Topic: Data Center Mining Garage and Man Mining Cave ( PART 2 !!!!!!!!) - page 20. (Read 106684 times)

Cant wait man, may the hash-force guide you.


Why are you so negative man? Did jstenfanop take money from you for doing this project? Did anyone ask him to develop such a solution? He is doing this out of his own free will and he has invested a significant amount of time and effort into it, if you can not appreciate it at-least don't take a piss on someone else just for the heck of it. And what do you mean "What if you cant deliver" he doesn't owe you jack shit, so stop feeling so entitled.
Let him complete what he is working on and deliver, and if people see value in it, which im sure they would unless they are blind, they would buy it. You on the other hand can completely avoid it and trust me no one would hold it against you. But don't be such a killjoy.


Ooo so here we see the genesis of the very idea. Thank you for sharing this kilo, now lets hope it becomes a reality.

Here is some info on doing multi GPU (8+) setups.  It has some great insight into what all is needed.  gstoner on the link is a Senior Director at AMD and is full of knowledge about this type of set up.

https://community.amd.com/thread/197524

this thread was started by jstefanop himself

so what your saying your going to be offering the motherboard for around 70 to 80 dollars? if not its hardly worth it. your still vetting out hardware so to say it will be cheaper in total supporting component costs per gpu is a bit premature. tbh I doubt this setup will be cheaper than compared to a typical 6 to 7 gpu setup not to mention your playin to a niche market and not everyone in that market will see that as viable, I see more points of failure being added etc, thats what I look at. but all in all I guess we all will find out

my question is what happens when you cant deliver on this promise your making that it will be cheaper and it turns out its not, you going to take a hit? doubt it lolz

Nope...skylake/kabylake pentium or i3 work just fine (my dev setup uses an i3 since double the threads is worth it when driving 16 GPUs....but I realized that CPU utilization even for 16 GPUs is so low that a pentium will probably work just fine as well. I will test this week). The motherboard IS a server based board with 1151 socket, but since I will buy these in bulk for this project it will cost less than a typical 6-7 GPU board you guys are using.

Sorry for the lack of info but I deliver what I promise (guys that bought my scrypt USB ASIC miner can vouch on that), so I don't like announcing things early since I'm still in the middle of vetting out the hardware.

What I can promise right now is that this solution will definitely be cheaper in total supporting component costs per GPU with the typical 6-7 GPU setup right now.

@yun9999 or anyone else with the Sapphire Fury cards.

I have 15 of them with all the same settings (-96mv) 900/400 15.12 driver and the last one I bought won't under-volt in Windows 10. Has anyone else ever experienced this before?
This is what I have tried;
moved it to a different rig with 5 other Fury's and it will run Optiminer but Claymore will return a error saying the GPU returned incorrect data.
moved it back to the new rig with 3 other Furys and it will crash the driver in about 30 seconds of mining with Optiminer
   reinstalled windows and new drivers - same
   installed 16.3.2 drivers - same
   UEFI Bios and standard - same
moved it to a independent computer, ran graphic benchmarks on default settings and no errors - but when I under-volt it, crashes the computer.

Every setup has more than enough power overhead, running TB 85 or H81 boards - except the independent PC (supermicro server)

I purchased it at Newegg but to send it back, I believe they would test it and it would pass there tests because how many gamers really under-volt that much.

Any suggestions

@ecohash @Marvell1 @yslyung

Regarding Gigabyte 470 4GB G1's just wanted to share some info on my experience - I had a similar issue with heat rising around 75-80c. After testing a few different settings and setups, I realized the issue in my case stemmed from having the cards in a case.

If the card was blowing up hot air against a card just under it, it would cause the top card to get hot, while the bottom card kept normal temps. Open air solved the heat issue and they run at around 62c now,

I only bought 10.  Yslyung emptied the Global supply of TB85.  =)  Regarding why Gigabyte over Gaming 5?  Desparation.  They were on sale for $89.99 refurb, they are not a good replacement as they only have 6 PCIE, the 7th slot is PCI.  It's a problematic board buying refurb.  I only have 3 out of 6 working well but even those can be flashing on and off when you make any changes to PCIE connections.  Very annoying.

Ha no wonder why there always out of motherboards lol

So YOU are the reason Newegg put a limit of 1 per customer on this board
Why Gigabyte? Isnt the MSI Gaming 5 a much better board?

Biostar became available on NewEgg so I quickly grab 10.  =)








Also looking at plan B, so I ordered a few Gigabyte MB.  These support 6 GPUs no issues but they are a pain in the butt at time with the multiple flashing of ON/OFF, when there's any changes.








ROM Station with Pico and HP PSU.








One of the Viewer recommended this ThermalTake case so I ordered to play with it.  It's may for Micro ATX not ATX motherboard that we use so it won't fit like you expected.  Since I hate doing return, I just have to do some creative changes and use it sideways vs front to back like it's intended to be used.   Unfortunately the max you can probably fit in here is 5 GPU.  This is a very small case.

I don't have a website, but my risers are on sale both on Amazon and eBay. I sell direct here too, payment through PayPal or btc, zec or eth. My fans are not on eBay yet, they are similar in specifications of scythe fans but built with double ball bearings for durability. Fans are $9.50 each plus shipping.  I will post picture of the small PCI-E piece later tonight. They are $1.50 each and can potentially save risers from being thrown away. Link to my Amazon listings is in my signature below. 6-pin risers will be available in a few days on Amazon but I have more in stock at home for immediate shipping.

Edit: pic of small PCI-E piece in stock.

whats hawkfishes website and does he have pictures

@ Tasty Potato  -  Those are difficult to troubleshoot as it could be risers, your ROM, your settings, etc. 

@ All - I was able to convince Hawkfish007 to get 1000 PCI-E Adapters for the risers and sell them separate from normal risers.  Here's the reason why this is important.

Most common failure with USB powered risers (The Common FAT board version)
1) MOLEX connector coming loose or the cable come loose causing a short or riser to fail - Over 60% of the issue
2) USB PCIE Adapter - These wear out very easily and are also the most likely to rub or get scrape during transit.  - Over 30% of the failure
3) The big board - 10% - Usually hardly failed unless damage during transit or capacitor is blown. 



Generally speaking the USB cable hardly ever fail.  I only had 1 bad one with over 1200 Risers, I bought.  I had to throw away many risers just due to the lack of the PCIE adapters, no one is willing to sell those separately so I'm very Happy that Hawkfish007 is able to provide this for us.  I just bought 50 as spares.    The cost is very cheap and this helps let me use many risers that otherwise would have been $7-8 gone to the trash each.

And here I thought everything was rock solid -,-

I need advise, if you have dealt with this.

5 rx 480, flashed bios 1750 to 2000 memory straps, the rest is stock. 1000W Corsair gold PSU, 90~100W per card, 1180 core clock, 2200 memory clock, undervolt -50.
Had stable 30+ Mh/s for each card (2 G1 Gigabytes, 3 Sapphire Nitro (switched into position to flash another bios)), total 150+ Mh/s. Used Msi AB. Temps 70-78.

Yesterday hash dropped to 140, like no OC. And even drops for 2 cards to 20-14-7 Mh/s. Had boot problem too, so reinstalled Windows 10 and AMD 16.9.2 drivers.
Now 140 Mh/s stable. Msi AB does not increase hashrate.

Where to dig?

@op, that's the past. Wait till u see my "nextgen" rigs & I bet ya it's really a 1 of a kind.as u know I'm tied up with rigs production & hosting for local community here but work is already in progress.

So far 1500's are working well for me.

Just be little patient I have the solution for hp "high pitch" ...

Stay tuned.

Some Fury's http://imgur.com/a/FnInT

Some Nvidia 1070's  http://imgur.com/a/hfN4G

Happy mining!

@ Tasty Potato -  Claymore doesn't really give you a true free option as he charges you by giving you slower hash penalty.  470 is definitely faster ROI.  480, you need to invest is SOLID ROMs which won't come cheap.  470s anyone can mod with ease

@  ecohash - I'm not a fan of any non reference 480s at the moment unless you're mining ZEC.  Too much work or cost to get that super 480.  I only like the old reference cards, the new one sucks too.

@ jstefanop - I believe the setup is using server CPU and other things that may make the setup a bit pricey?  I was excited about the config but the lack of info is making me lose interest, time to keep building rigs the normal way and MAKE SOME Money.................   I'm never a fan of constant teasers, I prefer the FULL SHOW.  =)

@ bhugatti - As a new miner, I would stick to tried and true solution vs bleeding edge stuff.  Let others suffer and pay the cost of being the Guinea pig.  Unless jstefanop solution is more cost effective, which I believe it won't as it's a focusing on a niche market, it's best to play it safe.  I don't think 100 hrs of stability means much.  Heck I bake my rigs for 7 days at 115F, however I'm not certain it will continue to be stable in that condition for 7 months.  Problems usually occurs as equipment ages and the more GPUs you have in a rig, the higher your instability rate is.  I don't care what anyone say about that, as it's just basic statistics.

@ yslyung - You MAD Scientist.  Love the progress and the LED SHOW!   Please let me know how the Noise level is on the HP 1500 before I waste more money to find out.  Appears to be a similar tiny fan so my guess is HP is still HP, meaning High Pitch?

@ deadsix -  I highly doubt the solution will be cheaper.  I think I've seen similar configs and the component cost adds up a lot, even your CPU required will be much high end meaning the cost of the CPU alone may be as high or higher than 3 x $65 ASROCK H81.  It's also targeting a niche market so there will be some premium in Play.  The cost of a 1000HP car is usually much more expensive then 4 x 250HP car.  Similar idea.

@ stoon - Very sexy and unique MONEY MAKER!  Thanks for sharing

@ Razziel - I'm guessing the Z97 is the replacement for the H97 anniversary.  I'm not a fan of the H97 anniversary as it requires connecting the molex to power the MB for a 3 plus GPU config.   The H81 you can press N to bypass that requirement.  I have no experience with the Z97 but if it's a direct replacement I would think it will be similar.  On the positive side, the H97 have no issues working with 6 GPU and you can't go wrong with ASROCK as their RMA is best in class.

@ p4n - I only tried the splitter board once.  Not sure if I got a bad one but it didn't work for me so I never tried it again as it waste $30 so I had to eat Ramen noodle for that night.  It was a sad night.

@ CryptoWatcher420 - I think you guys are saying the same thing.  He calls it a Nano PSU, other calls it PICO. 


@ALL - ETH $52!   HOT DAMN!!!!!!!!!!!!!!!!!!!! SHOW YOUR MONEY MAKER!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

you are incorrect sir, a baby psu's are not needed, grab a pico psu and you will see. as a matter of fact, lots of folks on here have posted about there rigs that use a server psu only, I myself am converting my gpu rig to use only a server psu, a good breakout board will atleast have a switch on the breakout board itself

That mainboard has only 4 PCI-express slots (1 x16 and 3 x1, there are 2 legacy PCI slots that are useless for mining). Z97 chipset should have enough PCI express lanes to run 6 to 8 GPU, but you will need some kind of PCI-e expansion board with more slots in order to get all the slots.  I don't know if those boards are realiable enough for a rig, maybe someone could tell us about those boards.



2000 watts should be enough for 6 gpus + a pico for the mainboard+cpu+hdd. Correct me if I am wrong, but i think you will need this.
You will need some 6pin to 8 pin adapters if your PSU has only 6pin connections, and each GPU will use 2: a 6(to 8 pin) cable for the GPU and a 6pin for the riser. So for 6 GPU you will need 12 PCI-e connectors in the server PSU breakout board plus one more for the pico. You could save some connectors using some 6pin PCI-e to Dual 6pin PCI-e cables, so you will end using 1 6pin PCI-e connector to power 2 GPU risers, so a 12 PCI-e 6pin breakout board will be enough


Has anyone tried to build their own 6 or 8pin PCI-e connectors? If so, where can I buy the connectors at a right price? Searching ebay bring me some expensive results

While the 4k PSU delivers most of the power for the system (namely, all the cards) each system still needs a mini PSU (basically just a low capacity regular PSU like 450W) which is what you use to turn on/off each system.
The alternative, and what everyone is switching to, is a nanoPSU which powers off of the 4K PSU and has a small on/off switch.