Why Supermicro for high end workstation?

Tengis

Supreme [H]ardness
2FA
Joined
Jun 11, 2003
Messages
6,578
I have a coworker that swears by Supermicro. We are building some Corei7 boxes and he claims that Supermicro is the only company that puts their designs through rigorous tests - give me a break :rolleyes:. We are having trouble sourcing a particular barebones by Supermicro so I suggested building our own boxes... he refused.

Ive NEVER had issues using quality motherboards and powersupplies to build systems before and have had just as many failures here and there with Supermicro as I do with anything else. This brought me to a question I couldnt find an answer to by Googling: As far as desktop/workstation level boards and/or barebones go - why would you pick Supermicro over anything else? It costs more and doesnt necessarily offer you anythiny extra.

Strictly talking high end desktop/workstation.

Thoughts?
 
The main reason for using supermicro is because they have alot of serverboards and dualsocket boards. As for their single socket boards, i really havent seen much of a difference in those. But its mainly just the fact that they only do workstation and serverboards is what makes them seem special other than tha i also look at tyan for serverboards as well.
 
The main reason for using supermicro is because they have alot of serverboards and dualsocket boards. As for their single socket boards, i really havent seen much of a difference in those. But its mainly just the fact that they only do workstation and serverboards is what makes them seem special other than tha i also look at tyan for serverboards as well.

I can understand the need for duel or quad processor boards and going with Supermicro but I shouldve mentioned that I was talking about single processor.

When comparing a Supermicro single processor barebones to a high end Asus board with a nice Corsair powersupply - why would you pick Supermicro? Assuming there is not a hardware failure then reliability/stability should be the same but you get more features, options, and better pricing by building your own.

Unless I was looking for something specific that only Supermicro carried I just dont understand why someone would go to Supermicro and spend more money to get less.
 
Having dealt with their support, it all kinda makes sense. You can't beat dealing with support that actually knows their stuff (and when they didn't know an answer, they consulted one of the engineers...at least that's what my experience was like) and is based out of California.
 
It's more of a reputation thing at this point as you wouldn't have problems with 99% of the motherboards out there. You might get a bad Supermicro board, and you might get a fantastic Asus or Tyan board for that matter.
 
I personally shy away from Supermicro boards -- pricier most of the time and I don't see any real quality/stability improvements vs. other boards. I have a rack full of Tyan boarded dual P-III, dual 940 Opteron and dual 1207 Opterons. Have not had a single issue with any of them at all. But I don't use SCSI on most of them or any higher end monitoring features, *shrug*.
 
I wouldn't bother with Supermicro for a desktop, but it's my first choice for a server or workstation (assuming you need a 2p or 4p system).

From a manufacturer's standpoint (I work for an OEM), Supermicro offers/excels in the following:
  • Product availability - frequent problems with supply/demand from other manufacturers, but not so much with Supermicro
  • Product longevity - Supermicro typically manufactures the board until the chipset is EOL. Most other manufacturers EOL their current lineup and make a new lineup (or make new revisions) whenever they feel like it. Platform stability is very crucial to most of our customers and therefore is a major factor in the decision to use Supermicro.
  • Support - Supermicro support is phenomenal from my experience at work.
  • Integrated component quality - For example: you won't see any Realtek crap NICs on their boards (in the server/workstation lines); Intel 82573 and 82575 NICs 4tw!

There are more, but those are the first to my mind.
 
Integrated component quality - For example: you won't see any Realtek crap NICs on their boards (in the server/workstation lines); Intel 82573 and 82575 NICs 4tw!
Even their AMD boards use Intel NICs (like mine). :)
 
Even their AMD boards use Intel NICs (like mine). :)

While I'm not terribly familiar with the AMD boards (like 99% of what we sell is Intel), I do know that some of their boards use the Nvidia NICs (for the ones that use the Nvidia chipsets) and some also use the Broadcom BCM5704.
 
While I'm not terribly familiar with the AMD boards (like 99% of what we sell is Intel), I do know that some of their boards use the Nvidia NICs (for the ones that use the Nvidia chipsets) and some also use the Broadcom BCM5704.
Some of their boards are really screwy. Mine uses AMD processors, nVidia chipset, ATI video, and Intel NICs.
 
Some of their boards are really screwy. Mine uses AMD processors, nVidia chipset, ATI video, and Intel NICs.

I think the ATI video (ES1000) is pretty much standard across all their server/WS boards. While that is a crazy combo (brand wise), I think it's also a good combination of parts (yeah yeah some people hate on nvidia chipsets...whatever).

I have the X7DBE in my fileserver (soon to be my virtualization server) and love it :)
 
Only thing that really sucks on mine is the nVidia RAID (not that I use it). I have a H8QME for my video encoding workstation. File server didn't need anything as extravagant, so I just have a single socket Intel board.
 
I only user SuperMicro and Tyan if i need a very specific set of components on the motherboard. Otherwise it doesnt make much of a difference. They are usually highly specialized boards that have specific connectors in places that are easier to work with or they support certiain peripherals, etc. but if those are not necessities i would much rather go with a consumer board.
 
I like the fact that they use Intel NICs, and also that they provide support for ECC memory even on uniprocessor boards. They also have some decent bare-bones systems, but they tend to be too noisy, even the supposedly quiet ones, and the power supplies they use aren't great.

Beyond that, I don't see much reason to go with them. In my experience, they cut corners on the uniprocessor boards - I like their dual processor boards better. Their support for individual buyers is not very good. It may be good for OEM partners or large customers (I have no experience with that), but not for individuals.

By the way, their uniprocessor boards are now made in China. I'm not sure where they make the dual-processor boards these days.
 
they do have more stability then a high end Asus board because they dont let you overclock on most of them, thus eliminating the chance for problems.

it is like asking why buy an Intel board vs an asus board, they are made to work out of the box, no tweaking and such, they do get tested rigerously (sp?)
 
they do have more stability then a high end Asus board because they dont let you overclock on most of them, thus eliminating the chance for problems.

Actually, in some sense it's the opposite of what you're saying. ASUS, Gigabyte, EVGA, etc. need to have a more robust design and better components to support overclocking, while Supermicro can get away with a very minimalist design.

This is obvious in their new Core i7 uniprocessor boards, for example. The VRM design and components are clearly geared toward minimizing cost, and are inferior to those of some competing boards. Supermicro didn't even bother to put heatsinks on the VRM components of these motherboards. This is unlike the case with their dual-CPU boards which have heatsinks on the VRMs, and also unlike all other Core i7 motherboards from other manufacturers - they all have heatsinks on the VRM components, and for a good reason - this stuff runs extremely hot - I wouldn't recommend touching anything in the VRM areas of the Supermicro Core i7 boards unless you want to get a nasty burn - this can't be good for the longevity of the components or the board. Maybe it's within spec strictly speaking, but do you really want to run any electronic components regularly at 100C even if it's within spec? I'm also not happy with their choice of components there.

Also, try putting any memory over 1333MHz in those boards. Significant issues even though 1600MHz is supposed to be supported.

Their boards have their advantages as I mentioned (Intel NICs and ECC support), but I really don't know what people are talking about in terms of quality and stability. If anything, their uniprocessor boards are at a disadvantage there in my opinion.
 
Another thing I don't like about Supermicro Core i7 motherboards is that they support fan speed control only for 4-pin PWM fans. Previous Supermicro motherboards had both voltage control (for 3-pin fans) and PWM control, but they removed the voltage control. The Intel Core i7 board has both, albeit using different fan headers.

If ASUS or Gigabyte started using Intel NIC chips and had ECC support, I'd probably start using their boards. As it is, I use Supermicro and Intel boards.
 
Actually, in some sense it's the opposite of what you're saying. ASUS, Gigabyte, EVGA, etc. need to have a more robust design and better components to support overclocking, while Supermicro can get away with a very minimalist design.

The X (for Xeon) series boards are NOT desktop boards, but rather they are workstation and server boards. This explains the lack of overclocking type settings in the BIOS. The C2 series are desktop boards, but I couldn't tell you if they have any clock adjustment BIOS settings or not.

This is obvious in their new Core i7 uniprocessor boards, for example. The VRM design and components are clearly geared toward minimizing cost, and are inferior to those of some competing boards. Supermicro didn't even bother to put heatsinks on the VRM components of these motherboards. This is unlike the case with their dual-CPU boards which have heatsinks on the VRMs, and also unlike all other Core i7 motherboards from other manufacturers - they all have heatsinks on the VRM components, and for a good reason - this stuff runs extremely hot - I wouldn't recommend touching anything in the VRM areas of the Supermicro Core i7 boards unless you want to get a nasty burn - this can't be good for the longevity of the components or the board. Maybe it's within spec strictly speaking, but do you really want to run any electronic components regularly at 100C even if it's within spec? I'm also not happy with their choice of components there.

We have some X8 series samples at work (X8STi for sure). I can see how they do without heatsinks, temperature wise (I keep a cheap Radioshack IR thermometer at work).

Also, try putting any memory over 1333MHz in those boards. Significant issues even though 1600MHz is supposed to be supported.

Supermicro only specs up to 1333MHz memory. Additionally, you should follow the tested memory list they provide.
 
For a high end intel based desktop/workstation supermicro would be high on my list. I'd also consider barebones intel setups.

All comes down to support. I've seen some big issues with asus workstation/server boards and just don't see them as the same as supermicro/intel.
 
We have some X8 series samples at work (X8STi for sure). I can see how they do without heatsinks, temperature wise (I keep a cheap Radioshack IR thermometer at work).

Try measuring the temperature of the MOSFETs and inductors while running a CPU stress test like Prime95, but make sure you don't touch them with your finger - I made that mistake once - I'm running a 130W 3.2GHz Core i7 965 which really puts some load on the VRM.

Supermicro only specs up to 1333MHz memory. Additionally, you should follow the tested memory list they provide.

The C7X58 is supposed to support 1600MHz. Similarly, the X8SAX was originally claimed to support 1600MHz, but the mention of that was removed recently (although they still have 1600MHz in the memory compatibility list for that board). The X8SAX and C7X58 are actually same board with the PCI-X slots and bridge depopulated in the C7X58. Also, the C7X58 has the nVidia SLI key, while the X8SAX doesn't.
 
Try measuring the temperature of the MOSFETs and inductors while running a CPU stress test like Prime95, but make sure you don't touch them with your finger - I made that mistake once - I'm running a 130W 3.2GHz Core i7 965 which really puts some load on the VRM.

Test setup:

Supermicro X8STI-F
i7 920 ES
Hacked up Supermicro 512F-520B

I had booted the system into our Gentoo PXE client and stressed all 4 cores (8 with hyper threading):
stress.jpg


Used this to measure the VRM temps:
IR_thermometer.jpg


And here are the results:
X8STI-F_VRM.jpg


There are 3x VRMs along the right side of the heatsink that you can't really see in the pic if you were wondering what it is I boxed on the right side. None of the VRMs are hot enough to burn your finger and in fact even the rear VRM, which ran at about 52C, was only warm to the touch.

Note that I had to remove the air shroud and chassis lid in order to measure temps and that these temps would be lower with the air shroud and lid on. Also the chassis I used for testing is meant for use with a DP Xeon board (the rear set of Xeon heatsink mounting holes are actually drilled out so that UP boards don't ground out) so the fan config could definitely be better. What I'm saying is: This isn't even an optimized config for cooling and the VRMs do not appear to need heatsinks.

P.S. I know I gave stats for the VRMs, but the MOSFETs actually ran cooler than the VRMs so I didn't see any point in posting that up as well.
 
Back
Top