dlmh Posted June 13, 2009 Share Posted June 13, 2009 I couldn't find anything here in the forum about this specific board from Supermicro. I was looking for a really low-power fileserver, so I started looking at Atom-based motherboards. Unfortunately, most Atom-boards based on Intel D945GC chip only have 2x SATA and a PCI-slot, but this one actually has 2x PCI-e 8x (one running at 4x) slots, enabling you to add 2 controller-cards. It costs about 150,00 EUR, or $200, but the CPU (ofcourse) is included. Here's the specs from the website: Form Factor Flex ATX Dimensions 9" x 7.5" (22.86cm x 19.05cm) Processor/Cache CPU Intel® Atom™ 330 Dual-Core 1.6GHz processor System Bus 533 MHz system bus System Memory Memory Capacity Two 240-pin DIMM sockets Supports up to 2 GB of unbuffered, non-ECC DDR2 533/400MHz memory (Single / Dual Channel) Memory Type DDR2 533/400MHz unbuffered, non-ECC SDRAM, 240-pin gold-plated DIMMs DIMM Sizes 256 MB, 512 MB, 1 GB, 2GB Memory Voltage 1.8V only On-Board Devices Chipset Intel 945GC + ICH7R Realtek RTL8111C-GR SATA Intel ICH7R SATA 3.0Gbps Controller RAID 0, 1, 5, 10 support Network Controllers 2x Realtek RTL8111C-GR Gigabit Ethernet Supports 10BASE-T, 100/1000BASE-TX, RJ45 output Graphics Onboard GMA950 video Super I/O Winbond 83627DHG-P chip Clock Generator CK505 Input / Output Serial ATA Four Serial ATA ports ( 3Gbps ) Four SATA hard drives supported IDE One EIDE channel supports up to two UDMA IDE devices Supports UDMA Mode 5, PIO Mode 4, and ATA/100 LAN 2x RJ45 LAN port USB 8x USB 2.0 / 1.1 compatible 2 rear + 3 headers (5 ports) + 1 Type A connector VGA Intel GMA950 Keyboard / Mouse PS/2 keyboard and mouse ports Serial Ports 1x Fast UART 16550 serial port 1x Serial Header System BIOS BIOS Type 8Mb Flash EEPROM with AMI BIOS BIOS Features Plug and Play (PnP) DMI 2.3 PCI 2.2 ACPI 1.0 / 2.0 / 3.0 USB Keyboard support SMBIOS 2.3 PCI-Express 1 (x8) PCI-Express Slot 1 (x4) PCI-Express (in x8 slot) PCI 1x 32-bit 33MHz PCI (5V) slots Management Software Super Doctor III Watch Dog Power Configurations ACPI/APM Power Management Main Switch Override Mechanism Suspend to RAM (STR) Power-on mode control for AC power recovery Internal / External modem remote ring-on PC Health Monitoring CPU Monitors for CPU Cores, +3.3V, +5V, +12V, +3.3V Standby, VBAT, Memory, Chipset CPU Core 1-Phase-switching voltage regulator with auto-sense from 0.8375V-1.6000V FAN Total of four 4-pin fan headers 4 x fans with tachometer monitoring Status monitor for speed control Status monitor for on / off control Support 3-pin fans (without speed control) Low noise fan speed control Pulse Width Modulated (PWM) fan conn. Temperature Monitoring for CPU and chassis environment CPU thermal trip support Thermal Control for 4 x Fan Connectors I2C temperature sensing logic Thermal Monitor 2 (TM2) Support LED CPU / System Overheat LED Other Features Chassis intrusion Detection Chassis intrusion Header SDDC Support I'm still hesitant considering all the great ION-based products to be released soon, but it seems like a low-power winner to me. Question remains if the Atom N330 will be powerful enough to do all the parity calculations, but as long as it's able to fill Gbit LAN bandwidth (or 60-80 MB/s using a cache disk), I'm a happy camper. Anyone got some more info to share about this board or other Atom-based ones? Thanks! Quote Link to comment
Joe L. Posted June 13, 2009 Share Posted June 13, 2009 Parity calculations take VERY LITTLE CPU. You could probably use a pentium 3 and keep up. The bottleneck is the PCI bus and the network. Your MB will probably have no problem with the PCI-e bus in keeping up with the disks at all. You do need to verify the SATA disk controller is supported. You also need to verify the network chipset is supported. If yes to both, then you will do just fine. With my old PCI based motherboard, with a mix of 10 IDE and SATA disks, I see the following during a parity check: top - 15:20:35 up 21 days, 12:22, 2 users, load average: 1.48, 0.67, 0.41 Tasks: 79 total, 3 running, 76 sleeping, 0 stopped, 0 zombie Cpu(s): 1.0%us, 16.6%sy, 0.0%ni, 82.4%id, 0.0%wa, 0.0%hi, 0.0%si, 0.0%st Mem: 498384k total, 484368k used, 14016k free, 70512k buffers Swap: 0k total, 0k used, 0k free, 318976k cached 16.6% of the Intel Celeron 2.26GHz CPU is being used to check parity... 82% is idle... Joe L. Quote Link to comment
dlmh Posted June 13, 2009 Author Share Posted June 13, 2009 Parity calculations take VERY LITTLE CPU. You could probably use a pentium 3 and keep up. The bottleneck is the PCI bus and the network. Your MB will probably have no problem with the PCI-e bus in keeping up with the disks at all. You do need to verify the SATA disk controller is supported. You also need to verify the network chipset is supported. If yes to both, then you will do just fine. Thanks for sharing. It's a ICH7R southbridge, so I'm guessing it will be supported. However, I will be using a Highpoint RocketRAID 2320 (8x SATA / PCIe x4), and I'm still able to add a second controller. Power consumption for this board is rated at 40W idle, that's 15W more than an ION-board. But then again, no ION-board will give me two PCIe x4 slots. There's also a different version of this board with only one GBit LAN-port and an Atom N230. You think that'll suffice too? (it's about 30,00 EUR cheaper). Quote Link to comment
enternoescape Posted July 17, 2009 Share Posted July 17, 2009 I currently am using this motherboard with an Adaptec 2241000-R 1430SA and a Supermicro AOC-SAT2-MV8. My hope was to be able to use 2 of the Supermicro AOC-SASLP-MV8 controllers and get 20 high speed connections, but due to the issues people have reported this setup will suffice for now. The parity calculation speed on 7 1.5TB WD green drives connected to the motherboard and Adaptec was somewhere around 50KB/s and dropped to 30KB/s towards the end of calculation. Considering I was using this on a platform with a 2GHz processor before I moved it over to this motherboard and the performance is identical, the Atom 330 is more than able for unRAID. On a side note, I did originally have an issue when I would pull from two computers at once saturating the gigabit connection, the entire server would completely disappear from the network after about 5 minutes of this traffic. There's nothing in the log indicating an error, I couldn't ping, telnet, access the web page or open a share. I ran ifconfig on the server revealed it still had an address and was connected, but from the server I couldn't ping anything on my local network. I relaunched emhttp and everything would magically come back to life. I saw a post regarding APIC issues and changed in syslinux.cfg the line: append initrd=bzroot to append initrd=bzroot noapic Now it seems to be working much more reliably. I personally would probably buy another one because the apic issue outlined here is a common one with linux apparently. Quote Link to comment
prostuff1 Posted July 17, 2009 Share Posted July 17, 2009 If you could take a looks at the Motherboard Rating System and the Hardware Compatibility Page that would be great. If I don't see the board added when I get home I will get it added and all set up. Thanks EDIT: Board has been added but if you can get the Motherboard Rating System stuff done that would be great!! Quote Link to comment
dlmh Posted July 20, 2009 Author Share Posted July 20, 2009 I currently am using this motherboard with an Adaptec 2241000-R 1430SA and a Supermicro AOC-SAT2-MV8. My hope was to be able to use 2 of the Supermicro AOC-SASLP-MV8 controllers and get 20 high speed connections, but due to the issues people have reported this setup will suffice for now. The parity calculation speed on 7 1.5TB WD green drives connected to the motherboard and Adaptec was somewhere around 50KB/s and dropped to 30KB/s towards the end of calculation. Considering I was using this on a platform with a 2GHz processor before I moved it over to this motherboard and the performance is identical, the Atom 330 is more than able for unRAID. On a side note, I did originally have an issue when I would pull from two computers at once saturating the gigabit connection, the entire server would completely disappear from the network after about 5 minutes of this traffic. There's nothing in the log indicating an error, I couldn't ping, telnet, access the web page or open a share. I ran ifconfig on the server revealed it still had an address and was connected, but from the server I couldn't ping anything on my local network. I relaunched emhttp and everything would magically come back to life. I saw a post regarding APIC issues and changed in syslinux.cfg the line: append initrd=bzroot to append initrd=bzroot noapic Now it seems to be working much more reliably. I personally would probably buy another one because the apic issue outlined here is a common one with linux apparently. Is the parity calculation generally this slow or can it be sped up by using a Quad core CPU (for instance)? Not that it really matters since I'll be using a cache-disk, but would you normally have 30-50 kbps throughput or is parity done AFTER files have been transferred to the disks? And what's this APIC issue? Is that specific for this board and can it be solved with a BIOS update? Quote Link to comment
RobJ Posted July 20, 2009 Share Posted July 20, 2009 ... somewhere around 50KB/s and dropped to 30KB/s towards the end ... I suspect you meant 50MB/s and 30MB/s? If it really was KB/s, then something is very wrong, should be apparent in your syslog. On a side note, I did originally have an issue when I would pull from two computers at once saturating the gigabit connection, the entire server would completely disappear from the network after about 5 minutes of this traffic. There's nothing in the log indicating an error, I couldn't ping, telnet, access the web page or open a share. I ran ifconfig on the server revealed it still had an address and was connected, but from the server I couldn't ping anything on my local network. I relaunched emhttp and everything would magically come back to life. I saw a post regarding APIC issues and changed in syslinux.cfg the line: append initrd=bzroot to append initrd=bzroot noapic Now it seems to be working much more reliably. I personally would probably buy another one because the apic issue outlined here is a common one with linux apparently. Since it seems related to the NIC under heavy load, another option would be to add an Intel Pro/1000 PCIe card. It is reported to be superior under heavy traffic. Quote Link to comment
RobJ Posted July 20, 2009 Share Posted July 20, 2009 Is the parity calculation generally this slow or can it be sped up by using a Quad core CPU (for instance)? Not that it really matters since I'll be using a cache-disk, but would you normally have 30-50 kbps throughput or is parity done AFTER files have been transferred to the disks? I suspect he meant 30-50MB/s. The bottleneck is not the CPU, but the I/O. Even a slow single core CPU is not likely to slow down the parity calculations. But you can be severely limited by your bus, or a faulty drive. And what's this APIC issue? Is that specific for this board and can it be solved with a BIOS update? The nForce boards were affected for awhile, but recent Linux kernel releases have apparently included special handling of them, so noapic may no longer be required. At least that is so with my board. I *think* that it is possible that a BIOS update could help too. Quote Link to comment
dlmh Posted July 21, 2009 Author Share Posted July 21, 2009 ... somewhere around 50KB/s and dropped to 30KB/s towards the end ... I suspect you meant 50MB/s and 30MB/s? If it really was KB/s, then something is very wrong, should be apparent in your syslog. On a side note, I did originally have an issue when I would pull from two computers at once saturating the gigabit connection, the entire server would completely disappear from the network after about 5 minutes of this traffic. There's nothing in the log indicating an error, I couldn't ping, telnet, access the web page or open a share. I ran ifconfig on the server revealed it still had an address and was connected, but from the server I couldn't ping anything on my local network. I relaunched emhttp and everything would magically come back to life. I saw a post regarding APIC issues and changed in syslinux.cfg the line: append initrd=bzroot to append initrd=bzroot noapic Now it seems to be working much more reliably. I personally would probably buy another one because the apic issue outlined here is a common one with linux apparently. Since it seems related to the NIC under heavy load, another option would be to add an Intel Pro/1000 PCIe card. It is reported to be superior under heavy traffic. Although this board has 2 PCI-e 4x expansions slots I plan to use them for SATA controllers. Is the PCI version just as good and will it improve much over the RTL8111c? I currently use the latter (on Intel D875XBX2) and usually get 70-90 Mbyte/s without issues (in OSx86, not unRAID). Quote Link to comment
RobJ Posted July 21, 2009 Share Posted July 21, 2009 Since it seems related to the NIC under heavy load, another option would be to add an Intel Pro/1000 PCIe card. It is reported to be superior under heavy traffic. Although this board has 2 PCI-e 4x expansions slots I plan to use them for SATA controllers. Is the PCI version just as good and will it improve much over the RTL8111c? I currently use the latter (on Intel D875XBX2) and usually get 70-90 Mbyte/s without issues (in OSx86, not unRAID). Just my personal opinion, but I think it would be almost as good, *IF* you do not have any drives on disk controllers on the PCI bus. In other words, it would be much better if your PCI NIC was using the PCI bus unshared with anything else. Quote Link to comment
dlmh Posted July 21, 2009 Author Share Posted July 21, 2009 Since it seems related to the NIC under heavy load, another option would be to add an Intel Pro/1000 PCIe card. It is reported to be superior under heavy traffic. Although this board has 2 PCI-e 4x expansions slots I plan to use them for SATA controllers. Is the PCI version just as good and will it improve much over the RTL8111c? I currently use the latter (on Intel D875XBX2) and usually get 70-90 Mbyte/s without issues (in OSx86, not unRAID). Just my personal opinion, but I think it would be almost as good, *IF* you do not have any drives on disk controllers on the PCI bus. In other words, it would be much better if your PCI NIC was using the PCI bus unshared with anything else. This board only sports one PCI-slot so it won't be shared with other devices. Onboard graphics, parallel and serial ports and audio will be disabled. Is there a specific Intel GBit LAN adapter that is better? Quote Link to comment
Slorcan Posted August 6, 2009 Share Posted August 6, 2009 I aswell currently am application this motherboard with an Adaptec 2241000-R 1430SA and a Supermicro AOC-SAT2-MV8. My achievement was to be able to use 2 of the Supermicro AOC-SASLP-MV8 controllers and get 20 top acceleration connections, but due to the issues humans accept appear this bureaucracy will answer for now. _________________ Refrigerator filter Quote Link to comment
WeeboTech Posted August 6, 2009 Share Posted August 6, 2009 I currently am using this motherboard with an Adaptec 2241000-R 1430SA and a Supermicro AOC-SAT2-MV8. My hope was to be able to use 2 of the Supermicro AOC-SASLP-MV8 controllers and get 20 high speed connections, but due to the issues people have reported this setup will suffice for now. I aswell currently am application this motherboard with an Adaptec 2241000-R 1430SA and a Supermicro AOC-SAT2-MV8. My achievement was to be able to use 2 of the Supermicro AOC-SASLP-MV8 controllers and get 20 top acceleration connections, but due to the issues humans accept appear this bureaucracy will answer for now. Hmmmmm..... Spammer? Quote Link to comment
RobJ Posted August 6, 2009 Share Posted August 6, 2009 Very well spotted, WeeboTech!!! I knew it was highly suspicious from the fake sig with product link, but it 'graded' as much more relevant than the others. And the awkward phraseology and word choices made me much more forgiving, thinking it *might* be a non-English speaker, or the badly botched product of an automated translator program. The apparently Alaskan timezone was not particularly consistent with a non-English speaker. I suppose you could call this the product of a plagiarizing program, a lousy one though. What is significant is these people are learning, improving their spam post relevance, and fixing some of the issues, better simulation of a real user. In the past, the writers of other posts like this had a total online time of about a minute, while this one has a somewhat longer accumulated time on the forums. The spam sigs still stick out though. Perhaps in their next attempts, they will change the formatting of these fake sigs, at least randomize their style ... Quote Link to comment
wjw Posted November 13, 2009 Share Posted November 13, 2009 I recently (ie, yesterday) switched to this mobo from a uATX/Celeron combo. Given, it's only been 12-18 hours, but I am -extremely- happy with it. With my old config, parity calculations across 5 drives (4d+1p) had my drive temps above 42degC, with a few drives finishing at 45degC. With this new mobo/proc, my drives didn't get above 33degC (parity was 33degC, data drives didn't go above 31degC). I fully endorse this mobo/proc. Very happy. I was worried that the GPU's tiny fan would be noisy at the stock 4500rpm, but it isn't. (side question: b/c of the lack of graphic calc/rendering, would it be "safe" to remove this fan?) [btw, I'm running 4.5beta8, and I have an Antec P180 with no additional cooling other than the 2 stock case fans (one is 210mm, high CFM, low noise)]. Quote Link to comment
WeeboTech Posted November 13, 2009 Share Posted November 13, 2009 I recently (ie, yesterday) switched to this mobo from a uATX/Celeron combo. Given, it's only been 12-18 hours, but I am -extremely- happy with it. With my old config, parity calculations across 5 drives (4d+1p) had my drive temps above 42degC, with a few drives finishing at 45degC. With this new mobo/proc, my drives didn't get above 33degC (parity was 33degC, data drives didn't go above 31degC). I fully endorse this mobo/proc. Very happy. I was worried that the GPU's tiny fan would be noisy at the stock 4500rpm, but it isn't. (side question: b/c of the lack of graphic calc/rendering, would it be "safe" to remove this fan?) [btw, I'm running 4.5beta8, and I have an Antec P180 with no additional cooling other than the 2 stock case fans (one is 210mm, high CFM, low noise)]. How long and how fast did it take to do the new parity calc. What controller are you using for the 5th drive? I'm pretty interested in this board for a low power machine. From what I've read the atom processor is equivilent to a PIII 900mhz, yet I do think that's enough for unRAID. Especially if you go with the dual core proc. I probably would not take off any fans unless I had a very large heat sink to replace it and I was absolutely sure there was airflow across this heat sink from the case fans. Quote Link to comment
wjw Posted November 13, 2009 Share Posted November 13, 2009 How long and how fast did it take to do the new parity calc. average 52kB/s. started around 57kB/s, though during the calc, I did copy over about 100gb of data and then a backup ran in the middle of the night (SyncToy) that "diffed" 60-80gb and copied about 5-10gb. ("diffed" is in quotes, b/c I don't know if SyncToy actually does any comparison beyond mtime, size, etc) What controller are you using for the 5th drive? Rosewill RC-218; all of my data drives are hanging off of this and parity is the only drive hanging off of the mobo. I do believe that I'd have faster throughput it I hung more off of my mobo, but -- frankly, I don't like how mobo manufacturers put all of the SATA heads in the same place making for a potential cable-bump problem. I'm pretty interested in this board for a low power machine. From what I've read the atom processor is equivilent to a PIII 900mhz, yet I do think that's enough for unRAID. Especially if you go with the dual core proc. I actually think it's more than enough, though I haven't done a lot of testing. I'm doing a backup right now of about 1.4T to an external drive [backups, backups, backups] via usb and i/o wait is fairly high (45% or so), cp cpu load is high (30%) and load is retarded (3.0-4.0) [most likely due to the high i/o wait]. I probably would not take off any fans unless I had a very large heat sink to replace it and I was absolutely sure there was airflow across this heat sink from the case fans. it does come with a small h/s, and my case has very good airflow (ref: parity calc drive temps), with a large "wind" across the mobo. again, it's not "doing" anything via the gpu b/c dmesg isn't spooled to console with this config. Quote Link to comment
WeeboTech Posted November 13, 2009 Share Posted November 13, 2009 Respectable numbers for a mini system. Quote Link to comment
wjw Posted November 13, 2009 Share Posted November 13, 2009 Respectable numbers for a mini system. i actually think those numbers are due to the RC-128. in my previous DQ45CB+Celeron config (5 onboard sata connections), i saw 70kB/s+ parity calcs when all drives were on the mobo. when i dropped the drives down to the RC-128, they dropped to the high 50s kB/s. so i don't think the atom is playing a part at all (i don't plan on hanging more drives off of the mobo for my aforementioned cabling desire). Quote Link to comment
WeeboTech Posted November 13, 2009 Share Posted November 13, 2009 It would be interesting to test the motherboard port speed out (hint hint). Also, there could be a difference with the FSB/memory speed. What celeron were you using before? Quote Link to comment
wjw Posted November 13, 2009 Share Posted November 13, 2009 It would be interesting to test the motherboard port speed out (hint hint). Also, there could be a difference with the FSB/memory speed. What celeron were you using before? 430. Quote Link to comment
enternoescape Posted November 29, 2009 Share Posted November 29, 2009 Since it seems related to the NIC under heavy load, another option would be to add an Intel Pro/1000 PCIe card. It is reported to be superior under heavy traffic. You are correct. I have since revised this system because actually noapic seemed to fix the issue, but didn't actually. I happened to have a PCIe Intel NIC and popped it in the case. The NIC issues went away. Then there was another problem. I only have 2 PCIe slots I can use, so I got an Intel PCI NIC. My logic was it's going to be the only PCI device. The PCI NIC does work better than the onboard NICs sadly. I also did some quick benchmarks between the PCIe NIC and the PCI NIC for typical file transfers and got identical transfer speeds. I didn't hit 500mbps once with either NIC which is probably a limitation of the settings on the NIC, not the actual card or bus. The computer I was testing with has an Intel NIC as well that is optimized and I know I can get throughput in the high 800mbps. These are the results I got from copying files from multiple hard drives to a Windows 2003 Server simultaneously: 1 Disk = 280mbps 2 Disks = 420mbps 3 Disks = 490mbps Just in case anyone else decides to use a PCI card (not PCIe) with this motherboard, you should know that unRAID will boot with eth0 and eth1 as the onboard adaptors. This means your PCI adaptor will be eth2 and will not be up or already configured. There is no option in the BIOS to disable the onboard NICs. I added these lines to the go script to make this work for me: # Removes the driver for Realtek Adaptors rmmod r8169 ifconfig eth2 up ifconfig eth2 <ip address> netmask <subnet> ifconfig eth2_rename up ifconfig eth2_rename <ip address> netmask <subnet> I've had the adaptor change names on me before which was confusing, so that's why it references two different adapters. I don't have a gateway defined in my script because my server is strictly for storage and there's no reason for it to know how to get outside my network. You would also of course need to change <ip address> and <subnet> to your unRAID server's IP address and subnet respectively. Quote Link to comment
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.