Hey Tex......What was that about a bus limit?

Mt_GoatMt_Goat Head Cheezy KnobPflugerville (north of Austin) Icrontian
edited May 2005 in Hardware
I am not sure if I can improve on this or not. But I would sure like to even things out between reads and writes. This is on my DFI NV-RAID controller. One thing is for sure. there doesn't seem to be much of a cap on the potential here. :eek:

Comments

  • lemonlimelemonlime Canada Member
    edited May 2005
    That's the great thing about chipset based controllers, they do not communicate over the traditional PCI bus. SiI 3x series controllers on the NF2 boards, for example, used the PCI bus, with around a 130MB/s limitation.

    Not too sure what the ceiling is on the NF4 nvraid, but I'm sure its pretty high :D

    Awesome atto btw, those new rev. raptors are fantastic..
  • RWBRWB Icrontian
    edited May 2005
    :p
  • Mt_GoatMt_Goat Head Cheezy Knob Pflugerville (north of Austin) Icrontian
    edited May 2005
    lemonlime wrote:
    That's the great thing about chipset based controllers, they do not communicate over the traditional PCI bus. SiI 3x series controllers on the NF2 boards, for example, used the PCI bus, with around a 130MB/s limitation.

    Not too sure what the ceiling is on the NF4 nvraid, but I'm sure its pretty high :D

    Awesome atto btw, those new rev. raptors are fantastic..
    lemonlime

    That is actually 3 of them. I got the one I'm going to be putting in my wife's new rig as a single frive yesterday and decided to try 3 in mine before I have all the parts for her machine. My 2 are still not performing like they should. I still have a HPT 1520 PCI controller that I tried my 2 on and it maxes it out around 115-120 with the reads and writes almost perfectly matched all the way down the chart. I gave up on trying to straighted my reads and writes out with this thing and am starting to consider another 2 drives for myself and a PCI-e controller card with 4 of these suckers on it. :eek:
  • TexTex Dallas/Ft. Worth
    edited May 2005
    There are very few PCI-E controllers out right now. I bought the DFI also and it won't even boot with my $700 pci-e scsi raid controller in it.

    Most the NF4 boards seem to be expecting only video cards in the pci-e slots right now. DFI is one of the worst it seems. Stick with the onboard raid.

    The supermicro and Intel dual xeons seem to function properly (with some bios's) using raid controllers in the pci-e slots but not many others.

    Tex

    And as others mentioned... Your not on the PCI bus?? And you mentioned yours on the HPT have matched reads all the way down the chart? In real life it might perform better then the ATTO you posted here off the onboard raid. Top end on ATTO isnt the key. Watch the low and middle section which indicate more about real life "feel" then the top end segment. And the OS drive (and most normal drives) have far more reads then writes. Your reads are whacked through the middle section bad. Your getting less then 30mb per raptor through the middle section. Does that even out more with two drives? Sometimes three is a bad number. Two or four work better.

    Whats the access time on that baby ? 6 or 7ms?
  • lemonlimelemonlime Canada Member
    edited May 2005
    I get a very similar ATTO pattern with my NF4 DFI. I didn't have this problem with the NF3 250GB 754 board, even though the nvraid is essentially the same. I started a thread about it a while ago here

    This was with two early rev. 36GB raptors, but similar pattern none the less..
  • TexTex Dallas/Ft. Worth
    edited May 2005
    Weird. Thats the pattern some of the WD, JB's used to always exhibit.

    I wonder if its a latency issue? You can't adjust it on these boards of course but I wonder??

    Tex
  • Mt_GoatMt_Goat Head Cheezy Knob Pflugerville (north of Austin) Icrontian
    edited May 2005
    Here is what I got w/ 2 of these drives on the onboard controller. I actually did pick up quite a bit in the low and middle, as you said "where it counts". ;) Also notice the access time on the Sandra from the 2 drive set up. I haven't run Sandra on here yet but will post one as soon as I get it on here. The real funny thing is that for the most part it feels pretty darn good right now.

    Here is the 2 drives. Looks like total suckage to me.:wtf:

    attachment.php?attachmentid=16169&stc=1


    attachment.php?attachmentid=16170&stc=1
  • TexTex Dallas/Ft. Worth
    edited May 2005
    You didnt pick up anything on the reads for sure. And the 17ms suxs big time. My HPT ide arrays used to adda ms or so on the raid arrays versus a single drive. Moved from 8 or 9 ms to 9 or 10. Which is why I posted my guess abaove. 17ms is just crappy.

    You would be better running them non raided with a 5ms access time. STR over access time is only an advantage in a very few applications.

    My scsi raid even with old controllers had access times like these. This is a 6 year old scsi controller. 2ms is so far advanced from 17ms even if the ATTO from peak STR is in the same ballpark.

    The drives had 4.5ms access as single drives. The access times for the array go down not up like IDE.

    Tex
  • Mt_GoatMt_Goat Head Cheezy Knob Pflugerville (north of Austin) Icrontian
    edited May 2005
    Tex

    The Sandra with accompanying ATTO was w/ 2 drives, which only got an av access of 17. Below is my Sandra of the 3 drives and my av access looks much better now but I am still in shock about this set-up not doing better.

    You certainly are more knowledgeable than I am on this subject. But I shure did pay carefull attention to everything you taught me and am extremely happy to have you as my teacher. I have been wondering about the latency s it looks exactly like several latency imbalances in regard to the way the dominance of reads and writes shifts back and forth with one end being much stronger than the other. When I tested my Hitachi drives in both RAID and as single drives the same pattern persisted every time! And in addition to the Raptors my Hitachi drives in RAID when hung off my HPT card were much closer and the difference in reads and writes was easy to close with a latency adjustment. :scratch:

    If there was some way to make a patch or adjust the latency by some back door method I think I would be willing to experiment with it.
  • TexTex Dallas/Ft. Worth
    edited May 2005
    I know of no way to adjust the laterncy with that controller not on the PCI bus.

    Powerstrip may work but I have serious doubts.

    you have probably tried more then me as you know how to tweak it. I have taught you well Obi-Wan

    Tex
  • Mt_GoatMt_Goat Head Cheezy Knob Pflugerville (north of Austin) Icrontian
    edited May 2005
    So I am ready to learn a new trick Master Yoda. :cool:

    I have;
    (2x) WD740 Raptors
    (1x) WD360 Raptors
    (2x) Hitachi 80 GB SATA
    (1x) Seagate 200 GB SATA
    (1x) HPT 1520 (can use for 2x RAID-0 or just to hang more drives as singles) for a totl of 6.

    What would be the best configuration for these or some of these to make the most of things. I have enough drives that I can spare one or two for swap files and or a scratch drive for Photoshop. Ideas Master Yoda????
  • lemonlimelemonlime Canada Member
    edited May 2005
    Sandra paints a pretty different picture for me, Access time is about 5ms, but I'm not sure about the STR, does this spot on for 1st gen raptors?
  • Mt_GoatMt_Goat Head Cheezy Knob Pflugerville (north of Austin) Icrontian
    edited May 2005
    Well I broke it up and now have;
    WD740 = C: (20GB) Windows and support progs / D: (56GB) My Documents
    WD740 = E: (20GB) Adobe scratch pad, etc. / F: (56GB) Progam Files
    WD360 = X: (10GB) 4GB Swap File
    Seagate 200 = L: (140GB) Backup / M: (50GB) Extra Storage

    I did discover that one of my original WD740's was borked w/ massive bad sectors and am RMA'ing it today. I tried the other 2 in a RAID-0 nd the ATTO was only slightly better than my original 2 but the accss times went down to 4ms. :thumbsup:

    I will see how I like running the above for a while and the RMA replacement will go to my wife's new machine as I originally got one for her as her new main drive and her Diamondmax 9 (80GB) will play backup for her. I also still have my HPT SATA controller card and the 2 Hitachi 80GB's in the drawer.
  • Mt_GoatMt_Goat Head Cheezy Knob Pflugerville (north of Austin) Icrontian
    edited May 2005
    lemonlime wrote:
    Sandra paints a pretty different picture for me, Access time is about 5ms, but I'm not sure about the STR, does this spot on for 1st gen raptors?
    Your access time looks spot on but I think the NV RAID has latency issues that may or may not be addressed in the future. I wish we had a patch like thones George Breeze did for the VIA chipsets. But that was a PCI bus issue and as my master has taught me those won't work here. It may take a bios tweek from the inside to resolve this or at least vastly improved drivers.
  • lemonlimelemonlime Canada Member
    edited May 2005
    mtgoat wrote:
    Your access time looks spot on but I think the NV RAID has latency issues that may or may not be addressed in the future. I wish we had a patch like thones George Breeze did for the VIA chipsets. But that was a PCI bus issue and as my master has taught me those won't work here. It may take a bios tweek from the inside to resolve this or at least vastly improved drivers.

    I just flashed the beta 5/10-2 beta bios, that was re-worked ground-up (based on 3/16 dram tables and the whole works) I'm gonna run some benchies tonight. It appears to have update the nvraid rom as well, to 4.84
Sign In or Register to comment.