Ninjastic
Search content
Sort by
Showing 4 of 4 results by kenpofighta
Post
Topic
Board Hardware
Re: HashFast BabyJet users thread
by
kenpofighta
on 12/02/2014, 02:11:21 UTC
Quote from: HF-Engineer on February 11, 2014, 10:44:20 PM
Quote from: Micky25 on February 11, 2014, 09:57:47 AM

Logistics is not my area, but I think that the delay might be the cooling system.  Apparently the supplier sent cooling systems with hoses that are too short to fit the BJ chassis second radiator location, so there's no way to cool a second board until new ones arrive.  That's probably why there is a delay, but again, not my department.

I know production is working smoothly now, so it shouldn't be long.

-Phil

That is nice to know this info. I have sent multiple requests in via email and phone calls and no one will answer me from the sales side. Even a quick response from them to this effect would have been better than no response at all.

I have been so angry I could literally cuss them out on the sales side, I am furious with the way we are being treated. I treat them with respect because I know it is the right thing to do but they are not earning any respect the way the customers are being treated. Phil, you are honestly the only positive right now with this company because you actually care enough to help. Whatever you are being  paid it is not enough. I appreciate all that you are doing for us.
Post
Topic
Board Hardware
Re: HashFast BabyJet users thread
by
kenpofighta
on 06/02/2014, 20:20:31 UTC
Quote from: HF-Engineer on February 06, 2014, 08:16:29 AM
To overclock (may void your warranty!), add the following parameter to the cgminer command line:
Code:
--hfa-hash-clock xxx
Where xxx is the speed in mhz you'd like to run.  Valid values are 125 on up, and default is 550.

If you are running a RPI with our image, you can add this line to the "Extra cgminer parameters" section in the settings page of the web interface.

In a perfect world the hashrate is mhz X 768 (the total number of hashing units built into the ASIC).  Each of the 4 die has 96 cores, and each core has two hashing units, for a total of 768.  Note: It's normal to have a few defective cores per die, and it's also considered normal for some cores to produce occasional errors.  This means the maximum possible hash rate is 422.4 Gh/s when running at 550mhz, but in the real-world it will typically be lower due to the reasons I just mentioned.

Some boards will like certain clock rates and some will like others.  If you do experiment with values, and I'm not recommending you do, try small increments and watch the error rates and/or if the ASIC stalls.  When it attempts to draw too much power, the power supply can momentarily dip and cause it to stop sending work.  This is when you'll see the watchdog timer get invoked.

On an average board, the buck converters (take 12V down to core voltage) are usually the limiting factor for ASIC performance.  They put out over 400 amps of low voltage power for the ASIC's cores to operate with.  Right now in the present version of the firmware and cgminer, the regulators cannot be adjusted, but soon we will release firmware that supports adjustment of these regulators as well as independent clock speed adjustment per die.

We are also presently performing lab qualification of the silicon, so I'll soon have some numbers for you guys as to what the silicon is capable of.

The on chip programmable PLL is capable of pushing the clock rate to well over 1ghz.  (You would need about 2X the power and cooling to hit this rate though!)

-Phil

Thanks Phil, I had actually found how to overclock last night by reading the latest cgminer documentation. Oddly enough I had to underclock the failing unit to get it up close to 400GH/s. I had to take it down to 525. I then bumped it to 535 and errors started back up so I brought it back down. It has been steady since about midnight last night.

I love having this info though because it means these machines are capable of much more.
Post
Topic
Board Hardware
Re: HashFast BabyJet users thread
by
kenpofighta
on 06/02/2014, 04:49:17 UTC
Quote from: HF-Engineer on February 06, 2014, 02:56:31 AM
,
Quote from: kenpofighta on February 05, 2014, 04:25:50 PM
I was getting great performance when my 3 BJ's arrived. At least on two of them, there was one that seemed a bit flakey and seems to be getting worse. All 3 seem to be getting worse. One of them keeps shutting down and restarting grabbing a new ID each time. I posted a small capture from my monitor so that maybe if others have had this and fixed it they can help. This one that is flaky was delivered in poor condition, all but one screw was sliding around the inside of the box and 4 risers that prop the board up were also floating around. I though the board was damaged and it may be. Worst assembly I have ever seen and I used to build PC's. That point is mute if they all work but they don't.

Here is what I need help answering.
1. How do I get back to my 400+ GH performance. All three of these were pushing +-420GH.
2. Worse case scenario is how do I raise the clock speed to get up to the 400GH range? I was promised by Erin at Hashfast that it would not void the warranty but the delivery paperwork stated otherwise. I will have to deal with that warranty issue separately, I know.
3. I seem to be getting huge error rates, I think it is on the flaky board but not 100% sure that is the only place. 70% to 80% errors are common. Elgius reports I am only getting 900 to 950 GH/s AVG with 3 boards and I should be getting closer to 1200. Not good for profits. Any help would be appreciated. I hate to try and send a board back under warranty since I might be stuck for months. I still am waiting on my 3 upgrades, I would hat to be down to 2 processors if I can avoid it.
Sorry about the loose parts.  The problem wasn't bad assembly, it was the parts coming loose during shipping.  They are using threadlock now to stop that problem.   It's unlikely your boards are damaged if they are hashing on all 4 die, which sounds like they are.

What are your die temps and voltages?  (if you run the "hf" version of cgminer it reports all dies/voltages)

Since your unit obviously had quite a ride during shipping, I'd look at each die temp and see if there is a discrepancy.  You might need to take the cooler head off, re-grease the thermal interface, and re-tighten the cooler.  Apparently some of those earlier orders that had loose standoffs have a high chance of having a loose cooler.

Also, what hash clock speed are you using?   Have you tired other speeds?

-Phil



Thanks for your time Phil, I know this is on your own and a favor to the newly create Hashfast community.

By the way, I would love to try other speeds but I am not sure how to adjust them and its all pretty hush, hush it seems. Before I ordered I was told I would be able to overclock without voiding my warranty but now I am worried about doing it based on paperwork received. I may try it on the machine that is struggling anyway if someone will tell e how. I run the machine headless so there seems to be little access. If you can email me instructions to overclock that would be great. kenpofighta at gmail dot com.

Details from the struggling miner: (You can see it stop an id and start new ones throughout the day, error rates stay around 80%. Seems 2 cores are much cooler than the other 2)
 cgminer version 3.9.0h2 - Started: [2014-02-05 16:45:10]
                                              
 (5s):322.7G (avg):367.5Gh/s | A:1038456  R:7040  HW:13622629  WU:5999.8/m                                                      
 ST: 2  SS: 0  NB: 38  LW: 1772615  GF: 0  RF: 0                                                                                
 Connected to stratum.mining.eligius.st diff 128 with stratum as user 1Bk9VSwaXn9UhpdowTwHEPEWw95i5F7N2_Miner2                  
 Block: c019bc52...  Diff:2.62G  Started: [22:23:14]  Best share: 699K                                                          
                                              
 [P]ool management Settings [D]isplay options [Q]uit                                                                          
 HFA 2: 61C/.23V796C/.23V1429C/.26V661C/.23V OFF   /40.21Gh/s | A: 96447 R:1472 HW: 1362431 WU:  650.5/m                        
 HFA 4:  75C/.80V 77C/.80V 62C/.82V 61C/.82V 368.7G/369.1Gh/s | A:810420 R:4800 HW:10748045 WU: 6021.6/m    

Here are stats from Eligius (I boke them down by naming the miners so that I could better analyse the data about 4 hours ago, Miner 2 is the struggling miner)
Miner2   3 Hours      214.79 Gh/s   540096   
           22.5 Minutes   196.28 Gh/s   61696   
Miner1   3 Hours      400.76 Gh/s   1007744   
           22.5 Minutes   416.19 Gh/s   130816   
Miner3   3 Hours      387.32 Gh/s   973952   
           22.5 Minutes   374.65 Gh/s   117760

I have the stats for the other two below for comparison. These units all sit side by side in the same room and run about the same regardless of how cool I keep the room.

I appreciate that the loose parts may not have been purely an assembly issue but I have never seen a shipped piece of hardware in this condition after being shipped. I hope the threadlock will stop the issue going forward. I am hoping what you say is correct and that the issue is due to loose parts and nothing more.

I am a software architect by profession and eventually I will port some of these features to Windows, I know there are many haters but it is a very stable platform, I am running Windows Server 2012 and can spin up multiple VM's to handle mining as I grow my operation. I can post on how that goes as I get to it.


 cgminer version 3.9.0h2 - Started: [2014-02-05 16:45:32]
                                      
 (5s):440.3G (avg):421.4Gh/s | A:1900827  R:17408  HW:7210  WU:11114.1/m                                                        
 ST: 2  SS: 0  NB: 38  LW: 2047136  GF: 0  RF: 0                                                                                
 Connected to stratum.mining.eligius.st diff 256 with stratum as user 1Bk9VSwaXn9UhpdowTwHEPEWw95i5F7N2_Miner1                  
 Block: 12f30e18...  Diff:2.62G  Started: [22:31:53]  Best share: 551K                                                          
                                            
 [P]ool management Settings [D]isplay options [Q]uit                                                                          
 HFA 0: 77C/.80V 82C/.80V 73C/.79V 72C/.79V 425.9G/421.4Gh/s | A:1899803 R:17408 HW:7215 WU:11117.8/m                          
                                              

 cgminer version 3.9.0h2 - Started: [2014-02-05 16:46:44]
                                          
 (5s):408.1G (avg):415.5Gh/s | A:1883755  R:16256  HW:6779  WU:10806.0/m                                                        
 ST: 2  SS: 0  NB: 38  LW: 2015649  GF: 0  RF: 0                                                                                
 Connected to stratum.mining.eligius.st diff 256 with stratum as user 1Bk9VSwaXn9UhpdowTwHEPEWw95i5F7N2_Miner3                  
 Block: 12f30e18...  Diff:2.62G  Started: [22:31:54]  Best share: 4.79M                                                        
                                            
 [P]ool management Settings [D]isplay options [Q]uit                                                                          
 HFA 0: 81C/.79V 76C/.79V 73C/.79V 73C/.79V 410.5G/415.6Gh/s | A:1884011 R:16256 HW:6780 WU:10808.2/m                          
--------------------------------------------------------------------------------    
Post
Topic
Board Hardware
Re: HashFast BabyJet users thread
by
kenpofighta
on 05/02/2014, 16:25:50 UTC
I was getting great performance when my 3 BJ's arrived. At least on two of them, there was one that seemed a bit flakey and seems to be getting worse. All 3 seem to be getting worse. One of them keeps shutting down and restarting grabbing a new ID each time. I posted a small capture from my monitor so that maybe if others have had this and fixed it they can help. This one that is flaky was delivered in poor condition, all but one screw was sliding around the inside of the box and 4 risers that prop the board up were also floating around. I though the board was damaged and it may be. Worst assembly I have ever seen and I used to build PC's. That point is mute if they all work but they don't.

Here is what I need help answering.
1. How do I get back to my 400+ GH performance. All three of these were pushing +-420GH.
2. Worse case scenario is how do I raise the clock speed to get up to the 400GH range? I was promised by Erin at Hashfast that it would not void the warranty but the delivery paperwork stated otherwise. I will have to deal with that warranty issue separately, I know.
3. I seem to be getting huge error rates, I think it is on the flaky board but not 100% sure that is the only place. 70% to 80% errors are common. Elgius reports I am only getting 900 to 950 GH/s AVG with 3 boards and I should be getting closer to 1200. Not good for profits. Any help would be appreciated. I hate to try and send a board back under warranty since I might be stuck for months. I still am waiting on my 3 upgrades, I would hat to be down to 2 processors if I can avoid it.

 [2014-02-05 10:15:06] HFA 21 HFHash usb write err:(-7) LIBUSB_ERROR_TIMEOUT                                                   
 [2014-02-05 10:15:06] HFA 21 attempted reset got err:(0) LIBUSB_SUCCESS                                                       
 [2014-02-05 10:15:07] Accepted 6b3cd65e Diff 611/512 HFA 1                                                                     
 [2014-02-05 10:15:08] HFA 21 HFHash usb write err:(-7) LIBUSB_ERROR_TIMEOUT                                                   
 [2014-02-05 10:15:08] HFA 21 HFGetHeader usb read err:(-4) LIBUSB_ERROR_NO_DEVICE                                             
 [2014-02-05 10:15:08] Accepted 4a23005d Diff 884/512 HFA 1                                                                     
 [2014-02-05 10:15:09] HFA 21 attempted reset got err:(-5) LIBUSB_ERROR_NOT_FOUND                                               
 [2014-02-05 10:15:10] Accepted 5c43a019 Diff 710/512 HFA 3                                                                     
 [2014-02-05 10:15:10] HFA 21: hfa_send_frame: USB Send error, ret -5 amount 64 vs. tx_length 64                               
 [2014-02-05 10:15:10] HFA21: send_packet: OP_USB_INIT USB Send error, ret -4 amount 0 vs. length 8                             
 [2014-02-05 10:15:10] HFA 21: Failed to reset after write failure, disabling                                                   
 [2014-02-05 10:15:11] HFA 21 failure, disabling!                                                                               
 [2014-02-05 10:15:11] HFA 21: hfa_send_frame: USB Send error, ret -4 amount 0 vs. tx_length 8                                 
 [2014-02-05 10:15:15] Accepted 1f29f70a Diff 2.1K/512 HFA 1                                                                   
 [2014-02-05 10:15:17] Accepted 1537e75f Diff 3.09K/512 HFA 1                                                                   
 [2014-02-05 10:15:21] Accepted 5bfd3c83 Diff 182K/512 HFA 3                                                                   
 [2014-02-05 10:15:22] Hotplug: Hashfast added HFA 22                                                                           
 [2014-02-05 10:15:25] Accepted 61ee0ec8 Diff 669/512 HFA 22                                                                   
 [2014-02-05 10:15:27] Accepted 7d126b85 Diff 524/512 HFA 22