mersenneforum.org  

Go Back   mersenneforum.org > Great Internet Mersenne Prime Search > Hardware

Reply
 
Thread Tools
Old 2020-02-21, 22:31   #12
kriesel
 
kriesel's Avatar
 
"TF79LL86GIMPS96gpu17"
Mar 2017
US midwest

66268 Posts
Default

Quote:
Originally Posted by docustompc View Post
Great stuff
...
It seems mfakto for amd is not complied and built? i have trailed through the forums and couldn't find a compiled / built version? ...
I have looked
...
Thankyou for your continued help!!
Computing applications: http://www.mersenneforum.org/showpos...91&postcount=2 Download and read the pdf attachment. Learn what you can run on your AMD gpus. And where the software mirror, etc are. (Downloads of precompiled)
Management applications: http://www.mersenneforum.org/showpos...92&postcount=3 Download and read the pdf attachment. Features and compatibility summary of software aids for handling "manual" (non-PrimeNet) assignments and results.

Note that both of these posts are within the dozen first reference links of https://mersenneforum.org/showthread.php?t=24607 I advised you to read the other day. OP = original poster = docustompc https://mersenneforum.org/showpost.p...63&postcount=8
Sheesh, this is sort of like having to stick the newly acquired puppy's nose in the food bowl, and in the water bowl, so he notices where the essentials are to be found. It's not enough to set him down in front of them, while he's distracted by an entire new to him house.

Last fiddled with by kriesel on 2020-02-21 at 22:32
kriesel is offline   Reply With Quote
Old 2020-02-21, 22:51   #13
docustompc
 
Feb 2020

23 Posts
Default

all is going well
i have a 'another' question
i have my gtx970 doing 520 odd ghz-d/day
and now
I have a gtx1650 doing almost 1000 ghz-d/day
using less than half the power
I understand the less power but not the higher work output?

Thanks guys!!
docustompc is offline   Reply With Quote
Old 2020-02-21, 22:57   #14
docustompc
 
Feb 2020

816 Posts
Default

Quote:
Originally Posted by kriesel View Post
Computing applications: http://www.mersenneforum.org/showpos...91&postcount=2 Download and read the pdf attachment. Learn what you can run on your AMD gpus. And where the software mirror, etc are. (Downloads of precompiled)
Management applications: http://www.mersenneforum.org/showpos...92&postcount=3 Download and read the pdf attachment. Features and compatibility summary of software aids for handling "manual" (non-PrimeNet) assignments and results.

Note that both of these posts are within the dozen first reference links of https://mersenneforum.org/showthread.php?t=24607 I advised you to read the other day. OP = original poster = docustompc https://mersenneforum.org/showpost.p...63&postcount=8
Sheesh, this is sort of like having to stick the newly acquired puppy's nose in the food bowl, and in the water bowl, so he notices where the essentials are to be found. It's not enough to set him down in front of them, while he's distracted by an entire new to him house.
SO Sorry I have a fair amount of puter horsepower available and not so much free time!
really sorry if you have already answered this

I have 4 children under 10 and me and my partner have full time jobs

Fitting in geeky s*** is difficult!!

Sorry if I have offended anyone with my lack of attention to your posts

I'm sure the info i need IS in the boards I just struggle to find it!!
docustompc is offline   Reply With Quote
Old 2020-02-22, 00:14   #15
kriesel
 
kriesel's Avatar
 
"TF79LL86GIMPS96gpu17"
Mar 2017
US midwest

2×37×47 Posts
Default

Quote:
Originally Posted by docustompc View Post
i have my gtx970 doing 520 odd ghz-d/day
and now
I have a gtx1650 doing almost 1000 ghz-d/day
using less than half the power
I understand the less power but not the higher work output?
1) https://www.techpowerup.com/gpu-spec...-gtx-970.c2620 2014 technology 28 nm
https://www.techpowerup.com/gpu-spec...gtx-1070.c2840 2016 tech 16 nm
https://www.techpowerup.com/gpu-spec...gtx-1650.c3366 2019 tech 12 nm
AKA Moore's law, the incredible shrinking transistor
There's nearly a 1000:1 range of GhzD/d/watt in https://www.mersenne.ca/mfaktc.php?sort=gpw

2) GTX x50 tends to be performance/power optimized;
x60, x70 x80, etc increasingly sacrificing power efficiency for more performance per gpu

Last fiddled with by kriesel on 2020-02-22 at 00:16
kriesel is offline   Reply With Quote
Old 2020-02-22, 14:08   #16
LaurV
Romulan Interpreter
 
LaurV's Avatar
 
Jun 2011
Thailand

100000110011112 Posts
Default

First generation steam engines consumed about two times more coal and ran at about half speed compared with diesel engines that followed. What is not clear? The OP don't need to read any post.

GPUs are *not* the same. Different builds, different architectures, Fermi vs Kepler vs Pascal vs Volta etc.

@OP, just in case you find the time, click here. It says everything.

Last fiddled with by LaurV on 2020-02-22 at 15:59
LaurV is offline   Reply With Quote
Old 2020-02-22, 14:20   #17
jwaltos
 
jwaltos's Avatar
 
Apr 2012

2·167 Posts
Cool

Quote:
Originally Posted by LaurV View Post
First steam engines consumed about two times more coal and run about half speed compared with diesel engines that followed. What is not clear? The OP don't need to read any post.

GPUs are the same. Different builds, different architectures, Fermi vs Kepler vs Pascal vs Volta etc.

@OP, just in case you find the time, click here. It says everything.
Are there any emulators you are aware of that will perform better than a cpu (within any operating system environment)? I came across an emulator reference that stated this claim and had bookmarked it about 2 years ago and I respect your views as a pro.

I'm a strong proponent of theoretical advances than pushing consumer grade techonology to its breaking point(s) which includes highly optimized code regarding the same.

Steam power was available during the Roman era but was not exploited and fusion is not yet a working reality.
jwaltos is offline   Reply With Quote
Old 2020-02-22, 17:10   #18
LaurV
Romulan Interpreter
 
LaurV's Avatar
 
Jun 2011
Thailand

37·227 Posts
Default

Quote:
Originally Posted by jwaltos View Post
Are there any emulators you are aware of that will perform better than a cpu (within any operating system environment)? I came across an emulator reference that stated this claim and had bookmarked it about 2 years ago and I respect your views as a pro.

I'm a strong proponent of theoretical advances than pushing consumer grade techonology to its breaking point(s) which includes highly optimized code regarding the same.

Silly me, I wanted to say gpu's are NOT the same. Because they are not.

I edited my own post, but I wouldn't dare to touch yours in which you quoted my full post, in spite of the fact that quoting the precedent post in full is not considered "good netiquette", and that's why my mistake, now, will last forever, I can't hide it under the carpet, and my grandchildren will make fun of me. Your fault

On the other hand, what do you want to say by emulators? Like "physical, tangible, hardware piece of junk", CPU emulators, done by FPGAs, or by other means? There are, for sure, and they all perform better than the CPUs they emulate, because additional of executing all (or almost all) functions the emulated CPU does, the emulator has a lot of more functionality for step-by-step debugging or tracing the code, seeing all the internals of the emulated CPU in real time and/or save them in files for later inspection, they have higher clocks, even up to 10 times the device they emulate, etc. So, yes, there are. But the two BIG problems with these emulators are: (1) they cost anything between 5 times and 5 thousand times the price of the device they emulate (this is not a figure of speech, an emulator for a Samsung MCU that costs $0.7, may be as expensive as $3500), and (2) they are power hungry in the same proportion (again, not a figure of speech, in the worst case scenario imagine one hypothetical CPU that consumes 10 mW, and contains a billion transistors, built with ten thousands discrete parts which consumes each of them a tenth of a mW, the power consumption would still be one thousand times higher - of course, this is not a real case, where FPGAs would be used, but those are still power hungry beasts, for such tasks).

Hardware emulators were very common 10 years ago and before, because the processors and controllers were not so "clever" and you needed an emulator to develop system code for them (i.e. in systems without any kind of OS, for tracing, step debugging, etc, ) or to develop them. Nowadays, even the cheapest chip (no pun intended) has a SWIM, or SWD, or JTAG, etc interface and you don't really need any kind of emulators to write code for the respective device from scratch. Complex MCUs offer full trace interfaces, where you can see any time what your MCU is doing and what he did in the last five years, and if his parents know.

But there is still a domain where you can't get rid of an emulator, even if that is not always called so, and this is when you make a new device. One of our customers developed some time ago their own CPUs which are 80 bits (don't ask me why!) and for that, they used some Xilinx'es mounted on some PCB, to do all the developments and tests, before going to foundry with it. That PCB, with the belonging firmware (called microcode), I would call "emulator".

If you think about software emulators, well, in theory, any device can emulate any other device, if you give up the requirement that the emulation be in real time.

To have real time emulation in software, the emulator's speed has to be one order of magnitude larger than the emulated device, in average. This highly depends of the instruction sets and abilities of both, like RISC vs CISC, caches, pipes, etc, but it is a good rule of thumb. Think about the fact that Sinclair Spectrum (Z80, 3.5 MHz clock) emulators existed on PCs from the very beginning when Z86 and 8086 appeared, but they didn't really look "real time" until PCs advanced over 30MHz, in spite of the fact that new processors were more powerful clock-per-clock.

So, with the machines we have today working close to 4 GHz, we could write some software to emulate machines having the power of Pentiums at 400MHz or so... this is a paradox that always fascinated me. And no, I don't talk about virtualization, and having 5 guest computers running in one host, from which 2 have faster clocks than the host, and 3 have more memory, and 4 have larger harddisks or other trickery you can do nowadays, I am talking about a machine with a totally different instruction set, computationally equivalent with a P400. Think about it.

Of course, the things work the other way around too. Give me enough time and I can use my old Cobra computer and a lot (and I mean A LOT) of cassette tape, to write a i7 emulator. Then I can run Win7 on it, start P95 and tell you if M31 is prime or not, in a couple of decades. Hope to live till then.

This in fact, we do everyday, with physics studies, where the scientists take the fastest supercomputers in the world, fill them with exabytes of data of some physical phenomenon, like particle kinetics for example, let them run for days and weeks, and at the end then come out with how each particle moved for 10 milliseconds. So out fastest supercomputers are just stupid slow emulators of a very small part of the nature... (yeah, yeah, I know, don't give me the counterexample with the colliding of the galaxies... )
Quote:
Steam power was available during the Roman era but was not exploited and fusion is not yet a working reality.
My steam-related riddle was the response to OP's question "why a Fermi/Kepler card takes two times more power but gives half performance compared to a Volta/Turing card". I guess I know that you needed to shovel the coal continuously under the steam boiler, and the proportion to Diesel was not only 2 to 1...

Last fiddled with by LaurV on 2020-02-22 at 17:26
LaurV is offline   Reply With Quote
Old 2020-02-22, 18:26   #19
Uncwilly
6809 > 6502
 
Uncwilly's Avatar
 
"""""""""""""""""""
Aug 2003
101×103 Posts

778710 Posts
Default

Quote:
Originally Posted by LaurV View Post
First generation steam engines consumed about two times more coal and ran at about half speed compared with diesel engines that followed. What is not clear? The OP don't need to read any post.
How much coal does a SD70ACe/LW burn?
Uncwilly is offline   Reply With Quote
Old 2020-02-22, 18:33   #20
LaurV
Romulan Interpreter
 
LaurV's Avatar
 
Jun 2011
Thailand

37×227 Posts
Default

Seeeee? I knew it !

I knew you will do this to me!

I go to sleep. It is 1:30AM here, and I am supposed to be sleeping for a while, as I am in convalescence after a week with fever 39... (this is not a joke, that is why noone saw me here for a week). Luckily SWMBO is sleeping upstairs, otherwise you could say bye to me forever...
LaurV is offline   Reply With Quote
Old 2020-02-22, 18:57   #21
kriesel
 
kriesel's Avatar
 
"TF79LL86GIMPS96gpu17"
Mar 2017
US midwest

2·37·47 Posts
Default

Quote:
Originally Posted by Uncwilly View Post
How much coal does a SD70ACe/LW burn?
In normal operation, about the same as a Westinghouse AP1000. The electrical output per unit input of coal is not well defined.

Last fiddled with by kriesel on 2020-02-22 at 18:58
kriesel is offline   Reply With Quote
Old 2020-02-22, 20:01   #22
jwaltos
 
jwaltos's Avatar
 
Apr 2012

1010011102 Posts
Default

Thanks LaurV. In the mid 90's I cobbled together some 4xx hardware/software (IBM clones /Windows/Linux) where I had to purchase a math co-processor emulator. Graphics card programming was not really on my radar but spreadsheets were. At that time I had some interesting F77 code that simulated colliding galaxies that partially worked on my home computer but I had to really pare it down and even then it would make Mandelbrot's initial printout of the Julia set look like a masterpiece..
I appreciated your comment regarding the complexity of modelling aspects of reality (as we perceive it).

Last fiddled with by jwaltos on 2020-02-22 at 20:03
jwaltos is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Many (gmp-)ecm questions... WraithX GMP-ECM 26 2014-04-03 15:54
GPU LL Questions Primeinator GPU Computing 40 2013-10-20 06:20
Some questions... OmbooHankvald PSearch 3 2005-09-17 19:29
5 questions OmbooHankvald Factoring 6 2005-08-28 19:31
Questions OmbooHankvald Prime Sierpinski Project 2 2005-08-01 20:18

All times are UTC. The time now is 06:29.

Tue Apr 7 06:29:40 UTC 2020 up 13 days, 4:02, 2 users, load averages: 2.06, 2.13, 2.05

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2020, Jelsoft Enterprises Ltd.

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.