Please help me with my Computer Science project...Check out my new, cross-platform compatible benchmark!

[adlep]

Hello all!
I need your help with evaluating and testing?.my benchmark
I was doing a coding project about six months ago about sorting algorithms. I had to practically implement seven most popular algorithms and find the fastest and the slowest one?
This excellent site about sorting algorithms was a great help

I have decided to create my own benchmark!

Here you can find a short explanation on how it works

Feel free to email your comments and results to:
sortmark@computerfix.org

If you would like to send your results please include:
a. Your Operating System (Linux, Windows XP, etc).
b. Your processor speed and type (P IV Norwood 1.8 GHz, AMD Athlon XP 2500+) Please feel free to include as many details about the chip as you can, overclockers are especially encouraged to send-in their results.
c. The motherboard you are using (as may details as you can).
d. The amount of RAM (as many details as you can).
e. Any other information about your system, like video card (not tested by the benchmark, but for example buggy drivers can affect the score), sound card, hard drive details, etc, etc.
f. You comments about the benchmark.

My results:

Windows XP; PIV Northwood CPU 1.8 GHz @ 18x142 Mhz fsb = 2.62 GHZ
mean of 3 runs = 70.92133 seconds.
Linux Mandrake 9.0
mean of 3 runs = 63.99 seconds

Please go here to download the benchmark if you would like to give it a try.
Any file listed on that site has a detailed instruction, but feel free to ask for help if some extra info needed.

During the run, program will create two files: data.dat, which contains the integers to sort, and SortOut.dat that contains arranged numbers. Both are created for verification purposes. Both are being overwritten on every run of the benchmark. They are safe to delete.
06/01/03 edited for clarity
Once again your feedback is important!

 

[adlep]

Going to sleep, be back tomorrow to answer any questions....

 

[adlep]

Bump?
No one is interested?

 

[BG4533]

You spelled random wrong uder step 1. The link in your post "This excellent site was a great help" also has an extra http:// in it. It took me 246.414 seconds to run this on my laptop while browsing Anandtech. It is a Gateway Solo 5300, 850P3, 256MB PC100, 4200RPM HD, WinXP, and some other stuff. I pray you didnt turn evil and just install 98304 virii on my system. I think a sorting benchmark isnt that bad of an idea, but I see a few problems with it. First, a sorting benchmark only will test a very small portion of a computers performance. Couple this with a few more mathematical cross platform benches and it might be a bit more successful. Second, the instruction sets are there to increase performance. I am not sure how these work, but I thought the compiler automatically optimized code for them. If it doesnt, you are leaving out vital parts of a computers performance. Good benchmarks test real life performance, in real life you should use those instructions. I might take a look at the code later if I have time.

Was this a HS project or college project or something else?

I think really long posts tend to scare people away.

Brian

 

[WarCon]

My junkpile here (2.4b @ 2.52gig running my memory (512megs) at 2.5-3-3-6 at 350fsb - not good memory as this is its max) ran the test in 76.4 secs while I did a little browsing.

I am running WinXP and the rest of my info is in my rig in my signature.

Edit: Ran a second time without browsing and the result is 75.3 secs.

To just benchmark the machine you might wanna make a self executing wrapper like memtest uses to just start the test from floppy or CD, to remove operating system variance and TSR variance.

Second Edit: Decided to do a little better test. I removed all TSR's
So for above conditions with no TSR's running - 72.9 secs

Then my processor at 2.79ghz and memory down to 310 but with 2.5 2 3 5 timings it ran a - 65.7 secs.

 

[adlep]

Thanks for this feedback, this is really what I have expected.
I am going to fix my long post, spelling errors ect, etc.
The floppy drive based test is a good idea. I will try to work on precise directions on how to do it...
Edit: Edited post above to make it shorter, easier to read...
Edit: Someone would like to help to make the code better? Open to offers and improvements..

 

[BG4533]

Tested it on my desktop.
Specs:
AMD Athlon XP1700+ @2GHz(166x12)
Asus A7N8X Deluxe Bios Uber1005
512MB Corsair PC3500 at 166MHz 3-2-2-2
WD Raptor
WD 400EB
MSI GF4 MX440VTP

Took: 89.906s

 

[amoralist]

like the earlier post said, benchmarks are considered flawed because they do not represent real world performance. computer architecture designers use benchmarks that run real application code. for example,
incorporate some code from free software projects that do alot of different things.
for example, use some of xvid encoder code, some ogg vorbis encoder code, etc.

 

[Jeff7181]

Sounds to me like your code is just running math problems, correct? That's hardly indicative of real world performance. That pretty much just tests raw Mhz computing power, which as we know, doesn't mean as much as some CPU manufacturers would like us to believe.
Cross-platform benchmarks would be hard to do to create a "whole computer" benchmark because they don't make a lot of 3D apps for Linux and Mac.
Maybe if you could write some code that would test the floating point capabilities of a processor... some that would test the integer capabilities... some that would test RAM performance (latency, bandwidth, etc.)... some that would test hard drive performance. THAT might be a better test of a "whole computer" ... and could probably be done across platforms.

 

[mroleg]

Step 5.
Timer stop!
Benchmark complete.
It takes 51.469 seconds to sort 98304 randomly generated integers.

System info is in the signature.

 

[Jeff7181]

Originally posted by: mroleg
Step 5.
Timer stop!
Benchmark complete.
It takes 51.469 seconds to sort 98304 randomly generated integers.

System info is in the signature.

That's one helluva overclock... 3.5 Ghz... nice... just out of curiosity, what score do you get in PCMark with that?

 

[adlep]

Ok,
Let me clarify, this never meant to be a "whole computer" benchmark.

Sounds to me like your code is just running math problems, correct?

Well, its not actually...My program uses very CPU intensive algorithm to test the sorting performance of the CPU, it checks how good your CPU is in comparing numbers....
But I would call it a math problem...It is just a procedure which helps computer to correctly arrange integers...
The idea behind the program is to give a ballpark way of comparing one CPU to another..
Thats it....

 

[adlep]

Step 5.
Timer stop!
Benchmark complete.
It takes 51.469 seconds to sort 98304 randomly generated integers.

:Q
Excellent!
You hold the best result in the entire world!
Congratualtions!
Edit: I guess I have created a monster....It is time to upgrade again...

 

[Jeff7181]

Originally posted by: adlep
Ok,
Let me clarify, this never meant to be a "whole computer" benchmark.

Sounds to me like your code is just running math problems, correct?

Well, its not actually...My program uses very CPU intensive algorithm to test the sorting performance of the CPU, it checks how good your CPU is in comparing numbers....
But I would call it a math problem...It is just a procedure which helps computer to correctly arrange integers...
The idea behind the program is to give a ballpark way of comparing one CPU to another..
Thats it....

One would question how valid the test is then in terms of real world performance. I guess it would be fine for comparing an XP2500 to an XP2600... or a P4 2.53 to a P4 2.4C. But not so good for an Intel vs. AMD battle because each company uses different methods of achieving higher performance. So a 2.08 Ghz P4 might look like it performs the same as an XP2500 in your test... but I don't think even Intel fanboys would argue that the XP2500 is undoubtedly faster.

 

[mroleg]

Originally posted by: Jeff7181

Originally posted by: mroleg
Step 5.
Timer stop!
Benchmark complete.
It takes 51.469 seconds to sort 98304 randomly generated integers.

System info is in the signature.

That's one helluva overclock... 3.5 Ghz... nice... just out of curiosity, what score do you get in PCMark with that?

PCMark2002 Results:
CPU-8804
Memory-11115
HDD-1129

 

[imgod2u]

As I recall, sorting algorithms are primarily bottlenecked by swap methods. Which involve a lot of memory referencing instructions. Wouldn't this mean your benchmark is primarily memory-bound?

 

[Jeff7181]

Originally posted by: mroleg

Originally posted by: Jeff7181

Originally posted by: mroleg
Step 5.
Timer stop!
Benchmark complete.
It takes 51.469 seconds to sort 98304 randomly generated integers.

System info is in the signature.

That's one helluva overclock... 3.5 Ghz... nice... just out of curiosity, what score do you get in PCMark with that?

PCMark2002 Results:
CPU-8804
Memory-11115
HDD-1129

Hmmm... I was expecting a higher CPU score. Memory is impressive though.

 

[68GTX]

Abit IC7
2.6C @ 3.4GHz
Radeon 9800 Pro
512mb Samsung PC3200 @ DDR420
Audigy 2
Maxtor 40GB 7200 RPM
WinXP Pro



Step 5.
Timer stop!
Benchmark complete.
It takes 52.547 seconds to sort 98304 randomly generated integers.

 

[adlep]

As I recall, sorting algorithms are primarily bottlenecked by swap methods. Which involve a lot of memory referencing instructions. Wouldn't this mean your benchmark is primarily memory-bound?

Well, I dont think so. I have done experiments a year ago, where my athlon 1.2 GHz was runnig first at 100 Mhz fsb, and then at 133 Mhz. There was a diffrinence in performance, but it wasnt great, maybe less than 10%. That would indicate that this particular algorithm is more CPU dependant than any other sorting algorithms...

 

[adlep]

One would question how valid the test is then in terms of real world performance. I guess it would be fine for comparing an XP2500 to an XP2600... or a P4 2.53 to a P4 2.4C. But not so good for an Intel vs. AMD battle because each company uses different methods of achieving higher performance. So a 2.08 Ghz P4 might look like it performs the same as an XP2500 in your test... but I don't think even Intel fanboys would argue that the XP2500 is undoubtedly faster.

I would say that you are exactly right, and I would love to find out if my benchmark is only CPU clock dependant, or there is something more to it....
IE Will peformance of the Athlon 1800+ would be equal to the performance of Pentium IV 1.8 Ghz?
Someone who owns XP 1800+, please post your results....
I can de-clock my CPU to the original speed so we can copmare results....

 

[Jeff7181]

Originally posted by: adlep

One would question how valid the test is then in terms of real world performance. I guess it would be fine for comparing an XP2500 to an XP2600... or a P4 2.53 to a P4 2.4C. But not so good for an Intel vs. AMD battle because each company uses different methods of achieving higher performance. So a 2.08 Ghz P4 might look like it performs the same as an XP2500 in your test... but I don't think even Intel fanboys would argue that the XP2500 is undoubtedly faster.

I would say that you are exactly right, and I would love to find out if my benchmark is only CPU clock dependant, or there is something more to it....
IE Will peformance of the Athlon 1800+ would be equal to the performance of Pentium IV 1.8 Ghz?
Someone onwning XP 1800+, please post your results....
I can de-clock my CPU to the original speed so we can copmare results....

Wow... I really screwed that up... what I meant to say is...

So a 1.8 Ghz P4 might look like it performs the same as an XP2500 (also running at 1.8 Ghz) in your test, but I don't think even Intel fanboys would argue that the XP2500 is undoubtedly faster.

 

[NesuD]

Benchmark complete.
Time to sort 65536 randomly generated integers is: 81.828987 seconds.
Win XP
Athlon XP 2100+ OC to 10x210mhz-2100mhz
Soltek SL-75FRN2-RL Nforce2 Ultra 400 memory controller
2x512=1gig Corsair XMS pc3200 timings 6-3-3-2 210/420mhz Dual channel

Benchmark complete.
Time to sort 65536 randomly generated integers is: 111.864080 seconds.
1.53 gig 1800+ using a 203mhz fsb

Benchmark complete.
Time to sort 65536 randomly generated integers is: 113.507581 seconds.
1800+ default 133 fsb

 

[adlep]

Ups
NesuD
It appears that you got a pre-release version of SortMark_Win.
Please download the final version off my web site right now. Sorry, I have uploaded the wrong version.
My mistake...

Also Jeff7181

So a 1.8 Ghz P4 might look like it performs the same as an XP2500 (also running at 1.8 Ghz) in your test, but I don't think even Intel fanboys would argue that the XP2500 is undoubtedly faster.

Well I understand what you mean.
Therefore, if P IV at 1.8 GHz will be slower than AMD XP 2500, then it would mean that SortMark is kind of working.
The result should depend not on CPU Mhz speed, but on processor effectiveness.
No one argues that AMD at 1.8 GHz is more effective, faster processor than P IV at 1.8 Ghz.
Edit: My results for PIV 1.8 GHZ.
Mean of 3 runs = 99. 583 seconds under Windows XP.

 

[NesuD]

Ok here is what i got with sortmark.exe that i just grabbed from your website.
Benchmark complete.
It takes 82.328 seconds to sort 98304 randomly generated integers.
this was taken at my normal overclock 210mhzx10=2100mhz
cramped for time but i will rerun it at the 1800+ settings later this evening.

 

[adlep]

Pleaze, I need some additional feedback...
Does anyone know, have any idea what would could I do to improve the benchmark?