Topic: GridSeed 5-chip USB miner voltage mod - page 69. (Read 156991 times)

I appreciate your skepticism.
Believe me, I have better things to do than start unnecessary and untrue rumors.

In any case, I do not apologize for the claims I am making. They are indeed true claims! I promise you!

Please READ the posts in this thread and in the tuning thread.
There are a few on here who have already achieved the same increase.

The evidence is adding up in favor of my claims and Sandor's claims who started this adventure in the first place with his mods and claims.
And, more favorable evidence is on the way!

Anyway.....
That is what you call, "verification".

Wolfey2014

Are you an electronics engineer? I'm interested in any hack that works well and is the easiest one to do.
I found one other shortcut that negates the need for the pencil mod part. I'm still testing its results.
So far, 135% to 150% stable gain in overall hash rate has been achieved at 1000MHz.
1100MHz throws less red nonce HW errors but it's still too much.
1200MHz stable is still pretty much impossible at this time with current mods.
PM me if you'd like to correspond and we can figure something else out together.
I wish I knew where Sandor went off to... He started this whole thing anyway
Wolfey2014

I will do a new picture, when i am done with my smd resistor replacement on the pencil side.
I will try 2 different SMD's (eg at 0.94V and 1.01V) and try to give you a 24h run for each PLL voltage. (Will probably happen first on the next week, because i needed to order parts.)
As for replication of results i would advise to use the 39kOhm 1% resitor because its more exact on the resistor value.

Lets see what stable resulst we all get : ].

Regards


Edit: @ Amix thanks for the reply. I tried to measure it myself, and also estimated that it would be the 0402 series. I ordered some of them, which will probably arrive next week. Then i will be able to confirm .
Thanks for your reply!

The fact is, you have no REAL results, only unverified, wrong results. Sorry.

so anyone tried swopping out the oscillator crystal (x6 tink its marked no?) for a higher frequency one?

alos whats wiht the reistor values, i mean where did the 39k reisitor value come from? could it be higher? (in line wiht chip u9 upo1111)

and laslty has anyone got or know where to get hold of the data sheet for the 2-bit VID buck converter chip u2, (up1509)

and for ref r211 is labeled a 3.65k -1% 04 and R212 10k 1% 04

And final why do the solder bridge to short out vid0 and vid1 ? then replace the matching reistor. Why not just replace the corresponding resistor to vid0 open and vid1 open? which i believe would be R139 default value set at about 22k. )

or did i miss something.

i try but i dont have the values and compents sizes -at the moment - to try.

Not interested in semantics.
Facts are facts, no matter how you try to rationalize them, REAL results speak for themselves.
Nuff said...
Wolfey2014

Thats a really good picture. Nice work .
I am looking for a way to make the "pencil" mode more time efficient. That would mean for me, that i will solder the correct resistor there that you reach 1.01V and the rest will be given by the pencil then. (lets assume you can go +- 0.01V with pencil mod, without changeing the voltage when occasionally sneeze on it.)].

OC Unit (2 bridge pin mod, 950mhz)
I measured 120,7 kohm on r211 and 0,3v
I measured  65.6 kohm between r211 and r212 1,09v
I measured  60kohm, 0,8v r212
OC+ Unit (1050mhz and 39k 1% resistor + pencil mod)
I measured 7 kohm on r211 0,03v
I measured 60.2 kohm between r211 and r212and 0,82v ( i measured it twice, i rubbed a little on the carbonglue and needed to recheck: 61kohm, 0,85v)
I assume    60kohm, 0,8v r212 (was not changed, so it should be the same)

So lets do some correlation:
65,6kohm / 1,09V  = 60,18349kohm/ v * 1.01v =  60.785kohm
60.2kohm / 0,82v =  73.41464kohm /v  * 1.01v =  74.149 kohm
(61kohm/0,85v = 71,7648kohm/v * 1.01v = 72,482 kohm)
(not sure where this offset comes from, i just assume that is just not a linear ?,(but i will try linear interpolation) )

Interpolation:
x = 1.01v; f(x)[kohm]; f0 = 60,2kohm,f1 = 65,6kohm, x0 =0,82v; x1 = 1,09v
f(x) = 64kohm
So my assumption would be, that ill get 1.01v when i can set the resistance between r211 and r212 to 64kohm.
Achievable that way:
x = (1.01 -0,8V ) = 0.21v; f(x)[kohm]; f0 = 7kohm,f1 = 120,7kohm, x0 =0,03v; x1 = 0,3v
= 82,8kohm

So in short: i need to replace the R211 with 82,8kohm and i would get a steady 1.01V.
Please confirm if correct.

On my currently modded unit, i remeasured things. Looks like as i did a little overshoot with the carbonglue ... As it got hard, it was more conductive and my pll voltage is now 0,82V:
I know that it is not a 24h run, and i am scientifcly very sorry for that and will post one the next days. In our first run it was much more stable, when the voltage was at 0.94V.
I will reduce som of the carbon glue and will rerun a new test.


Is there actually a scientific proven reason, why 1.01V shouild be the sweatspot ? In our sample, i could not measure any differences.
Can also someone specify which size the smd resistors are currently ? I mean which type of smd resistors (to be exact: which size?) are they ?

regards,
nemercry

still digesting all this just in from long day at work. but i believe the SMD series is 0403. i could be mistaken though. i did have a parts list for the gridseed, compelte with size, value, series and footprint, but i cant find it now. so im going on memory.

2. There are a theoretical problem: you will never get higher average pool hashrate than local average hashrate in long term. Cgminer's hashrate calculation based on how many jobs are finished in a second, whatever its result(you got a share or it has no solution). Pool calculation based on how many shares are submitted in a second. It depends on luck, but as Law of Large Numbers says, in long term it has to be same as your local hashrate(in an ideal environment: no network latency, no stale share, no rejected share, etc.).

Not interested in semantics.
Facts are facts, no matter how you try to rationalize them, REAL results speak for themselves.
Nuff said...
Wolfey2014

1. Pool hashrate reports are inaccurate and its have high variance. You reported a peak hashrate, not an average. Maybe try it on ghash.io, there are 6 hour, 12 hour and daily average values.

2. There are a theoretical problem: you will never get higher average pool hashrate than local average hashrate in long term. Cgminer's hashrate calculation based on how many jobs are finished in a second, whatever its result(you got a share or it has no solution). Pool calculation based on how many shares are submitted in a second. It depends on luck, but as Law of Large Numbers says, in long term it has to be same as your local hashrate(in an ideal environment: no network latency, no stale share, no rejected share, etc.).

The results I speak of and have been speaking of ALL ALONG are POOL SIDE!
I don't care what local hash rate says, I can't see it anyway using cpuminer. It's pool side that matters! AND That is where the money is made!
And by your own calculations / admission , it IS A 150% increase - client side - AT LEAST - moron!

SEE THAT! It's RIGHT IN YOUR FACE!!!

Wolfey2014

So what?

I believe that's possible with NO mods!

You're wasting your time not doing all 4 mods. They MUST be done together or modding is a waste of time.
IF you can solder 2 tiny bridges, you can easily replace the resistor and do the pencil mod - provided you have a dvm on hand to check the PLL voltage.

Anyway, it's all on you!
Remember, you cook it, you eat it

Wolfey2014

It didn't happen. If you double the frequency, then you get double hashrate. At 1050MHz you will get 425kH/s. There are no magic...

So what?

I believe that's possible with NO mods!

You're wasting your time not doing all 4 mods. They MUST be done together or modding is a waste of time.
IF you can solder 2 tiny bridges, you can easily replace the resistor and do the pencil mod - provided you have a dvm on hand to check the PLL voltage.

Anyway, it's all on you!

Wolfey2014

2 bridge modded all my gridseeds, 7 of them.
running at 925 seems to be stable (none to 1 or 2 errors / 12 hours). its giving few H/W errors at 950.
most GSs can handle 938 and even 950 without H/W errors so im using the modified cgminer to set different unique clock speeds to each of them.
http://cryptomining-blog.com/1688-alternative-cgminer-3-7-2-for-scrypt-mining-on-gridseed-5-chip-gc3355-asics
want to do the pencil+SMD replace mod as well but too afraid to risk 250$ for few kilo hashes.