Topic: The Habanero - 650GH/s - OOS - page 17. (Read 96051 times)

Damn. I am at 900-970 right now and getting 677 GH/s. Hash rate dips at 925-970 and I can see temp drops in the app. Temps are at 99 on both cards with my current settings. Oh well, no complaints here other than my OCD wanting to get 700 GH/s.

You need to up the voltage.  I've hashed as high as 735-740 GH/s at 962Mhz, with high hw errors and bouncing off the 104 mark.

I'm getting a pretty solid 700-710gh/s at 925-950.

Sold out...Good luck...

This looks like a very sexy and fierce miner!!! Definitely going to order one ASAP!

I agree, under load I typically measure 12.55v and 12.8 max when not loaded and with no connectors attached.  Should be well below 13.2v though.  Unfortunately I don't have a Habanero to play with so it is hard to diagnose what the issue maybe remotely.

Edit: To close the loop on this I'm told a firmware update fixes this issue.

ATX spec is +/-5%, up to a maximum of 10% at peak loading. The VRM will shut down after I believe 5ms above 13.2V.

This sounds like the most likely candidate to me.  In my testing these supplies operate at the high end of the ATX accepted range though still within it.  What is the exact voltage cut off for the habanero to shut itself down if it is less than the accepted range it could be causing the shut down.

The current sense wire that people are talking about is already connected on these boards.  What is the voltage range that the Habanero accepts is there any way to adjust that to see if that is the issue?

If you haven't increased the voltages, you'll probably need to in order to get above 900MHz. You can do that using the tool in the thread gateway posted.
https://bitcointalksearch.org/topic/hf-tool-to-set-voltages-639810

What is the highest clock rate that has been achieved?  I find anything above 900MHz causes the hash rate to drop in half even when my temperatures are below 80C.  I may be power limited or can I play with the voltages somehow?

I have tried multiple psus (I purchased 11 of them from gigamps). All of them trip when cgminer is fired up. I have measured on the back of the hab board at the power sockets and they are 12.7v..

I am using 18AWG cables and have tried ones I made and ones purchased from a retailer.

Nexus: Your running the gigampz + HP psu setup.. anything I should know? Are you on 120v or 240v power?

I have measured the plugs and sometimes they are showing 13-14 volts. Most of the time they show 12.7.

When plugged in they are 12.7v measured from the back of the hab board.

When you say add a wire.. Does that mean I am soldering something?

Ah, groovy. You guys thought of everything.

There is input voltage monitoring, and an overvoltage shutdown if the input voltage goes over spec. It could be possible that if your supply is idling quite high and there is some voltage sensing circuitry on the supply that isn't connected, you could have enough overshoot to trip that protection. It would be hard to see with a multimeter, but you could be going over 13.2V.

Try connecting the voltage sense lines as klondike_bar suggested to get the voltage back to 12V, and see if that fixes the issue.

Does the Habanero have input voltage monitoring? 12.8V is less than 7% out of spec for 12.0V, which most things should handle at least 10% tolerance.
I'd consider it might be an undervolt issue, either from bad or insufficient output capacitors in the PSU or from a poor connection in the edge connector. I prototyped a DPS-800 board but always had trouble keeping the edge connector in solid contact with the PSU output contacts. Could be when the VRMs fire up they're draining their input caps and putting a huge sudden current load on the DPS-800, which is undervolting and tripping from one of those reasons. It could be the PSU is providing ample current but the cables are either too thin or too long, or poor connector contact, such that line impedance (either from high resistance or inductance) is not allowing sufficient sudden current changes to make up the instantaneous load. I'd suggest tucking an at-least-1000uF capacitor on each 6-pin connector (slide it in to the contacts on the back of the cable-side connector) which will help buffer out brief current spikes like that at the device instead of at the PSU. If that's the issue it ought to help, otherwise I don't really know.

Not to be a shameless plug, but that's why we design our PSU boards with additional output capacitance and low-impedance screw terminals, and supply 16-AWG cables for good current handling.

could be an issue with the higher voltage (12.4-12.8v) that often is produced by server supplies. Check the below image and try adding a wire to the gigampz board for  voltage sense - that should cause the PSU to correct the output back into the proper 11.9-12.2V range that perhaps the habernaro requires

where can i find this for sale?

More on this, I tried with new cables (pcie extensions cut down and attached to the gigampz breakout board). Same problem.

I even tried with 2x Gigampz and HP psu for the one hab board and exactly the same happens even to 100Mhz

 Now I'm getting somewhere.. It appears to be the gigampz and HP psu setup. Possible my cables as I made them myself.

I tested with a corsair 760i PSU and got it working.

Anyone using the gigampz and HP PSU combo? How many psu did you use per hab board I'm on 240v power.?

did you try the 25MHz clock rate?

Have tried a different PSU, but same gigampz breakout board (will try another breakout board now).

Mining platform is Ubuntu running cgminer 4.4.1.

cgminer --hfa-fan 100 --hfa-temp-target 0 --api-listen --api-network --hfa-temp-overheat 106 --hfa-hash-clock 100