Archive : : Unit of Measurement! or... WATTS HAPPENING!
Posted By: bbrv. on 2004/2/21 16:59:27
We are looking to all you smart people out there to help us come up with a new concept of measurement to describe "PegPower." :-)

BBRV give further comments on this item (see comments!)

(example: Now, coming back around to the basics, because it has to be simple. The idea of an online Power Testing Center (site) could be interesting.)

We all know the PowerPC is an efficient CPU and that the Pegasos has a much better MIPS per watt number than today's x86 CPUs. Heat dissipation also becomes a real problem for PC's now, as air cooling is not likely to be not enough for a greater than 4 GHz clock speed. We could measure the power consumption with external equipment, but that does not tell the whole story either.

Then, there is the efficiency of the code itself, so just watts per number of digits processed is not the whole story either. For example, our distributed.net client beats even the 3.2 GHz x86 on a 1 GHZ G4.

We need to something catchy to dispell the MHz myth. We might as well throw the my memory is bigger than your memory misnomer out too. It is partly a marketing thing and partly a genuinely accurate indication that PegPower is really something!

We need way to describe the efficiency, speed and work accomplished in a way that puts out out front in terms of performance where we are.

Thanks to Krashan today for the great conversation about this! Please post your ideas here!

R&B :-)
 
  • Acolyte of the Butterfly
    Acolyte of the Butterfly
    Donar
    Joined: 2003/12/27
    Posts: 142
    From: Germany
    First i want to say that we should take already available measurement methods like MIPS/MFLOP or something like that. If we make our own measurement we will be accused of posting marketing crap.

    The 1 GHz G4 is faster than the 3.2 P4 in distributed.net?
    Fine! If we take this to the "Battlefront" on Ars Technica, there will be one guy who knows an application that is faster on the P4. OK the P4 has much more GHz but we come to this later....

    For me take MIPS/MFLOPS (no Hz in that :-) ). Maybe set it in association to MIPS per Watts consumed.
    So you can say that the Pegasos is cheaper and more powerfull as an home server running 24/7. (Electricity bill).

    For the people who after the mention of MIPS/MFLOPS say "This can't be true the P4 has much more GHz"
    You could state that there is a difference between
    the two architectures. In work done per clockcycle.

    I can imagine that the distributed.net client is so much faster on the G4 because of AltiVec, and this should be part of the MFLOPS measurement, right?
    So i can't state if it is because the code is so much better.
    Be aware my knowledge of this is limited to things i read in my spare time. If something i posted is not 100% feel free to set it right.

    Tried to make this short, if i made it too short to understand. I will come back later and have a look.
    :-D
  • »2004/2/8 9:47
    Profile
  • Order of the Butterfly
    Order of the Butterfly
    liquidbit
    Joined: 2003/10/12
    Posts: 404
    From:
    Many schientists have tried to do that before.
    But not all the people are schientists thats why they failed to convince the average users.
    Calculations per second is important for mathimaticians but not for 3D designers.
    Today speed is the most common middle of measurement and the most important factor of purchasing a computer.
    Maybe because people seeing computers as cars.
    But cars have only one way of use: Moving in one direction to another.
    Measurement is easy because it has one dimension of measurement, and that is speed.
    Computers have *n dimensions depending of the requirements of the user. Thus multiple dimensions,measurements and needs.
    It is too difficult to define because of the variation of users requirements.
    My types of measurement looking more on the side of an average user.

    Factors of purchasing a computer system by an average user:
    1.Price.
    2.Variation and plethora of applications.
    3.Speed of applications & job processes.
    4.Trend.
    5.Design & Look.
    6.Easy to use.

    I hope you got my point.

    Thanks for listening,

    Steve
  • »2004/2/8 12:42
    Profile
  • Cocoon
    Cocoon
    palpatine
    Joined: 2003/3/7
    Posts: 49
    From: Rotterdam, The...
    Hi,

    First thing that comes to mind is that it should be a general term that sounds catchy like MIPS or FLOPS, i.e. easy to pronounce (in all languages).

    Secondly, it should basically describe the total computing efficiency of a computer solution.

    Something like Index of Computer Efficiency (ICE).
    "What is your PC's ICE score? Only 100? and it's a 3 Ghz.? Mine is a G3/600 and it scores 250!!"

    Cheers,
    Ron
  • »2004/2/8 12:47
    Profile
  • Cocoon
    Cocoon
    palpatine
    Joined: 2003/3/7
    Posts: 49
    From: Rotterdam, The...
    Continuing on the above, if there was to be a new "index" that described the total power of a platform/machine, it should be supported by a big IT brand name that has something to gain by showing off the PowerPC CPU (not necessary to mention the obviuous names here).

    In addition, people should be able to test their own configuration and get an index number for it, by internet. Surf to www.ICEtest.org (or whatever) and see how your machine performs. And at the same time you can check the best scores on the site and compare to what you got.
  • »2004/2/8 12:53
    Profile
  • Priest of the Order of the Butterfly
    Priest of the Order of the Butterfly
    Stevo
    Joined: 2004/1/24
    Posts: 839
    From: #AmigaZeux
    Hmmm,

    The way I see it is that it is next to impossible to create a standard to express the (objective) perfomance of a computer. Different CPU's, different OS'es, different programs etc. Instead, I propose SAPS (Subjective Allround Performance Score), a bit like the ICE proposal by palpatine.

    Different variables should be created for every aspect of computing, ranging from speed to graphical capabilities to user-friendlyness to software availability etc. All known computer setups should be scored on those variables by three groups of test-subjects: experts, users and "aliens" (people who have never worked with the given computer). Then, a big group of random test subjects should be given the same variables and then be asked to give weights to the variables. Some will find graphical abilities more important than speed and some will find it very important to be able to play the latest games.

    Mean-SAP scores can now be calculated for the different computer setups or specialized SAP-means for diffents areas of computing (like graphics. games, etc). Futhermore, individuals can culculate a personal SAPS profile by giving weights to earlier mentioned variables and find a matching computer setup.

    Purely theoretical, and by all means not a "serious" proposal, but it is in the subjectivity of computer usage and expectations that a "Unit of Measurement" should be found and not in the "objective" figures like speed etc.
    ---
    http://www.iki.fi/sintonen/logs/its_only_football.txt
  • »2004/2/8 15:32
    Profile
  • Priest of the Order of the Butterfly
    Priest of the Order of the Butterfly
    Robin
    Joined: 2003/2/24
    Posts: 741
    From:
    Well, compare it to cars ... ppl dont care that a
    porsche needs more fuel, is more expensive.
    Everyone want to drive one ... (well ... ;) )

    They dont care that Toyota is leading in all
    security statistics. They dont care that
    peugeot has great engines.

    They get the car they can afford and where
    they can visit the LOCAL dealer. Where they
    get the most out of their money ...

    Make a winbox next to the abox supporting
    the standard apis of windows. Thats the
    only way to get more ppl to look into
    the pegasos. Today ppl steal their software
    way too often. And the biggest source to steal
    software currently is the win-market.

    Thats why gaming systems with cd's won over the
    one's with cartridges.

    I know that sounds a little frustrated ... but hey
    I am frustrated atm ;-)
  • »2004/2/8 16:07
    Profile Visit Website
  • Moderator
    gunne
    Joined: 2003/2/26
    Posts: 441
    From: Sweden
    Hi

    I believe 'Risc' is a very good word.

    My points for this is;

    1; Its something some people know about.

    2; Its something that show that there is difference between different CPU's eg -> x86

    3; It can be something that more people (who have a common interrest in computers but not for now know exactly what it means) want to know 'what it means'.

    4; Because of three its something that can be used to explain 'what it is and the differences with other CPU's in various ways.

    5; Because of three again, it have some kind of 'want to discover factor'.

    5; It can be combined with different 'slogans'.

    6; Its common for many platform that uses the same CPU, except them that uses another CPU = no cool computing :-).

    Well, try to take a spin on it, and smell for a while...

    Greetings
    Gunne
    Best wishes, Gunne
  • »2004/2/8 23:04
    Profile Visit Website
  • Moderator
    gunne
    Joined: 2003/2/26
    Posts: 441
    From: Sweden
    ... and

    Perhaps PegPower should be turned the other way around... PowerPeg :-)

    About Risc

    About Cisc
    Best wishes, Gunne
  • »2004/2/8 23:12
    Profile Visit Website
  • Priest of the Order of the Butterfly
    Priest of the Order of the Butterfly
    bbrv
    Joined: 2003/2/14
    Posts: 750
    From: Earth
    Those are all good comments. Thanks!

    The idea of making a CISC vs. RISC comparision by elimination might be a good idea, Gunne. If the first world is RISC, CISC is irrelevant. On the other hand, by omitting CISC you have eliminated 95% of the installed base you want to compare yourself to in the first place...:-)

    Keep thinking!

    R&B :-)

    P.S. Ron, link does not work here...?
  • »2004/2/9 12:41
    Profile Visit Website
  • Moderator
    gunne
    Joined: 2003/2/26
    Posts: 441
    From: Sweden
    Well, I was thinking something like this way...

    First;

    If people are considering a non-wintel based solution. What is there to choose from ?

    Well, then you could perhaps think about that there is two solutions to choose from out of today - Apple or Pegasos. Soon also perhaps AmigaOne, which then will make it three, but remember right now sells as earlybird.

    On this level Risc is what it is based around, and whats the benefit by using a Risc-based system comparing with a Cisc-based.

    Then next;

    You come to where to choose between different risc-based solutions.

    Then when I try to look on it from the outside, I can perhaps see something like this.

    What is Apple ? well, I got the feeling its most about MacOS. There is no info on Apples webpage except about MacOS. If you want to look for other systems, to run on your Apple computer, you have to find this in other places.

    Also when looking at the AmigaOne, you might get the impression it seems to be mostly about AmigaOS.

    Then when you look at the Pegasos, you will see that it not only seems to be about MorphOS, but also about that there is much info around about other solutions. Also when you, as coming from the outside, and get in contact with people in the community, you might perhaps also got the impression that people are using different systems, and also talk about different solutions.

    This is then more somewhat like what the x86 stands for, you can have different solutions, using nearly exact the same hardware.

    My feelings is that this is what it is a about for Genesi, and also for the Pegasos, and also what it is the benefit with the Pegasos.

    I dont know, whats right or wrong to do, but as from speaking with people, and then trying to hear what they say and ask for this is the impression I get.
    Best wishes, Gunne
  • »2004/2/9 13:43
    Profile Visit Website
  • News Moderator
    News Moderator
    Darth_X
    Joined: 2003/2/10
    Posts: 571
    From: Vancouver Isla...
    I think security and reliability are the 2 most important things.

    for example, a few days ago a virus/trojan was discovered on my PC, which caused reliability problems. Now its gone, thanks to good virus scanning software... but still.. this is a very serious issue on Windows.

    And Microsoft for all their billions does not seem to care about this threat. Do you see updates for Outlook Express or MSN.com for improved security and email filtering to remove the trojans embedded in emails? Not so far...
    When you have eliminated all which is impossible,
    then whatever remains, however improbable, must be the truth!!! - Sherlock Holmes
  • »2004/2/9 15:57
    Profile
  • Just looking around
    yyoru
    Joined: 2004/1/19
    Posts: 13
    From: Turkey
    There must be a global benchmark apllication which runs on many OS's, specially Windows, MacOS, AmigaOS, MorphOS, Linux ..This benchamrk utility should be opensource. All of algorithms anda data (e.g. textures, matixes etc. ) should be the same. These algorithms must include 2Dgraphics, 3Dgraphics, sound, office tools, desktop, and math tests.

    And then this program will ask you what kinda user u are ? I think there are four kind of users.
    These are;

    Mathematicians (Science, Eng, Math etc..)
    Graphicer (2D & 3D game players,3D solid & vector deginers, analyzers)
    Officers (Office tool users, e.g. secretaries)
    Network users, (wastes much more time & internet, e.g. servers)

    So you shuld select what you are?
    e.g. I am:
    Math : %80
    Game: %60
    Office: %40
    Graph: %80
    Network: %40

    Within these parameters there must be translation formula for tests. So benchmark application will display Benchmark result which are also mutipleyed buy its percent.

    By the way user can select what kinda platform is best for him.

    Todays benchmark tools are not honest. E.g. Mark3D, says that some cards much better then my GF4 MX440, but we calculated that FPS rate pf my card is larger in many cases. These are also could be marketing reasons.

    On the toher hand, for the best marketing of MorphOS & Pegasos I think these features must be perfect. In order;

    1. Latest technolgy support. Memorysticks, Palms, usb hardrivers, digital cams etc.

    2. Latest update support in many cases. e.g. DivX,
    Mp4,OS update, application upda, security update etc. And Multimedia aplications like Kaya.

    3. Office Tools, specially Word & Excell like tools.
    These are important for many users. And these tools must be free, e.g. openoffice. And it must be strongly compatible with docs, xls etc.. (most used & latest document types)

    4. Fully Internet Support, (Java, Asp, Cgi, etc.) & Browsers, eMail, FTP, IRC, TCP-IP utilities, PDF reading, network compatibility.
    (e.g. playing doom with a Peg & PC via network)

    5. Security & Latest Security update support

    6. Games & Tools for understanding power of OS.


    The latest and largest opration systems, WindowsXP, Mac-OS, Mandrake , Redhat includes all of these steps in many ways..

    So we have some missings...

    :)

    YILMAZ YORU
  • »2004/2/10 15:38
    Profile Visit Website
  • Targhan
    Joined: 2003/2/8
    Posts: 2833
    From: USA
    I've thought of this before, and the answer is still "marketing." We just need to create something that can record the amount of time it takes to do the following:

    1. open an application
    2. open a directory via the icon with a given number of icons.
    3. how long it takes to close an application
    4. how long it takes to decode a given animation/movie file.

    Maybe more ideas to throw in there. I mean REAL-TIME, not clock cycles but seconds. Add the number of seconds together, divide by "activities," and we call the method of measurement SPA, or "Seconds Per Activity." The lower the number, the better the machine. This can be used to compare OS's on the same hardware platform, the same OS's on different hardware, is generic, and is completely "top level" and understandable.

    SPA: How low can you go?
    :idea:Targhan

    MorphOS portal? www.MorphZone.org
  • »2004/2/10 19:50
    Profile Visit Website
  • Order of the Butterfly
    Order of the Butterfly
    cecilia
    Joined: 2003/8/30
    Posts: 459
    From: universe, milk...
    yeah, "real world use" is what is important!
    and before anyone forgets: how about how long it takes to boot up?

    anytime i show people this peg i have, they look like they have been hit over the head with a brick as they see how quickly it boots. cold OR warm!

    hm...the Brick Effect.

    8-D
    "if you ever slam anyone, for anything, somehow you always end up eating shoe" Targhan
  • »2004/2/10 20:33
    Profile Visit Website
  • Acolyte of the Butterfly
    Acolyte of the Butterfly
    downix
    Joined: 2003/2/10
    Posts: 105
    From: Lightning capi...
    @Palpatine

    ICE... I like it.
  • »2004/2/13 15:51
    Profile Visit Website
  • Priest of the Order of the Butterfly
    Priest of the Order of the Butterfly
    bbrv
    Joined: 2003/2/14
    Posts: 750
    From: Earth
    These are all good ideas...

    Now, coming back around to the basics, because it has to be simple. The idea of an online Power Testing Center (site) could be interesting.

    We are trying to eliminate CPU speed from the marketing discussion/buying decision and actually come up with our own paradigm in which we reign supreme. :-) The issue is what WORK (activity exerted through power or faculty applied to perform a task) can be accomplished in the time allotted -- not how just fast we ran to accomplish the work!

    Work is the force applied to an object to create a displacement. Power is the work done in terms of the time required. Power provides a measure of both the amount of work that is done (or, equivalently, the amount of energy expended), and the time it takes to do it.

    WATTS HAPPENING! PegPower! That next iteration of watts happening is what we are looking for!

    Perhaps, think about the progressive orbitals of molecular construction. The orbital names s, p, d, and f stand for sharp, principal, diffuse, and fundamental.

    The orbital letters are associated with the angular momentum quantum number, which is assigned an integer value from 0 to 3. s correlates to 0, p = 1, d = 2, and f = 3. The angular momentum quantum number can be used to give the shapes of the electronic orbitals. s orbitals are spherical; p orbitals are polar. It may be simpler to think of these two letters in terms of ORBITALS (d and f aren't described as readily). Which is sort of our problem now!

    OK, forget the last one and the Chemistry lesson! :-P (but you get the idea! No?!)

    R&B :-)
  • »2004/2/13 21:08
    Profile Visit Website
  • JKD
  • Order of the Butterfly
    Order of the Butterfly
    JKD
    Joined: 2003/4/4
    Posts: 448
    From: South of heaven
    I might have said this sort of thing before, but the only possible way to make headway is to talk about Applications. Recent benchmarks have begun to focus more on what you can do with your computer (witness Apple's megahertz myth campaigns.) Although such campaigns are intevitably doomed I thin there is a valid point in there somewhere and it is actually Workflow!

    Workflow comparisons allow the following:

    1. Prove that the platform is capable of doing similar tasks to the bigger players
    2. Can play to the strengths of a system with a quick/efficient UI.

    A sample worklflow might be importing a TIFF file, scaling, color correcting exporting to xxx and then saving to a networked resource.

    The workflows can be different based on availability of :

    1. Keyboard shortcuts
    2. UI Apps e.g. Docks, DockIcons that facilitate it.

    Workflow operations can be quoted in quantities per unit of time etc. and then converted into a quantity per GHz type measurement or 'pseudo ROI' of intital cost of platform

    Now I have no ide whether a Pegasos might win any of these and it's close to your infamous Power Analogy....just need to identify what certain Applications need and come up with an appropriate workflow bechmark (instead of singling out obscure ones like MIPS, dnetc etc...see next paragraph.)

    I think any arguement based around 'megahurts (sic) myth' is doomed to failure (a bit like Apple or whomever singling out the dnetc benchmark or some obscure PS operation)...I wouldn't touch CISC/RISC either nor any utility comparison e.g. cost of electricity to run an Athlon box vs a low power PPC....utility bills are a fact of life. The cost saving alone of rendering on a 3GHz box would probably far outweight the pennies saved on electricty. BUT....for certain APPLICATIONs...lower power, heat and silence is very desirable right...I think you already have the right focus here.

    OK, I rambled (as usual...) :-D
  • »2004/2/16 20:08
    Profile
  • Paladin of the Pegasos
    Paladin of the Pegasos
    Jupp3
    Joined: 2003/2/24
    Posts: 1192
    From: Helsinki, Finland
    Quote:

    our distributed.net client beats even the 3.2 GHz x86 on a 1 GHZ G4.

    Is that with or without Altivec support, btw? If it isn't, it should be even faster with it...

    Then...

    About software...

    I wouldn't say that we don't need any "big name" GPL ports (such as Quake 2) but to be really impressive, they should be highly optimized (AND stabile...)

    More we'd need good original software / games. Problem is, todays world much rates games from their graphics. Instead we should favour "playability over graphics".
  • »2004/2/17 6:43
    Profile Visit Website
  • Paladin of the Pegasos
    Paladin of the Pegasos
    Jupp3
    Joined: 2003/2/24
    Posts: 1192
    From: Helsinki, Finland
    One idea of what could be shown in public Pegasos demonstrations...

    How anout some device, that shows total power consumption for Pegasos?

    That would likely be interesting to some people. (Even Amiga users, as "Likely AmigaOne would get quite similar results")
  • »2004/2/17 6:59
    Profile Visit Website
  • Order of the Butterfly
    Order of the Butterfly
    liquidbit
    Joined: 2003/10/12
    Posts: 404
    From:
    What about to be an official member to the Standard Performance Evaluation Corporation?

    http://www.spec.org/
  • »2004/3/17 0:34
    Profile