Bitcoin Forum

I couldn't find screenshots to compare to but I have two abd their voltage are very different! Chips aren't supposed to vary this widely.

http://imgur.com/uRVktSd

Guess they are not supposed to, but then again it may be chips with two slightly different designs.
AFAIK there are fairly large quality-differences in the chips, the jalapenos might even have been individually tuned to deliver around 5 Gh/s.

Could be fun to see if you get different hashrates with/without the box and flipping the fan (apparently they suck the air up, whereas some says it's better to have the fan blow down at the heatsink.)

Testing different heatsinks might also be fun.

Thanks. I guess I have to wait for more people to post pictures. Man imagine an Intel varying like that. I don't expect quality but I don't expect some to fail very quickly.

If you check in the hardware-section, you will see that GPU's also vary in delivered speed, not only between different models, but exact same cards with several % variance.
Not sure how it's done now, but in the olden days Intel tested each CPU, and branded them according to stable running speed. Hence overclocking got started. Actually moderate overclocking is about "using up" the safety-margin that is normally built into chips.

When HP was known for excellent and stable computers, they would frequently underclock models, and sell say 25Mhz computers with 33Mhz CPU's. Doubt they do that today. ;)

True the speed changes but a 25% difference in vrm voltage is insane when a chip needs that dead stable.

Something is way way way way way wrong there.

What about time series on the two chips themselves?  Variation between different revs (subject to any requirements elsewhere in the circuit) is not necessarily a problem.

It's pretty clear that BFL has put some significant redesign B,S,&T into the Jally, no?   

I looked at the info from BFL and the chips themselves only use 1V, so I'm not sure what that's reading.

Edit:  Just looked at the schematics again and the 3.3V power is for the AVR etc, and I think it's just created with a voltage divider, so I wouldn't expect much power regulation from the wall warts.  The 1V chip power comes from a regulator, so that should be good.  I'm not sure why they didn't use a 3.3V regulator for the other ICs.

Good catch.  I was actually wondering from what point in the circuit that reported voltage was being read.

I switched power supplies around and same result. Vrm inside shouldn't vary this much. Imagine what the chip is getting if the vrms are this bad.

I'm wondering what that voltage is also. Mine is in the 3.8v range in cgminer. It's not any of the voltage rails because I measured them all with a multimeter and they were pretty much spot on.

Where's Sherlock

You probably also remember that, in the days before functional graphics IDE's, vendors provided apps using underclocking and nop's to get things slow enough that you could watch what was happening, even in the days when the answer to that was "not very much."

EDIT: Actually, I think that machine in bitpop's avatar might have one of them apps.

Yes I use this app on my vintage computer

Yeah, but that CRT isn't CGA, is it?  It looks like text-only to me.

No you can see MySpace if you look closely. I just upgraded to ega.

Help! I cut my finger and broke the fan. What size is the fan? !

Looking at the source code, the MCU reads three different voltages: 3.3v, 1v, and PWR_MAIN (12v). The 3.3v is being fed to the MCU through a voltage divider (/2) and also with the 12v (except with a /10 divider). I wonder which of these voltages is cgminer trying to report.

Thanks.

Ps the fan is 92mm https://forums.butterflylabs.com/jalapeno-single-sc-support/2932-jalapeno-fan-stopped-working-4.html

80, 92 or 120mm. Just measure across one dimension, external edge to edge.

Man who came up with 92

my jalapeno voltage varied with temp.

3.4v @ 44 C
3.1 v @ 30 C

You mean "looking at" cgminer's source code?  Somewhere there is the printf that outputs that metric, stored in a variable being fed the reference value upstream from there.  Given BFL's doc push of the last days, couldn't it be determined that way?

Damn I got a big problem, 1000 hardware errors! http://imgur.com/dX8Hlnc

Ignore the rejects, I was testing p2pool

And that's with the cooling mod

http://imgur.com/a/HFMbb

Yeah, I glanced at the BFLSC driver in cgminer yesterday and found the voltages but I didn't quite know what was going on there.

I have > 4K errors since last 12hrs or so :)

2013-06-18 16:26:06] 5s: 5.38 avg: 5.45 u: 5.19 Gh/s | A:19519 R:42+0(.21%) HW:4586

my 2 do not have a large variance.

stock cooling/case, running 24 hours straight, in a warm room (80F).  40C 41C, 3.0v 3.1v, 7.35Gh/s  7.23Gh/s, HW errors 64 122.

These were delivered about 1 week apart.  Hope the data helps you.

Spooky, could this mean that a lot of the jalapenos will fail within a short timeframe?
Did you change the fan? Could that make any difference(normally it should not)?

Is it possible that the heatsink have been moved around or something like that, or that you have shorted something?

Possibly, sadly I didn't really test it before modding it. But the temp is great. I do remember the temp before was +10

Crazy, well i guess 1k is better but my other is nill.

Seems Jalapenos do indeed vary quite a bit, here's a screenshot from mine:

https://dl.dropboxusercontent.com/u/1181169/jaladerp.png

Yeah but mine is very low compared to yours and my other one. And it's also getting 1,000 hw errors

Pretty in blue

http://imgur.com/a/kUyKv

They are really common, a lot of the GPUs use 92mm fans.