Topic: CCminer(SP-MOD) Modded NVIDIA Maxwell / Pascal kernels. - page 863. (Read 2347659 times)

if this is true about pascal - i will setup an entire 'subfarm' to be utilized for the dev that wants to optimize for the cards ...

not only to optimize and improve the code - but also to take a percentage of all mining that comes through it for an agreed period of time ...

im looking forward to the pascal series of gpu ...

#crysx

IBM(Xilinix) and Intel(Altera ) are both working with FPGA makers to produce CPUs with FPGAS built in.

Nvidia Pascal looks to be 10 times faster than Maxwell and is expected to be released in 2016

If the Pascal specs are true, it would breathe some new life into GPU mining

FPGA's were predecessors to ASICs. It was a short lived interm between GPU mining and ASIC development when SHA256 first moved over. Scrypt skipped the FPGA step and went right to ASICs.

https://en.bitcoin.it/wiki/Mining_hardware_comparison

Sometimes they were faster then GPUs, sometimes they were slower. It's like a gateway to ASICs that allow developers a lot of flexibility without the cost associated with them. A FPGA is like a programmable CPU.

While ASICs require a huge investment to get off the ground and GPU algos move around too fast, FPGAs are more like GPUs and are pretty darn flexible. So it's highly likely that cloud mining services and huge farms may be using them, especially for a bread and butter algo like x11. Basically anyone that already is putting a ridiculously huge investment in.

FPGA CHIP DEVELOPMENT--

Ok, thanks.  An ARTIX-7 FPGA kit is available for $99, and there is a training package for $49.  Reasonably priced modern boards, priced less than used obsolete systems on E-bay.  The KINTEX-7 kits cost a lot more, thousands.  A diligent individual might be able to produce embedded code that does any algo, I guess.

The Intel Xeon Phi was billed as a co-processor, and it could be mounted in a PCI-e slot.  I was thinking that it could be used to amplify the results of available CPUmining code.  The Tesla was a USB-connected board, I believe.

--scryptr

NEW CHIPS--

OK, but not with off-the-shelf, used chips.  Chip design, prototyping, and production requires financing and sophisticated equipment.  Development would involve communication.  Do you know of a new FPGA chip architecture capable of the cryptomining process?  What about the Xeon Phi that CBuchner1 was toying with?  The Tesla purportedly had too few CUDA cores to compete with off-the-shelf video cards.

--scryptr

UNICORNS PASS WIND--

Ok, I spent 2 hours of my life researching this FPGA thing.  Here are two solid links on the subject:

1)  https://github.com/kramble/FPGA-Litecoin-Miner

2)  http://wiki.ztex.de/doku.php?id=en:ztex_boards:ztex_fpga_boards:litcoin_mining_on_fpga

FPGA modules were obsolete for Bitcoin in 2013 and discontinued.  You can buy used Lancelot modules on E-bay for $160.  They are programmable by j-tag.  Note that the best speeds obtainable with the FPGA technology produced, and now available used, was about 400MH/s SHA256, and 5 to 8kh/s scrypt.

--scryptr

Yeah also pretty certain there are x11 asics and have been for quite awhile (six months maybe longer). x11 by no means has been the money making algo for GPUs for awhile so it's not far fetched to believe there are ASICs there. x13 and x15 died with x11 mining.

I think whoever invested in ASICs thought x11 was going to take off like SCRYPT, but it never did so they decided not to sell and start a arms race. Also keep in mind there are cloud mining services that have hordes of GPUs, specifically for x11.

More likely is they're FPGAs... which are predecessors to ASICs and a lot more flexible.

Thanks for testing. Something happed between release 66 and 69.
Looks like 66 is the fastest. but the clocks are higher.

lyra2v2:

release 66: 9805
gtx 970 (core 1354Mhz, mem 1502MHz)

release 69: 9282
gtx 970 (core 1328.5Mhz, mem 1502MHz)

release 73: 9204
gtx 970 (core 1328.5Mhz, mem 1502MHz)

release 74: 9550
gtx 970 (core 1316.3Mhz, mem 1502MHz)

I forgot to say that the 2015 bios wasn't loaded..it was a 2014 bios. It should have loaded but now when done loading everything I will make sure it's upto date.

>so why the fan at 100% would kill the card is beyond me ...<  I'm not sure the card is dead yet.

>motherboards are rarely the reason for gpu failure or recognition ( though that does happen at times ) ...< it seems like it is the mb..it wont let me get pass the warning .... add the extra power cable....which is on there of course.

>as im looking at putting together a 6 x gigabyte 980ti g1 system to see how it performs<  that was what I was going to do. But I stopped at 2 980ti and 2 750ti because the new technology will be out soon plus I need some fiat for bills. As far as the psu it should take 6 980ti but a psu at 1500watts  mine is 1300watts ..depends on the algo too. 1300 should run 6 but on low power consumption algos. Was thinking of 5 980ti and one 750ti but. One thing if I needed it was to sli the psu's.
It's an altec platinum  continuous power psu..it will deliver 1300 watts continuously.

As far as oc  nvidia's sweet spot is 1345 max on stable oc ing. I have that setting on all my 750ti's..I have every make of gigabyte there is.
Their standard ..small oc..oc...and their super oc... the super oc is a card that I don't oc because it stays at 1345 max during the hash cycles. A rare card.

EDIT: My 980ti's were in the P2 state  and my 750ti are in the P0 state..I don't understand why. Or how yet to change this.
And with the new mb I will be using nvidia inspector.

1.5.66(sp-MOD) - Lyra2v2 - GTX750Ti  - 1431/1440 - 5177kh/s
1.5.69(sp-MOD) - Lyra2v2 - GTX750Ti  - 1431/1440 - 5160kh/s
1.5.73(sp-MOD) - Lyra2v2 - GTX750Ti  - 1431/1440 - 5100kh/s
1.5.74(sp-MOD) - Lyra2v2 - GTX750Ti  - 1431/1440 - 5145kh/s

1.5.66(sp-MOD) - Quark - GTX750Ti  - 1431/1440 - 7195kh/s
1.5.69(sp-MOD) - Quark - GTX750Ti  - 1431/1440 - 7251kh/s
1.5.73(sp-MOD) - Quark - GTX750Ti  - 1431/1440 - 7238kh/s
1.5.74(sp-MOD) - Quark - GTX750Ti  - 1431/1440 - 7190kh/s

thats actually quite strange ...

i dont have a great deal of experience with oc or with gpu control over nvidia ... the gpus thefarm has are all running at stock clocks and stock default values ...

i tried messing with oc for about 15mins a while back and found that with nvidia - it is very difficult ...

so why the fan at 100% would kill the card is beyond me ...

motherboards are rarely the reason for gpu failure or recognition ( though that does happen at times ) ...

let us know how you go ... i am quite interested as how this pans out - as im looking at putting together a 6 x gigabyte 980ti g1 system to see how it performs ... though it is still a while off before its built - as its a very expensive piece of hardware to put together ... what psu do you have to run the 980ti cards? ... which is the best psu to run for 6 of them? ...

#crysx

Yes exactly.
 

and then we change algos to mine something different again ... the beauty of gpu ...     yes 

But in the testing phase ...they maybe mining with lots of them right now.
For me x11 had a good profit for about 2 months. Now it's dead too me.

But crysx I got my new rig going, the one that I put precision x 16 on , and it finally recognized the 4 card.
It has 2 750ti's and 2 980ti's.
It ran for a week then I was checking my hashrate to operating temp. so I put one 980ti fan at 100%  and then the other 980ti fan at 100%. What I saw for 1 second on the fan speed on the second 980ti was 4,800 rpm    And then the system didn't recognize that card...fan run away? I don't know..but I wasn't able to lower the fan speed.
I tried updating the bios which I thought was latest but not.
The mb was stuck at " you need to put in the extra power cables "which where there of course.
So I ordered another mb and see what happens.
The 2nd 980ti maybe stuck at fan run away I dont know.

I'm fairly confident that there are x11 ASICs of some sort for a while now but the race haven't started yet so of course the first owners aren't going to sell their golden egg laying goose.
By the time we'll see the first advertisements for x11 ASICs will probably be the time their profitability dropped a lot and the race begins.

there are companies out there that have been wanting to do this for a very long time - and some companies actually succeeded from what ive read in the past ..,

but none have actually gone into production with it due to the cost of the asic systems themselves ... so far - almost all calculations based on asic design is one hell of a non-profitable ( yes - this time i mean fiat ) method of manufacture ... if it was worth the effort - it would have hit productions systems already ...

this doesnt mean they are not going to do it - it just simply means that when they can see a nice big fat fiat prifit at the end of the manufacturing cylce for x11 - they will do it ...

and then we change algos to mine something different again ... the beauty of gpu ...

...

#crysx

There is rumor of asics comig out soon that will do x11 so I consider x11 a dead algo for gpu's.  But just a rumor.

Well I can't compile and my new rig with the 2 980ti's died will have a new one today.

No ..I never made any profit on that because other algo's  are always paying more.

we dont oc the cards - so an average of 2800KH is what we get per card is what we get ...

so even a nice increase would be a really good thing ...

#crysx

Yep, you right. But if 750 Ti can achieve a 4mh\s this will be great deal.
Because current speed of 3.4 mh (overclocked) is not good for competition with amd, especially after recent DASH network hashrate raise.
Btw, noboby knows root cause of such 2x dash netwrok raise?