CCminer(SP-MOD) Modded GPU kernels.

[scryptr]

AFTER A CLEAN REBOOT, 5.5% REJECTS--

This morning, I closed many programs and rebooted my Win 7 box.  It is an old mule, and overloaded.  On restart, I only set SPminer#4 in motion.  After a few hours, the reject rate was ~5.5%.

I then opened my CPU miner, and my web pages.  The miner ran for another hour or so with consistant performance.  But, I had to close SPminer in order to complete this post.

The point is that my Win 7 box is not a miner, it is a workstation.  I am able to run SPminer, and I mined 1 VTC+ overnight (24hr+).  But SPminer will not run efficiently on this box.  I was not able to run VertHashMiner on it at all, also.

When I get a Linux binary able to run on Linux kernel 4.15, CUDA 9/10, AMD 18.1, I'll test SPminer on a clean, dedicated mining rig.  The Win 7 box is a poor environment for the miner.

A mining rig with SPminer will likely bring 5-10 dollars a day.  There are many miners with older, smaller memory cards that won't mine ETH any longer.  Given a proper Linux binary, I could set several rigs online.       --scryptr

[sp_]

Fix 6 has been released.


fix 7. Lower the mem usage so that we can mine with xx gpu's on 4gb rigs

https://github.com/sp-hash/SPMiner/releases/tag/SPMiner

[moderator's note: consecutive posts merged]

[scryptr]

FIX #7 WAS LOOPY--

It never exited the batch loop.  It tried, but the pool connecton never really engaged.  My Win 7 box was under full load.

Fixes #4 and #6 worked, but were no big improvement.  Under full load on my PC, still about 15% rejects.

I'll keep checking in.       --scryptr

[sp_]

Submitted SPminer for the verthash algo fix 8, removed hang at startup.

It's using too much cpu on some rigs. If you get alot of rejected shares you can try the new option --skip-stale-work.  See in start_p2p-spb.bat. But some pools pay for stale work, so hashrate can drop by using this option. There is also an intensity option that can improve the speed on NVIDIA cards. i.e -i 19. -i 20. or -i 19,18,20,20,20,20,18,18 comma seperated for each card. You can also use. -i 18.9

[scryptr]

FIX #8 IS ALSO LOOPY--

But I get long strings of "skipped-stale-work" warnings prior to re-entering the batch loop.  And, is that VTC address in the examples yours?  I may try the other pools, but I don't want 2-3 VTC.  I still have 3+ from T. Pruvot's YIIMP, when the algo went from "rev2" to "rev3", and the ASICs were just waiting.  Ask TBear, he was there also.

It took 2-3 weeks to get them.

--scryptr

[sp_]

I may try the other pools, but I don't want 2-3 VTC. 

VTC is halving in 2021, might be a good idea to mine some more and hold them..

FIX #8 IS ALSO LOOPY--

Fix 9 released

Should be stable and good.

https://github.com/sp-hash/SPMiner/releases/tag/SPMiner

[moderator's note: consecutive posts merged]

[scryptr]

FIX #8 IS ALSO LOOPY--

Fix 9 released

Should be stable and good.

https://github.com/sp-hash/SPMiner/releases/tag/SPMiner

DOWNLOAD PROBLEM--

My Win 7 box upchucks, I can't download it.  It gets deleted before I can make an exception.       --scryptr

[sp_]

Fix 10  has been released. 0.3% faster

https://github.com/sp-hash/SPMiner/releases/tag/SPMiner

[djeZo]

Yo sp_ my nigga! You came back?

You have any private optimzation for any algorithm used in ethash? I am buying. Interested for cards pascal, turing and ampere only  (no maxwell shit anymore).

Also if you have any private memory timings modification code I would buy. Just send me PM if you have anything like that for sale.

Oh and Octopus algo, you have clean CUDA C++ code for it? Cos I was looking, there is only some weird unknown language stuff that I've never seen before in public miners. Also interested in speed opts regarding this algo for pascal, turing, ampere.

Basically, anyone who can handle CUDA and want to earn big bucks and not only pocket change, PM me and show me your skills

[chrysophylax]

Yo sp_ my nigga! You came back?

You have any private optimzation for any algorithm used in ethash? I am buying. Interested for cards pascal, turing and ampere only  (no maxwell shit anymore).

Also if you have any private memory timings modification code I would buy. Just send me PM if you have anything like that for sale.

Oh and Octopus algo, you have clean CUDA C++ code for it? Cos I was looking, there is only some weird unknown language stuff that I've never seen before in public miners. Also interested in speed opts regarding this algo for pascal, turing, ampere.

Basically, anyone who can handle CUDA and want to earn big bucks and not only pocket change, PM me and show me your skills

hahaha ...

You are definitely asking the wrong person there mate.

Long time no see or speak MreZo

I hope NiceHash is still treating you well!

#crysx

[sp_]

I am working on a fast ethereum fee miner.


spminer (1% fee)  for the verthash algo (vertcoin) fix 12 released with more speed on AMD cards.

Sapphire rx 580 pulse elpida 1450/2050 ~604khash -15% powerlimit
Sapphire rx 580 pulse hynix  1450/2200 ~604khash -15% powerlimit

Strap modded bios

https://github.com/sp-hash/SPMiner/releases/tag/SPMiner




[moderator's note: consecutive posts merged]

[mirny]

hmm, hmm, Sappshire it's some new vendor? lol...

[sp_]

I managed to reach 570khash stable on the rx 570 now with hynix memory. With latest spminer and bios patch(polarisbioseditor 1 click patch, and AMD/ATI Pixel Clock Patcher 1.4.9). The card is applied with rx 580 straps and the ATI Pixel Clock Patcher 1 removes the signature checking so the drivers don't fail.

Windows 8.1
core/mem 1370/2150
pl -15%

Stock clocks without biospatch is only 320KHASH

[ruplikminer]

I managed to reach 570khash stable on the rx 570 now with hynix memory. With latest spminer and bios patch(polarisbioseditor 1 click patch, and AMD/ATI Pixel Clock Patcher 1.4.9). The card is applied with rx 580 straps and the ATI Pixel Clock Patcher 1 removes the signature checking so the drivers don't fail.

Windows 8.1
core/mem 1370/2150
pl -15%

Stock clocks without biospatch is only 320KHASH

When a super fast version of ETH?

[pallas]

I managed to reach 570khash stable on the rx 570 now with hynix memory. With latest spminer and bios patch(polarisbioseditor 1 click patch, and AMD/ATI Pixel Clock Patcher 1.4.9). The card is applied with rx 580 straps and the ATI Pixel Clock Patcher 1 removes the signature checking so the drivers don't fail.

Windows 8.1
core/mem 1370/2150
pl -15%

Stock clocks without biospatch is only 320KHASH

When a super fast version of ETH?

Never, unless you can break the algo.

[ruplikminer]

I am working on a fast ethereum fee miner.

This is what I meant

[pallas]

I am working on a fast ethereum fee miner.

This is what I meant

Ether hashrate is memory bound, so unless you can break the algo there's no way to speed it up further.
Also, many have tried for years, imagine how much money there's into play.
Now, I don't mean sp_ is not a good miner coder because he is, but on eth I think there's very little chance he can do anything.

[ruplikminer]

I am working on a fast ethereum fee miner.

This is what I meant

Ether hashrate is memory bound, so unless you can break the algo there's no way to speed it up further.
Also, many have tried for years, imagine how much money there's into play.
Now, I don't mean sp_ is not a good miner coder because he is, but on eth I think there's very little chance he can do anything.

Ok then. let's see if SP can surprise us this time.

[scryptr]

I am working on a fast ethereum fee miner.

This is what I meant

Ether hashrate is memory bound, so unless you can break the algo there's no way to speed it up further.
Also, many have tried for years, imagine how much money there's into play.
Now, I don't mean sp_ is not a good miner coder because he is, but on eth I think there's very little chance he can do anything.

Ok then. let's see if SP can surprise us this time.

A GOOD LINUX BINARY--

If sp_ would only publish a good Linux binary of his VertHash code, I'd be happy.  I have a boxload of dusty R9 and RX cards that are 4GB and less.  I also have the spare motherboards to put them in play.  Right now, some of them are mining the newer CryptoNight variants.

I'd like to see if mining the VTC is more profitable.

--scryptr

[sp_]

Ok then. let's see if SP can surprise us this time.

Goals of my ethereum miner.

1. Hash at full speed on 3060 rigs
2. Create the best runtime strap mods of the memory timings. GDDR6 support.  No bios flashing needed (Nvidia/AMD supported)
3. Try to remove the decline of hashrate on older cards like the 1080ti.
4. Linux support

Create a stable and new stratum implementation from scratch with improved fee code inside. (This work is currently in progress by a friend of mine)

I will probobly not add api and monitoring.