Let's Benchmark Our SMP Systems

T

Tim_axe

Gawd
Joined
Dec 12, 2003
Messages
946
The Overclocking & Cooling forum has its own benchmarks, but those are geared towards single CPU systems. A few months ago when I built my first new SMP system (Had a Dell PowerEdge Dual P3 Server before it) I found the clibench SMP benchmark.

So, let's put together our System Specs & run this benchmark a few times so we can at least see how our different systems handle different multi-threaded benchmarks.


On 06/20/2005, these were the Top 20 scores sorted by Dhrystones:



------------------------------------------------------
Edit: I have setup a database on my website where I am listing & sorting the results from this thread. You can access it via the web here: CliBench Results Database - http://files.timaxe.com/projects/clibench/

Edit 2/7: Results are now avaliable in a CSV (Comma Seperated Values) file that can be imported into most spreadsheet software and graphed. The download can be found on my page, or you can direct download all of the results with this link: Download All Results -- http://files.timaxe.com/projects/clibench/export.php?type=all
(I may eventually edit a few graphs into this first post)
------------------------------------------------------


Here is what I would like you to do:
  • Go to the clibench home page, and download the CliBench Mk III SMP 0.7.16 binary only file.
  • Run CliBench ( clibench-mkiii-0.7.16.exe ). It should have a large empty window.
  • Go to Test -> Select All CPU Tests
  • Go to Test -> Run Selected Tests
  • On the Window that pops up, Fill in # of Threads To Run and click Ok (begin)
    Threads To Run = Number of (Virtual) Processors Recognized By OS
  • At the end of the benchmark session, your computer should be responsive again.
  • Download my results, and do a File -> Compare to load them into CliBench.
    (Zxcs has kindly mirrored the comparison file here for me incase my server goes off-line)
  • Do a File -> Export and save a text file.
  • Copy & Paste The Results into a reply, along with information such as your Processor Architecture & Speed, RAM (& ratio if not 1:1), FSB/Multiplier settings, Motherboard, and whatever else you deem important


You may also repeat the test with a single CPU, and do a File -> Compare. But to make it easier to read/scroll through the results, please post each benchmark compared to my results. (I guess mine should be the "base" system :D)


Base System Specs in Comparison File (See next post):

  • Dual AthlonXP-M 2400+ @ 2.448GHz (18x136) (1.75v) on Gigabyte GA-7DPXDW-P
  • PC3200 512MB Generic Kingston RAM @ 136MHz (Don't remember timings)
  • Windows 2000 SP4



------------------------------------------------------

Edit: If you can, please use the [ code] [/ code] tags for the results so formatting/spaces can be preserved.


[ code]

Results

[/ code]


Thanks :)
 
Here are my results. Single Thread compared to Double Thread.

System:

Dual AthlonXP-M 2400+ @ 2.4GHz (18x136) (1.75v) on Gigabyte GA-7DPXDW-P
PC3200 512MB Generic Kingston RAM @ 136MHz (Don't remember timings)
Windows 2000



Code: 
                                this      other              relative
                            computer   computer           performance
 
Dhrystone 2.1                   5876      11682  kDhryst.  50 percent
Whetstone                       1273       2522  MFLOPS    50 percent
Eight queens problem            8884      17665  pps       50 percent
Matrix operations              81734     148309  k ops     55 percent
Number crunch                 185624     370078  k ops     50 percent
Floating point                 19776      39423  k ops     50 percent
Memory throughput             349562     603339  kB/sec    57 percent

                                           Total CPU       52 percent
                                           Total FPU       50 percent
                                             Average       51 percent
                                                                     
                             Application Performance       52 percent

benchmark%20clibench.png


Well, I ended up comparing these scores backwards when I took the screen shot a few months ago :p
 
this other relative
computer computer performance

Dhrystone 2.1 5360 n/a kDhryst. 0 percent
Whetstone 1081 n/a MFLOPS 0 percent
Eight queens problem 6946 n/a pps 0 percent
Matrix operations 186051 n/a k ops 0 percent
Number crunch 156458 n/a k ops 0 percent
Floating point 16616 n/a k ops 0 percent
Memory throughput 616714 n/a kB/sec 0 percent
 
If you can, please use the [ code] [/ code] tags for the results so formatting/spaces can be preserved.


[ code]

Results

[/ code]


Thanks :)



Man, that single Opteron murders my dual memory busses and the matrix operations :eek:
(760MPX gives each processor a dedicated bus)
 
Code: 
                                this      other              relative
                            computer   computer           performance
 
Dhrystone 2.1                   5341      11682  kDhryst.  45 percent
Whetstone                       1079       2522  MFLOPS    42 percent
Eight queens problem            6943      17665  pps       39 percent
Matrix operations             186839     148309  k ops    125 percent
Number crunch                 153997     370078  k ops     41 percent
Floating point                 16509      39423  k ops     41 percent
Memory throughput             616983     603339  kB/sec   102 percent
                                                                                                          
                                           Total CPU       70 percent
                                           Total FPU       41 percent
                                             Average       62 percent
                                                                     
                             Application Performance       68 percent

Single Opteron 246 (can't sell enough blood yet to buy the other one)
Tyan Thunder K8W
1GB Corsair PC3200

Even though that is single threaded with one processor vs. your duals, I would have figured the opteron to fair better than that. I didn't do any bench preparing, not even closing any running apps or rebooting, but that still isn't what I expected.
 
Code: 
                       this      
                                computer   
 
Dhrystone 2.1                   9738        kDhryst.   
Whetstone                       2103        MFLOPS     
Eight queens problem           14722       pps        
Matrix operations             136551       k ops      
Number crunch                 308398        k ops      
Floating point                 33376        k ops      
Memory throughput             592015       kB/sec
Dual Athlon XP 2400 @ 2.0GHz on Asus A7M266-D
PC2100 1024MB Crucial
Windows XP Pro SP2
 
Code: 
                                this      other              relative
                            computer   computer           performance
 
Dhrystone 2.1                   8631      11682  kDhryst.  73 percent
Whetstone                       1744       2522  MFLOPS    69 percent
Eight queens problem           11193      17665  pps       63 percent
Matrix operations             213962     148309  k ops    144 percent
Number crunch                 254054     370078  k ops     68 percent
Floating point                 26600      39423  k ops     67 percent
Memory throughput             989128     603339  kB/sec   163 percent
                                                                                                          
                                           Total CPU      102 percent
                                           Total FPU       68 percent
                                             Average       92 percent
                                                                     
                             Application Performance      100 percent
clibenchies.jpg

compunet.jpg


On an Iwill DK8X
 
Code: 
                                this      other              relative
                            computer   computer           performance
 
Dhrystone 2.1                  11032      11682  kDhryst.  94 percent
Whetstone                       2380       2522  MFLOPS    94 percent
Eight queens problem           16682      17665  pps       94 percent
Matrix operations             139158     148309  k ops     93 percent
Number crunch                 349182     370078  k ops     94 percent
Floating point                 37443      39423  k ops     94 percent
Memory throughput             557096     603339  kB/sec    92 percent
                                                                                                          
                                           Total CPU       93 percent
                                           Total FPU       94 percent
                                             Average       93 percent
                                                                     
                            Application Performance       94 percent

Gigabyte GA-&DPXDW-p
dual 2200 35 watt mobile cpus @ 2.2ghz @ 1.75v
512mb crucial pc 2700 ECC
80gb WD HDD
 
Code: 
                                this      other              relative
                            computer   computer           performance
 
Dhrystone 2.1                  11847      11682  kDhryst. 101 percent
Whetstone                       2392       2522  MFLOPS    94 percent
Eight queens problem           15366      17665  pps       86 percent
Matrix operations             413140     148309  k ops    278 percent
Number crunch                 347890     370078  k ops     94 percent
Floating point                 36807      39423  k ops     93 percent
Memory throughput            1274129     603339  kB/sec   211 percent
                                                                                                          
                                           Total CPU      154 percent
                                           Total FPU       93 percent
                                             Average      136 percent
                                                                     
                             Application Performance      150 percent
msi k8t master far2
2x248 opterons
2gb Corsair xms ddr400 (2x1gb).

Waiting for my Iwill DK8ES, I will post new scores then.
 
Code: 
                                this      other              relative
                            computer   computer           performance
 
Dhrystone 2.1                  12997      11682  kDhryst. 111 percent
Whetstone                       2623       2522  MFLOPS   104 percent
Eight queens problem           16850      17665  pps       95 percent
Matrix operations             452087     148309  k ops    304 percent
Number crunch                 375699     370078  k ops    101 percent
Floating point                 40383      39423  k ops    102 percent
Memory throughput            1455060     603339  kB/sec   241 percent
                                                                                                          
                                           Total CPU      170 percent
                                           Total FPU      103 percent
                                             Average      151 percent
                                                                     
                             Application Performance      165 percent
I shouldnt really post because mine is under 2 Vapochills but there ya have it, rig is in sig, ram is in dual channel, running at 2703mhz at time of this test, I did have 2 FAH consoles open too but they shouldn't have affected it any :D
 
Code: 
                                this      other              relative
                            computer   computer           performance
 
Dhrystone 2.1                   9788      11682  kDhryst.  83 percent
Whetstone                       2115       2522  MFLOPS    83 percent
Eight queens problem           14797      17665  pps       83 percent
Matrix operations             145972     148309  k ops     98 percent
Number crunch                 310231     370078  k ops     83 percent
Floating point                 33547      39423  k ops     85 percent
Memory throughput             624952     603339  kB/sec   103 percent
 
                                           Total CPU       90 percent
                                           Total FPU       84 percent
                                             Average       88 percent
                                                                     
                             Application Performance       90 percent

MSI K7D Master-L
2 * Barton-core Athlon XPs @ 2.0 GHz (133MHz fsb * 15)
1GB Samsung ECC Registered PC2700 @ 266 MHz; 2:2:2:2 latencies

Haha, I just realized that I downloaded this same program years ago to benchmark my system after I upgraded my processor. I had an Asus super socket 7 motherboard, and went from a Cyrix MII 233MHz to an AMD K6-2 500MHz. The K6-2 murdered the MII in everything... except floating point, where the MII was like 50% better, even though it was more than twice as slow. Weird. :p
 
Well here's my results. I had some other background stuff open and I havn't rebooted in a few days, but i don't know that that would make much of a difference.

Iwill DPI533
Intel Prestonia Xeon @ ~2.9Ghz @ 180/720Mhz FSB (2 Processors)
1024mb (2x512) Corsair Value Select PC3200 running at 360Mhz

Code: 
                                this      other              relative
                            computer   computer           performance
 
Dhrystone 2.1                   8397      11682  kDhryst.  71 percent
Whetstone                       1869       2522  MFLOPS    74 percent
Eight queens problem            9807      17665  pps       55 percent
Matrix operations             137657     148309  k ops     92 percent
Number crunch                 284503     370078  k ops     76 percent
Floating point                 29982      39423  k ops     76 percent
Memory throughput            1035864     603339  kB/sec   171 percent
                                                                                                          
                                           Total CPU       93 percent
                                           Total FPU       75 percent
                                             Average       87 percent
                                                                     
                             Application Performance       92 percent
 
2x246 opteron, 2gig pc2700 in dual channel on a tyan thunder k8w. THough, I'm curious, what is the base compare system running? I haven't seen it posted anywhere, either that Or I missed it.

Code: 
                                this      other              relative
                            computer   computer           performance
 
Dhrystone 2.1                  10719      11682  kDhryst.  91 percent
Whetstone                       2164       2522  MFLOPS    85 percent
Eight queens problem           13901      17665  pps       78 percent
Matrix operations             232908     148309  k ops    157 percent
Number crunch                 312603     370078  k ops     84 percent
Floating point                 33223      39423  k ops     84 percent
Memory throughput            1201387     603339  kB/sec   199 percent
                                                                                                          
                                           Total CPU      121 percent
                                           Total FPU       84 percent
                                             Average      111 percent
                                                                     
                             Application Performance      118 percent
 
defakto said:
THough, I'm curious, what is the base compare system running? I haven't seen it posted anywhere, either that Or I missed it.
Click to expand...

It's in Tim_axe's signature, and also in the 2nd post.

2 * AthlonXP-M 2400+ @ 2.4GHz (18 * 136fsb) (1.75v)
Gigabyte GA-7DPXDW-P
PC3200 512MB Generic Kingston RAM @ 136MHz (Don't remember timings)
Windows 2000
 
The other thing that worries me is I don't think CliBench is reporting correct Mflops, Under sandra I hit 5k-6k Mflops not, 2k-2.5k
 
From what CliBench says, its Whetstone test fits entirely into the L1 cache of the processor. But it does mention that some of the tests were rewritten...which might explain why SiSandra & CliBench come up with different results. I'm not sure how CliBench was compiled, but I'll look into taking the sources and trying to get them to compile under BloodShed DevC++ to see if there are differences from the compiler used (I'm not sure if I can convert some MSVC++ specific code though or even make sense of most of it though -- I might look at his GliBench, which is the Linux port of it).

Overall though your observation is true, it does look like Whetstone is reported way lower on CliBench than SiSandra, and I'm not sure why. This version of the program was released ~3.5 years ago, so perhaps there are fewer optimizations for newer instruction sets/processors? (SSE 2/3?)

Also I think Dhrystones are not directly compareable because it isn't measured/calculated in VAX MIPS. I have no idea how/if the kDhrystones reported here can be converted into a compareable MIPS value to be "checked" against SiSandra. (For what it is worth, according to a comment on the site regarding this, the VAX 11/780 is considered 1 MIP, and can do ~1757 Dhrystones per second)



CliBench is a suite of synthetic benchmarks

It includes the following benchmarks. They are partially self developed or rewritten industry standard benchmarks.
  • Dhrystones: An industry standard benchmark which is an indicator for integer performance
  • Whetstones An industry standard benchmark which is an indicator for floating point performance
  • Eight queens problem: A famous algorithm that depends on the latency time of the CPU. (The longer the pipeline, the bigger the latency time)
  • Matrix operations: By calculating matrices you'll get a sign for the CPUs level 1 performance
  • Number Crunching: Raw integer performance
  • Floating point: Raw floating point performance
  • Memory throughput: This shows the transfer speed from the CPU into the memory
  • The hard disk's throughput and the CPU usage
Click to expand...


From the above description though, I think we'll find that most of Intel's processors will lag behind in the Eight Queens Problem (which places 8 queens on a chess board so they don't check each other) -- something we're beginning to see from GotNoRice's Xeon setup. Figuring out what these results mean and why they differ is a good lesson in different processor architectures. The person put together a graph of what is today much older processor families to see how they scale -- check it out here http://clibench.daemonware.ch/index.php?read=8 and do a View-Image (new Window) on the graph to see it full size.



At least we can compare results within the program. I think that after we get a variety of systems here we'll be able to compare it better. I'm currently building a database with all of this information so it could possibly be graphed later on, and when that happens I'll choose a different "base score" so that the results fall over the 100%+ mark nicer. Because comparing 90% to 140% is *very* confusing.

If you want to see what I have so far in this database, here it is -- http://files.timaxe.com/projects/clibench/
 
Code: 
                                this      other              relative
                            computer   computer           performance
 
Dhrystone 2.1                   3791      11682  kDhryst.  32 percent
Whetstone                        904       2522  MFLOPS    35 percent
Eight queens problem            5179      17665  pps       29 percent
Matrix operations             210882     148309  k ops    142 percent
Number crunch                 169675     370078  k ops     45 percent
Floating point                 14170      39423  k ops     35 percent
Memory throughput             617851     603339  kB/sec   102 percent
                                                                                                          
                                           Total CPU       70 percent
                                           Total FPU       35 percent
                                             Average       60 percent
                                                                     
                             Application Performance       68 percent

Not SMP, 1.7 Dothan, 1GB 2700
 
Hmm I can't copy mine :confused:

Xeons.png


System's in the sig, HT enabled....this is 2 threads. 4 is marginally faster at a couple, slower at a couple but close all around.
 
Tim_axe said:
At least we can compare results within the program. I think that after we get a variety of systems here we'll be able to compare it better. I'm currently building a database with all of this information so it could possibly be graphed later on, and when that happens I'll choose a different "base score" so that the results fall over the 100%+ mark nicer. Because comparing 90% to 140% is *very* confusing.

If you want to see what I have so far in this database, here it is -- http://files.timaxe.com/projects/clibench/
Click to expand...

Just a correction on my information, I have 2 Opterons and ran CliBench with 2 threads. Also, I have 1 GB of memory (2x 512 modules of PC3200 @ 3:3:3:8 timings).
 
eighteen_psi said:
Hmm I can't copy mine :confused:

...img...

System's in the sig, HT enabled....this is 2 threads. 4 is marginally faster at a couple, slower at a couple but close all around.
Click to expand...


To copy, you need to do a File -> Export to make a text file that you can open in NotePad for a Copy/Paste.



Also, feel free to post multiple scores if you changed something. When I'm adding these to the database I keep track of how many processors and # of threads used, and there isn't a problem with multiple submissions from a person. (The "primary key" to the array/database is just the submission number when I add it)



With the Intel setups, I'd love to record the differences with stuff like HyperThreading On/Off (if there are any) and the # of threads used. If you have time to kill, I'd be interested in extra numbers :)
 
I also have some interesting observations that explain how this benchmark is multi threaded and how it scales with different # of threads. In summary, run threads as a multiple of avaliable processors. The rest just explains why, and is very boring. ;)



If I try 3 threads though (with Options->RealTime Class enabled -- by default), the score is ~75% what it usually is with 2 processors. 5 threads is about ~83%. It follows the function:

f(x) = (1 - 1/(x+1))

And sure enough with 7 threads my scores are about ~87%. Also, with a single thread, my score is of course 50%. It all follows that function...which happens to be useful for figuring out how the internals of this benchmark work...


That function leads me to believe that the benchmark fires off seperate/independant threads for each benchmark, and then averages them all together compared to the time it took to execute.

This design allows it to scale well since in a given time frame; the processors will usually be used to their full potential; not being limited by having to exchange data between different threads -- they're independant. But it makes an assumption - given a period of time, everything is being processed to full capacity.

That is true when the number of threads is a multiple of the number of avaliable processors. That assumption isn't true when it isn't a multiple, such as having 3 threads on a 2-CPU machine. This results in having the last thread using only half of the full capacity during the time that executes. Since we're looking for full capacity processing power, a calculation using half-capacity is a no-no, and is misleading. In short, keep it a multiple.



This benchmark tells you how much work your processors can get done when they do their own thing as fast as they can when threads is a multiple of processors keeping them busy the entire time. (The processors don't have to talk to each other when this is multi-threaded -- only report their times/results back to the parent process) So be sure to use a multiple of the processors to get accurate results, and wake up because this post is finally over :)
 
Can't benchmark my dually right now. Too busy folding for DR_K13's friend!

But should be about the same as AMBIT's. Same chips / mobo, but 2GB PC2100 ECC REG.
 
Code: 
                                this      other              relative
                            computer   computer           performance
 
Dhrystone 2.1                   6262      11682  kDhryst.  53 percent
Whetstone                       1542       2522  MFLOPS    61 percent
Eight queens problem            9980      17665  pps       56 percent
Matrix operations             225341     148309  k ops    151 percent
Number crunch                 230505     370078  k ops     62 percent
Floating point                 21307      39423  k ops     54 percent
Memory throughput            1558607     603339  kB/sec   258 percent
                                                                                                          
                                           Total CPU      116 percent
                                           Total FPU       57 percent
                                             Average       99 percent
                                                                     
                             Application Performance      112 percent
 
                             Drive C        
Read max                       43573  kB/sec 
Read average                   28246  kB/sec 
Read min                       12809  kB/sec 
Write max                     179637  kB/sec 
Write average                  77789  kB/sec 
Write min                      43225  kB/sec 
CPU usage                          1  percent


P4 3.4 @ 3.58 using DDR2-533. While not an SMP system, I ran the benchmark with 2 threads to see where this machine sits.
 
Tim_axe said:
To copy, you need to do a File -> Export to make a text file that you can open in NotePad for a Copy/Paste.



Also, feel free to post multiple scores if you changed something. When I'm adding these to the database I keep track of how many processors and # of threads used, and there isn't a problem with multiple submissions from a person. (The "primary key" to the array/database is just the submission number when I add it)



With the Intel setups, I'd love to record the differences with stuff like HyperThreading On/Off (if there are any) and the # of threads used. If you have time to kill, I'd be interested in extra numbers :)
Click to expand...

I must say I was pretty impressed with how well mine fared, especially compared to rice's far faster Xeon rig. Anyway...I'll run through whatever you want, when I'm not a week and a half prior to graduation at a design/media school where my box is rendering 24/7 for my grad reel :D

It should be noted that I paused my renders but i still had 4 folding threads going at the time of that benchmark.
 
I'm posting this as a heads up for everyone: the HDD write test is not accurate. What happens is that the benchmark is tricked by the Write Buffer on the HDD, and also Window's own file/write-back buffer. Which is why Jebus's Write Max exceeds SATA 150 spec. So just don't rely on the HDD/Storage tests with CliBench ;)



Also, I have figured out code to sort by Dhyrstone scores, so you can go to my CliBench results database (http://files.timaxe.com/projects/clibench/) and see who is on top of that test (Zxcs). I'll expand that code to handle the other scores when I'm not studying for finals, so stay tuned.

Also, I've filled in all of the information I was able to deduce from the descriptions of the setups given. If they're incorrect and you want them corrected, PM me or post corrections and I'll get to them when I'm not studying for finals.


Cheers. And for everyone going into Finals (such as myself), good luck. :)
 
Code: 
                                this      other              relative
                            computer   computer           performance
 
Dhrystone 2.1                   8010      11682  kDhryst.  68 percent
Whetstone                       1714       2522  MFLOPS    67 percent
Eight queens problem            9139      17665  pps       51 percent
Matrix operations             178464     148309  k ops    120 percent
Number crunch                 276145     370078  k ops     74 percent
Floating point                 28803      39423  k ops     73 percent
Memory throughput            1323623     603339  kB/sec   219 percent
                                                                                                          
                                           Total CPU      106 percent
                                           Total FPU       70 percent
                                             Average       96 percent
                                                                     
                             Application Performance      104 percent
HyperThreading: on
Threads used: 4
Memory: Dual channel DDR400 (2-3-3-8)
Processors: Xeon 1.6LVs @ 2.8GHz (Prestonias, 200x14)
Mobo: ASUS PC-DL
OS: WinXP Pro SP2
 
Code: 
                                this      other              relative
                            computer   computer           performance
 
Dhrystone 2.1                   3389      11682  kDhryst.  29 percent
Whetstone                        762       2522  MFLOPS    30 percent
Eight queens problem            3811      17665  pps       21 percent
Matrix operations             113172     148309  k ops     76 percent
Number crunch                 132245     370078  k ops     35 percent
Floating point                 12975      39423  k ops     32 percent
Memory throughput            1037836     603339  kB/sec   172 percent
                                                                                                          
                                           Total CPU       66 percent
                                           Total FPU       31 percent
                                             Average       56 percent
                                                                     
                             Application Performance       64 percent

Asus P4P800 with 2.4GHz Pentium 4
 
Code: 
                                this      other              relative
                            computer   computer           performance
 
Dhrystone 2.1                  11795      11682  kDhryst. 100 percent
Whetstone                       3994       2522  MFLOPS   158 percent
Eight queens problem           19919      17665  pps      112 percent
Matrix operations             256641     148309  k ops    173 percent
Number crunch                 400170     370078  k ops    108 percent
Floating point                 40546      39423  k ops    102 percent
Memory throughput             430607     603339  kB/sec    71 percent
                                                                                                          
                                           Total CPU      112 percent
                                           Total FPU      130 percent
                                             Average      117 percent
                                                                     
                             Application Performance      114 percent

Dell PowerEdge 6850 with eight Xenon Pentium III processors at 700 MHz.
 
Code: 
                                this      other              relative
                            computer   computer           performance
 
Dhrystone 2.1                  10601      11682  kDhryst.  90 percent
Whetstone                       1933       2522  MFLOPS    76 percent
Eight queens problem           15804      17665  pps       89 percent
Matrix operations             289532     148309  k ops    195 percent
Number crunch                 313542     370078  k ops     84 percent
Floating point                 28871      39423  k ops     73 percent
Memory throughput            1321909     603339  kB/sec   219 percent
                                                                                                          
                                           Total CPU      135 percent
                                           Total FPU       74 percent
                                             Average      118 percent
                                                                     
                             Application Performance      131 percent


Dell Precision 650, Xenon at 3.2 GHz. Hyperthreading disabled, two processors, two threads.
 
Code: 
                                this      other              relative
                            computer   computer           performance
 
Dhrystone 2.1                   5805      11682  kDhryst.  49 percent
Whetstone                       1111       2522  MFLOPS    44 percent
Eight queens problem            9820      17665  pps       55 percent
Matrix operations             146521     148309  k ops     98 percent
Number crunch                 181047     370078  k ops     48 percent
Floating point                 16846      39423  k ops     42 percent
Memory throughput             900002     603339  kB/sec   149 percent
                                                                                                          
                                           Total CPU       79 percent
                                           Total FPU       43 percent
                                             Average       69 percent
                                                                     
                             Application Performance       77 percent

Dell Precision 530, Dual Xenon at 2.0 GHZ. 2 gis memory. 2 threads, hyperthreading disabled.
 
Code: 
                                this      other              relative
                            computer   computer           performance
 
Dhrystone 2.1                   7075      11682  kDhryst.  60 percent
Whetstone                       1342       2522  MFLOPS    53 percent
Eight queens problem           12052      17665  pps       68 percent
Matrix operations             168226     148309  k ops    113 percent
Number crunch                 220140     370078  k ops     59 percent
Floating point                 20289      39423  k ops     51 percent
Memory throughput             979345     603339  kB/sec   162 percent
                                                                                                          
                                           Total CPU       92 percent
                                           Total FPU       52 percent
                                             Average       80 percent
                                                                     
                             Application Performance       89 percent

Dell Precision 530. Dual Xeons at 2.4 GHz, hyperthreading disabled and two threads.
 
Code: 
                                this      other              relative
                            computer   computer           performance
 
Dhrystone 2.1                   5260      11682  kDhryst.  45 percent
Whetstone                       1150       2522  MFLOPS    45 percent
Eight queens problem            7951      17665  pps       45 percent
Matrix operations              75013     148309  k ops     50 percent
Number crunch                 165726     370078  k ops     44 percent
Floating point                 17944      39423  k ops     45 percent
Memory throughput             406097     603339  kB/sec    67 percent

                                           Total CPU       50 percent
                                           Total FPU       45 percent
                                             Average       48 percent
                                                                     
                             Application Performance       50 percent

AMD Athlon XP 2700+ (2.17 GHz) in a ASUS A7V333 motherboard. Just one thread.
 
Code: 
                                this      other              relative
                            computer   computer           performance
 
Dhrystone 2.1                   4126      11682  kDhryst.  35 percent
Whetstone                        706       2522  MFLOPS    27 percent
Eight queens problem            5978      17665  pps       33 percent
Matrix operations             106478     148309  k ops     71 percent
Number crunch                 106721     370078  k ops     28 percent
Floating point                 10305      39423  k ops     26 percent
Memory throughput             819107     603339  kB/sec   135 percent

                                           Total CPU       60 percent
                                           Total FPU       26 percent
                                             Average       50 percent
                                                                     
                             Application Performance       58 percent

Dell 600SC with 533MHz FSB and Pentium 4 at 2.4 GHz. One thread, one proc, hyperthreading not available.
 
Code: 
                                this      other              relative
                            computer   computer           performance
 
Dhrystone 2.1                   5590      11682  kDhryst.  47 percent
Whetstone                       1350       2522  MFLOPS    53 percent
Eight queens problem            8881      17665  pps       50 percent
Matrix operations             218302     148309  k ops    147 percent
Number crunch                 200768     370078  k ops     54 percent
Floating point                 18886      39423  k ops     47 percent
Memory throughput            1193677     603339  kB/sec   197 percent

                                           Total CPU       99 percent
                                           Total FPU       50 percent
                                             Average       85 percent
                                                                     
                             Application Performance       96 percent

IC7-MAX3 with Prescott 3.2GHz Pentium 4. Hyperthreading enabled, 2 threads used.
 
Code: 
                                this      other              relative
                            computer   computer           performance
 
Dhrystone 2.1                   5580      11682  kDhryst.  47 percent
Whetstone                       1293       2522  MFLOPS    51 percent
Eight queens problem            6642      17665  pps       37 percent
Matrix operations             166191     148309  k ops    112 percent
Number crunch                 170537     370078  k ops     46 percent
Floating point                 15120      39423  k ops     38 percent
Memory throughput             867292     603339  kB/sec   143 percent

                                           Total CPU       77 percent
                                           Total FPU       44 percent
                                             Average       67 percent
                                                                     
                             Application Performance       75 percent

Asus P5AD2 Deluxe Edition with Pentium 4 3.2GHz LGA775 processor. Hyperthreading enabled, two threads run.
 
Code: 
                                this      other              relative
                            computer   computer           performance
 
Dhrystone 2.1                   7506      11682  kDhryst.  64 percent
Whetstone                       1643       2522  MFLOPS    65 percent
Eight queens problem           11351      17665  pps       64 percent
Matrix operations             117712     148309  k ops     79 percent
Number crunch                 237613     370078  k ops     64 percent
Floating point                 25715      39423  k ops     65 percent
Memory throughput             534562     603339  kB/sec    88 percent

                                           Total CPU       71 percent
                                           Total FPU       65 percent
                                             Average       69 percent
                                                                     
                             Application Performance       71 percent


Tyan Tiger MP with dual AMD Athlon MP 1800+ (1.53GHz) processors. Two threads running.
 
Code: 
                                this      other              relative
                            computer   computer           performance
 
Dhrystone 2.1                   1940      11682  kDhryst.  16 percent
Whetstone                        468       2522  MFLOPS    18 percent
Eight queens problem            3289      17665  pps       18 percent
Matrix operations              39711     148309  k ops     26 percent
Number crunch                  65223     370078  k ops     17 percent
Floating point                  7161      39423  k ops     18 percent
Memory throughput             163236     603339  kB/sec    27 percent

                                           Total CPU       20 percent
                                           Total FPU       18 percent
                                             Average       20 percent
                                                                     
                             Application Performance       20 percent
Asus CUV4X-EA with 933 MHz Pentium III processor. One thread.
 
Code: 
                                this      other              relative
                            computer   computer           performance
 
Dhrystone 2.1                   4087      11682  kDhryst.  34 percent
Whetstone                        811       2522  MFLOPS    32 percent
Eight queens problem            5872      17665  pps       33 percent
Matrix operations             105850     148309  k ops     71 percent
Number crunch                 142683     370078  k ops     38 percent
Floating point                 10201      39423  k ops     25 percent
Memory throughput             744774     603339  kB/sec   123 percent

                                           Total CPU       59 percent
                                           Total FPU       28 percent
                                             Average       50 percent
                                                                     
                             Application Performance       57 percent

Dell 600SC. This one's at home and running Windows 2003 Server Standard Edition. 533 MHz FSB with a Pentium 4 2.4GHz processor. No hyperthreading, so just one thread.
 
Code: 
                                this      other              relative
                            computer   computer           performance
 
Dhrystone 2.1                   4191      11682  kDhryst.  35 percent
Whetstone                       1425       2522  MFLOPS    56 percent
Eight queens problem            7116      17665  pps       40 percent
Matrix operations              62882     148309  k ops     42 percent
Number crunch                 144171     370078  k ops     38 percent
Floating point                 14397      39423  k ops     36 percent
Memory throughput             194747     603339  kB/sec    32 percent

                                           Total CPU       37 percent
                                           Total FPU       46 percent
                                             Average       39 percent
                                                                     
                             Application Performance       38 percent
Supermicro P6DGU with two Penitum III 1GHz parts. Two Threads.
 
Thanks mikeblas -- once I counted how many systems you added I decided to write a script to help me add to the database easier. :p

Would you mind my using the "Supermicro P6DGU with two Penitum III 1GHz parts" as the weighted score on my website's database? I was waiting for some P3 scores, and Dual 1GHz is a nice speed. I just have to write the code to weigh the scores. :)



That 8-way P3-700 Xeon system is one mean machine! :eek: I can't really imagine how much bite it has in regular usage, but it sure made a mark in the benchmarks! I don't think anything will knock it loose from the 8 queens test (short pipelines, lots of horsepower), and it doesn't look like it will give up that Whetstone benchmark either.

I'm seeing that CliBench's Whetstone doesn't scale as well with extra MHz as SiSandra does -- more processors, even if they're slower, seem to produce better performance than a single *much* faster one on CliBench. I'm not sure why, and haven't had time to look at any code yet or ask any questions.



BTW, results (sorted) can be found here: http://files.timaxe.com/projects/clibench/
 
You must log in or register to reply here.
Back
Top