Hacker News Reader

Comment by skullone | Hacker News Reader

Supermicro is definitely a "you get what you pay for". We bought thousands of servers from their vertical integrations partners, had massive board and backplane problems. Took a few years but they eventually took back over $30 million dollars worth of servers, which were scrapped ultimately because the rework on them was so cost prohibitive. We lost $30M on that even after the $30M in good will refunds. Supermicro also has the lowest bios/efi/bmc/ipmi/redfish out of any vendor we have seen. Just low tier cheap ass shit by a company who can barely survive quarter to quarter without running some new scam on customers, investors, and even governments.

View on HN

Pretty much the same experience (on a much smaller scale). And just open up one of their servers and compare the engineering to a Dell or HPE server. Anything that can be cheaped out is. Corrugated plastic for cooling air channels, FRU assemblies held in place with sheet metal screws, all very bargin basement.

They look cheap even from the outside. They all look like they last went through a chassis redesign in 2002.

I haven't worked with anything at that scale, but the little bit that I was SuperMicro adjacent I was always unimpressed by the "fit and finish" of the entire experience, as compared to Dell and HP. (Having said that, the entire x86 commodity server experience is shitty anyway. I had a brief time, early in my career, when I did work with DEC Alpha machines. Man, they had their shit together. Stuff was expensive as sin, but stuff worked together and worked well. Build quality was tank-like.)

When Compaq servers were still a thing it was the same with those. You could drop them two stories and they'd probably continue playing if the cable was long enough ;)

Oh and you'd get fined for damage to the pavement.

Pretty much. But at one point you could buy 2 to 3 units to every equivalent Dell or HP unit unless you had enough scale to get volume discounts. At $30M I expect the price to be a lot closer though.

Then it’s a matter of how well your engineering/ops org is setup to deal with silly hardware issues and annoyances. Some orgs will burn dozens of hours on a random failure, some will burn an hour or treat the entire server as disposable due to aforementioned cost differences. If you are not built to run on cheaply engineered gear that has lots of “quality of life” sharp edges (including actual physical sharp edges!) then you are gonna have a bad time. Silly things like rack rails sucking will bite you and run up the costs far more than anyone would expect unless you have experience to predict and plan for such things beforehand.

Of course you do have the risk of a totally shit batch or model of server where all that goes out the window. I got particularly burned by some of their high density blade servers, where it was a similar story to yours. Total loss in the 7 figures on that one!

Totally agreed on their BMC/firmware department. Flashbacks to hours of calls with them trying to explain the basics. My favorite story from that group is arguing with them over what a UUID is - they thought it was just a randomly generated string. Worked until one didn’t pass parsing on some obscure deeply buried library and caused mysterious automation failures due to being keyed against chassis UUID… and that’s when they’d actually burn one into firmware in the first place.

It was also always a tradeoff of having to deal with cheaped out hardware engineering with supermicro or with some horrible enterprise quarterly numbers driven sales process with Dell.

> unless you had enough scale to get volume discounts

Volume discounts from the big American brands are at least sometimes available at volumes that are remarkably close to one unit.

That happens when they hope that you will place other bigger orders.

If you are unknown to them, you will not get discounts unless you order a quantity big enough.

This is not consistent with me experience.

If you are buying a material amount of gear, and have your own internal accounting/finance people to establish proper etiquette in purchasing, Dell will never "lose" based on price. As customers, you can say what you're willing to spend on a solution, if you understand the BOMs and all the details and "the business rules" then Dell can be a good enough partner. We operated a hundred petabyte sized warehouse, and did it at the fraction of the cost of cloud services even factoring in engineers, networking, power and redundancy and DR.

I've ordered from Dell and Supermicro (and others, but not in volume or often enough to have much data) for a couple decades now.

Supermicro you consistently get a good/best price. It's already pretty low, so going from a 10 unit order to 1,000 unit order gives you some discount but nothing crazy most of the time.

Dell it's basically based on phase of the moon. I typically tried to time my smaller purchases to coincide with end of quarter. Wait for my rep to call me and ask if I had anything for him. If my little 6 unit order helped them or their boss hit whatever target you could get amazing deals. Stuff where they wouldn't budge 2 months earlier could be had at below supermicro pricing. More than a few times they would give such ridiculous pricing that I considered ordering way more than I needed to part out the components on eBay and just recycle the sheet metal and motherboard.

Other times/years (and different reps too!) would give a budgetary quote for a build and say they'd match a Supermicro quote which was nice, but took extra work and never feels great when you know you're using a vendor just to get pricing down with the one you actually want to go with.

I just hate dealing with that junk, so I tended to prefer Supermicro to reliably get decent pricing when I needed the gear vs. having a whole strategy around order timing.

Then you get into stuff like hard drives/SSD and that was a whole thing w/ Dell for quite some time. These days they are competitive but they were not always.

Right, and it's always fun trying to figure out which chassis with serial number 123456789 is the one you are taking to.

God, SuperMicro... never change. (But also, please go bankrupt and stop making trash.)

Anyone know if MSI are better?

No, MSI is not better. It pains me to say it, but Dell has the best post-purchase support on data center gear. Their servers "just work" for commodity needs, even into HPC and custom gear. Once you exceed that, there's a half dozen opencompute vendors that will go deeper with customizations and requirements.

Reply on news.ycombinator.com