Let's Benchmark Our SMP Systems

Tim_axe said:
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

Sorry to blast you. I'm not even done!

I haven't seen an easy benchmark that I can quickly install, run, scrape the results, and then uninstall. I like having a vareity of tests, but all the other tools I've seen either suck, don't come with source, or require a lot of time to install.

Because of this, I'm afraid I'm actually abusing your thread just to get a comparative list of all m machines. I've never had a good opportunity to do it, and you've create the perfect storm.

Anyway, if you need more details about each rig, please just ask. Everything is running Windows XP Pro except for: the 8-way Dell (Windows 2003 Server Enterprise), the Supermicro (Windows 2003 Server Standard), and the second Dell 600SC (2003 Server Standard, too).

Tim_Axe said:
Would you mind my using the "Supermicro P6DGU with two Penitum III 1GHz parts" as the weighted score on my website's database?

Of course, it's fine ... as long as you don't call it "Mike B's stupid SMP Dinosaur because he's too stupid and cheap to upgrade it", or something like that. That machine just dropped another hard drive last night, BTW. It's just getting too old.

Tim_Axe said:
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!

It's an interesting machine to use. The "slow and wide" machines of that era present interesting challenges when writing software. Memory bandwidth for machines like that certainly limits their scaling potential, so you have to think very creatively to get the most from them when writing applications.

Tim_Axe said:
I'm not sure why, and haven't had time to look at any code yet or ask any questions.

Me, neither. I'm very eager to tear into the sources.

Tim_Axe said:
BTW, results (sorted) can be found here:

Thanks!

.B ekiM
 
Code:
                                this      other              relative
                            computer   computer           performance
 
Dhrystone 2.1                   4215      11682  kDhryst.  36 percent
Whetstone                       1012       2522  MFLOPS    40 percent
Eight queens problem            7115      17665  pps       40 percent
Matrix operations              88383     148309  k ops     59 percent
Number crunch                 141247     370078  k ops     38 percent
Floating point                 15497      39423  k ops     39 percent
Memory throughput             171614     603339  kB/sec    28 percent

                                           Total CPU       40 percent
                                           Total FPU       39 percent
                                             Average       40 percent
                                                                     
                             Application Performance       40 percent

Intel S450NX motherboard with four Xenon Pentium III processors at 500 MHz and 512 kilobyte cache. Four threads. Windows 2000 Advanced Server.
 
This is another Dell. I don't have the model number, but it's a PowerEdge server. Four 1.4 GHz Xenon processors, which support hyperthreading. 512 kilobyte L3 cache. Hyperthreading is on, and I ran the benchmark with four threads and with eight threads.

Four threads:
Code:
                                this      other              relative
                            computer   computer           performance
 
Dhrystone 2.1                   8583      11682  kDhryst.  73 percent
Whetstone                       1801       2522  MFLOPS    71 percent
Eight queens problem           14029      17665  pps       79 percent
Matrix operations             122798     148309  k ops     82 percent
Number crunch                 341074     370078  k ops     92 percent
Floating point                 13990      39423  k ops     35 percent
Memory throughput             648720     603339  kB/sec   107 percent

                                           Total CPU       86 percent
                                           Total FPU       53 percent
                                             Average       77 percent
                                                                     
                             Application Performance       84 percent

Eight threads:
Code:
                                this      other              relative
                            computer   computer           performance
 
Dhrystone 2.1                   7269      11682  kDhryst.  62 percent
Whetstone                       1876       2522  MFLOPS    74 percent
Eight queens problem            8800      17665  pps       49 percent
Matrix operations              26150     148309  k ops     17 percent
Number crunch                 352919     370078  k ops     95 percent
Floating point                 28093      39423  k ops     71 percent
Memory throughput             624520     603339  kB/sec   103 percent

                                           Total CPU       65 percent
                                           Total FPU       72 percent
                                             Average       67 percent

                             Application Performance       66 percent
 
this other relative
computer computer performance

Dhrystone 2.1 10718 11682 kDhryst. 91 percent
Whetstone 2166 2522 MFLOPS 85 percent
Eight queens problem 13905 17665 pps 78 percent
Matrix operations 293623 148309 k ops 197 percent
Number crunch 315436 370078 k ops 85 percent
Floating point 33102 39423 k ops 83 percent
Memory throughput 1183653 603339 kB/sec 196 percent

Total CPU 129 percent
Total FPU 84 percent
Average 116 percent

Application Performance 126 percent


Specs of Dually in sig

Also, how are you guys posting this with the "code tags." I'm confused :confused:
 
Tim_axe said:
I'm not sure why, and haven't had time to look at any code yet or ask any questions.

The code is very disappointing. It's pretty sloppy, and has a few pretty bad bugs.

One of the worst is that the guy defines his own maxthreads constant; it's set to 128. Asking the program to use more than 64 threads will give unpredictable rseults, though, because he's trying to do WaitForMultipleHandle on that array of 128 handles ... and he can't, since WaitForMultipleHandles can't wait for more than 64 objects in one call.

Another issue for larger numbers of threads is that he starts the threads running in a loop. Since they aren't all started at the same time, he's timing his houskeeping calls --including CreateThread() itself, which isn't cheap.

Some of the sloppy coding might hinder performance, and certainly makes the measurements a bit goofy. But if we're all using the same program with the same problems, I guess it just comes out in the wash.

.B ekiM
 
Code:
                                this      other              relative
                            computer   computer           performance
 
Dhrystone 2.1                  12873      11682  kDhryst. 110 percent
Whetstone                       2601       2522  MFLOPS   103 percent
Eight queens problem           16695      17665  pps       94 percent
Matrix operations             369567     148309  k ops    249 percent
Number crunch                 379032     370078  k ops    102 percent
Floating point                 39102      39423  k ops     99 percent
Memory throughput            1460721     603339  kB/sec   242 percent
                                                                                                          
                                           Total CPU      159 percent
                                           Total FPU      101 percent
                                             Average      142 percent
                                                                     
                             Application Performance      155 percent
Dual 250's.
Tyan K8W.
4x1 Gig DDR400ECC memory.
Windozes XPpro SP2.
Two threads.
 
Code:
                                this      other              relative
                            computer   computer           performance
 
Dhrystone 2.1                   8656      11682  kDhryst.  74 percent
Whetstone                       1885       2522  MFLOPS    74 percent
Eight queens problem            9836      17665  pps       55 percent
Matrix operations             248226     148309  k ops    167 percent
Number crunch                 340333     370078  k ops     91 percent
Floating point                 32534      39423  k ops     82 percent
Memory throughput             750574     603339  kB/sec   124 percent
                                                                                                          
                                           Total CPU      102 percent
                                           Total FPU       78 percent
                                             Average       95 percent
                                                                     
                             Application Performance      101 percent

Dual 3.06Ghz Xeons w/HT
4 threads run
supermicro x5dal-g
533mhz fsb
1mb l3 cache
 
I've written some code to export this into CSV (Comma Seperated Values) format that can be imported into most spreadsheet software. This data can be fairly easily graphed. I may eventually graph a few scores and edit it into my first post.


The links can be found on the database page. Or, you can download the files from here (Note they'll save with the same name, so rename them if you want seperate AMD/Intel scores):

All Scores - http://files.timaxe.com/projects/clibench/export.php?type=all
AMD Scores - http://files.timaxe.com/projects/clibench/export.php?type=amd
Intel Scores - http://files.timaxe.com/projects/clibench/export.php?type=intel



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


mikeblas - That's dissapointing news to hear on this benchmark. :( I was expecting it to be more polished since it was a fairly well used benchmark in reviews a few years ago. Of course at that time I think it was single-threaded... Anyways, I also noticed weird results from the implementation when running the test with an "odd" number of threads. IE, 3 threads on a 2-CPU system results in scores ~75% lower than usual, while 4 threads results in fairly usual scores. Hopefully the fact that everyone is using the same benchmark can keep our results somewhat meaningful, although the exact scores are useless outside of this benchmark...

You can keep firing away with the results, I think I'm ready for a few more... :p

And I'll get around to the "average" scores sometime with the P3 1GHz system (and figure out where/how it would make sense). I need to figure out where to put it in, but I also need to clean-up my database-results-printing code a little bit. (And find a table color for Pentium 3 / P3 Xeon systems)
 
My new box Dual 2.2 Xeons with HT 1024 Pc800 RDRAM ECC
4 Threads run

Dhrystone 2.1--------------------------- 5846
Whetstone--------------------------- 1225
Eight queens problem--------------------------- 6945
Matrix operations--------------------------- 144958
Number crunch--------------------------- 215053
Floating point--------------------------- 20940
Memory throughput--------------------------- 908544
 
Tim_axe said:
You can keep firing away with the results, I think I'm ready for a few more... :p

I'm out of machines -- I mean, I have a few more, but they're all single-proc boxes.

Tim_axe said:
Hopefully the fact that everyone is using the same benchmark can keep our results somewhat meaningful, although the exact scores are useless outside of this benchmark

Yeah, I guess that's the saving grace of benchmark software. I'm a little worried that there are a couple of spots where the code in the benchmark is so bad that a processor/system that is truly more efficient is not able to show its true colors because of the code being so bad, though.

.B ekiM
 
Code:
                                this      other              relative
                            computer   computer           performance
 
Dhrystone 2.1                   9895      11682  kDhryst.  84 percent
Whetstone                       2261       2522  MFLOPS    89 percent
Eight queens problem           14601      17665  pps       82 percent
Matrix operations             131854     148309  k ops     88 percent
Number crunch                 354182     370078  k ops     95 percent
Floating point                 36203      39423  k ops     91 percent
Memory throughput            1301559     603339  kB/sec   215 percent

                                           Total CPU      112 percent
                                           Total FPU       90 percent
                                             Average      106 percent
                                                                     
                             Application Performance      111 percent

Iwill DH800
Xeon 3.06GHz @ 3.28GHz (156x21) x 2
2GB PC3500 CL2
WinXP Pro + SP2
 
This is on an Asus PC-DL Deluxe with 1.6LV's running @ 12X200 (2600MHz) at stock voltage. :) 4 threads
2_1024.ts1108270333483.jpg
 
Sorry for the size, it is the wacky website I am using to host them. :rolleyes:
 
here

Code:
               this      other              relative
                            computer   computer           performance
 
Dhrystone 2.1                   7091      11682  kDhryst.  60 percent
Whetstone                       1959       2522  MFLOPS    77 percent
Eight queens problem            8496      17665  pps       48 percent
Matrix operations             135505     148309  k ops     91 percent
Number crunch                 379951     370078  k ops    102 percent
Floating point                 30090      39423  k ops     76 percent
Memory throughput            1152499     603339  kB/sec   191 percent
                                                                                                          
                                           Total CPU       98 percent
                                           Total FPU       76 percent
                                             Average       92 percent
                                                                     
                             Application Performance       97 percent
 
Got some new D1 stepping 1.6LV Xeons, between that and a Vdimm mod to my PC-DL, I can't be happier with the results. :)

Code:
                                this      other              relative
                            computer   computer           performance
 
Dhrystone 2.1                   9304      11682  kDhryst.  79 percent
Whetstone                       2156       2522  MFLOPS    85 percent
Eight queens problem           10602      17665  pps       60 percent
Matrix operations             272343     148309  k ops    183 percent
Number crunch                 360706     370078  k ops     97 percent
Floating point                 36253      39423  k ops     91 percent
Memory throughput            1467282     603339  kB/sec   243 percent
                                                                                                          
                                           Total CPU      132 percent
                                           Total FPU       88 percent
                                             Average      119 percent
                                                                     
                             Application Performance      129 percent


ASUS PC-DL R1.05, Vdimm modded to 2.85V
Intel Xeon 1.6LV D1s @ 3285 MHz (219x15)
1GB PC3200 CL2.5 Kingston HyperX 2x512M
WinXP Pro + SP2
 
Code:
                                this      other              relative
                            computer   computer           performance
 
Dhrystone 2.1                  10767      11682  kDhryst.  92 percent
Whetstone                       2854       2522  MFLOPS   113 percent
Eight queens problem           13984      17665  pps       79 percent
Matrix operations             199089     148309  k ops    134 percent
Number crunch                 316072     370078  k ops     85 percent
Floating point                 31345      39423  k ops     79 percent
Memory throughput            1167054     603339  kB/sec   193 percent

                      
                                           Total CPU      116 percent
                                           Total FPU       96 percent
                                             Average      110 percent
                                                                     
                             Application Performance      115 percent


This is my new dual-Opteron rig. It has two 246 processors and is on a Tyan S2882 motherboard. I ran two threads.

.B ekiM
 
Code:
                                this      other              relative
                            computer   computer           performance
 
Dhrystone 2.1                  10720      11682  kDhryst.  91 percent
Whetstone                       1881       2522  MFLOPS    74 percent
Eight queens problem           15415      17665  pps       87 percent
Matrix operations             282886     148309  k ops    190 percent
Number crunch                 304601     370078  k ops     82 percent
Floating point                 28391      39423  k ops     72 percent
Memory throughput            1715967     603339  kB/sec   284 percent
                                                                                                          
                                           Total CPU      146 percent
                                           Total FPU       73 percent
                                             Average      125 percent
                                                                     
                             Application Performance      141 percent

Xeon 2.66 @3.1 HT/off
Asus PC-DL
2gb PC3200 Kingston
Win XP Pro w/sp2
 
System:
Iwill MPX2
2 X Athlon XP-M 2600s @ 2500MHz
2 X 512 Crucial PC2700 ECC

Code:
                                this      other              relative
                            computer   computer           performance
 
Dhrystone 2.1                  12141      11682  kDhryst. 103 percent
Whetstone                       2662       2522  MFLOPS   105 percent
Eight queens problem           18389      17665  pps      104 percent
Matrix operations             145415     148309  k ops     98 percent
Number crunch                 384930     370078  k ops    104 percent
Floating point                 41646      39423  k ops    105 percent
Memory throughput             716686     603339  kB/sec   118 percent
                                                                                                          
                                           Total CPU      105 percent
                                           Total FPU      105 percent
                                             Average      105 percent
                                                                     
                             Application Performance      105 percent
 
I don't want to make anyone cry, but here are my results:

Code:
                                this      other              relative
                            computer   computer           performance
 
Dhrystone 2.1                  96141      11682  kDhryst. 822 percent
Whetstone                      25888       2522  MFLOPS   1026 percent
Eight queens problem          123144      17665  pps      697 percent
Matrix operations            2978970     148309  k ops    2008 percent
Number crunch                2795539     370078  k ops    755 percent
Floating point                276436      39423  k ops    701 percent
Memory throughput             846563     603339  kB/sec   140 percent
                                                                     
                                           Total CPU      884 percent
                                           Total FPU      863 percent
                                             Average      878 percent
                                                                     
                             Application Performance      883 percent

Specs: 8xAMD Dual Core Engineering Sample (2.21ghz) [16 Cores]
128GB of RAM
Sitting in an IWill H8501 Barebones package.

Complete benchmarks and a writeup will be available on Extended64.com soon (tomorrow probably), but I was just too excited after running this benchmark that I had to post it!
 
Extended64 said:
I don't want to make anyone cry, but here are my results:

Code:
                                this      other              relative
                            computer   computer           performance
 
Dhrystone 2.1                  96141      11682  kDhryst. 822 percent
Whetstone                      25888       2522  MFLOPS   1026 percent
Eight queens problem          123144      17665  pps      697 percent
Matrix operations            2978970     148309  k ops    2008 percent
Number crunch                2795539     370078  k ops    755 percent
Floating point                276436      39423  k ops    701 percent
Memory throughput             846563     603339  kB/sec   140 percent
                                                                     
                                           Total CPU      884 percent
                                           Total FPU      863 percent
                                             Average      878 percent
                                                                     
                             Application Performance      883 percent

Specs: 8xAMD Dual Core Engineering Sample (2.21ghz) [16 Cores]
128GB of RAM
Sitting in an IWill H8501 Barebones package.

Complete benchmarks and a writeup will be available on Extended64.com soon (tomorrow probably), but I was just too excited after running this benchmark that I had to post it!

:eek: :eek: :eek: :eek:
 
Well, I decided to try making another graph, and I had to clip your scores to even see the rest of them... :eek: :eek: :eek:




(I can safely say the 8x P3 Xeon system was crushed in all CPU intensive tasks, along with everyone else. I think you should have also dominated memory bandwidth, but this test can't take advantage of how your machine sees RAM so you were out of luck there. Overall very impressive results there :))

Download and fire up Nasa's World Wind application and watch that beast scale it across those processors with ease...it's one of the few applications I've had that can take good advantage of SMP machines...though the most I've run it on is my dually...

I hope to read about this machine soon :drool:
 
the first result is with World Community Grid on. the second is with agent in snooze mode.

this other relative
computer computer performance

Dhrystone 2.1 8093 11682 kDhryst. 69 percent
Whetstone 1753 2522 MFLOPS 69 percent
Eight queens problem 12239 17665 pps 69 percent
Matrix operations 123960 148309 k ops 83 percent
Number crunch 256410 370078 k ops 69 percent
Floating point 27697 39423 k ops 70 percent
Memory throughput 534478 603339 kB/sec 88 percent

Total CPU 75 percent
Total FPU 69 percent
Average 73 percent

Application Performance 75 percent


this other relative
computer computer performance

Dhrystone 2.1 8082 11682 kDhryst. 69 percent
Whetstone 1757 2522 MFLOPS 69 percent
Eight queens problem 12239 17665 pps 69 percent
Matrix operations 123075 148309 k ops 82 percent
Number crunch 255991 370078 k ops 69 percent
Floating point 27683 39423 k ops 70 percent
Memory throughput 529496 603339 kB/sec 87 percent

Total CPU 75 percent
Total FPU 69 percent
Average 73 percent

Application Performance 75 percent
and this is my pc:

Number of CPUs 2

CPU #1
Name AMD Athlon MP
Code name Palomino
Specification AMD Athlon(tm) MP 2000+
Family/Model/Stepping 662
Extended Family/Model 7/6
Package Socket A
Core Stepping A5
Technology 0.18µ
Instructions Sets MMX, Extended MMX, 3DNow!, Extended 3DNow!, SSE
Clock Speed 1666.8 MHz
Clock multiplier x12.5
Front Side Bus Frequency 133.3 MHz
Bus Speed 266.7 MHz
P-Rating 2000+
L1 Data Cache 64 KBytes, 2-way set associative, 64 Bytes line size
L1 Instruction Cache 64 KBytes, 2-way set associative, 64 Bytes line size
L2 Cache 256 KBytes, 16-way set associative, 64 Bytes line size
L2 Speed 1666.8 MHz (Full)
L2 Location On Chip
L2 Data Prefetch Logic yes
L2 Bus Width 64 bits

CPU #2
Name AMD Athlon MP
Code name Palomino
Specification AMD Athlon(tm) Processor
Family/Model/Stepping 662
Extended Family/Model 7/6
Package Socket A
Core Stepping A5
Technology 0.18µ
Instructions Sets MMX, Extended MMX, 3DNow!, Extended 3DNow!, SSE
Clock Speed 1666.8 MHz
Clock multiplier x12.5
Front Side Bus Frequency 133.3 MHz
Bus Speed 266.7 MHz
L1 Data Cache 64 KBytes, 2-way set associative, 64 Bytes line size
L1 Instruction Cache 64 KBytes, 2-way set associative, 64 Bytes line size
L2 Cache 256 KBytes, 16-way set associative, 64 Bytes line size
L2 Speed 1666.8 MHz (Full)
L2 Location On Chip
L2 Data Prefetch Logic yes
L2 Bus Width 64 bit
Memory Type DDR
Memory Size 1024 MBytes
Memory Frequency 133.3 MHz (1:1)
CAS# 2.5
RAS# to CAS# 3
RAS# Precharge 3
Cycle Time (tRAS) 7
Bank Cycle Time (tRC) 10
DRAM Idle Timer 8
DIMM #2
General
Memory type DDR-SDRAM
Manufacturer (ID) Fairchild (837F7F7F00000000)
Size 512 MBytes
Max bandwidth PC2100 (133 MHz)
Number of banks 2
Data width 64 bits
Correction None
Registered no
Buffered no
Timings table
Frequency (MHz) 100 133
CAS# 2.0 2.5
RAS# to CAS# delay 2 3
RAS# Precharge 2 3
TRAS# 5 6
DIMM #4
General
Memory type DDR-SDRAM
Manufacturer (ID) Melco (7F7F7F8300000000)
Size 512 MBytes
Max bandwidth PC2700 (166 MHz)
Attributes
Number of banks 2
Data width 64 bits
Correction None
Registered no
Buffered no
Timings table
Frequency (MHz) 133 166
CAS# 2.0 2.5
RAS# to CAS# delay 3 3
RAS# Precharge 3 3
TRAS# 6 7
System #1
manufacturer TYAN
product PAULANER
version X2.0
serial 0123456789
UUID 00000000-00000000-00000000-0000000
DMI Physical Memory Array
Physical Memory Array #1
location Motherboard
usage System Memory
correction None
max capacity 4096MBytes
max# of devices 3
Software
Windows Version Microsoft Windows XP Professional Service Pack 2 (Build 2600)
DirectX Version 9.0c

sr for long big post:)
 
Tim_axe said:
Download and fire up Nasa's World Wind application and watch that beast scale it across those processors with ease...it's one of the few applications I've had that can take good advantage of SMP machines...though the most I've run it on is my dually...

Heh, problem is it only has an 8mb ATI Rage XL video card, and I'm using it all via Remote Desktop :)

Forgot to mention, it's running Windows Server 2003 Enterprise x64 Edition.

Oh, and I figured out that this processor is the Opteron 875. So it has 8xOpteron 875 which are all Dual Core, making it have 16 cores.
 
Extended64 said:
Heh, problem is it only has an 8mb ATI Rage XL video card, and I'm using it all via Remote Desktop :)

Forgot to mention, it's running Windows Server 2003 Enterprise x64 Edition.

Oh, and I figured out that this processor is the Opteron 875. So it has 8xOpteron 875 which are all Dual Core, making it have 16 cores.
How much did it all cost? ;)
 
Extended64 said:
.

Oh, and I figured out that this processor is the Opteron 875. So it has 8xOpteron 875 which are all Dual Core, making it have 16 cores.

...drools uncontrollably :eek:
 
Extended64 said:
Specs: 8xAMD Dual Core Engineering Sample (2.21ghz) [16 Cores]
128GB of RAM
Sitting in an IWill H8501 Barebones package.

Complete benchmarks and a writeup will be available on Extended64.com soon (tomorrow probably), but I was just too excited after running this benchmark that I had to post it!
THIS is why we need a :drool: smiley.
 
Extended64 said:
You can't buy this stuff. Assuming you could, I'm willing to guess it would be well over $10,000.
Happen to have any close approximate? thanks
 
Tim_axe said:
(I can safely say the 8x P3 Xeon system was crushed in all CPU intensive tasks,

I would hope so! My 8x P3 machine is more than four years old, and you're comparing it to a machine that isn't available retail yet.

Based on the pricing I've seen for the dual core parts, I think the $10,000 estimate is very conservative. $10,000 won't even pay for the processors, and you'll still need to buy 128 gigs of memory and an operating system.
 
Anyone have any other multiprocessor aware benchmarks that I can throw at this system?

What about a *decent* 64-bit memory benchmark? The only one I can find, in SiSoft Sandra locks the machine up (I don't think it can handle 128gb of ram!)

Btw, to whoever runs that page: can you change the OS from Windows XP SP2 to Windows Server 2003 Enterprise x64 Edition? XP SP2 isn't 64-bit, and I think it only handles 2 cpus.
 
Zxcs said:
Happen to have any close approximate? thanks

This is a very rare class of machine. Right now, few boxes comes close because Intel quit making 8-way machines; they said 4-way boxes with HyperThreading were enough. To have more than 4-way Xeon means you need to do build a node-oriented box.

The 4x Xeon machines are not hard to find. HP Proliant DL760 and DL730 boxes are easily available. The cheapest preconfigured DL760 on their site right now is $25,000, and has only two gigs of memory:

• Four (4) Intel® Xeon™ MP 2.20GHz/2MB Processors
• 2GB Addressable Memory (8x256MB + 2x256MB Redundant Memory)
• Integrated Smart Array 5i Controller (Ultra3 support)
• Integrated HP NC7170 Dual Port PCI-X 1000T Gigabit Server Adapter
• Two (2) Redundant Hot Plug 1150W/500W Power Supplies
• Two (2) Redundant Hot Plug Fans
• SmartStart, Insight Manager 7 & ROM-Based Setup Utility
• Warranty - 3 year next business day onsite

Getting to eight Xeon processors gets you over $36,000, and you've still only got 2 gigs of memory. The box can only hold 64 gigs.

To get bigger than that, you have to start looking at machines like the HP SuperDome or the Unisys ES7000 series. The Aries 510 has two, three, or four nodes (for four, six, or eight processors), up through the Orion line, which will have up to 32 processors.
 
Btw, to whoever runs that page: can you change the OS from Windows XP SP2 to Windows Server 2003 Enterprise x64 Edition? XP SP2 isn't 64-bit, and I think it only handles 2 cpus.

I'll be sure to change that later tonight. Right now I can't SSH into my server to change that value (or add MPX's results), but I'll be sure to do so once I get home.

I don't know of many multiprocessor benchmarks myself; CliBench was the only one I came across that seemed to use more than one processor. (And we've come to the conclusion that the way it operates isn't very good, but at least we all run the same program/benchmark) When I get home tonight I'll check to see if there are any other benchmarks I can find on my machine... (Have you tried CPU-Z's L1/L2/L3 Latency Benchmarks? Results would be average depending on how much cache the cores have, but you'll have more some numbers to look at and perhaps compare to regular Opteron machines to see if there is any difference, though I predict none)


Edit: Your information has been updated. Thanks for pointing out it was wrong :)
 
Extended64 -- Try loading your processors with Cinebench. ( http://www.cinebench.com/ )

See if you can beat what is on this page -- http://techreport.com/reviews/2005q2/opteron-x75/index.x?pg=8

My machine does the single CPU test with a score of 289 CB-CPU, and with 2 processors I hit 533 CB-CPU. At least I'm holding my ground against Dual Opteron 246's (528 CB-CPU @ TechReport)... I'd love to see what 16 processors can do (which, happens to be the most CINEBench application can scale to). Chances are you'd finish the benchmark in under 5 seconds...
 
hey guys, i was just googling for some SMP benches since i have this dual core 4400+ A64 and happened upon this thread ;)

been a while since I been around these parts. Some of you know me from DFI-Street but for those that don't, I'm a lamer yes...so ignore me heh

here's the system that I am using:

http://www.dfi-street.com/forum/showpost.php?p=113497&postcount=9 (fully benched in this link)

DFI LANPARTY nF4 SLI-DR NF44 SLI-DR - 5/10/2005-2 BIOS

AMD Athlon 64 X2 4400+ Dual Core Socket 939 Processor Model

2x512MB OCZ Gold Edition 4000VX - (PN: OCZ5001024ELDCGEVX-K)

Nvidia 6800GT PCI-E engineering sample, 76.41 beta drivers - 401/1100

3x HITACHI Deskstar 7K80 HDS728080PLA380 -0A30356 80GB 7200 RPM Serial ATA II Hard Drive - OEM (RAID-0)

ASUS CD-S520/A5 Black IDE CD-ROM Drive

OCZ PowerStream OCZ600ADJ ATX 600W Power Supply - Retail

Thermalright XP-90 aluminum heatpipe heatsink + Enermax 92mm adjustable fan




Code:
                                this      other              relative
                            computer   computer           performance
 
Dhrystone 2.1                  14476      11682  kDhryst. 123 percent
Whetstone                       3915       2522  MFLOPS   155 percent
Eight queens problem           18519      17665  pps      104 percent
Matrix operations             461390     148309  k ops    311 percent
Number crunch                 420769     370078  k ops    113 percent
Floating point                 41863      39423  k ops    106 percent
Memory throughput            1658082     603339  kB/sec   274 percent
                                                                                                          
                                           Total CPU      185 percent
                                           Total FPU      130 percent
                                             Average      169 percent
                                                                     
                             Application Performance      181 percent


let me know if I missed any info!

ps: Tim, where in Idaho? I live in Boise =/
 
Angry_Games -- That system pushes Zxcs's Phase-Change Cooled Dual Opteron 250 @ 2.703GHz down a spot in this benchmark (probably because your memory timings are much tighter)... He seems to be running it at 2.8GHz now, so I'll see if I can have him retest it. Either way though, Dual Core really changes the Dual CPU landscape to make it much more affordable.


BTW, I won't be able to update the page on my server for a while. The file ownership (and permissions) were taken away from my (limited) user account so I can't edit it at all anymore :(


ps - I'm near Moscow, ID -- University of Idaho.

pps - Try cinebench out too. Cheers.
 
Cinebench results, same system as before.
Code:
CINEBENCH 2003 v1
****************************************************

Tester           : Moog

Processor        : Opteron
MHz              : 1600 x2
Number of CPUs   : 2
Operating System : Win XP SP2

Graphics Card    : 9800 Pro
Resolution       : <fill this out>
Color Depth      : <fill this out>

****************************************************

Rendering (Single   CPU): 222 CB-CPU 
Rendering (Multiple CPU): 413 CB-CPU 

Multiprocessor Speedup: 1.86

Shading (CINEMA 4D)                : 261 CB-GFX 
Shading (OpenGL Software Lighting) : 1230 CB-GFX 
Shading (OpenGL Hardware Lighting) : 2724 CB-GFX 

OpenGL Speedup: 10.43

****************************************************
 
thx Tim (wasnt sure if you guys were still following this thread heh)

I ran the 64-bit Cinebench but I'm new at that bench so not familiar with the scores etc. I will run it again on the 64-bit (RC1433) version I have on the rig again and post the scores, then see if i can reload again with 32-bit (running 3x SATA II drives in RAID so it wont take that long)

ill check through this thread at anyone else who might have posted Cinebench scores too just for personal comparison
 
My Dually low voltage ES Xeon 2.4's running at 3.4 Ghz. Running with 2GB Corsair XMS and an Asus NCCH-DL motherboard.

Code:
Results

                                this      other              relative
                            computer   computer           performance
 
Dhrystone 2.1                   9599      11682  kDhryst.  82 percent
Whetstone                       2131       2522  MFLOPS    84 percent
Eight queens problem           10815      17665  pps       61 percent
Matrix operations             241137     148309  k ops    162 percent
Number crunch                 371101     370078  k ops    100 percent
Floating point                 36341      39423  k ops     92 percent
Memory throughput            1342638     603339  kB/sec   222 percent
                                                                                                          
                                           Total CPU      125 percent
                                           Total FPU       88 percent
                                             Average      114 percent
                                                                     
                             Application Performance      122 percent
 
loafer87gt - your score is in.

Feel free to submit more scores. I'd love to see where the new Dual Core systems fall into on this list as they become avaliable, and I think it may help people choose Dually vs Dual Core as time progresses. Of course if they go single socket dual core they would miss out on some 4-way action :p

Results database (with sorting functions) so you don't need to dig through the thread for the top-performing systems (and systems similar to your own) -> http://files.timaxe.com/projects/clibench/



I also took some time to optimize the database code on my page. I created some functions to handle the general tasks of displaying results and sorting them, which cut the size of my PHP code from 22KB to 8KB. It's also much easier to read/edit/update, so before long I'll write a function to compare the results to a "base" score provided by one of mikeblas's Dual 1GHz P3 systems
 
Code:
 			 	this	 other 		 relative
 			 	computer computer	performance
                   
 Dhrystone 2.1			 5832 	11682 kDhryst. 49 percent
 Whetstone			 1436 	2522 MFLOPS	 56 percent
 Eight queens problem 		 9871	17665 pps	 55 percent
 Matrix operations		 87977 	148309 k ops	 59 percent
 Number crunch			 201284 370078 k ops	 54 percent
 Floating point			 21916 	39423 k ops	 55 percent
     Memory throughput		 224083 603339 kB/sec	37 percent

Dual Pentium III-s 1.4ghz (stock)
MSI Pro266TD Master-LR
2GB PC2100, CL2
Maxtor 60GB 7k2 IDE drive

edit: why is my code dealie so messed up? i copied it striaght from the txt file...
 
Back
Top