Check your crystal!!! just inspected mine and found a small dent in the top - caught possible from when fitting the top and maybe overtighted a bit too much. the dent fits exactly where the edge of the central heatsink block is. darn it. time to get the drimmel out again. hopefully it aint damaged it too much. oh well looks like crystal swap is in the pipline - upgrade the crystal anyone? thinking maybe a 33mhz in there .

the frequency wont have fried it i had mine at over 2200MHZ back in the ealry days of this thread - got too many hadware errors though!

Actually you stand more chance of damaing your miner using a soldering iron - even 30w - than a hot air rework station. With a soldering iron you need to be very careful that you dont apply just a bit too much heat and lift the track off the board. Not gonna kill it little bit of green jumper wire can correct it easily. but it easily done. even wiht skill and experience. Done several in the past myself -years ago. Since i started using hot air ive never damaged a single piece or lifted a track. the hot air melts the solder and its not direct heat on the track. the componetents can take the heat. most SMDs are mounted with anything from 120,180 up to about 230 c. As long as the hot air isnt applied for more than about 10 mins youll be fine. And if after 10 minutes youve not attached it or removed give up your doing somethign wrong.

Either way it done carefully and properly you want damage it. but think the risk is greater using a soldering iron.

That blog is a joke. Even if it was "jamie", they didn't even cite her.

Also, Everyone is saying to use a 47kOhm resistor. Can I use a 56kOhm at 1100mhz? Does it need to be 1/4W or 1/8W? I mean, I know I can increase the mhz but will it be good at 1100mhz?

Thanks nemercry!

@SVK:
Thank you.
It should be this repository: https://github.com/girnyau/cgminer-gc3355 and i found it in this thread: https://bitcointalk.org/index.php?topic=477709.msg5752544#msg5752544

Give me a little more time and ill give you a run of my 1250mhz and of my 1150mhz units. 24h each. I was able to setup a headless machine now and can let them run 24/7 now.
Probably i'll get a higher stable rate. Lets see

Stable 510kH/s@1200MHz mod of Reggie0, _only_ by replacing R52 with 100k Resistor and C32 with 120k? On the first view this sounds good!
@Reggie0: You've said GS takes 28W, do you have measured only the GS with fan or including the power supply?
Additionally I'm a bit confused. Here you talk about 510kH/s:
https://bitcointalk.org/index.php?topic=519112.msg5954771#msg5954771
But later, when it comes to proof stability you screenshot talks about 473kH/s
https://bitcointalk.org/index.php?topic=519112.msg5971887#msg5971887
Are we talking about different mods?

@nemercy: What are your modifications to get 489kH/s@1150MHz stable? It consumes 21W (with fan and PSU) you wrote, so I think its maybe the better choice.

So that's what's been mystifying us all - lost 'Completed Work' which is caused by a False Negative, of which the GS isn't programmed to detect and flag or display in the print out?
It just goes un-noticed in real time locally - then one sees the result pool side which ends up as Stale Shares or just No Shares of any kind reported at all.
Lost Completed Work = Lost Money = Throwing Coin Away!....Pssssshhhhhhhhhhhhhhhhh!
Is this correct?

Still, I see my Reward$ caufers filling up nearly twice as fast as before I made the mods.
Something still isn't being caught or explained here. Maybe I'm just hallucinating!
Better explanation. But the complete 'why' still eludes, I fear.

I wonder if an oscilloscope connected to the right spots on the card will give a better if not the indication of what's going on and even tell one when they've hit the real 'sweet spot' and not go any further. That would be some serious fine tuning. I happen to have a CRO. I guess I'm going to have to start probing and see what I can find out.

Or, is it all calculable (math) and the end result of the equations ARE the REAL answer and THAT"S THAT no matter what hardware changes are made or probing is done?

False Positive's SUCK!  

But....

The fact is, the original 4 mods have caused hardly any increase in power consumption while having reduced if not negated and stabilized an otherwise very unstable 'red noncey' stock miner at 850MHz or higher clock rate. So at this point, I still believe the original 4 mods are beneficial. I still want to circumvent the pencil mod without losing its positive results and I may have achieved this already. Still testing.

The thing is to get the clock and resulting counting / calculations in the chips to be stable and positive at higher clock rates so as to produce more valid shares i.e. $Moolah$! at the pool. I still believe this is achievable and we're on the right track.

All in the name of progress....right?  

Wolfey2014

Sorry to disappoint, but that's not yet stable by my standards, which is less than 10 HW error in 24h. But obviously the highest stable frequency wins, so it will be you if you manage to get it stable...
Now that, folks, is called pushing it to the limit, great work mate.  

Exactly.


There are an equalation for cmos logic IC power consumption:
P=c*f*V^2+Pleakage, where c is a constant.

Gridseed uses 45nm technology, so Pleakage is high, but you see P~V^2.
If you divide it with the nominal values, you will get P/Pn=1200/850*2^2/1.2^2 -> P=4*Pn plus the leakage, but it is not easy to calculate.

There is obviously a point where overclocking no longer yields positive results at the pool, not to mention making magic black smoke . Seeing positive results locally 'client side' means nothing if it isn't reflected pool side!

You keep saying that. However so far the reports from the modified miners don't back that up.

I can overclock my unmodified miner to get 500 kh/s, but that doesn't mean I'm getting 500 kh/s worth of completed work. Due to rejects/HW errors I'd be getting a lot less than that. You posted a screenshot showing 500kh/s but you had 429 HW errors over the course of about two days. For the same period of time, I have 9 HW errors. Those HW errors hurt and lower your effective hashrate.

It doesn't matter what hashrate your miner is reporting. What matters is how much work the pool sees you are doing. If, as we've seen, your miner is reporting 450 kh/s but the pool is reporting a 24 hour average of 350 kh/s then something is off.

We need OBJECTIVE tests to verify how effective the mods are. Testing shouldn't be run against pools or solomining anyway. There's too much variance. A real test would be run against test blocks. Remove the pools and coins from it entirely. Send the same workunit over and over again. Take the stats from that and calculate the effective hashrate. I think cgminer has a test mode that does something like this.

So, you're saying to put a 120k resistor across C32 in parallel with it, right?

If one chooses the single resistor method of 54.54k instead, it is to replace R52 instead, right?

No longer any need for the 4 previous mods i.e. 2 jumpers, 39k and pencil mod, correct?

The ultimate option / mod is:
Just a single resistor (54.54k) solution and that's it?

If so,,,,
How sweet it is!  

Wolfey2014

28W?? how?

Wow, 2V DVDD is impressive - thanks for taking the risk

I have achieved 1401mV DVDD and 1204mV VDD_PLL (measured) with this approach:

- Enable VID1: Short R336 (original VMOD)
- Increase DVDD to 1.4V: Cut R46 (0R jumper) and replace with 2.94k
- Increase VDD_PLL to 1.2V: Add 27.4k on top of R212 (27.4k || 10k = 7326R)

For reference, the default VDD_PLL is 0.8 x (R211 + R212) / R212, or 1.092V.

Not getting HW errors at 1000 MHz, but I haven't tested long term because I only got my units yesterday. I'm sure 1100+ could be achieved by increasing DVDD using the above method.

Sorry to disappoint, but that's not yet stable by my standards, which is less than 10 HW error in 24h. But obviously the highest stable frequency wins, so it will be you if you manage to get it stable...
Now that, folks, is called pushing it to the limit, great work mate. 

Awesome! I'll gonna try to beat that !

Solder bridges + resistor + pencil ?

Tell me your secret!

I am giving out 0.6BTC to the first to get his Gridseed miner stable at 1100 MHz (<10 HW error in 24h) and post the steps to mod the miner. So far I have managed to get it stable at 1013 MHz, but I feel we can push it further.

Disclaimer: as always, I am not responsible for damage to your miner, do this at your own risk!

I want to share some pics with you guys. Yet not determined to be 24h stable. But actually more than 1100mhz, so i probably get some reward here ?

Gridseed porn (:])

That isn't what the results indicate so far. If you're running at 1000 MHz, you should be hashing at around 420 kh/s. The stats so far indicate that even with no errors the performance isn't noticeably better than an unmodified unit. There could be a number of reasons for that.

That being said, you can't make the claim that there are no other issues when overclocking without knowing how every part of the system works. If any part of the system is being saturated or starved, increasing chip speed isn't going to help (and may even make things worse depending on the software/firmware/architecture).

Again, hashrate does NOT equal performance. Hashrate is a measure of how fast the chips are hashing, not how much work is actually being done. That's why you need to look at the whole system when you're trying to improve performance. More hashing power is pointless if the problem is actually IO, for example.

If after another 24 hours elapse and the performance is still about the same, then other parts of the system may need to be investigated for potential bottlenecks.

@wolfey2014.
As in black square, not only the 24 hr hash rate is same as running at 850Mhz, but also the pay out is same for LTC. ~0.06/24 hr.
Coz i also mined past 2-3 days litecoin & my pay out is similar.

I respect you wolfey for doing great work, helping others.
But one thing i cannot understand why u hate cgminer or bfgminer & like cpuminer too much.

Since i read your past posts also, i came to know that it took you too much time to know how to run cpuminer, so now u dont want to run any other miner, as you fear, it will waste your time.
Running cgminer is very simple, i dont know abt bfgminer.
but cgminer requires zadig_2.1.0 usb driver. you can install the usb driver in a SEPARATE system & run just ONE miner in cgminer.

You can download zadig from here
http[Suspicious link removed]

& over clock supported up to 1300Mhz with increment of 25 Mhz cgminer here
http://cryptomining-blog.com/1662-updated-cgminer-3-7-2-oc-for-mining-on-gridseed-5-chip-gc3355-asics/
http://cryptomining-blog.com/wp-content/files/cgminer-3-7-2-gridseed-windows-oc.zip

To run cgminer in windows mode, this is i am using in bat file.
cgminer.exe --scrypt -o stratum+tcp://litecoinpool.org:3333 -u YOURWORKERNAME -p PASSWORD --gridseed-options=baud=115200,freq=850,chips=5 --hotplug 5

Run cgminer first & then connect miner to pc, so cgminer detects it correctly & mines.
I hope it will help you & also me & others as every one can see the hash rate in client side also & also HW errors, rejects.....

Thanks