Topic: Fury/Blizzard tuning and mods - page 30. (Read 115281 times)
Hi guys, i wonder if you are runing this mods with the stuck power supply (The 5amp brick) or you are using other psu?
I've not had time to read the entire thread since finding it...but I wanted to ask for those folks who applied new TIM to their Blizzards, are you seeing any benefits other than temps? I see some folks are pushing the clocks pretty high (~380-ish)...but is that with volt-mods or just new TIM applied.
Cheers!
Cheers!
Yes, you can increase your clocks quite a bit from my experience just from using good thermal paste and applying heatsinks to the tops of the chips. Adding a small fan to blow through the chamber over the circuitboard also helps.
Without the addition of the extra fan, I was able to raise my clocks from 333 with a 4%-5% hw error rate to 350 with a 4%-5% hw error rate. (roughly about 1.4Mhs).
The addition of the voltmod does allow all the way up to the max clock of 381Mhz though. We're hoping we can actually overcome that 381Mhz barrier.
I looked at one of my Blizzards and the entire back side of the circuit board looks like it was coated in TIM...doesn't seem quite right. Should I clean that off and apply some high quality TIM just behind the 6 chips or spread it over the whole board like Zeus did?
Definitely clean the old stuff off first.
You can either cover the entire back with a nice coat (not really needed in my opinion) or just do the spots with the chips. Personally, I put three lines across the back of the board under the chips (2 rows for the chips, 1 row for the rest of the components). I however have a huge 30gram tube of Prolimatek pk-2...so 3 thin rows wasn't really a lot. Whether you do three rows, cover the whole back, or just do dabs on the backsides of the chips...it will make a huge difference because the stock setup was done for cost efficiency...not thermal transfer efficiency.
Thanks for the clarification...I've already spent way too much on these with the very strong likelyhood they'll never break even....so I'm debating whether to pop more $ for a big tube of TIM or just use the bit I have left to place 6 dots on key areas...probably the latter.
I had some old Zalman VGA heatsinks like someone posted earlier...was thinking I may put those on whichever of the 7 Blizzards has the most HW issues (assuming heat was causing it)...I don't have a means to measure the temps so I'm guessing it could help for that one.
Just for the hell of it I changed that line of code in my last build of bfgminer to
so that it was .8 instead of .75...didn't notice much of a difference really. I only let it run for half a day but the poolside numbers didn't really look much higher than normal, the local hashrate was in the 1.7-1.8 range though. Maybe higher gives better results?
Code:
info->read_count = info->read_count * 8/10;
The Zeus doc indicated it was 3/8 and maybe 3/4 later.
All the code I looked at showed 3/4.
So how do we figure out what it does? Which is faster readcount 3/8 or 7/8?
I guessed the smaller fraction would be faster readcount timeout but don't know for sure.
Newer code lets us specify readcount like we do chips and clock etc. Default is 300.
So would 250 be a quicker readcount? How does diff affect readcount?
Lots of questions ...........
Edit: just started Darkwinde's new build with default rc at 300 and will see how it does.
New BFGMiner 4.2.2 using new Zeus code https://litecointalk.org/index.php?topic=20477.0
I'll have to try this as 4.2 never ran for me.
New BFGMiner 4.2.2 using new Zeus code https://litecointalk.org/index.php?topic=20477.0
Damnit...seems like every time I clone and compile a new build another one comes out.
I'm glad it's actively being worked on though.... better than the stagnate state things were in when we first got the hardware
New BFGMiner 4.2.2 using new Zeus code https://litecointalk.org/index.php?topic=20477.0
I've not had time to read the entire thread since finding it...but I wanted to ask for those folks who applied new TIM to their Blizzards, are you seeing any benefits other than temps? I see some folks are pushing the clocks pretty high (~380-ish)...but is that with volt-mods or just new TIM applied.
Cheers!
Cheers!
Yes, you can increase your clocks quite a bit from my experience just from using good thermal paste and applying heatsinks to the tops of the chips. Adding a small fan to blow through the chamber over the circuitboard also helps.
Without the addition of the extra fan, I was able to raise my clocks from 333 with a 4%-5% hw error rate to 350 with a 4%-5% hw error rate. (roughly about 1.4Mhs).
The addition of the voltmod does allow all the way up to the max clock of 381Mhz though. We're hoping we can actually overcome that 381Mhz barrier.
I looked at one of my Blizzards and the entire back side of the circuit board looks like it was coated in TIM...doesn't seem quite right. Should I clean that off and apply some high quality TIM just behind the 6 chips or spread it over the whole board like Zeus did?
Definitely clean the old stuff off first.
You can either cover the entire back with a nice coat (not really needed in my opinion) or just do the spots with the chips. Personally, I put three lines across the back of the board under the chips (2 rows for the chips, 1 row for the rest of the components). I however have a huge 30gram tube of Prolimatek pk-2...so 3 thin rows wasn't really a lot. Whether you do three rows, cover the whole back, or just do dabs on the backsides of the chips...it will make a huge difference because the stock setup was done for cost efficiency...not thermal transfer efficiency.
I've not had time to read the entire thread since finding it...but I wanted to ask for those folks who applied new TIM to their Blizzards, are you seeing any benefits other than temps? I see some folks are pushing the clocks pretty high (~380-ish)...but is that with volt-mods or just new TIM applied.
Cheers!
Cheers!
Yes, you can increase your clocks quite a bit from my experience just from using good thermal paste and applying heatsinks to the tops of the chips. Adding a small fan to blow through the chamber over the circuitboard also helps.
Without the addition of the extra fan, I was able to raise my clocks from 333 with a 4%-5% hw error rate to 350 with a 4%-5% hw error rate. (roughly about 1.4Mhs).
The addition of the voltmod does allow all the way up to the max clock of 381Mhz though. We're hoping we can actually overcome that 381Mhz barrier.
I looked at one of my Blizzards and the entire back side of the circuit board looks like it was coated in TIM...doesn't seem quite right. Should I clean that off and apply some high quality TIM just behind the 6 chips or spread it over the whole board like Zeus did?
looks like Zeus has changed readcount in the latest code
https://github.com/zeusminer/cgminer_zeus
no more
I don't understand it well enough to know if we can optionally set readcount in the new code or not.
There's other new stuff in there also.
I'm hoping this new code lets us push through the .7 freq to kh/s ratio.
https://github.com/zeusminer/cgminer_zeus
Code:
if(info->read_count>opt_zeus_readcount){
info->read_count = opt_zeus_readcount;//send a new work every 10 secondsno more
Code:
info->read_count = info->read_count * 3/4;
I don't understand it well enough to know if we can optionally set readcount in the new code or not.
There's other new stuff in there also.
I'm hoping this new code lets us push through the .7 freq to kh/s ratio.
Just for the hell of it I changed that line of code in my last build of bfgminer to
Code:
info->read_count = info->read_count * 8/10;
looks like Zeus has changed readcount in the latest code
https://github.com/zeusminer/cgminer_zeus
no more
I don't understand it well enough to know if we can optionally set readcount in the new code or not.
There's other new stuff in there also.
I'm hoping this new code lets us push through the .7 freq to kh/s ratio.
https://github.com/zeusminer/cgminer_zeus
Code:
if(info->read_count>opt_zeus_readcount){
info->read_count = opt_zeus_readcount;//send a new work every 10 secondsno more
Code:
info->read_count = info->read_count * 3/4;
I don't understand it well enough to know if we can optionally set readcount in the new code or not.
There's other new stuff in there also.
I'm hoping this new code lets us push through the .7 freq to kh/s ratio.
I've not had time to read the entire thread since finding it...but I wanted to ask for those folks who applied new TIM to their Blizzards, are you seeing any benefits other than temps? I see some folks are pushing the clocks pretty high (~380-ish)...but is that with volt-mods or just new TIM applied.
Cheers!
Cheers!
Yes, you can increase your clocks quite a bit from my experience just from using good thermal paste and applying heatsinks to the tops of the chips. Adding a small fan to blow through the chamber over the circuitboard also helps.
Without the addition of the extra fan, I was able to raise my clocks from 333 with a 4%-5% hw error rate to 350 with a 4%-5% hw error rate. (roughly about 1.4Mhs).
The addition of the voltmod does allow all the way up to the max clock of 381Mhz though. We're hoping we can actually overcome that 381Mhz barrier.
I've not had time to read the entire thread since finding it...but I wanted to ask for those folks who applied new TIM to their Blizzards, are you seeing any benefits other than temps? I see some folks are pushing the clocks pretty high (~380-ish)...but is that with volt-mods or just new TIM applied.
Cheers!
Cheers!
hi sir,
sorry i want to ask you about this post,
because I do not understand this,
This tutorial to maximize the performance of fury?
if there is damage that will we get?
thank you for your answer
sorry i want to ask you about this post,
because I do not understand this,
This tutorial to maximize the performance of fury?
if there is damage that will we get?
thank you for your answer
If your know what you are doing. If your know how to solder a 0603 SMD resistor, there is almost no risk with the 7.5k resistor instead of the 9.1k. The chip will run around 1.35v. That is the upper limit in Zeus datasheet.
Extra cooling of the chips will help also.
As usual, if your aren't sure you can fix it if you fail, don't do it. Ask someone with good soldering knowledge to do the resistor swap.
hi sir,
sorry i want to ask you about this post,
because I do not understand this,
This tutorial to maximize the performance of fury?
if there is damage that will we get?
thank you for your answer
sorry i want to ask you about this post,
because I do not understand this,
This tutorial to maximize the performance of fury?
if there is damage that will we get?
thank you for your answer
This is a total NOOB question, but I am brand new to working in LINUX and have switched my miners to being controlled by RPIs.
I just need a little help with installing CGMINER 3.1.1 (for my Fury's) onto a new install of Raspbian. If someone can help and post the instructions I would be very appreciative.
Thank you in advanced.
I just need a little help with installing CGMINER 3.1.1 (for my Fury's) onto a new install of Raspbian. If someone can help and post the instructions I would be very appreciative.
Thank you in advanced.
Zeusminer makes the chips and the devices.
GAWMiners would be an OEM distributor (Fury)
ZoomHash would be a distributor (Blizzard)
Zeusminer also sells direct to anyone (Blizzard)
GAWMiners would be an OEM distributor (Fury)
ZoomHash would be a distributor (Blizzard)
Zeusminer also sells direct to anyone (Blizzard)
No, it's Darkwinde code, Cryptomining Blog build.
I think the newest nwoolls code is: https://github.com/nwoolls/bfgminer/tree/feature/zeusminer-support-cleaner
Next oldest: https://github.com/nwoolls/bfgminer/tree/feature/zeusminer-support-clean
I think the newest nwoolls code is: https://github.com/nwoolls/bfgminer/tree/feature/zeusminer-support-cleaner
Next oldest: https://github.com/nwoolls/bfgminer/tree/feature/zeusminer-support-clean
1,602 kH/s 24hr avg on pool now. 60 hours running, no restarts etc. The longer it runs the better it gets.
Is that Nwoolls build? I cloned and compiled it last night and started it running.
1,602 kH/s 24hr avg on pool now. 60 hours running, no restarts etc. The longer it runs the better it gets.
So for the time being we've reached a limit. J4bberwock is going try to break through on the hardware side. A few guys are working on software.
Can we improve on the kh/s to freq ratio?
Can anyone explain the effect of raising or lowering readcount timeout?
Current software builds use 3/4 for 6 chips or 96 chips etc. Can this be optimized for 6 chips?
Small improvements with the kh/s to freq ratio would produce substantial gains.
381 * .7 * 6 = 1600
381 * .75 * 6 = 1714
Can we improve on the kh/s to freq ratio?
Can anyone explain the effect of raising or lowering readcount timeout?
Quote
4.3 Readcount Timeout
After we send one command packet to the chained chips, that BIG chip will cover all the 32 bits nonce in a certain TIME. So we must send a new work to it before that, otherwise its hash power is wasted.
Cgminer’s readcount timeout is used for that. Every count means 0.1 second. We now set the readcount to be 3/8 of the covering time. It may be 3/4 later.
After we send one command packet to the chained chips, that BIG chip will cover all the 32 bits nonce in a certain TIME. So we must send a new work to it before that, otherwise its hash power is wasted.
Cgminer’s readcount timeout is used for that. Every count means 0.1 second. We now set the readcount to be 3/8 of the covering time. It may be 3/4 later.
Current software builds use 3/4 for 6 chips or 96 chips etc. Can this be optimized for 6 chips?
Small improvements with the kh/s to freq ratio would produce substantial gains.
381 * .7 * 6 = 1600
381 * .75 * 6 = 1714
Jump to: