Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

I am trying to do comparision of xeon processor and utrasparc IV+

for the xeon, I have this result:

Cisco UCS B200M3 Intel Xeon Processor E5-2660 (Sandy Bridge)

Hardware Vendor  System Result Baseline # Cores # Chips  # Cores Per Chip  Published Disclosure 
Bull SAS NovaScale R440 F3 (Intel Xeon E5-2660, 2.20 GHz) 47.6  44.5  16 2 8 Apr-2012 HTML CSV PDF PS Text Config 
Bull SAS NovaScale R460 F3 (Intel Xeon E5-2660, 2.20 GHz) 47.7  44.5  16 2 8 Apr-2012 HTML CSV PDF PS Text Config 
Bull SAS NovaScale T840 F3 (Intel Xeon E5-2660, 2.20 GHz) 47.4  44.3  16 2 8 Mar-2012 HTML CSV PDF PS Text Config 
Cisco Systems Cisco UCS C220 M3 (Intel Xeon E5-2660, 2.20 GHz) 48.3  45.0  16 2 8 Jul-2012 HTML CSV PDF PS Text Config 
Cisco Systems Cisco UCS C240 M3 (Intel Xeon E5-2660, 2.20 GHz) 48.5  45.2  16 2 8 Jul-2012 HTML CSV PDF PS Text Config 

for ultrasparc IV,

Model   Processor Type  Processor   CPU Cores   No of Processors    Memory
    SUN FIRE V490   SUN ULTRASPARC IV+  1500    8.00    4.00    32768

Hardware Vendor System Result Baseline # Cores # Chips # Cores Per Chip Published Disclosure Sun Microsystems Sun Fire V490 78.0 71.7 8 4 2 Apr-2007 HTML CSV PDF PS Text Config

Looking at these results, which one performed better, how do you read this output?

this is the spec output: Intel Xeon Processor E7-4860

Benchmark   Base # Copies   Base Run Time   Base Rate   Base Selected   Base Status Peak # Copies   Peak Run Time   Peak Rate   Peak Selected   Peak Status Description         
400.perlbench   80  988.746037  790.49624   1   S   80  827.236707  944.83232   1   S   SelectedIteration (base #3; peak #2)            
401.bzip2   80  1401.357446 550.8944    1   S   80  1329.663649 580.59792   1   S   SelectedIteration (base #3; peak #1)            
403.gcc 80  821.134215  784.28104   1   S   80  818.863431  786.45592   1   S   SelectedIteration (base #1; peak #1)            
429.mcf 80  627.881517  1162.00264  1   S   80  596.869458  1222.37784  1   S   SelectedIteration (base #3; peak #3)            
445.gobmk   80  929.814428  902.54568   1   S   80  883.392809  949.97376   1   S   SelectedIteration (base #1; peak #3)            
456.hmmer   80  606.72107   1230.21936  1   S   80  481.870461  1548.964    1   S   SelectedIteration (base #3; peak #1)            
458.sjeng   80  1111.901674 870.58056   1   S   80  1040.358199 930.4488    1   S   SelectedIteration (base #2; peak #2)            
462.libquantum  80  366.940844  4517.34936  1   S   80  366.940844  4517.34936  1   S   SelectedIteration (base #2; peak #2)            
464.h264ref 80  1437.900502 1231.2396   1   S   80  1444.884847 1225.28792  1   S   SelectedIteration (base #1; peak #1)            
471.omnetpp 80  900.233245  555.4116    1   S   80  840.44434   594.92336   1   S   SelectedIteration (base #2; peak #1)            
473.astar   80  974.344331  576.3876    1   S   80  974.344331  576.3876    1   S   SelectedIteration (base #3; peak #3)            
483.xalancbmk   80  575.587683  959.01984   1   S   80  575.587683  959.01984   1   S   SelectedIteration (base #1; peak #1)            

Intel Xeon Processor E5-2660 (Sandy Bridge)                                                     
Benchmark   Base Ref Time   Base Run Time   Base Ratio  Base Selected   Base Status Peak Ref Time   Peak Run Time   Peak Ratio  Peak Selected   Peak Status Description         
400.perlbench   9770    363.71793   26.861475   1   S   9770    308.821887  31.636359   1   S   SelectedIteration (base #1; peak #3)            
401.bzip2   9650    481.936537  20.023383   1   S   9650    473.177143  20.394054   1   S   SelectedIteration (base #2; peak #2)            
403.gcc 8050    275.243939  29.246784   1   S   8050    273.476945  29.435754   1   S   SelectedIteration (base #1; peak #2)            
429.mcf 9120    152.092584  59.963476   1   S   9120    152.092584  59.963476   1   S   SelectedIteration (base #3; peak #3)            
445.gobmk   10490   492.109626  21.316389   1   S   10490   455.186823  23.045483   1   S   SelectedIteration (base #3; peak #2)            
456.hmmer   9330    202.295578  46.120632   1   S   9330    197.804691  47.167739   1   S   SelectedIteration (base #3; peak #3)            
458.sjeng   12100   484.218283  24.98873    1   S   12100   486.068138  24.893629   1   S   SelectedIteration (base #2; peak #2)            
462.libquantum  20720   8.110392    2554.747046 1   S   20720   8.110392    2554.747046 1   S   SelectedIteration (base #2; peak #2)            
464.h264ref 22130   567.298128  39.009471   1   S   22130   463.737891  47.720923   1   S   SelectedIteration (base #3; peak #3)            
471.omnetpp 6250    252.260035  24.776021   1   S   6250    185.447273  33.702302   1   S   SelectedIteration (base #1; peak #2)            
473.astar   7020    262.741702  26.718256   1   S   7020    262.741702  26.718256   1   S   SelectedIteration (base #3; peak #3)            
483.xalancbmk   6900    151.761465  45.466087   1   S   6900    144.520671  47.744035   1   S   SelectedIteration (base #2; peak #3)            


    SUN FIRE V490   SUN ULTRASPARC IV+                                                  
Benchmark   Base # Copies   Base Run Time   Base Rate   Base Selected   Base Status Peak # Copies   Peak Run Time   Peak Rate   Peak Selected   Peak Status Description         
400.perlbench   8   1224.828996 63.81298961 1   S   8   891.155322  87.7063718  1   S   SelectedIteration (base #2; peak #3)            
401.bzip2   8   1580.039751 48.85953024 1   S   8   1318.300807 58.56023116 1   S   SelectedIteration (base #1; peak #3)            
403.gcc 8   1186.319051 54.28556504 1   S   8   1118.707572 57.56642899 1   S   SelectedIteration (base #2; peak #2)            
429.mcf 8   715.023723  102.0385725 1   S   8   714.701975  102.0845087 1   S   SelectedIteration (base #3; peak #2)            
445.gobmk   8   1181.380568 71.03553442 1   S   8   1001.299034 83.8111265  1   S   SelectedIteration (base #1; peak #3)            
456.hmmer   8   1061.728244 70.30047512 1   S   8   1055.574115 70.71033567 1   S   SelectedIteration (base #3; peak #1)            
458.sjeng   8   1527.577971 63.36828747 1   S   8   1415.486478 68.38638271 1   S   SelectedIteration (base #1; peak #1)            
462.libquantum  8   1289.56188  128.5397797 1   S   8   1278.533905 129.6484977 1   S   SelectedIteration (base #2; peak #1)            
464.h264ref 8   1757.491311 100.7344952 1   S   8   1543.499756 114.7003745 1   S   SelectedIteration (base #1; peak #2)            
471.omnetpp 8   1007.226126 49.64128581 1   S   8   947.26431   52.78357843 1   S   SelectedIteration (base #2; peak #3)            
473.astar   8   871.046418  64.47417593 1   S   8   871.046418  64.47417593 1   S   SelectedIteration (base #2; peak #2)            
483.xalancbmk   8   677.614178  81.46228605 1   S   8   677.614178  81.46228605 1   S   SelectedIteration (base #3; peak #3)            
share|improve this question

closed as off-topic by voretaq7 Aug 15 '13 at 16:28

  • This question does not appear to be about professional server, networking, or related infrastructure administration within the scope defined in the help center.
If this question can be reworded to fit the rules in the help center, please edit the question.

    
This question appears to be off-topic because it is about interpretation of benchmark results - This is the sort of thing you should talk to the benchmark maker about (spec.org would be the resource for SPEC benchmarks). With no information about your environment or application we can't conclusively say what is "better" or "worse" for your specific application. Benchmarks are just raw data - you need to synthesize information by matching them with your use case. –  voretaq7 Aug 15 '13 at 16:28

2 Answers 2

It's not that simple a question, different chips will do better at different tasks. Generally for moderate use pure compute Intel will come out better price/performance but Sparc can beat it, especially at high end multithread processes that require high concurrency/low latency.

If you don't know you need otherwise you probably want to stick with Intel and before you start looking at exotic hardware (ok, Sun is not exactly exotic...) you need to identify where your bottleneck is. If your are bandwidth & latency limited on your storage subsystem all the processing power in the world won't make you go faster.

share|improve this answer
    
@TheFiddlerWinds, I am looking at the spec results. there is base and ratio. Looking at these results for both chips, I think I can be very close. I need to understand how do they calculate base and ratio? –  user1471980 Aug 15 '13 at 14:19
1  
I understand but artificial benchmarks are very hard to line up with real world apps. Well, usually, if you're running brute force decryption or modeling atomic bombs it may be different. This is old but a great write up that might be worth reading: archive.arstechnica.com/cpu/2q99/benchmarking-1.html –  TheFiddlerWins Aug 15 '13 at 14:37
    
thank you for this article. I have one question if base is 50 and bencmarked is 60, does that mean your benchmark result is worse or better than reference by 10 points? –  user1471980 Aug 15 '13 at 15:06
    
Don't ask us how to read a benchmark result that you found. Please do some of your own work. You didn't even provide the URL for us to know what the column headings mean. Unless the author of that report is on here, how do you realistically expect an answer to that question? –  mfinni Aug 15 '13 at 16:26

The only way to know how this will impact your app is to actually benchmark your app on the different platforms. It's not like you'll be running the same binaries anyway. Recompile your app for Intel arch, and run it on a workstation to get some performance numbers. Then compare the workstation to some server specs.

share|improve this answer
    
yes you are 100% right but that is not the case now. I think looking at the spec result and comparing the base and result could be great indication. –  user1471980 Aug 15 '13 at 15:11
    
It's almost entirely as if you don't understand what TheFiddlerWins and myself are telling you. –  mfinni Aug 15 '13 at 15:34
    
I understand it, 100%. But I have no choice to go by spec results. –  user1471980 Aug 15 '13 at 15:48
    
He didn't seem to understand anything others said when he asked virtually the same question yesterday either: how do you convert solaris ultrasparc cpu utilization to xeon –  yoonix Aug 15 '13 at 15:49
1  
OK - then you need to determine which one of those benchmarks is closest to what your application does, and then just compare only that benchmark for the two architectures to see which is better. Hope you've profiled your app to know where it spends most of its time. Assuming CPU is even the bottleneck.... –  mfinni Aug 15 '13 at 16:17

Not the answer you're looking for? Browse other questions tagged or ask your own question.