Topic: [ANN] [QRK] Quark | Core 0.10 upgrade - page 390. (Read 1031025 times)

I have an AMD Sempron 145 Sargas 2.8GHz (single core) w/ "core unlocked" and overclocked a bit (with that motherboard's "OC Genie") that gets about 220khash..  my i7-960 (it's not overclocked, but I have turbomode enabled, so it does that automatically, goes to 3.35ghash or so) can get about 600-620khash/s tops..  i7-4770k is 1mhash+, but well, it costs a lot more than what you'd need for equivalent from AMD CPU

oh, *intel e6850 gets around 200khash w/o overclocking

The hashrate is a linear frequency function (what a rocket science). For quark algo just divide frequency by 27.5 for no HT, or by 40 for HT to estimate Khashes PER THREAD. If you run less threads than CPUs on HT, the number is somewhere in between. E.g. for 3 threads on 4-thread CPU or 6 threads on 8-thread CPU it is ~35, see https://bitcointalksearch.org/topic/m.3076346. It would be nice if various CPU owners will try the formula and prove or disprove it.

The hashrate is a linear frequency function (what a rocket science). For quark algo just divide frequency by 27.5 for no HT, or by 40 for HT to estimate Khashes PER THREAD. If you run less threads than CPUs on HT, the number is somewhere in between. E.g. for 3 threads on 4-thread CPU or 6 threads on 8-thread CPU it is ~35, see https://bitcointalksearch.org/topic/m.3076346. It would be nice if various CPU owners will try the formula and prove or disprove it.

and i'm telling you AGAIN it crashes on my Intel CPU (both versions in the archive)
i downloaded it before and tried them all and i re-downloaded just now and tried them all..
arguing doesn't make code run lol

Intel Core 2 Duo E7500 = fail / crash on start up - win 7 x64
and yeah i checked and these are 64bit builds not x86

and your welcome for the constructive feedback and time and effort i took to test this everyone

edit:
I'd love to know how that guy is compiling code from GitHub.. all its giving me is compiler errors with mingw on windows
i can get various previous version of the code to run just fine..
i often see "get the tarball" but how am i suppose to find that ?
all i can see on the github page is the button that says download zip or that other page with the .zip + targz
and that can't be the tarball because i see these devs constantly telling people asking for help compiling it code to not use it and use the tarball
when autogen.sh fails etc.. so obviously the tarball should have more files inside the archive in it so you don't have to run that command right ?
this quark miner stuff is severe pain in the ass for a noob to come along and get the newest version and compile.
these miners should have dedicated topic or do they already ?

bottom line: I'm figuring it all out on my own anyway, so whatever keep scrolling guys lol
started in on linux and c/c++ well over a decade ago and its sucks balls i can't use VS as an ide
so yes i am a noob but no i am not.. i may start a new thread soon with a tutorial on building this on windows for people..

Did some benchmarks of the sse4 miner (just took this version https://bitcointalksearch.org/topic/m.3047868) on a bunch of Intel Core2 Duo and Ivy bridge, and here comes the great equation.

The hashrate is a linear frequency function (what a rocket science). For quark algo just divide frequency by 27.5 for no HT, or by 40 for HT to estimate Khashes PER THREAD. If you run less threads than CPUs on HT, the number is somewhere in between. E.g. for 3 threads on 4-thread CPU or 6 threads on 8-thread CPU it is ~35, see https://bitcointalksearch.org/topic/m.3076346. It would be nice if various CPU owners will try the formula and prove or disprove it.

I do not have any AMDs at hand, but it seems they follow "no HT" path for ALL cores (https://bitcointalksearch.org/topic/m.3021817). Thus, high-frequency AMD FXs are winners in terms of kh per CPU. I am getting 780kh on i7 3770k at 4GHz with HT (8 * 98), while FX 8350 gives 1100kh at 4 GHz (8 * 140). Here power consumption comes into play. For Intel it is ~8kh/watt for IB and ~2kh/watt or less for C2D. Can anyone provide wattage for AMDs (real-world numbers, not from online benchmarks)? Sandy Bridge and Haswell numbers would be usefull as well. Lets make quark hardware comparision table 

you should be able to get faster on the i7-3770s

my i7-960 gets 600khash on 8 threads (on turbomode), i7-4770k gets like 1mhash

if it gets to 100oC it'll start to throttle itself

Sorry, I'm not quite the wizard with autotools. Anyway, compile it on either linux, or someone has been successful with mingw or cygwin. Autotools doesn't do MSVS. SO HERE:

git clone https://github.com/uncle-bob/quarkcoin-cpuminer.git  
cd quarkcoin-cpuminer  
sh autogen.sh  
./configure CFLAGS="-msse2"  
make  

-msse2 is not optional, should be built into the build scripts. However, I've avoided getting good with autotools this long, might as well not start now. Really anyone have a simple examples for autotools? They seem to all turn into a huge pita quickly. Step two might be cmake or anything else.

really though,  the code at:  

http://stonefoz.myfastmail.com/  

is probably the best, first release where fast is first.
I've been monkeying around with stuff on github, should be faster later, just not anytime soon. Rewriting large bits and it's going to be broken some times.

Using: https://github.com/uncle-bob/quarkcoin-cpuminer

E3-1230 (quad-core, HT)
90khash per thread, 540khash on six of eight threads.

i5-3570S (quad-core)
130khash per thread, 400khash on three of four threads.

i7-3770S (quad-core, HT)
80khash per thread, 520khash on six of eight threads.

Xeon X3440 (quad-core, HT)
66khash per thread, 400khash on six of eight threads.

i3-2130 (dual-core, HT)
100khash per thread, 300khash on three of four threads.