Topic: Experimenting with Jalapeno firmware... - page 17. (Read 62561 times)

Does cgminer crash without the Jally plugged in? If so it's probably trying to use the laptop GPU and failing. Tell it not to.

I'm using bfgminer 3.1.1.  I tried to use cgminer but it didn't work.  Just crashed on start up (basic Asus laptop, Win8).  I have restarted both units several times and apart from a 5% or so speed difference at start up, one unit always reports far more errors than the other.

I have noticed hardware errors are almost zero when bfgminer starts with difficulty 1, and only when it switches to 4 do the vast bulk of errors appear.  I'm mining on Slush's pool.

It doesn't look like bfgminer reports voltages.  Here are the other stats from the main screen.

ST: 2  NB:603  E:43.53  GF:2  AS:0  U:51.4/m  BS:2.39M

What are your voltages?

Looks like your 7.6GH/s is having 1.414% HW error and the other is having 7.578% HW error.
The second chilly is having a high HW error rate. Have you tried restarting it or upgrading the miner software? I am currently running on cgminer 3.3.1 with 0 HW errors.

IMHO, flashing the firmware to the latest is not really overclocking the chillies but I do agree some of the mods here are overclocking it. But that explains why flashing the firmware voids the warranty

Interesting overclocking results and great thread.

My two recently received Jalapenos are quite different beasts, with one performing much better than the other.

Both upgraded to 5+2Gh/s, 37 celsius operation.  Unmodified firmware straight from BFL.

7.6Gh/s, A: 120,291,  HW: 1701
7.1Gh/s, A: 112,529,  HW: 8528

Mining is occurring at difficulty 4 but I understand the hardware error check is run at difficulty 1, so the error rate isn't as terrible as it looks.

Overclocking usually comes with the risk of ruining the product in the long term due to effects such as electromigration, but with my Jalapenos I'm a little concerned the higher error rate from the second unit suggests it's already running at the upper boundary of overclocking.  Thoughts?

Nice thread. where's the Like button?

After I ROI from my jallies, I will be flashing them. Anyone selling their dragons?

So ... either you got a Jalapeno or you got a Little Single.

Which one?

If you ordered a Little Single and they gave you a Jalapeno instead then obviously you need to sort that out with them.

The Jalapeno doesn't earn much any more (5.0GH/s is less than 0.12BTC a day) so sending it back to be fixed shouldn't be an issue.

If they gave you a Jalapeno and they are still going to send you a Little Single and you are complaining about delivery times ... well go complain about that somewhere else.
There are plenty of places to complain about delivery times and this isn't one of them.

  sorry that I was thinking you got Jal and you might got bad LSC or damage by shipping.

I just state that i fear if i try this with the allready bad performing device, it will make things worse instead it gives better hashes

Also tom99 you assumed incorrect my order was a sc single which would do 30Gh not a jalapeno and i paid enough to get 10 jalapeno's
Second its no use to contact them they simply refuse to listen nor answer and the costs for sending it back will up the losses even more
 

If you have a warranty issue, this is not the thread for it, contact BFL support, or post a question on their forums.

remember you ordered for 4.5Gh/s for older order and new order after April 1st is 5.0Gh/s.

Well the crap jala i received from those bfl people is prolly not even worth trying to pump up

Which placed the very bad experience with BFL to an absolute -251 kelvin in my view

When it came out of the box it ran for a few minutes at 6000 Mh then all of a sudden it kept going down steadily
While the hashrate is going down the error rate keeps creeping up the scale

At the moment i get between 3900 to 4700 Mh, mostly around 4300

So even on stock this thing produces enough amounts of errors

While the temp according the readout not exceed 31 c, which looks to me cool enough to get decent results

Nevertheless to buy a programmer and spend lots of money just to try up one bad jala a bit might even spoil the short life this thing probably have even more
Which look to me the same choice as doing business with BFL.

I just got through spending most of the day messing with the jalapeno.

I made a custom peltier assembly and a temperature controller that regulated the temperature to around 65-68 F. I didn't want to go colder due to risk of condensation.

I also modified the firmware to make the same changes as ckolivas did in his, except you can just comment out in std_defs.h... ...and it should have the same effect. I also reduced ...down to just 1. After these changes, I was able to get 30 engines running from 28 previously.

I also increased the voltage from 1.016v to 1.047v by changing the resistor at R11 to 7.5kohm. This resulted in the chips automatically clocking to a higher frequency. At frequency index 7, I was previously attaining 291MHz/271MHz on the chips. After the voltage increase, it was clocking at 305MHz/291MHz.

After these changes, my hash rate went up from ~8.2 to 8.3 GH/s at frequency index 7, to ~8.85 GH/s. I was able to achieve ~9.05 GH/s with index 8 (~4.5% errors), and ~9.23 GH/s with index 9 (~7.5% errors).

Unfortunately, the peltier assembly didn't seem to have much effect on the error rate. I put the old heatsink on and it handled the heat just fine. The only thing the peltier was good for was getting a high initial frequency setting. With the peltier drawing ~90w, it wasn't really worth it to continue using.

Right now I'm running frequency index 8 @ 8.93 GH/s (~2.6% errors). It is possible for me to go higher at that index, but that was when I was using the peltier. I stuck the unit in the freezer, but I guess I didn't get lucky.

Some pictures of the peltier assembly...

Well they did over delivery 1 ghs and as much as I hate bfl, they even allowed further free overclocking.

I see your point. My point had been that with the frequent 10% mismatch there would be a fair (5%) gain overall and the double loss at LP wouldn't be more then the gain. Even then anything less then 20 seconds would be a loss, over 40 seconds I would be ahead.

I also see how with trivial maths I could keep both chips un-hobbled and a low time for work to finish.

I guess what bothers me is that it could have shipped ~5% faster on 7 of 8 Jalapeno's and for no cost to the manufacturer. Just flash the current firmware and life is good. Instead of keeping connectors uncovered BFL built the heatsink over them. Then to make matters worse I could flash on a 5% gain but it wouldn't keep the warranty. I intended to keep the tests on and likely keep the units running around 5.5Ghash as I want them quiet. But there isn't a good way to get everything. Now I have to figure out what the warranty is worth. So far I have 1 failed fan and 1 failed power supply. I emailed and haven't heard anything back.

Except the 1 job per chip is bad due to the latency it incurs.
A single 4GH/s chip takes ... 0.93s to complete a nonce range, and the BFL MCU (with the same design flaw as the BFL FPGA) doesn't return results until it completes the work item.
(Note the latency/work loss factor in the 1 job per chip design is multiplied by the number of chips you have - more chips, the worse it is)

It's also not necessary.
(though I've no idea if Luke will take notice of how I suggested it's easy to fix this ... divide the work up so they all complete at roughly the same time - very simple maths)

So, although the gain from 1 job per chip is greater than the loss it incurs on any pool other than p2pool, it's still far from optimal (and optimal is not difficult at all)
However what this means for you is you would again want to re-flash it later when it's fixed properly

See I see some of your statement as wrong.
No one said I was going to overclock it. The thermal protections should avoid issues with running too fast. But even then what I want is 1 job per chip. It is faster that way as all but one of my units (8-1=7) have mismatched chips. The worse they mismatch the more benefit from the newer firmware.

PS. Firmware was released 10 days before mine shipped. I have the old firmware. I suppose it being available and installed don't have to be the same thing.

That's pretty much the case with every device, so it is not a surprise at all.
There is a big fat warning about loosing your warranty when your overclock your GPU too