simple CPU Benchmark with comparisons + 5%FB speed boost tip

Post your FreeBASIC tips and tricks here. Please don’t post your code without including an explanation.
TESLACOIL
Posts: 1769
Joined: Jun 20, 2010 16:04
Location: UK
Contact:

Re: simple CPU Benchmark with comparisons + 5%FB speed boost

Postby TESLACOIL » Dec 08, 2012 17:50

@ MichaelW

good point ref bottlenecks....you free one up and another domino falls instead

btw CAT 6 is worth 0.5% increase over CAT5e ( in my nothing special LAN ) but i went with CAT6 just for belt n braces. CAT6 is a bit stiff, so i may go far CAT5e for the short bendy patch leads and use CAT6 for anything over 3foot.

Giglan cards and a Gigabit switch made a huge difference. As did an SSD, puter boots up in 30 seconds now and its a bottom of the range SSD.

Native USB3 ports with all the right drivers are good for extra speed. But its still a bit of hit and miss buying wise as we cross the bridge between USB2 and USB3
dodicat
Posts: 6155
Joined: Jan 10, 2006 20:30
Location: Scotland

Re: simple CPU Benchmark with comparisons + 5%FB speed boost

Postby dodicat » Dec 09, 2012 1:40

That is my processor TESLA~ (your link)
Looks like Hyperthreading plus 2Mb cache.
I also note that it can use 64 bit processing, IF I got XP 64.

It's a marvel that you can pin point my xeons at 400 miles away , with the scant information I have provided, faster than I can at 2 feet away.
I've got a tube of contact grease somewhere, so I think I'll get to work tomorrow.
I'll stop short of drilling holes in the case though.
Thanks again.
TESLACOIL
Posts: 1769
Joined: Jun 20, 2010 16:04
Location: UK
Contact:

Re: simple CPU Benchmark with comparisons + 5%FB speed boost

Postby TESLACOIL » Dec 09, 2012 6:23

No Problem

How thermal compound spreads using glass
http://www.youtube.com/watch?v=EyXLu1Ms-q4

Ps i found just the thing for you :-p...the kids these days huh
http://quietpc.co.uk/reserator1-v2

Or if plumbing isn't not your scene and your a fan of quiet computing
http://quietpc.co.uk/nof-cr-100a


You probably have thermal throttling on your Xeon CPU so it may run a few % quicker when cold vs hot. You can run speed test from cold boot and compare it to post ' Bingo ' results. Eg my surfer system = 377 mips speedtest9 @ 31deg c and 353mips @ 51deg c after stress tests. A 7% difference in CPU performance due to temperature. ( I cant get my cpu much hotter than 51 deg c under heavy every day loads.) The highest ever recorded with my surfer at stock settings is 391mips, it seems that even an idle internet/WLAN connection chews up 15 million FB loops worth of processing power. I should imagine the typical background load on a home PC is prolly 25 to 50mips if you include idle OS overheads too. Having lots of windows open, passive apps chews about 5 to 10mips.

My speedtest programs may seem oversimplified but when you have comparative data their true value emerges.



@ dodicat , if you download and run the Fritzbenchmark ( its used to gauge chess strength and runs on all cores and threads ) you should see a Fritzbench of approx 2500kilonodes assuming 2 cores + hyper-threading. Im unfamiliar with old Xeons if there architecture is good you might see a Fritzbench score hit 3000. Let me know your Fritz results, im intrigued.

My Fritz estimator predicts 2670 but your chips are slightly older so ill make a brave prediction and call out 2175 to 2350kn


Fritzbench conversion / elo estimator / CPU architecture comparison

Code: Select all


    'Fritzbench conversion / elo estimator / CPU architecture comparison
   
    dim i as ULONGINT
    dim mem as integer
    dim st as  double
    dim fin as  double
    dim mips as double
    dim q as integer
   
    dim  as integer mpint,cores,fritzb,elo
     
    color 13
    ?
    ?" Fritzbench conversion / elo estimator / CPU architecture comparison"
   
   'goto skip
     
    mem = FRE
    color 12
    ?:?" Free memory:" ;mem  \ (1024 * 1024); " megabytes"
    color 7
    ?:?" starting Free Basic speedtest9.exe....approx 10 seconds":?:sleep 100

    st = TIMER :?" start   "; st , timer

    for i = 1 to 2500000000:next i

    fin = TIMER :?" finish  "; fin , timer

    ?: color 13: ?" time = " ;fin-st;" i = ";i;" how many loops done in total"

    i=i/1000000
    mips = i/ (fin-st)

    ?:?" mips = "; mips; " FreeBASIC for next loop instructions X 1 million"
   

'skip:

'mpint = 425
'* 1.30
'? 3.9*1.30
'sleep
'elo = 0


screen 19
mpint = mips

color 13
?
?"  Fritz Benchmark Conversion & Comparison Table  including ELO aproximation"
color 12
?
?"  Your core speed   ="; mpint ;" mips    using Speedtest9 test as the CPU mark "
color 7
?
?"  Applying your single core speed to different architectures & converting to Fritzbenchmark"
?
?"  No hyper threading       Fritz bench"
?
fritzb = mpint *6*1*.95
?"  Single            =     "; fritzb
fritzb = mpint *6*2*.925
?"  Dual              =     "; fritzb
fritzb = mpint *6*3*.895
?"  Triple            =     "; fritzb
fritzb = mpint *6*4*.867
?"  Quad              =     "; fritzb
fritzb = mpint *6*6*.790
?"  Hex               =     "; fritzb
fritzb = mpint *6*8*.690
?"  Octo              =     "; fritzb   
?
?
?"  With hyper threading     Fritz bench"
?
fritzb = mpint *6*1*.99*1.65
?"  Single            =     "; fritzb
fritzb = mpint *6*2*.86*1.45
?"  Dual              =     "; fritzb
fritzb = mpint *6*3*.86*1.35
?"  Triple            =     "; fritzb
fritzb = mpint *6*4*.86*1.31
?"  Quad              =     "; fritzb
fritzb = mpint *6*6*.86*1.30
?"  Hex               =     "; fritzb
fritzb = mpint *6*8*.69*.95
?"  Octo              =     "; fritzb   
?
?"  Notes:"
?"  Multi core CPU's suffer performance penalites"
?"  Hyper threading cores tend to suffer even more"
?"  This means a significant % of perfomance is lost"
?"  In general Ghz is king, then cores, then threads"
?
?"  Calibrated with CPU's 2007 to 2012ad [spread +/- 10%]"
?"  * Indvidual cores turboboost higher when tested in isolation"
?
?"  hit Enter to exit"
color 6


locate 8,55:? "Fritz    ELO   HUMAN"
locate 10,55:? "    1     20   Beginner"
locate 11,55:? "    5    500   Casual player"
locate 12,55:? "   50   1200   Weak Club player"
locate 13,55:? "  500   1900   Strong Club player"
locate 14,55:? " 5000   2500   Grandmaster"
locate 15,55:? "10000   2750   World Champion"
locate 16,55:? "20000   3000   Greatest Human Possible"
locate 18,55:? "Note: Very rough aproximation"

sleep: end
'    ?"  500 Mhz Pent III    =   68"
'    ?"  Celeron 2600        =  180"
'    ?"  Sepmron 2600+       =  305"
'    ?"  Athalon 2000+       =  330"
'    ?"  Atom N450 2 core    =  235 * task manager cpu usage was showing 50%   "
'    ?"  Atom N450 2 core    =  335 * running 2 progs at once cpu usage 100% "
'    ?"  AMD XII 250         =  375"
'    ?"  AMD FX-8 8350       =  400 * estimated"
'    ?"  i5 2500k stock      =  475"
'    ?"  i5 2500k mild OC    =  503"
'    ?"  i7 3770k mild OC    =  525 * estimated "
'    ?"  Best CPU 5Ghz OC    =  590 * estimated available in 2012ad"
'    ?"  6 core 12 thread    = 6250 Total @5ghz OC summing all cores\threads"
'    ?
'    ?"  Dual cores total    = 400  to  900 typical"
'    ?"  Quad cores total    = 1200 to 2000 typical"
'    ?"  Hex + HT   total    = 1500 to 4500 typical"
'    ?"  Death Run on L'N    = 4500 to 6500 typical"

dodicat
Posts: 6155
Joined: Jan 10, 2006 20:30
Location: Scotland

Re: simple CPU Benchmark with comparisons + 5%FB speed boost

Postby dodicat » Dec 09, 2012 23:53

Hi TESLA~
From Fritz12 app I get 2030 Kilo nodes per second.
TESLACOIL
Posts: 1769
Joined: Jun 20, 2010 16:04
Location: UK
Contact:

Re: simple CPU Benchmark with comparisons + 5%FB speed boost

Postby TESLACOIL » Dec 10, 2012 0:19

Q how many logical processors does fritz see dodi 2 or 4 ?

Its the tiny grey box with up and down arrows. If im correct ? you have two separate CPU's on your Xeon motherboard and they are hyper-threading making a total of four logical processors.

Your mips score might be a bit wimpy but 2000kilonodes is enough to swat aside most human chess players even if you take the queen off the computers side and play 'queen odds'
http://www.jens-hartmann.at/Fritzmarks/
dodicat
Posts: 6155
Joined: Jan 10, 2006 20:30
Location: Scotland

Re: simple CPU Benchmark with comparisons + 5%FB speed boost

Postby dodicat » Dec 12, 2012 2:30

Hi TESLA~
2 logical processors, 2 cpu's in the top box.
4 where the arrows are.
I have a choice of selecting 1 to 4 for testing.
TESLACOIL
Posts: 1769
Joined: Jun 20, 2010 16:04
Location: UK
Contact:

Re: simple CPU Benchmark with comparisons + 5%FB speed boost

Postby TESLACOIL » Dec 12, 2012 13:22

sorry, im getting a bit confused with what your saying, this is where a screenshot sometimes helps

so the 2030 Kilo nodes result you got was with 2 or 4 ? 'processors to use' in the arrowed box

just to clarify 100%
albert
Posts: 5486
Joined: Sep 28, 2006 2:41
Location: California, USA

Re: simple CPU Benchmark with comparisons + 5%FB speed boost

Postby albert » Dec 23, 2012 20:55

I got 252 MIPS

On my AMD quad core A8 with 6Gig ram and windows 8 64bit. without messing with priority or affinities..
TESLACOIL
Posts: 1769
Joined: Jun 20, 2010 16:04
Location: UK
Contact:

Re: simple CPU Benchmark with comparisons + 5%FB speed boost

Postby TESLACOIL » Dec 26, 2012 12:35

252mips is really low , you didn't mention the ghz so unless its a very low ghz A8 chip you may have some settings wrong as you should be well over 300mips.


CPU identity and Fritz bench
http://asimov1.wikispaces.com/Download+Benchmarks

Searchable CPU list
http://www.cpubenchmark.net/cpu_list.php

please let me know what CPU have, the full name of it
TESLACOIL
Posts: 1769
Joined: Jun 20, 2010 16:04
Location: UK
Contact:

Re: simple CPU Benchmark with comparisons + 5%FB speed boost

Postby TESLACOIL » Dec 27, 2013 21:11

Almost every CPU listed by single thread performance
http://www.cpubenchmark.net/singleThread.html

If anyone has an i7 4770k i would be very interested to know how many mips you get in speedtest7.exe or speedtest8.exe

I have an i5 2500k that gets 500 mips and i was wandering if an upgrade is gonna be worth it hence my ask.
TESLACOIL
Posts: 1769
Joined: Jun 20, 2010 16:04
Location: UK
Contact:

Re: simple CPU Benchmark with comparisons + 5%FB speed boost

Postby TESLACOIL » Apr 07, 2014 15:57

Been doing some research into future CPU speeds. as in the next 5,10 and the next 20 to 50 years

Looks like we will be stuck with less than a billion instructions a second per core speedtest7.exe for the foreseeable future. Most computers today will hit between 100 million (clunker) and 750 million (heavily overclocked 5ghz+) per core.

This means that having huge amounts of ram, 128gig or even a terabyte of ram will be the next thing to look at if your shooting for maximum code performance for serial computation. Im now seriously looking at buying a commuter that can support a terabyte of ram. A 12 core CPU running at 5ghz is going to need something to chew on & hammering an SSD is not wise ref lifespan.

In the past we spent more on hardware than electricity, in the future that trend looks likely to be reversed so spending $1000+ just on RAM over a period of time starts to make sense....a lot more sense than paying $1000+ for a Pentium 133mhz that ended up as giveaway junk at a car-boot sale just a handful of few years later.

http://en.wikipedia.org/wiki/Moore%27s_ ... of_the_law



ultimate limits of the law

On 13 April 2005, Gordon Moore stated in an interview that the law cannot be sustained indefinitely: "It can't continue forever. The nature of exponentials is that you push them out and eventually disaster happens". He also noted that transistors would eventually reach the limits of miniaturization at atomic levels:
albert
Posts: 5486
Joined: Sep 28, 2006 2:41
Location: California, USA

Re: simple CPU Benchmark with comparisons + 5%FB speed boost

Postby albert » Apr 18, 2014 20:42

I got 233 mips , when i set the priority to "RealTime" on my AMD A8 Vision 4 core processor running at 1.9GHz

without setting priority, i got 187 mips
TESLACOIL
Posts: 1769
Joined: Jun 20, 2010 16:04
Location: UK
Contact:

Re: simple CPU Benchmark with comparisons + 5%FB speed boost

Postby TESLACOIL » Apr 20, 2014 3:35

187 mips is about the same a Celeron 2.6ghz

While clocks speeds have stalled or in some case been dialed lower for power saving/heat reduction other improvements in architecture have helped performance.

Because moarh! ghz was, in the past, fairly easy to come by, hit it with big $ hammer AND its an easy marketing number other parts of the computer system didnt get so much attention. There is now a concertina effect, the front end of the train has hit the ghz brick wall but the carriages at the back are still maintaining speed...for now.

Its a bit like dropping an egg on a concrete floor, the computer system ends up spreading itself out like an omelet, forced by the laws of physics to take up a parallel architecture.

Moore's Law Dead Within A Decade Says Former Chief Architect At Intel - Analysis Of Computing Future
http://www.youtube.com/watch?v=xYj2fe4DZfY




I think desktop computer will eventually be able to make use of around 16 cores (once software developers are forced to go parallel ) Graphics cards have heavy parallel workloads, so the more cores or micro cores the better. Maybe one day each pixel of your display screen will have a cluster of micro cores, that would be cool.

Get cores (x=1080,y=720)
Pass complex algorithm
Graphical Job done

The future of computing might be two computer cases, one for your every day computing and one just to house your GFX card.

As commutation generates waste heat the phrase 'Compuheater' will enter the dictionary. My PC case already lives in the room next door.(dedicated holes for computer cables in the wall) thus i have two benefits, 100% silent computing for me, and a heater for the utility room next door to keep the damp at bay...Maybe in the future, when we make a cup of tea, boiling the kettle will involve switching on your computer and contributing to BOINC. That makes a lot of energy sense.

http://en.wikipedia.org/wiki/Berkeley_O ... _Computing

Return to “Tips and Tricks”

Who is online

Users browsing this forum: No registered users and 1 guest