Many companies make motherboards. As far as I know the big computer companies such as Dell, HP, Compaq, Gateway, etc. don't make their own motherboards, but buy somewhat customized ones from the motherboard companies. Generally, motherboards are pretty boring. They are a big board that lots of stuff gets plugged into. Well, there are a few reasons to care about the motherboard.
I have used many brands of motherboards and I have found some companies make very reliable boards, with good support. By far the best company for support and documentation is Intel. They have hundreds of pages of documentation for their boards. They boards may not have many features, but they are documented and well supported. If there is a new BIOS, you can find out exactly what fixes are in the BIOS. Other good motherboard manufacturers are Asus, Tyan, Supermicro. Others have a reasonable amount of support, and work reasonably well, such as Abit. Some have very poor support such as Iwill (which no longer makes motherboards for individual sale).
One of the biggest motherboard companies is Foxconn. They make boards under their own name, as well as for Dell, HP, and no doubt others. I have heard they make MSI motherboards, but I am not sure about that. They have a BIOS in some of their motherboards that is known to be poorly written, and to not support Linux, despite claiming ACPI support (which is operating system independent). What is worse, they don't seem to care that their BIOS is poorly written, and that it isn't ACPI compliant. See details here and here
Purchased mar 2009. It supports AM2, AM2+, and AM3 cpus. The reason I bought it was it supported ECC memory, including memory scrubbing and chipkill. It has 'solid' capacitors (conductive polymer), express gate (a small linux that boots in 5 seconds), Q-Fan2 (intelligent fan speed control), onboard video with DVI and HDMI, and the usual features. Overall, it is a very good motherboard.
There are a few things I don't like. SATA is a hot swap technology. If you want to enable the hot swap feature you have to set the 'OnChip SATA type' to ACHI. Though it isn't documented in the manual, doing this makes only the first 4 SATA devices visible. This 'feature' is documented in the BIOS. If you use a windows driver, you can see the other 2 devices. You can't boot the computer off of them, however. If you use something other than windows, you are screwed. I consider this unacceptable. There are 5 SATA ports on the motherboard, and an e-SATA port on the back. Not being able to add or remove a e-SATA drive without rebooting is unacceptable. At least this stupidity should be documented in the manual. Another small nit is there is only one PS/2 connector. Nobody uses them anymore, right? Well some people have KVM switches that are PS/2 based (like me). I ended up buying a USB to PS/2 adapter. Some don't work, however.
In 2014, I started having odd issues with the motherboard. Most of the disks attached to the motherboard are SATA, but the boot disk and optical drive are PATA, as I have lots of them. I am 99% sure that the PATA controller has partially failed. I can read cd's and dvd's with the optical drive, but I was having a hard time installing linux on the PATA hard drive (which has worked in the past). A little later, the BIOS failed to recognize PATA hard drives. I tried 4 different PATA drives. I can't easily access the motherboard directly, but the optical drive on the PATA bus works fine. After I realized what the problem was, I successfully installed several operating systems on the new SATA boot disk. The motherboard is 5 years old, and everything else seems to be working fine. It is certainly an odd failure.
Purchased Jul-2015 after my M3N WS motherboard failed. It is a AM3+ micro atx board. It has what I need, which is ECC memory support, AM3+ CPU support, 6 SATA-2 connectors, and a PCI-E 2.0 x16 and onboard graphics. The x16 slot is for a Highpoint 2720-SGL SAS controller. It is a very boring motherboard, which I view as a good thing. The only downside I have noticed is the 3 pin CPU fan that I have doesn't seem to spin down very much under low loads. I suspect things would work much better with a 4 pin PWN fan. It would be nice to have more PCI-E slots, but I suspect that is limited by the cheap, old chipset. I paid $54 dollars for it, and had a $10 rebate. It is hard to imagine a less expensive motherboard that supports ECC memory. There is a really cool low power embedded 8 core Intel atom that supports ECC, but with the motherboard it costs around $400 (the ASRock C2750D4I). There is also a low power embedded 4 core Intel atom in the Asus P9A-I motherboard which includes 2 SATA-3 ports and 4 MiniSAS connectors which supports 16 drives, but it costs $380. If price was no object, I would probably get the Asus P9A-I motherboard, but the M5A78L-M does the job and is far more flexible.
This is my file server motherboard. It supports AM2, AM2+, and AM3 cpus. The reason I bought it was it supported ECC memory, and has a PCI-X slot (64 bit PCI). It has 'solid' capacitors (conductive polymer), Q-Fan2 (intelligent fan speed control), onboard video with DVI and HDMI, and the usual features. Overall, it is a decent motherboard.
There are a few things I don't like.
I bought this motherboard Aug-2009 for use in a fileserver. It has a PCI-X slot, which my 8 port sata controller needed. It worked until Jul-2015. No obvious errors while running Windows XP. However, running Linux if you start X-Windows, the computer freezes. It is quite repeatable. Even without starting X-Windows, when using it as a fileserver and copying a 50gb file, it repeatedly failed during the copy. Fortunately for me, I had a Phenom II 710 processor, which is socket AM3, and will work in a new socket AM3+ motherboard. It did have pretty heatpipes which were quite sharp and good at cutting things like wires and fingers.
I had read about a tour of ECS and how they had improved their motherboard manufacturing process. When I needed a new motherboard, I decided to get another ECS motherboard because it was very inexpensive. This motherboard was a basic modern motherboard, with SATA, a 24 pin ATX power connector, DDR-2 dram, and the like. It didn't have gigabit Ethernet, nor did it have firewire, but I didn't really need those features. I added an Intel processor, and two sticks of memory, and booted it up. I usually run memcheck+ for a while on new hardware, and I did so without incident. I added an Asus video card, and booted the computer. When I loaded up windows to check that all the hardware was working, I got a blue-screen of death. Using the Microsoft crash dump analysis tool, I found that an Asus program that ran in the kernel was causing the crash by trying to page in a non paged area. I contacted Asus and they told me to remove a stick of ram. I was quite dubious. Nonetheless, I removed a stick of memory, and was unable to get the computer to crash. I added the stick back in, and ran memcheck+ for several days without any problem. I now believe that the memory wasn't seated well, and was causing problems. I ended up running some cpu and memory loading programs for several days and the system didn't crash.
One odd thing, was the device manager showed one unknown PCI device. I had installed the motherboard software cd, so I was puzzled. It turned out that what is necessary is to have the motherboard cd in the computer, and to click on the unknown device. When I did that, It found some kind of high performance audio processor. It installed a driver, and seven more unknown PCI devices popped up. I had to click on each device, and using the motherboard cd, a driver was installed for each unknown device. It was a bit time consuming...
I bought this motherboard as a combo with an AMD Athlon-II-250, for a screaming price of $50. I knew it was a very basic motherboard, but I bought it rather than an ECS motherboard (at the same price) because of the reputation of Gigabyte. The motherboard supports AM3 processors, and DDR2 memory. It is a micro-ATX motherboard with one PCI-E x16 interface, one PCI-E x1 interface, and 2 PCI slots. It has onboard nvidia video, audio, Ethernet, and usb connectors. The back panel features a parallel port, a serial port, two PS/2 connectors, a VGA socket, and 3 audio plugs. It is nice to have some of these older style connectors, as some peripherals that I have still use them (like my KVM switch, my printers, and my UPS). If you want the rest of the audio ports, you have to buy a backplate. This isn't a big deal, as it is an entry level motherboard. It would be nice if there was gigabit Ethernet, and a DVI port, but I have cards to support those ports. There is a PATA connector on the motherboard, and only 2 SATA connectors. The BIOS is pretty basic, but does the job.
There are a few issues however. For some reason the BIOS disables SMART by default, though you can enable it. In the BIOS you can set the memory to 'ganged mode', which uses both memory slots at the same time to access memory. This doubles the memory bandwidth. As I have 2 identical 1gb memory sticks, I enabled this mode. When the motherboard boots, it reports the memory mode, and reports the memory as un-ganged. I contacted Gigabyte support, clearly explaining the problem. Their reply seemed to not understand my message, and said I had to have 2 DIMMs that were DDR2, and stuff like that. I replied that I had mentioned that I had that in my first message to them. They next told me to flash to the latest BIOS, and kindly included a link to the BIOS. Unfortunately, they pointed me to a BIOS for a different motherboard, despite my first message plainly stating the motherboard I had. Fortunately, the BIOS flash utility told me it was the wrong BIOS. I updated to the newest BIOS for my motherboard, but it had no effect on the 'un-ganged' display during boot. I can excuse the poor english of the replies, but their support seems quite confused, and has not been able to fix the issue so far. I am used to much better technical support from motherboard companies.
I now have exchanged 8 messages with tech support. I ran cpuid which verified that the memory is not in ganged mode. They now understand which motherboard I have. They seem to think that the memory isn't standard ddr-2 memory even though it is. The memory is HP labeled memory, made by Micron, and the exact pert number is 377726-888. Gigabyte suggests that I test the mb with different memory. I told them that I was happy to, and supplied my address so they could send memory. The claimed they couldn't do it, and guess that my memory isn't 1.8v memory even though it is. Now there is some 'overclocking' memory that requires more voltage to work reliably. I told them that I ran memtest86+ for about 24 hours without any problems. This doesn't prove the memory is perfect, but it is pretty strong evidence that there aren't any serious issues. I always do this when building a new computer so I know the basic mb, cpu, memory and ps are ok. I fear Gigabyte is pointing the finger elsewhere and seem unwilling to deal with the issue, or even admit that their motherboard or BIOS could be less than perfect. I won't be buying any more gigabyte motherboards, nor will I be recommending that anyone buy one. I am truly saddened, as I thought they were a first rate motherboard company.
Well, I dug out some other DDR-2 memory I had, and the motherboard acted the same way. I told tech support, and they finally said I should exchange the motherboard. I suspect the problem is the BIOS rather than the motherboard. As the motherboard is in a case, it is a hassle to remove it for exchange. I think I will hold off doing it, and hope a new BIOS fixes the issue. I also read that ganged mode wasn't so much faster than un-ganged mode, so it isn't as important as I first thought it was. It is unfortunate that tech support required me to find some other memory, and it took 8 messages before they suggested exchanging the motherboard.
Purchased in Jul-2016. This is a micro ATX motherboard for AM3+ processors. It has built in AMD graphics (from the chipset not the CPU), 2 memory slots, a pci-e x16 slot, a pci-e x1 slot, and a PCI slot. Being a newer motherboard it has 2 rear USB 3 ports as well as 2 internal USB 3 ports. It uses a traditional BIOS which is a bit odd given how new it is. So far everything is working smoothly. The manual is really quite terse however. For example, there is a JSPI1 9 pin connector which is documented on the motherboard layout on page 13. However the text makes no mention of it all. No pinout, no purpose. What is worse is the lack of BIOS documentation. Each menu in the BIOS is not documented. For example, there is no documentation on the graphics memory settings. It turns out that the graphics on the chipset is defaulted to the 'auto' setting. You can set it to something else, such as 128mb. What this means is that if you use a modern OS, and you have 8gb of RAM, then by default over 4gb of RAM is allocated to something else other than OS usage (which I presume is graphics usage). After I set it to 128mb, the OS reported 7.9gb of memory was usable. If you search for 'graphics' or 'video' in the manual, you will find nothing. I know paper is expensive, but bits are free and it should be documented. It is curious that there is paper included to show how to install the CPU heatsink, but it shows an Intel heatsink, not an AMD heatsink.
This is an E-ATX server motherboard. It accommodates two Xeon processors and up to eight fbdimm memory sticks. I put in two L5420 xeons, which are the low power type. Each xeon has 2 dies, and each die has 2 cores. They run at 2.5ghz and Intel's Thermal Design Power is rated at 50 watts per xeon. The motherboard uses the Intel 5000P (blackford) chipset. I am pretty sure this was the last generation of Xeons that hung the memory off of the chipset, rather than directly connecting the memory to the processors. Of course, as AMD realized, attaching the memory through the chipset shows down memory performance. In order to drive lots of memory sticks, Intel came up with the clever idea of fully buffered DIMMs (FB-DIMM). Each DIMM has a buffering chip and each DIMM draws roughly 10 watts at idle. With eight 2gb sticks, I measured an idle power consumption of about 180 watts using an Antec EarthWatts 500w 80+ power supply. When I reduced the memory to four 2gb sticks, the power consumption went down to about 140 watts, measured at the wall plug. This is using two L5420 low power Xeons. Having memory take so much power is very bad in a datacenter, and is also very bad in my home environment. No doubt this is why Intel got rid of FB-DIMMs. The rest of the board is a typical server motherboard with 2 pci-e x8 slots, 1 pci-e x4 slot, 2 pci-x 133mhz slot, 1 pci-x 100mhz slot, a PATA connector and 6 SATA connectors. The Xeon heatsink are designed to bolt to the server chassis, so there are big holes in the motherboard. It does provide 8 cores at 2.4ghz, which is quite nice. Of course, I got a good deal on the motherboard, memory and processors.
Well, the Supermicro monitoring utility does show CPU temperature, but not the FB-DIMM temperature. Since they run so hot, they each have a temperature sensor. Fortunately, gkrellm and lm_sensors on linux will show the temperature. Using two 120mm exhaust fans, my four FB-DIMMs were running just below 80C which I consider too hot. I switched the two fans for powerful Panaflow fans, and the temperature dropped to about 75C (at idle). That was still hotter than I was comfortable with. I fabricated a shroud for one of the 120mm fans to cool the four FB-DIMMs. The shroud is made out of cardboard and held together with tape. It is also taped to the fan and the case. The small end on the right side of the screen lets air in, and makes it flow over the FB-DIMMs. With the shroud in place, the two central FB-DIMMs run at 66C, which is hotter than I like, but far better than before. Also note the small fan on the chipset heatsink. I added it, as the chipset also runs hot.
This is an ATX server motherboard. It is about 0.2 inches wider than the ATX standard, so it will fit in most ATX cases. It accommodates two Xeon processors and up to 6 sticks of ddr3 registered or unregistered memory. If you use registered memory, you can use 16gb sticks. Unlike the Supermicro x7dbe, the memory hangs directly off of each processor. This greatly increases bandwidth. More important, the motherboard uses normal ddr3 ECC memory, not the power sucking FB-DIMMs, which will literally burn you. I bought a used motherboard along with two L5520 processors, which have 4 native cores and hyperthreading.
My motherboard arrived DOA. It seems that the seller on ebay said it was working, and had a 30 day warranty. They also said that they would ship me a replacement motherboard which never arrived. I made a complaint to ebay, and ebay found in my favor. They said, just send it back with a tracking number, and after we verify that arrived, you will get your money back. Sounds reasonable, right? Unfortunately, the motherboard cost me $150 or so, including shipping. Shipping it back with a tracking number would cost me about $215 (because it was overseas). If I did that, then I would lose about $65. That is not such a good deal. I recommend not buying stuff when it would expensive to ship it back.
So ebay won't help me get my money back from the seller. It turns out that I used paypal to pay for the item. I also cleverly used american express to pay paypal. I contacted amex and disputed the charge. I explained that ebay wanted me to ship it back, which would cost more that I paid. Amex is much more user-friendly than ebay/paypal. I recommend that if you use paypal, then you should always pay with a credit card in case there is an issue with your purchase.
Fortunately Supermicro has great tech support as well as a great warranty. When you call them, you get a real person who understands problems and can make sensible suggestions. The motherboard was covered under warranty, so I am waiting for it to come back fully operational. Well it turns out that the Supermicro warranty does not cover physical damage. The cpu1 socket pins are damaged. I was very careful putting the cpu in the socket, so it is likely that the socket was broken before it was shipped. The cpu socket has 1366 pins, which are all very small, and it would seem, not too difficult to damage. Supermicro charges $50 to replace it, which isn't very expensive considering they have to unsolder all 1366 pins and then replace the socket itself (which isn't cheap). It turns out that even after Supermicro replaced the cpu socket, the board was still broken. They decided it would be a good idea to send me a replacement motherboard instead. So the board I bought on ebay had two separate problems. I am eagerly awaiting getting a working motherboard.
One odd feature of the motherboard is that it uses proprietary Supermicro heatsinks. This is not such a bad thing, as the motherboard has a very robust backplate attached to it, and heatsinks are cheap on ebay. I bought the 4U heatsinks which have 4 heatpipes and come with a nice 4 pin PWM fan. They are as big as the motherboard will tolerate; in fact you have to mount them 90 degrees from each other, with one fan blowing up and one blowing back. That is what happens when you have 2 sockets and 6 DIMMs on a mostly standard sized ATX motherboard.
Well, the motherboard is installed and running. I found a few issues. The first one I noticed is that 2 mounting screws are in non-standard positions. This was an issue, as the standard location screws were permanently attached to my Antec p180 case. I had to use a bit of force to remove them. They are the upper center screw and the central screw. The motherboard has holes for these, but in non-standard locations.
The bigger issue is the northbridge heatsink. It is the largest heatsink on the motherboard. Noticing it was running warm, I added a 60mm fan using some drywall screws. Problem solved - until I added a video card to the pci-e x16 slot. The first issue was the heatsink got in the way of the silly tab at the back of the graphics card. The tab was easily removed. The big problem was my fan got in the way of the graphics card. Since the on-board graphics is a 32mb matrox solution, I really wanted a graphics card. I picked a small, cheap, passively cooled one, but it still overhung the northbridge heatsink. I removed the 60mm fan. I also added a host based adapter (hba) on one of the x8 slots. It also overhung the northbridge heatsink.
I tried to run the Supermicro 'super-doctor' program, but it only works with windows xp. So I installed linux and ran gkrellm. The northbridge idles at 80C, and gets hotter under load. I decided I needed to cool it. I thought a bit and there was no obvious way to mount a fan of any size in my case. After much thought and hacking, I added a 40mm fan that hangs off the edge of the northbridge. It is quite ugly, but it reduces the temperature by 20C. Under load, I have seen temperatures of 73C with the fan running. If the board is to be used in a server case with a lot of airflow, it probably won't be a problem, but in a normal case, the northbridge runs much too hot for my liking. It is cleverly close to the 4 pci-e slots, which means the heatsink has to be very low profile. Ideally, it would be moved back about 2 inches so it wouldn't get in the way of the pci-e slots. I am sure there is a reason it is where it is, though it gets in the way of the pci-e slots.
With my two L5520 CPU's, two 16gb sticks of memory, 4 hard drives, a platinum 450w power supply, and a bunch of cards, the system idles at about 100w under mageia 3 or knoppix linux. With windows idle is around 120w. That is acceptable to me, but not ideal.
The 'sleep' mode for the motherboard puts the CPU's in S1, which isn't a very deep sleep. It also keeps the cpu fans and the chassis fans spinning. Though the manual says 'CPU fan auto-off in sleep mode', it is incorrect. Supermicro says their 'workstation' motherboard will put the CPU's in S3, but it is an E-ATX motherboard. Another 'feature' is sometimes the motherboard does not come out of sleep mode, and I have to power cycle the computer. I think that sleep mode isn't well thought out for this motherboard. I can 'hibernate' the system, but that means writing out as much as 32gb of ram to disk. When powering up the system, all that data has to be read from disk and copied back to ram. Not very quick.
For some reason, the motherboard doesn't like my Belkin OmniView Pro2 KVM switch. (This isn't a consumer KVM, but a rack mounted unit with on screen display and configuration.) My kenboard and mouse work fine after the bios finishes, but I can't enter the bios using a keyboard attached to the KVM. Other motherboards work fine. Supermicro says this is a known issue with some KVM switches, and directly attach a keyboard to the motherboard. This works, but is less than ideal. Either a PS/2 or USB keyboard works fine when directly attached.
Well, after about 3.5 years, the motherboard died. I ordered a replacement from ebay. The price has gone down, as they are now older. $50 shipped including 2 nice passive heatsinks for a 1U server. It was dead. The seller is a large seller of used supermicro stuff, and sent a replacement board which works great. I did learn a few things. You may have to clear the cmos memory to get the motherboard to boot. Never needed to do this with another motherboard. Pull the power cord, pull the battery, and short some terminals on the mb. Also if you want to use a video card, you have to un-jumper the onboard video. The mb manual says how to un-jumper the video but doesn't bother mentioning that external video cards won't work unless you do that. Another thing I don't really like is the board takes roughly 30-40 seconds after powering up to display any video at all. This can be a bit frustrating at times. Overall, a great board.
While debugging the dead new motherboard, I bought two 6 core Xeon 5640L (low power). They were about $50 total from ebay. Screaming fast if you use multi-threaded software. I also ended up using some generic DDR-4 memory (which the board supports). As I suspected, my old memory was fine as well as my old processors. If you want to build an inexpensive computer - $50 for a mb, $50 for processors, you can add some active cpu coolers for about $30, and 32gb of ECC registered memory for roughly $100, and end up with a screaming fast computer.
This is a E-ATX high end server and/or workstation motherboard. It supports two 200 series socket 940 dual core Opteron, 4 sticks of memory for each socket, dual PCI-E 16 bit sockets (full width), a PCI-X socket at 133mhz, two PCI-X sockets at 100 mhz, and a PCI 33mhz socket. It has the usual firewire, usb, dual gigabit Ethernet, and the like. There are only 4 SATA II connectors and one UDMA connector although the chipsets (for the two CPU's) support eight SATA II connectors and two UDMA connectors. It also has optional dual channel ultra-320 SCSI, which my motherboard doesn't have :-(
I bought this motherboard, processors and memory used from a reputable source, along with two other motherboards. The other two motherboards worked fine. This one was DOA. I know it was working before it was shipped. I tested it thoroughly. When it booted, all it would do is make a BIOS beep code of 2-2-2-2-2. Tyan didn't know what the code meant (which I should have realized was a bad sign), and they eventually said send in the board, along with $75 to fix the board. They returned it fixed, saying the BIOS was corrupted and one heat sink was loose.
I powered up the board, and it shut itself off in a few seconds. No matter what I did, it wouldn't boot. I was perplexed, as I presumed they tested it before they shipped it to me. There is no way it could have passed any testing. I told them about and sent it back. This time I didn't have to send them any money (other than my shipping costs). At least this time I didn't take my old computer apart first to mount the motherboard.
They said the board was indeed broken, and they sent me a different board. I know it is a different board, because the first one had a 6 pin mystery power connector that had the same mechanical pins as a 6 pin PCI-E video card power connector, though different electrical wiring. Tyan admitted this wasn't a real good idea, and they removed them on newer motherboards. This motherboard did not have the 6 pin power connector. I powered it up (I learned from before not to remove my old motherboard first), and it did... Nothing. No fans spinning at all, no nothing. 100% dead. I have no idea how Tyan could send out a board untested, and 100% broken. I will contact them as soon as possible, and ask them to resolve this issue.
Well, I sent the replacement board back, and had them cross-ship me a new replacement motherboard. They did a fine job cross shipping me a board, but it also seemed to be dead on arrival. I called their tech support. They asked me what type of power supply I had. I told them it was a Seasonic Super Silencer 460w rev A3. It has a 12v 25a output. He suggested even though it has a 8 pin SSI connector, it might not be SSI compliant. I thought he was nuts. He suggested trying another power supply. I took an Antec EarthWatts 500w power supply from my fileserver (and replaced it with an EarthWatts 380w with a 4 pin to 8 pin SSI adapter), and tried it out. The Antec has two 12v rails, each one rated for 17a. The motherboard booted successfully. I guess my Seasonic didn't have enough current (it also had 6 hard drives attached to it, which can suck a fair amount of current at 12v during startup.) I now suspect the motherboard I recently returned to Tyan worked fine, though not with my Seasonic power supply which has inadequate current at 12v. I really respect the Tyan tech support folks, for helping find the source of the problem.
It is interesting that the EarthWatts 380 works fine on my Asus NCCH-DL system, with two lv 2.8ghz Xeons, as well as 7 hard drives (without any delay spinning up while booting). Also the Seasonic works fine on my Asus PC-DL with two lv 2.6ghz Xeons, as well as 6 hard drives. The Opterons are rated as having a TDP of 95w each, which is much higher than my lv Xeons (@40 and @55 watts)...
If you have comments or suggestions, Email me at firstname.lastname@example.org
Created with gnu emacs and template-toolkit, not some sissy HTML editor.
both have significant security issues.