The FastTrak SX4000 card has 256MB of cache installed and is on its own IRQ; it has a RAID 0 array with stripe width 4. The performance of this array is slower, both in synthetic benchmarks, as well as in day to day usage, when compared to the FastTrak100 array that is in this same system, and that array only has a stripe width of 2 (dual IBM 75GXPs, 15GB capacity each). I understand that the onboard processor of the SX4000 is not as powerful as the system CPU which performs striping calculations for the FastTrak100, but that should not be such an advantage that the SX4000 array, with twice as many drives (being 400BB are individually faster than the 75GXPs that are on the FastTrak100) as well as 256MB of hard cache, is actually still slower than the FastTrak100 array. I am very disappointed in the performance of my new array from the standpoint of comparison to an older, and what, on paper, ought to be slower, array. Both are on the same machine, so system configuration cannot be an issue. I have tried adjusting write cache between write-thru and write-back but it hardly yields any different results. I did notice, however, that in the synthetic benchmark, HDTach, the drive map for the dual 75GXP on FastTrak100 array shows a normal graph for hard drives, with a gradually increasing downward taper towards the end of the array; it starts around 43MB/s and tapers down to about 28MB/s. When performing the same benchmark on my quad 400BB on SX4000 array, the graph starts at around 32MB/s and flatlines right to the end of the array. This suggests to me that there is some sort of bottleneck; could it be my drivers or my software? Obviously, I am not hitting the maximum performance of the array because there is no taper. I have tried all sorts of Windows cache settings and it did not help. The fact that it flatlines at about (a mere) 32MB/s tells me that my PCI bus is not the culprit, as that's way below the barrier of about 133MB/s. Will there be newer BIOS releases or drivers to rectify the problem? Any idea on what could be going on? Thank you very much in advance for any help you can provide me; the situation is quite frustrating as I have invested a hefty amount of money into this new array only to find that not only does it cost significantly more than my old array, but it performs worse! My system configuration follows?
-Edward Ng
13 Campus Pl./SL
Scarsdale, NY 10583
(914) 725-2638
PS BTW, the SX4000, as I said earlier, shares no IRQs, while the FastTrak100 actually shares an IRQ with the Audigy2 board (IRQ11) and the only other IRQ shared on this system is IRQ5, and it is shared by the video card and the IEEE1394 controller on the Audigy2 board. So I know it's not a resource conflict issue. Finally there has been no shadowing of any memory addresses in the BIOS, and I have also experimented with the PCI latency patches and BIOS settings, to no avail.
System configuration:
Abit KT7A with BIOS revision A9 (dated 7/11/02), VIA KT133A-based Socket-A motherboard, ATX
AMD Athlon (C-core Thunderbird) 1000 cooled by Thermaltake Volcano 7+ with fan set to Median
Antec TruePower550 ATX power supply
Enlight 7237 steel ATX system enclosure with additional cooling (multiple blowholes, Thermaltake Smart Fans)
512MB PC133 SDR SDRAM
Microsoft Windows XP Professional, Build 2600
FastTrak SX4000, 256MB cache, PC133 SDR SDRAM, IRQ10:
Channel 1: Western Digital 400BB, standard single-device, 80-conductor ribbon IDE cable provided by Promise
Channel 2: Western Digital 400BB, standard single-device, 80-conductor ribbon IDE cable provided by Promise
Channel 3: Western Digital 400BB, standard single-device, 80-conductor ribbon IDE cable provided by Promise
Channel 4: Western Digital 400BB, standard single-device, 80-conductor ribbon IDE cable provided by Promise
Southbridge IDE Controller, IRQs 14 & 15:
Channel 1, Master: Lite-On LTR-40125S 40X CD-RW drive, dual-device, 80-conductor rounded copper mesh shielded IDE cable
Channel 1, Slave: none
Channel 2, Master: Lite-On LTD-163D 16X DVD-ROM drive, dual-device, 80-conductor rounded copper mesh shielded IDE cable
Channel 2, Slave: Kenwood UCR-421 CD-ROM drive, dual-device, 80-conductor rounded copper mesh shielded IDE cable
FastTrak100, IRQ11 (shared with Sound Blaster Audigy2):
Channel 1, Master: IBM Deskstar 75GXP (15GB model), standard, dual-device, 80-conductor ribbon IDE cable provided by Promise
Channel 1, Slave: none
Channel 2, Master: IBM Deskstar 75GXP (15GB model) , standard, dual-device, 80-conductor ribbon IDE cable provided by Promise
Channel 2, Slave: none
Sound Blaster Audigy2, IRQs 11 (shared with FastTrak100) & 5 (onboard IEEE1394, shared with video card)
Visiontek Xtasy nVIDIA GeForce4 Ti4200-based card, IRQ5 (shared with onboard IEEE1394 on Audigy2)
NO IDE cables exceed 18? in length, nothing on this system is run out of spec (no CPU overclock, no video overclock, no aggressive RAM settings) and it has very recently been fully reformatted (reformat performed to install OS onto new array, old array fresh formatted empty). I doubt that there is a power supply issue, as this supply is plenty strong to drive all the devices in the system. Also, I seriously doubt an overheating issue because my system probe reports a case temperature range of 24-28 degrees Celcius, which is perfectly fine; the drives are mounted in well ventilated positions.
-Edward Ng
13 Campus Pl./SL
Scarsdale, NY 10583
(914) 725-2638
PS BTW, the SX4000, as I said earlier, shares no IRQs, while the FastTrak100 actually shares an IRQ with the Audigy2 board (IRQ11) and the only other IRQ shared on this system is IRQ5, and it is shared by the video card and the IEEE1394 controller on the Audigy2 board. So I know it's not a resource conflict issue. Finally there has been no shadowing of any memory addresses in the BIOS, and I have also experimented with the PCI latency patches and BIOS settings, to no avail.
System configuration:
Abit KT7A with BIOS revision A9 (dated 7/11/02), VIA KT133A-based Socket-A motherboard, ATX
AMD Athlon (C-core Thunderbird) 1000 cooled by Thermaltake Volcano 7+ with fan set to Median
Antec TruePower550 ATX power supply
Enlight 7237 steel ATX system enclosure with additional cooling (multiple blowholes, Thermaltake Smart Fans)
512MB PC133 SDR SDRAM
Microsoft Windows XP Professional, Build 2600
FastTrak SX4000, 256MB cache, PC133 SDR SDRAM, IRQ10:
Channel 1: Western Digital 400BB, standard single-device, 80-conductor ribbon IDE cable provided by Promise
Channel 2: Western Digital 400BB, standard single-device, 80-conductor ribbon IDE cable provided by Promise
Channel 3: Western Digital 400BB, standard single-device, 80-conductor ribbon IDE cable provided by Promise
Channel 4: Western Digital 400BB, standard single-device, 80-conductor ribbon IDE cable provided by Promise
Southbridge IDE Controller, IRQs 14 & 15:
Channel 1, Master: Lite-On LTR-40125S 40X CD-RW drive, dual-device, 80-conductor rounded copper mesh shielded IDE cable
Channel 1, Slave: none
Channel 2, Master: Lite-On LTD-163D 16X DVD-ROM drive, dual-device, 80-conductor rounded copper mesh shielded IDE cable
Channel 2, Slave: Kenwood UCR-421 CD-ROM drive, dual-device, 80-conductor rounded copper mesh shielded IDE cable
FastTrak100, IRQ11 (shared with Sound Blaster Audigy2):
Channel 1, Master: IBM Deskstar 75GXP (15GB model), standard, dual-device, 80-conductor ribbon IDE cable provided by Promise
Channel 1, Slave: none
Channel 2, Master: IBM Deskstar 75GXP (15GB model) , standard, dual-device, 80-conductor ribbon IDE cable provided by Promise
Channel 2, Slave: none
Sound Blaster Audigy2, IRQs 11 (shared with FastTrak100) & 5 (onboard IEEE1394, shared with video card)
Visiontek Xtasy nVIDIA GeForce4 Ti4200-based card, IRQ5 (shared with onboard IEEE1394 on Audigy2)
NO IDE cables exceed 18? in length, nothing on this system is run out of spec (no CPU overclock, no video overclock, no aggressive RAM settings) and it has very recently been fully reformatted (reformat performed to install OS onto new array, old array fresh formatted empty). I doubt that there is a power supply issue, as this supply is plenty strong to drive all the devices in the system. Also, I seriously doubt an overheating issue because my system probe reports a case temperature range of 24-28 degrees Celcius, which is perfectly fine; the drives are mounted in well ventilated positions.
Facebook
Twitter