *[

    • Anonymous
    • 18 years ago

    *[

    • Anonymous
    • 18 years ago

    *[http://www.promise.com<]§ ) I\'ve got 2*120 GB WD 7200 rpm 2 MB harddisks. But still looks HdTach like hell... trust me. Pictures can be found here. §[<http://home.no.net/rvenes/raid/<]§ Contact me if ya need more info. (just write a post or send an email nfp1338@tiscali.no )

    I think the PCI BUS Utitization did much of the inprovment here…

    Average speed is 37661

    With the last bios 1003.2

    64k:Write:46806 Read: 41156
    128:Write:46834 Read: 41630
    256:Write: 46978 Read: 41839
    512:Write: 42757 Read: 41943
    1024:Write: 46909 Read: 39229

    With 1003nc001 bios

    64k:Write:27456 Read: 25887
    128:Write:27840 Read: 26799
    256:Write: 27824 Read: 30690
    512:Write: 27844 Read: 30900
    1024:Write: 27784 Read: 30868

    • Anonymous
    • 18 years ago

    *[

    • Anonymous
    • 18 years ago

    *[

    • Ryu Connor
    • 18 years ago

    [q]Is the last beta bios better then the the other one?[/q]

    Do a before and after benchmark.

    • Anonymous
    • 18 years ago

    *[

    • Anonymous
    • 18 years ago

    Not so good….

    MSI K7N420 Pro, Bios 2.4
    Promise FastTrack100 TX2, bios 2.00.0.24, driver 2.0.0.25
    2x IBM IC35L040AVVA07-0 (40Gb 120GXP, 7200 rpm)
    Windows XP
    64K : Write : 20145 Read : 110087
    128 : Write : 20204 Read : 109825
    256 : Write : 20164 Read : 110074
    512 : Write : 20136 Read : 110048
    1024: Write : 20219 Read : 110254

    • Anonymous
    • 18 years ago

    Further confirmation. Following up on my post #49 switching to the Asus 1003.2 Beta BIOS improves writes to expected levels for a single drive.

    Asus K7N266-E, BIOS 1003.2 beta
    Promise Ultra ATA133 TX2, 2.20.0.12 and 2.00.0.29
    Maxtor D740X 40G, NTFS
    Windows XP Home Edition
    64k:Write:30855 Read:31223
    128:Write:30626 Read:84984
    256:Write:31223 Read:84822
    512:Write:31250 Read:84822
    1024:Write:31304 Read:31457

    I assume the Read inconsistencies are because of testing from the XP boot partition.

    • Anonymous
    • 18 years ago

    Hi Guys!

    Here is the proof:

    System: ASUS A7N266-C ACPI b[

    • Ryu Connor
    • 18 years ago

    Gerbil #70 the scores create more questions for me.

    Is 30MB/s really the best burst writes value a Atlas 10K III on a Adaptec 29160 can give?

    Shouldn’t it have been closer to your read values? I’m just wondering if were seeing a different shade of the same bug.

    • Anonymous
    • 18 years ago

    Sorry Win2k SP2 NTFS

    • Anonymous
    • 18 years ago

    NV7-133R

    I dont seem to be affected by this.
    Adaptec 29160 U160 scsi controller
    Maxtor Atlas 10K III

    Write Read
    64 34779 76546
    128 30962 88906
    256 30765 90737
    512 30426 94268
    1024 30442 96909

    • bugsy
    • 18 years ago

    Damn it, wrong one. Its the other one.

    • bugsy
    • 18 years ago

    This article is on the frontpage on §[<http://www.techbargains.com<]§ so expect more visitors.

    • Doc Oc
    • 18 years ago

    Here are my results.

    MSI K7N 420 pro. BIOS 2.4, reference drivers 1.05
    Adaptec 19160
    Quantum Atlas 10kIII

    Between U160 SCSI and PCI, PCI is the limiting factor. Reads eek very close to the maximum possible speed. Writes, obviously, don’t.

    [64] – 17,953 – 106,430
    [128] – 17,902 – 107,479
    [256] – 17,939 – 108,256
    [512] – 17,884 – 108,240
    [1024] – 17,936 – 108,464

    • Ryu Connor
    • 18 years ago

    §[<http://cweb.msi.com.tw/eforum/?target=article&tid=538984640<]§ A report of writes difficulty using a hard drive connected to the VIA USB 2.0 controllers on the Abit NV7-133R. I'm going to need more results before I can confirm that the Abit boards are or are not impacted by the bug.

    • Anonymous
    • 18 years ago

    *[

    • Ryu Connor
    • 18 years ago

    [q]if the file size in the benchmarking program you use does not exceed 8MB in size (Cache size on the JB), the writes will be insanely high.[/q]

    Even with a 32MB file length the write scores edge out the read scores ever so slightly.

    • Anonymous
    • 18 years ago

    *[

    • Ryu Connor
    • 18 years ago

    drruhe:

    Your quote from c’t made me curious. I own a nForce board from the late November, early December 2001 timeframe. Seeing as this is when boards first started shipping, if c’t was correct I should have one of the impacted steppings.

    MCP-D
    F46078.1 0144C3
    S Taiwan

    I’m guessing C03 became C3. It would seem that none of the retail boards shipped with the C2 stepping.

    • Anonymous
    • 18 years ago

    *[

    • Anonymous
    • 18 years ago

    *[

    • Anonymous
    • 18 years ago

    Asus A7N 266-C Bios: 1003.001 Beta
    Dawicontrol PCI-Controller with Highpoint 370 chip
    2 IBM ICL35 40GB @Raid 0

    test results:

    write read
    64 48637 85721
    128 48898 85245
    256 49956 65895
    512 41479 85391
    1024 51307 73328

    * i was running some background apps while benchmarking,
    test results seem to be ok though

    • Anonymous
    • 18 years ago

    udma100 that was supposed to be 🙂

    • Anonymous
    • 18 years ago

    just postet those news at planet3dnow.de. there’s quite a bunch of A7n266 users there that will surely benchmark their systems to help.
    rumours in the forums @planet3dnow say that pci-performance issuies on the A7n266 have been solved with the lates bios revisions. (tested with digital video equipment).
    I personally don’t own an duma100/raid card and can’t test but I’ll keep you informed of news in the forums over there.

    Best regards
    Korben

    • Forge
    • 18 years ago

    Tanj! – I believ I’m the other JB/ICH2 user Ryu is referring to, and I can confirm that the JB series does write much faster than it reads. Baffles the piss out of me, too.

    Does anyone here have Win2K handy to test with? I see a lot of WinXP, but not a lot of 2K. I’m sure it’s the same story, but it’d be interesting to see the effects of the newest beta BIOSes and patches on 2K vs. XP.

    • Anonymous
    • 18 years ago

    [b]graft[/b] [i]verb[/i]

    1. a. To unite (a shoot or bud) with a growing plant by insertion or by placing in close contact.

    1. b. To join (a plant or plants) by such union.

    2. To transplant or implant (living tissue, for example) surgically into a bodily part to replace a damaged part or compensate for a defect.

    3. To join or unite closely.

    [b]graph[/b] [i]verb[/i]

    1. What Tech Report has plenty of.

    • Anonymous
    • 18 years ago

    *[

    • Anonymous
    • 18 years ago

    *[

    • Ryu Connor
    • 18 years ago

    Gerbil #51

    Yes, those scores appear correct. So it seems that Asus has corrected the low PCI writes in their current beta BIOS.

    I still need more results, but that is very encouraging.

    • Anonymous
    • 18 years ago

    Asus A7N266-EA 1003 Beta 002
    Adaptec 19160 with Seagate X15-36LP 18GB

    Write Read
    64.0 49944 50855
    128.0 50477 50477
    256.0 44422 42306
    512.0 40664 41431
    1024.0 40158 41397

    seems ok?

    • Ryu Connor
    • 18 years ago

    Gerbil #49

    Would you be willing to apply the newest Beta BIOS from Asus and run the test again for us?

    • Anonymous
    • 18 years ago

    “Houston, we have a problem.”

    Asus K7N266-E, BIOS 1001.D
    Promise Ultra ATA133 TX2, 2.20.0.12 and 2.00.0.29
    Maxtor D740X 40G, NTFS
    Windows XP Home Edition
    64k:Write:3261 Read:60624
    128:Write:3245 Read:32443
    256:Write:3240 Read:84822
    512:Write:3255 Read:84822
    1024:Write:3257 Read:31457

    vs: Onboard IDE
    64k:Write:26933 Read:31223
    128:Write:27840 Read:30962
    256:Write:27194 Read:31145
    512:Write:27893 Read:92368
    1024:Write:27995 Read:31457

    Seems PCmark 2002 also confirms a problem with HDD scores of 491 for the controller card and 859 without. OOUCH!

    • Anonymous
    • 18 years ago

    AG28, They are a tad slower, well within the ‘margin of noticability’ though.

    and, well, hell, they also have a rather noticable price difference going for them.

    and, get this.

    THEY ACTUALLY WORK
    🙂

    • Mr Bill
    • 18 years ago

    [q]rturner, “…I also get a huge disparity with/without direct IO checked…”[/q]

    Without direct IO checked, the OS will use system ram to buffer reads and writes, which hides the true values.

    • Anonymous
    • 18 years ago

    *[

    • Anonymous
    • 18 years ago

    *[

    • Ryu Connor
    • 18 years ago

    [q]I also get a huge disparity with/without direct IO checked[/q]

    You have to use Direct I/O.

    Take a look at your results:

    [q]Without Direct I/O checked :
    64k:Write:497549 Read:579600
    128:Write:505675 Read:590348
    256:Write:381681 Read:385089
    512:Write:389283 Read:389808
    1024:Write:390856 Read:393740[/q]

    500MB/s? I don’t think your PCI bus is quite that fast. 🙂

    As for your Direct I/O results, it continues to confirm the 20MB/s write limit.

    • Anonymous
    • 18 years ago

    *[

    • Anonymous
    • 18 years ago

    *[

    • Mr Bill
    • 18 years ago

    Ryu, I apologize to you and everyone on this thread for wasting thread space and your time because I did not read the discussion carefully enough. Sorry.

    • Ryu Connor
    • 18 years ago

    Tanj: Other benchmarks have confirmed that my 1000JB writes faster than reads. It appears to have been a firmware optimization by Western Digital and I have seen other JB drives on different platforms (Intel ICH2) perform the same way.

    This is getting off-topic though. Really need some more submissions.

    • Anonymous
    • 18 years ago

    [quote]
    What is interesting is that I’ve got a number of boards running on a setifarm. Looking at two boards from the same vendor, one being a P4 (i845) and a P3 (i815), I get very different results with the onboard IDE controllers.

    With ATTO tools, the i815 will actually go from ATA100 mode down to PIO4 mode in the duration of the test, whereas the i845 board does just fine. The i815 will stay in PIO4 until the computer is rebooted.

    Both under Win98, same “south bridge” and the same OS and ATA drivers installed. This only happens with ATTO, other disk benchmarks are ok…

    Uninstalling the Intel App Accelerator, the i815 actually performs better with this benchmark with the MS drivers being used. ATTO is the only one that sees this type of behavior.

    So… is this an application issue with ATTO, a driver issue with the IDE vendor drivers, or a mainboard chipset issue?[/quote]

    lwd_iceman, I’ve gotten results similar to what you’ve described.
    I posted them here:
    §[<http://www.sharkyforums.com/showthread.php?s=&threadid=136780<]§ but my drives would stay in PIO mode for i[

    • Anonymous
    • 18 years ago

    *[

    • Ryu Connor
    • 18 years ago

    Take a look at post #15 and #17’s scores.

    Different controllers, different driver, different hard drives, and even a different OS (2K vs XP).

    Yet their write results are the same. Trapped at the 20MB/s range.

    • Mr Bill
    • 18 years ago

    Oh. Can’t do “dynamic” in XP only in XP Pro. My bad.

    • Ryu Connor
    • 18 years ago

    See the second link in the first paragraph Mr. Bill.

    • Mr Bill
    • 18 years ago
    • Mr Bill
    • 18 years ago

    Have any of you tried converting the XP file system to “Dynamic” as suggested in the storagreview thread? It seems to do the trick.

    • TheCollective
    • 18 years ago

    [quote]Remember, this problem will not occur with the nForce’s on-board IDE controller.[/quote]

    Ryu is correct as the IDE controller is on the MCP and communicates through the HyperTransport link to the IGP. Therefore it is NOT on the PCI bus.

    • Ryu Connor
    • 18 years ago

    [q]I ran the program WINMSD to extract the following information about the IDE drivers that my K7N420 uses. Looks like PCI IDE to me.[/q]

    It’s graphed to the HT bus though, not the PCI bus. What’s the point of having a 800MB/s connection if you are going graph all your on-board devices to a 133MB/s PCI bus?

    • Anonymous
    • 18 years ago

    *[

    • Mr Bill
    • 18 years ago

    Interesting benchmark. This is not an Nvidia chipset mb but maybe useful for comparison?

    1. MSI K7D Master-L Bios 1.1
    2. Adaptec 3950U2B server controller (not raid) 3 x IBM UltraStar 18ES SCSI LVD software raid 0 using NT defaults on setup of stripe set
    3. WinNT4.0 SP6a SMP kernel
    4. 64k: Write 31640 Read 42099
    5. 128: Write 33741 Read 41630
    6. 256: Write 33316 Read 42356
    7. 512: Write 33857 Read 42045
    8. 1024: Write 35034 Read 40901

    single IBM UltraStar 8ES drive for comparison:
    4. 64k: Write 12787 Read 12749
    5. 128: Write 12725 Read 12410
    6. 256: Write 12702 Read 12582
    7. 512: Write 12573 Read 12573
    8. 1024: Write 12582 Read 12723

    • TheCollective
    • 18 years ago

    #23 The SiS chipsets don’t perform that well against the likes of nVidia and VIA.

    • Anonymous
    • 18 years ago

    AG23: I second this.

    Actually, I’m giving strong consideration to shuffling the ECS K7S5A I’m using now into another box and slapping a Duron on it, and giving this AthlonXP an Asus A7S333 to sit on (SiS 745 based)

    This has LITERALLY been the most troublefree board I’ve ever owned, whilst I generally dislike ECS hardware, they did good with the K7S5A 🙂

    • TheCollective
    • 18 years ago

    The onboard controller is not using the PCI bus therefore will not show any PCI issues.

    • TheCollective
    • 18 years ago

    Looks like the MSI boards are the problem not the chipset.

    • Anonymous
    • 18 years ago

    I don’t run PCI Raid card in this machine anymore (I did but lost one of the drives). However, I found the results strange enough to post.

    1. Asus A7N266-C Bios Rev 1003.2
    2. Maxtor 5400 RPM 40GB generic POS /w onboard controller
    3. Windows XP
    4. 64k:Write: 48276 Read 25052
    5. 128:Write 57906 Read 25124
    6. 256:Write 57852 Read 25063
    7. 512:Write 57940 Read 24966
    8. 1024:Write 57696 Read 25137

    I find it very strange that my WRITE speed is more than double the read speed.

    • Anonymous
    • 18 years ago

    #12, TheCollective:

    There is one platform available that will probably meet your requirements: SiS735

    • Anonymous
    • 18 years ago

    MSI K7N420 Pro BIOS 2.3
    AMI Megeraid 428, driver 2.68.0.0 (3 Quantum Atlas 10k’s)
    Windows 2000 SP2

    I get a huge disparity with/without direct IO checked.
    With Direct I/O checked :
    64k:Write:4042 Read:8398
    128:Write:3966 Read:8318
    256:Write:3943 Read:8238
    512:Write:3945 Read:8233
    1024:Write:3940 Read:8153

    Without Direct I/O checked :
    64k:Write:497549 Read:579600
    128:Write:505675 Read:590348
    256:Write:381681 Read:385089
    512:Write:389283 Read:389808
    1024:Write:390856 Read:393740

    • kraquen
    • 18 years ago

    [q]The test was at 1MB length for the 10,000 RPM Cheetah? [/q]

    i’m pretty sure i ran it at 1mb for the 10k cheetah.. i’ll make sure after work..

    [q]It should be noted that the Adaptec 2100 sucks donkey balls, and is just slow in general, so don’t take results from it as indicative of Nforce performance. [/q]

    my 2100s reads average 94MB/s my average writes are 16.7MB/s..

    that’s obviously a problem

    kraq

    • Ryu Connor
    • 18 years ago

    [q]This is using 2 WD 60Gig Caviar drives. Don’t know what hey should be? But if someone can give me an idea I would appreciate it. Thanks[/q]

    Your writes should be at parity with reads or faster. It seems the WD firmware is more optimized for writes.

    §[<http://www.tech-report.com/etc/2002q2/example.gif<]§ [q]It should be noted that the Adaptec 2100 sucks[/q] That may be true, but the performance it gaves fits other adapters, even non-SCSI ones. A pattern is begining to form.

    • Anonymous
    • 18 years ago

    It should be noted that the Adaptec 2100 sucks donkey balls, and is just slow in general, so don’t take results from it as indicative of Nforce performance.

    • Ryu Connor
    • 18 years ago

    [q]I still recommend Gigabit ethernet because they are not affected by Windows caching and it is not yet clear that there are NO issues in the Windows-Raid-Driver configuartion. Also run a Gigabit copying operation during a HD test.[/q]

    Asking for gigabit ethernet results would be like me asking for you to test a solid state disk.

    [q]I know, I know. I would be also pissed, because my board is brand new and I have bought it to throw all the buggy VIA crap out of my box. I would not like it, If the nForce PCI bus is also buggy.[/q]

    That much is obvious. Just chill and stop being a back seat driver. The results I have asked for will bear out the problem once some volume has been established.

    • Anonymous
    • 18 years ago

    *[

    • Anonymous
    • 18 years ago

    *[

    • Anonymous
    • 18 years ago

    *[

    • lwd_iceman
    • 18 years ago

    BTW – the i815 is Soyo SY7-IS2, the i845 is P4-IS2 – both are basic boards, no RAID controllers, both have PC133 installed. The P4 is 1700MHz Willy, the P3 is 866

    OS – Win98SE

    iceman

    • lwd_iceman
    • 18 years ago

    I’m not too sure this is an issue with nForce.

    Much depends on what the drive performance is, the mainboard drivers, the drivers for the card, etc, etc, etc…

    What we’ve seen with the examples is that nForce might have a problem. It might not be a “bios” problem with the PCI bus, could be a driver issue with the add-in card.

    What is interesting is that I’ve got a number of boards running on a setifarm. Looking at two boards from the same vendor, one being a P4 (i845) and a P3 (i815), I get very different results with the onboard IDE controllers.

    With ATTO tools, the i815 will actually go from ATA100 mode down to PIO4 mode in the duration of the test, whereas the i845 board does just fine. The i815 will stay in PIO4 until the computer is rebooted.

    Both under Win98, same “south bridge” and the same OS and ATA drivers installed. This only happens with ATTO, other disk benchmarks are ok…

    Uninstalling the Intel App Accelerator, the i815 actually performs better with this benchmark with the MS drivers being used. ATTO is the only one that sees this type of behavior.

    So… is this an application issue with ATTO, a driver issue with the IDE vendor drivers, or a mainboard chipset issue?

    iceman

    • TheCollective
    • 18 years ago

    Arrahhhh!!!! Aren’t there ANY AMD platforms out there that deliver supurb performance without any quirks? This is nothing like the VIA PCI bug but annoying at the least.

    • Forge
    • 18 years ago

    Darph Bobo!

    • Anonymous
    • 18 years ago

    Good god. Darth nVidia?

    • Ryu Connor
    • 18 years ago

    [q]my mylex acceleraid has a stripe size of 64k..
    all drives (the 2 15k cheetahs and the 10k cheetah) negotiate at 160..[/q]

    The test was at 1MB length for the 10,000 RPM Cheetah?

    • kraquen
    • 18 years ago

    my mylex acceleraid has a stripe size of 64k..
    all drives (the 2 15k cheetahs and the 10k cheetah) negotiate at 160..

    kraq

    • kraquen
    • 18 years ago

    This is poster #1.. i have a 2nd computer w/an MSI nForce..
    stats as follows

    i’m not sure what’s wrong w/this one.. it’s stats suck read and write..

    1. MSI K7N420 Pro, BIOS 2.4
    2. mylex acceleRAID 170 .. 1 10k rpm cheetah (obviously no raid)
    3. xp
    4. 64k:Write: 11822 Read 22783
    5. 128:Write 12157 Read 14100
    6. 256:Write 11991 Read 15450
    7. 512:Write 12663 Read 13076
    8. 1024:Write 12732 Read 15660

    • Forge
    • 18 years ago

    drruhe – I could be wrong, but are you testing a drive on the nForce IDE, or on the onboard HPT372? Only the HPT372 would show the problem, as the nForce native IDE is NOT on the PCI bus…

    • Ryu Connor
    • 18 years ago

    [q]Ryu, I strongly recomment to run tests with gigabit ethernet PCI cards which should reach 60-70MB/s for a final confirmation that there exist a nForce write weakness.[/q]

    Not enough proliferation. High-end SCSI and IDE like in post #1 are more taxing anyways.

    [q]it could also be a bug in ATTO when using partion #1 and #2[/q]

    ATTO is working correctly.

    §[<http://forums.storagereview.net/viewtopic.php?t=1758<]§ [q]No PCI write weakness yet proven on my system[/q] Abit may have fixed the issue in firmware like Asus. That is what I hope this thread will determine. As you can see from post #1, some people are in serious pain.

    • Anonymous
    • 18 years ago

    *[

    • Anonymous
    • 18 years ago

    *[

    • Forge
    • 18 years ago

    Owww Jebus, that hurts me just to read.

    Cheetahs… RAID0… 15MB/s write. Your seek time is really good, at least, right?

    • Anonymous
    • 18 years ago

    1. MSI K7N420 Pro, BIOS 2.4
    2. adaptec 2100 raid (2 15k rpm cheetahs in raid0)
    3. xp
    4. 64k:Write: 15200 Read 78708
    5. 128:Write 17130 Read 105642
    6. 256:Write 17646 Read 99614
    7. 512:Write 17801 Read 116624
    8. 1024:Write 17712 Read 83322

    obviously my writes suck

Pin It on Pinterest

Share This