Bitcoin Forum

best aspect of server? Most or all have B.M.C. aka baseboard management control (separate crash-proof CPU that runs out of band) that lets you remotely manage your server if it crashes or freezes. WOL does not work if OS crash. Otherwise you have to go to site to do a maintenance.
Is anyone doing on cheap server from HP, Lenovo or Cisco etc?

For server form factors, there are 3 mostly:
- tower
- rack
- blade.
For the purposes of mining I think blade is out as it is too space constrained and tower and rack is best bet.

i do not have much experience with server motherboard
but it think recognizing more than six gpus may be a problem with them
and even system stability
i will still recommend you to go with atx motherboards for mining

I have only ever seen a few pictures of 6+ GPU server builds , the information regarding comparability  is really hard to find if even up to date.
The best way to learn the specs if good server builds today would most likely be to ask someone doing it.

there are some Intel reference but that is still hundreds of dollars.
I think servers are out of question.

Intel CPU/chips seems to have remote management capability vPro? probably path to explore. However words on the internet that it hosts http which make whole system vulnerable to attack.

Simple solution i see is if there is ANY switch on the market to plug into the PWR pin of the motherboard, daisy chain it and that itself can be controled through IP stack. Then it can at least give remote reset/power on and off capability.
Crazy thought.

I'll just leave this here:

https://i.imgur.com/ssXi3uX.jpg

There's no such thing as a crash proof CPU. An out of band management system has very limited utility.
It's more for routine management than crash recovery. Once a system is up and running stable most
crashes either recover on their own or require a site vist. The only scenario where I see OOB management
being useful is when you need to give access to an expert to remotely troubleshhoot a dead system. That would
probably also involve someone on site to help coordinate.

Im using a HP dl385 gen 7 !!!  and its more then a motherboad.. its on a 2U HP case and power supplies which haves plenty of watts and ILO to manage your server over the ip's 

are you using that to mine also?

Server hardware is too expensive. It will ruin your ROI and gives basically 0 features that another machine does.

If you want to be able to remotely restart crashed rigs just get something that will flip the power off and on. There are plenty of managed power receptacles that do this.

There's plenty of cheap used server hardware out there that you could mine with, and more often than not you can run lots of GPUs with those boards (especially dual socket boards, since there's a lot of pci-e lanes with 2 CPUs). I wouldn't say that IPMI is a big advantage, as was mentioned before it's great for actually using the server, but there are simpler ways to remotely reboot a crashed mining rig. The main downside of server boards and why I personally don't use them is the power consumption (that is if we're talking about older lga2011 and lga1366 boards, which are the only affordable platforms, anything newer is just too expensive) — they're very inefficient compared to modern intel desktop hardware.

i can 24 x cpu and 6 x gpu mine on my own wallets and also its a 42-coin node.  I had the cut out the ends of the pci-e x 8 slots  so the the pci-e x 16 cards could fit in the pci-e x 8 slots .. it works though!!




and I forget to mention , yes its takes a lot of power but you know that was a issue in the first point



I bought my Hp last year used for 300 dollars and it came with 32 G/b of ram and 2 x 160 G/b ssd in raid 1 and 2200 watts power supply.  this computer have been running everyday since i bought it and since it a HP server and HP is king for servers .. you will never be let down since reliability is No 1 and parts are cheap at ebay

and a link on ebay to build and buy the hp dl385 gen7 online.. to give you an idea on cost.

https://www.ebay.com/itm/Build-Your-Own-HP-ProLiant-DL385-G7-SFF-8B-32-Core-2-30GHz-AMD-6276-Wholesale/182933018957?hash=item2a97a8594d:m:mxMJ5b8swrrEZeIDaHAjSbw


most blades only have one pci-e slot per blade..  :(

if you want a server that is more impressive and cheap to buy used is the HP dl585/dl580 gen 7 servers or the all mighty hp dl980

this link on ebay is the first cheapest dl 585 onthe list

 https://www.ebay.com/itm/HP-Proliant-DL585-G7-48-CORES-AMD-6172-2-10GHZ-NO-RAM-NO-HDD-P410i-4x-1200W-PS/282643015997?hash=item41ced6293d:g:Z6cAAOSw23lZsYhP

price for a used is US $339.95 avg and a few hundreds more you can upgrade to your dream server.. It can have 12 pci-e slots and 64 cores cpu and a gig of ram..  this system comes with 4 x 1200 watts of power!!!! but no hard drive or ram :(  you can search in ebay for better deals and parts for your server..  I own a few hp dl385 and im very happy with the 1st year operations and looking forward to invest into a HP dl585 since the software and parts can be transfer from the current setup i have so it makes upgrading alot better ..  keep in mind that gen 7 and get 8,9 ect ect will differ in technologies and the rule is to only gen 7 parts will only work in other gen 7 systems and only the hard drive and video card's can be transfer to newer gen 8 or 9 since SATA and Video Cards on the Pci-e bus havent changed between gen 7 to gen 9.. 

yes there power rails with IP but how about you turn the flip back on and motherboard does not power up?. Being able to reboot will not troubleshoot everything that happens but at least get out of crashed state.

good idea but going into more specific. hp's bmc does not come free. Their update mechanism sucks and it is KVM is based on javascriprt and crashes a lot. However that is with my old HP server. Otherwise HP is pretty good and stable. If  I remember they were charging about 150$ for 2 years for LiO access.

I also used to work with Cisco servers and its bmc is free however aspects I aint got no time to compare. Not sure of other vendors that easy access to bmc. But for sure, HP's really sucked when I used it. If only they made if free it would be almost perfect.

I am thinking to investigate vPro of i-5 series. There might be some promise, I am not sure where it will lead to but I will keep it updated here.

I've done some digging on this vPro thing some time ago, and in the end came to the conclusion that it's not suitable for mining rigs. This tech is supported only by a very limited number of chipsets, it's basically only the rare "Q"-series (and most of those boards, if not all, are matx) and some server chipsets. I just couldn't find any full size atx motherboards with a reasonable number of pci-e slots among those. It was a year ago, but I doubt that something's changed (q370 isn't out yet, and I don't see anything good for mining with q270).

that is one thing, i just bought z270 asus so you think it will not work? I havent gotten yet, because i purchase i3 not i5.
For limited No. of slot may be you can pcie riser card and that should solve the problem.

No, I'm pretty sure it won't work with z270. As far as I can tell, out of all recent intel chipsets, only Q270, Q170, C(M)236 and CM238 support vPro.

and to be honest, power usages is actually low even though the motherboard is on high performance "mode" and that was a trick to learn !! also this is required due to hardware limitations that the compact design of a 2U gives a odd arrangements of pci-e slots on risers boards that allows to use 4 x single slot, full length video cards allotments and 2 x low profile, single slot and half length video cards allotments for a total of the 6 pci-e slots.. due to server case being low profile  Using 4 x XFX single slot 460 and 2 x MSI single slot micro badass 460 would be the best setup to mount them properly and and no external power source is required to plug into these video cards  and that increased power savings efficient!!!  There is a major flaw with with BMC controller that prevents the motherboard from disabling  the onboard AMD gpu . So any AMD Gpu video card that I installed will always become the secondary video dispaly under the windows or linux environment. Even if i disabled the onbaord video in the BIOS setup.. all six gpus plug into a Video screen just goes blank and you can hear the server ramp up the fans which is not good!!  Your forced to reset the jumper on the motherbaord in order for the defaults to set the onboard video back on!! and thats the only way to get all 6 gpu to work under secondary display. if the bios rest didnt work, removing all 6 gpu is the only solution Then by accident after doing a bios reset and cards removal , I install a Nvidia card in the low profile, half length slot and to my surprise The server bios started on the Nvidia screen and the onboard gpu was disable which none of the amd video cards I've install can never do before and the nvidia is actually is siting on a HP/AMD motherboard.   Then I allowed windows to install the nvidia drivers and then asked to restart windows which i allowed and I seen the fastest reboot and re start like it did never before. Somehow The Nvidia video card did the trick and I installed all 5 x amd with the nvidia being the main dispaly which allow more resources to mine cryptocurrenecy on the amd cards.. to be honest , im actually mining on all gpus and cpus , the fans are low and not ramping up..  when I removed the nvidia card, that caused the onboard video to kick back on and  youll noticed how much slower the server is running under the windows/linux environment and the fans are ramping again..  

damn thats a brilliant quick fix. how did you cut them and was it clean?

dude you aint playing game you dont need gpu to main display, just need its computing power to do a hash. i actually have z270 right now to which i connected my display to onboard video and mining on two gpu. Initially I set one of the gpu to monitor and set in the bios an external video as main but it just burped. My suspicion is usb3.0 connector. So i decided not to bother with this and went for onboard for main display. Actually I dont really need display at all, once it is setup and running, i am using ssh (linux) and rdp (windows) to log on and interact with the miner.
10yrs experience with bios dev and test. AMA and I will try my best! :)

wow hopefully he did cut it just right.

Regarding another throwback on remote management:
I worked with client sometime which is called Amulet Hotkey.
They actually sell a console that sends video and usb signal over the IP stack so you can control it remotely.
However I havent investigated the price, i think it might be pricey. There was something over ebay. I would not call it just solution but perhaps path to exploration.

And I am not sure if those are out-of-band, if i remember correctly, they are.

I'm doing 90% of my CPU mining on dedicated servers, most of them are 10+ years old. Credits(CRDS) can only be mined with a CPU so you're not competing with massive hashpower from GPU's. If you have servers already I would highly recommend you take a look.

check out this single slot video card!!! 

https://www.newegg.com/Product/Product.aspx?Item=N82E16814487391




and if any one is curious, I used a butter knife and heated the tipped till it glow from over the gas range. Used the hot knife to curve out the back end of the pci-e x8 slot like how you insert a card into the slot. straight down and pull out away threw the back. if your quick enough, you can do 2 slots before knife cools down. 

Pretty much all cards with full-cover waterblocks are one-slot. Buying EVGA kingpin for mining is a pretty stupid way to spend money. :)

For mining, what I look for is efficiency, not the power and obviousl they are not necessarily same thing.
This seems pretty decent entry level article, as mentioned in the article, I am looking for a way to find a "sweet spot" where the sol/s OR H/S is maximized per /w. Basically increase Hashes/watt, Sol/watt as much as possible.
http://cryptomining-blog.com/tag/gtx-1080-ti-equihash/

I predict that video cards will shirnk in size and the video cards I use are single slot design which incorporates smaller size to fit in a 2u or half height slot.. Have you ever held a video card that is the size of the palm of your hands and the power of a nvidia 1050t ??  , less power usages  since No power cables are required to plug into video card. I believe my hp server setup is the right direction from future of mining and using 2u space with more power efficiency since im only one who pays the electric bill at my business and its a concern  , half height , half length video cards

Nothing is efficient about the watercooled 1080 Ti Kingpin. :D It's a card made for breaking overclocking records. It's not power-efficient, and it's ridiculously expensive. There are a lot cheaper 1080 Ti's with full-cover waterblocks out there, including some other cards from EVGA as well.

I have. R9 Nano. A lot more powerful than gtx 1050 ti actually.

Is your nano on a half height half lenghth card? Plus that 1080 I posted would be more efficient then a two slot fan cooling version of the 1080

So far most r9 nanos I googled are full height , half length .. Doubled slot .. A waste of slot space ! But it's not half height, single slot ..    I believe that is the way of the future .  

Wacko, this is a true half length, half height single slot card link i found and this card is tiny in size compared to an avg R9, i can't find any R9 at half height , single slot. Can you share more info on your video card??  The R9 is actually very good..  its heading in the direction to be miniaturize

 A half height, half length single slot card

GeForce GT 1030’s TDP is a mere 30W and 10 years ago this chipset would been strong as the Nvidia GT 8800 @ 75/125 watts .. I believe by the year 2020 the power reduction  of the 1080 will be tomorrows 1030



https://www.newegg.com/Product/Product.aspx?Item=N82E16814125972&cm_re=1030-_-14-125-972-_-Product


 A half height, half length single slot card @ 75 watts AMD rx460

https://videocardz.net/msi-radeon-rx-460-4gb-lp/



A half height, half length but double slot card


 https://wccftech.com/msi-gtx-1050-ti-low-profile/

You either have some serious TDP adjustments, or have a 999999 CFM fan in there. >.>

Whats the temps?

Right now i appears to be heading to other direction. More connectors, bigger size.

I building one now all i need is the case and PSU with four CPU's for xmr mining I might build two more after i see how this one works out.. an gonna keep miing with open air rigs when i use GPU's, i did think about using a server MB and Case for that ..

 Im an AMD fanatic but your showing inferior 2 slotted space wasting design compared to the champ of software mining is the Nvidia 1080 series.
  
I have a few 2U and a 4U for mining operations and beside cost , space is a factor to setup and run your soon to be mining farm. You want the best of course going ASIC and a mix of servers would over cover different algos which would be important

https://linustechtips.com/main/topic/821849-amd-vega-64-mining-performance-below-nvidia-gtx-1080-ti/


check out ebay on used servers..  !!

SIngle slot width FULL size cards are very restrictive on their cooling - reference the PhillipMa1957 thread where he talks about using some of the Katana 1070 cards and their heat issues.
Low profile cards are EVEN MORE restrictive on the cooling they can manage.
Low profile single-slot width cards are a lot worse yet.

 THAT is why nothing with high capability ever shows up in those form factors.

 Do note that EVERY 1080 IS 2 SLOT WIDE AT LEAST (a very few like the Aorus are *3* slot wide) - which means it is just as much a "space wasting" design as high-end AMD cards.
 Also, the 1080 is not "the champ" at anything, though it is competative with the rest of the NVidia 1070/1070ti/1080ti cards on most algos.

 2U for a mining setup is a waste of time and effort, you are WAY too limited on the cards you can use due to the "low profile" only limit of the case heights - or you have to use 90 degree risers and can normally only mount 1 or 2 cards TOTAL which is also a Bad Thing.
 4U is viable, but you WILL have cooling issues even if you use high-power fans to toss air THROUGH the case if you run a lot of cards in the case - though it IS a lot more compact than most riser setups.

well my equipment dont over heat and the 20 years of repairing HP servers I only remember one case of a clients server of over heating due to a mouse got trapped inside a server.  

  PhillipMa1957 is a pioneer and i due listen to his advice and he even did a review of The Panda Miner and even though he doesn't recommend the product (due to overheating,ect,ect),  he even agrees that miniaturizing is the way of future and he shares his experience with products like the Panda Miner or the Nvidia 1070 by Galax?? ..   The technology isnt perfect and a few years it will get better.. so either jump abroad or smell my rocket fumes as I go to the moon!!!

Recently, I just rented @ AWS a virtual server with 8x Nvidia tesla v100 !!!   A example of 4U computing and you think they run their servers cases air open??  I dont think so!!

https://aws.amazon.com/blogs/aws/new-amazon-ec2-instances-with-up-to-8-nvidia-tesla-v100-gpus-p3/

Not all 1080"s are doubled slot!!  There is a retail version the EVGA GeForce GTX 1080 Ti K|NGP|N Hydro Copper single slotted

Water cooled are a special case - yes, there are few where the CARD is single slot - but when you add in the space eaten by the RADIATOR, you don't gain any space overall.

He was using some of the "Katana" model 1070 cards for a short while - I don't remember if those are branded Galax or KFA but it's the same company behind both brands - and the cooling on them was very poor.

 Note that the Tesla is SPECIFICALLY DESIGNED for use in server cases - it has BOTH ENDS OPEN FOR FLOW-THROUGH AIR unlike comsumer-type GPUs (or it uses a "mezzanine" design in some specific servers that doesn't interface via PCI-E in it's TOP model).

 There are other workstation/server specific GPUs with that design - quite a few of the AMD Firepro line, and some older Nvidia cards as well.
 They are "passive cooled" relying on the CASE to provide lots of airflow through the GPU area.
 They also have LOWER CLOCKS due to heat dissipation limits where there is a "consumer-grade" version AND a workstation-version - like the AMD WX series (which are RX series GPUs but in a single-wide workstation card format).

I love this - you basically went full macgyer - do you have a write up and pics of your build?

no i dont have pics or a write up. When i upgrade to a 4U HP server, i might do a write up..  

On AWS, I do rent for mining purpose and overall the rental prices are to high and dont expect a healthy return. I only can get one mining software to work while most of the poplar  software didnt work..  either AWS is limits what software can be used or maybe its Cuda 9 being touchy. You can burn $200 usd easy renting a spot instance  in a short evening and still didnt give me time enough to master the equipment..  


I encourage for anyone to do a OEM server mining setup and share your results here on this thread