Page 2 of 3 FirstFirst 123 LastLast
Results 26 to 50 of 60
  1. #26

    Red face

    Dude.. Did you look at what results REAL people are getting? I don't seem to understand how a 2 people with identical setups get the same performance, and your tests results are quadruple the speed results with SATA drives and a Slower processor. Am I missing something? The key isn't a one time benchmark setup, but consistent repeatable results.

    It has already been referenced that a person with similar hardware as I, have results similar to mine.

    Taken from the very same post you reference from. Does 64bit OS really help increase speed 4x with slower hardware? or are your results flawed?


    @justbenice : your result is same with me. same server conf (dual 244, 2gb ram, scsi) os centos 3.4 32 bit , Linux version 2.4.21-27.0.2.ELsmp


    sign verify sign/s verify/s
    rsa 512 bits 0.0004s 0.0000s 2249.2 23532.7
    rsa 1024 bits 0.0023s 0.0001s 427.5 7800.1
    rsa 2048 bits 0.0093s 0.0003s 107.1 3464.6
    rsa 4096 bits 0.0675s 0.0012s 14.8 850.3
    From these results from 2 people could you deduce that a single Opteron could sign/s 215 rsa 1024 bits? That would mean the Sempron 3100+ 1gb is ~50% faster than a single 244 opteron 2gb ecc. and that a Sempron 3100+ is only 15% slower than a dual opteron setup with twice the memory. Please explain???

    wm2100 results

    Opteron 242 2P, 2GB ram, 4x 200GB SATA

    openssl speed rsa -multi 2

    sign verify sign/s verify/s
    rsa 512 bits 0.0002s 0.0000s 5952.6 83333.3
    rsa 1024 bits 0.0007s 0.0000s 1532.6 30303.0
    rsa 2048 bits 0.0038s 0.0001s 260.1 8948.6
    rsa 4096 bits 0.0260s 0.0004s 38.4 2590.7

  2. #27
    Join Date
    Aug 2000
    Location
    Sheffield, South Yorks
    Posts
    3,627
    Two CPUs are not twice as fast, never has been, never will be, you're just showing a little bit of ignorance now, first comparing systems not CPUs in benchmarks - If everything isn't the same, except the CPU/Mobo, then you're comparing a system, not a CPU and now you're still going for the 2 CPUs = Twice as fast, even though it has been pointed out earlier that this is not the case. SMP does not scale linearly - Add a 2nd CPU and get maybe 60-70% improvement, not 100%, add 2 more CPUs and you might see 59-69% improvement, it isn't linear, improvements have been made, but it still isn't linear.
    Karl Austin :: KDAWS.com
    The Agency Hosting Specialist :: 0800 5429 764
    Partner with us and free-up more time for income generating tasks

  3. #28
    Join Date
    Aug 2004
    Posts
    371
    Originally posted by ikeo
    Dude.. Did you look at what results REAL people are getting? I don't seem to understand how a 2 people with identical setups get the same performance, and your tests results are quadruple the speed results with SATA drives and a Slower processor. Am I missing something? The key isn't a one time benchmark setup, but consistent repeatable results.

    It has already been referenced that a person with similar hardware as I, have results similar to mine.

    Taken from the very same post you reference from. Does 64bit OS really help increase speed 4x with slower hardware? or are your results flawed?




    From these results from 2 people could you deduce that a single Opteron could sign/s 215 rsa 1024 bits? That would mean the Sempron 3100+ 1gb is ~50% faster than a single 244 opteron 2gb ecc. and that a Sempron 3100+ is only 15% slower than a dual opteron setup with twice the memory. Please explain???

    wm2100 results
    Ikeo, clearly, there are some problems with the two people's results of extremely low scores. In benchmarks you compare optimal or near optimal settings. I believe that Anandtech, as a very experienced lab, setup the sempron in near optimal configuration. Tomshardware, an experienced lab, also setup their xeon 2.8 pretty well. So the mysql results from these two labs should be rather conclusive.

    We see high flying openssl scores for Opteron 242s, and two machines are showing rather low scores. You have to investigate why these two machines are slow. You will need to check motherboard and other settings to see why.

  4. #29
    Join Date
    Aug 2004
    Posts
    371
    Originally posted by KDAWebServices
    Two CPUs are not twice as fast, never has been, never will be, you're just showing a little bit of ignorance now, first comparing systems not CPUs in benchmarks - If everything isn't the same, except the CPU/Mobo, then you're comparing a system, not a CPU and now you're still going for the 2 CPUs = Twice as fast, even though it has been pointed out earlier that this is not the case. SMP does not scale linearly - Add a 2nd CPU and get maybe 60-70% improvement, not 100%, add 2 more CPUs and you might see 59-69% improvement, it isn't linear, improvements have been made, but it still isn't linear.
    In this particular test, which the openssl speed test, it mostly CPU bound, there is little disk activity. You can see the P4 2.8GHZ score is almost exactly 50% of a dual xeon 2.8GHZ. P4 2.8 and Xeon 2.8 is basically the same chip.

  5. #30
    Join Date
    Aug 2000
    Location
    Sheffield, South Yorks
    Posts
    3,627
    Xeon != P4, therefore you can't base your guestimated scores for uni-processor setups by dividing by 2, I don't care what you say, ask anyone worth talking to, and they will tell you, adding a 2nd processor does not increase performance by 100% because you have issues with multiple caches, locking etc.
    Karl Austin :: KDAWS.com
    The Agency Hosting Specialist :: 0800 5429 764
    Partner with us and free-up more time for income generating tasks

  6. #31
    Join Date
    Aug 2004
    Posts
    371
    Ok, let's stop arguing, we already got pretty reliable mysql scores from tomshardware and anandtech showing sempron 3100+ faster than xeon 2.8.

    Also, we find sempron 3100+ substantially fasters than p4 2.8GHZ in openssl tests.

    Can anyone post the openssl scores for a single xeon 2.8GHZ?

    The command is
    openssl speed rsa

  7. #32
    You're asking people for their results, then bashing the results they get or making excuses as to why they are slow. Then point back to anandtech or tomshardware as definitive, which defeats the purpose of you asking people for results.

    My point was Opteron scores in your other thread are all over the place. Many are slower than the Sempron 3100+ you compare to here. Then you use my results as 'optimal'. You can't pick and choose which data to put in your results. All the Intel chips have performed consistently. 27000+ for 533Mhz fsb 35000+ for 800Mhz fsb.

    Also, the way you are determining your conclusion by choosing the best case scenario is misleading because in the real world, nothing is best case scenario. There'll always be multiple processes running. You also don't include OS processing to include for running multiple processors actual 2nd processor utilization is 70-80% the rest goes to managing it.

    From your reasoning, does that mean that opteron chips take more work to optimize than intel chips? because of the large gap in performance, check mobo settings etc. (there's a reason its slow) The other intels results don't seem to have that problem of 4x detriments in performance results. Please explain?

    I don't think you can read the answer on Anandtech or Tomshardware due to the fact that they have a lab and perfect conditions.

  8. #33
    Join Date
    Aug 2004
    Posts
    371
    Ikeo, many serevrs actually use cheap workstation boards, those boards won't perform. Also, it is not clear how much load the servers are under load when the openssl tests were run. The scores I got were from servers that have no traffic.

    I think if someone has a brand name single xeon 2.8GHZ(so we know it got decent components), an openssl test result will be very helpful.

  9. #34
    Originally posted by PixyMisa
    No you don't.

    Celerons and Xeons are the same chip, just with varying levels of cache and FSB speed (and hyperthreading disabled on the Celeron).

    Semprons (the socket 754 ones) and Opterons are likewise the same, with varying levels of cache and FSB speed (and 64-bit mode disabled on the Sempron).

    Larger cache is very necessary in dual/multi-processor systems because it frees up the FSB (particularly for Xeons with their shared FSB). How much it helps for a single processor depends on the application.

    If you benchmark a Sempron and a Xeon with your application (or something similar to it) and the Sempron is faster, then the Sempron is faster. The fact that it is much cheaper is irrelevant.
    You CAN'T compare the performance by running just two tests.

    Sempron is not even Athlon64 and obviously has lower performance than that mainstream desktop processor.
    Opteron is even more powerfull than the Athlon64 processor.

    If Sempron was really faster than Xeon for most applications then AMD had no reason to release Athlon64/Athlon64FX and Opteron.

    This not to say that for some applications even Sempron may outperform Xeon - however for most general applications and for heavy DB usage the cache size matters.

    Comparing the performance of the system by just two tests is similar to comparing the performance by checking the time it takes for the system to boot...

  10. #35
    Join Date
    Aug 2004
    Posts
    371
    Originally posted by secureserver
    You CAN'T compare the performance by running just two tests.

    Sempron is not even Athlon64 and obviously has lower performance than that mainstream desktop processor.
    Opteron is even more powerfull than the Athlon64 processor.

    If Sempron was really faster than Xeon for most applications then AMD had no reason to release Athlon64/Athlon64FX and Opteron.

    This not to say that for some applications even Sempron may outperform Xeon - however for most general applications and for heavy DB usage the cache size matters.

    Comparing the performance of the system by just two tests is similar to comparing the performance by checking the time it takes for the system to boot...
    We already showed that Sempron 3100+ outpeforms Xeon 2.8GHZ in MySQL insert and select tests by a wide margin. If you are familiar with the mysql tests, they are very heavy inserts and selects involving inserting 100K rows and then select them back by various means. Xeon has bigger cache, but it has to fill a 31 stage pipe line, meaning one instruction in xeon takes 31 steps and 31 clock cycles, sempron has only 12 stages.

    Opteron has 940 pins, and it uses slower but more reliable(and expensive) registered ram.

    Athlon 64 has 939 pins, and it uses faster unbuffered ram.

    AMD is rumored to introduce dual core Opteron based on the 939 pin platform, which can use cheaper unbuffered ram. It wll the same as athlon 64, just a different name.

  11. #36
    Join Date
    Jan 2004
    Location
    North Yorkshire, UK
    Posts
    4,164
    Tell you what, instead of arguing here, why don't people who have these spec machines do some real benchmarks, and we'll find out for *ourselves*, rather than playing with other peoples.
    █ Dan Kitchen | Technical Director | Razorblue
    █ ddi: (+44) (0)1748 900 680 | e: dkitchen@razorblue.com
    █ UK Intensive Managed Hosting, Clusters and Colocation.
    █ HP Servers, Cisco/Juniper Powered BGP Network (AS15692).

  12. #37
    Join Date
    Jun 2002
    Location
    sydney.au
    Posts
    248
    Originally posted by secureserver
    You CAN'T compare the performance by running just two tests.
    Yes you can, if those tests represent your application accurately.
    Sempron is not even Athlon64
    The socket 754 Semprons are Athlon 64s with 64-bit mode disabled.

    Performance is identical to an Athlon 64 running a 32-bit OS.
    and obviously has lower performance than that mainstream desktop processor.
    Nope. Exactly the same performance in matching configurations.
    Opteron is even more powerfull than the Athlon64 processor.
    Wrong.

    Athlon 64s are Opterons. Some versions of the Athlon 64 have smaller cache. They aren't configured for registered memory, and don't have coherency enabled on the hypertransport channel, so you can't run multi-processor Athlon 64s - unless you get the dual-core version.

    It's all the same core. AMD don't have the resources to run around creating new and different cores all the time.
    If Sempron was really faster than Xeon for most applications then AMD had no reason to release Athlon64/Athlon64FX and Opteron.
    It's marketing.

    An 8-series Opteron is identical to a 1-series Opteron except that HT coherency has been turned off on the 1-series. AMD charge far more for the 8-series.
    This not to say that for some applications even Sempron may outperform Xeon - however for most general applications and for heavy DB usage the cache size matters.
    Maybe. Cache can certainly have an effect - if it's too small, your app can run significantly slower. Having extra cache beyond a certain point yields diminishing returns, but that certain point varies depending on what you're doing.
    Last edited by PixyMisa; 05-05-2005 at 08:02 PM.

  13. #38
    Join Date
    Jun 2002
    Location
    sydney.au
    Posts
    248
    Originally posted by KDAWebServices
    Xeon != P4
    Wrong.

    Xeons are P4s.

    Some Xeons have extra cache added.

    Some don't. The ones you find in reasonably-priced servers don't.

  14. #39
    Join Date
    Aug 2004
    Posts
    371
    Durons have 128KB cache I believe, but some Russian folks found a way to solder some connectors on the back of CPU and got 380+ kB of cache enabled.

    There isn't a way to enable the 3 hypertransport links within a 1xx opteron to make it 8way opteron, the packaging does not allow any tricks to be played.

    SeeSevers bought a quad 32bit 400MHZ FSB 3GHZ Xeon with 4MB of level 3 cache, those chips costs $3000 each. That kind of money can buy 2 dual core 8 way opteron 865 procs.

    Customers pay a big premium for INTEL technology, said INTEL CEO.

    You pay $90 for a sempron 3100+ and probably $200 for a xeon 2.8GHZ. but a sempron is faster.

    Goging forward, INTEL roadmap looks like a desert...INTEL won't have anything similar to Direct Connect Architecture until 2007.

  15. #40
    So what's the point of this thread? Are you asking people for real performance results or spitting out marketing material from Intel/ AMD websites and anandtech/tomshardware? No one has stated their actual results from a Sempron 3100+, yet you state your conclusion as FACT. Others have posted their results where the Opteron is SLOWER than the Sempron and you glaze over that instead of asking about their setup. You have results that show that Xeons are FASTER and SLOWER than Opterons but discount those that are faster and reference the slower Xeons.

    INTEL won't have anything similar to Direct Connect Architecture until 2007.
    And you constantly repeat the same line over and over HyperTransport. You sound like a broken record.

    Ikeo, many serevrs actually use cheap workstation boards, those boards won't perform.
    Isn't Dell just a screwdriver company that uses cheap parts? Also, aren't there very limited providers motherboard providers for the opteron platform?

    The results I posted were from a Dell SC420 and a Dell 1850. I use Dell equipment extensively and will be getting a server certification shortly, so that I can dispatch service technicians for clients.

    If you look back at your other thread, a Dual Xeon 3.2 is on par with the higher performing opterons. Unless your calling the guy a liar, when will you accept actual REAL-LIFE data not tested in a lab?

    AMD/Intel both make chips. One is faster than the other, depending on the application. ATI likes intel, Nvidia likes AMD they write code optimized for those platforms. They're in competition to push each other forward, and lowers the price for us. Brand Loyalty is one thing, what you are doing is borderline eccentric.

    I originally was planning to build a opteron server but the cost stopped me. In the end, I chose Dell because of their 3 year on-site support. Later down the line, i'll probably pick up an opteron server, but your attitude is one of extremism and narcisism.

    Now do tell, what will you be using your Uber powerful server for? Lots of power without any practical use isn't the greatest use of that 1337 AMD Opteron. I just setup a CS server and a WebHost on my equipment and I feel I'm getting the most for my money. I don't have to worry about the hardware for 3 years and if there's a problem I call and they come out and fix it. BTW The openssl test was run while the CS Server was running.

    Also, if you believe those results so much, why didn't you build a sempron server?

  16. #41
    Join Date
    Aug 2000
    Location
    Secaucus, NJ
    Posts
    44

  17. #42
    Join Date
    Aug 2003
    Location
    East Coast
    Posts
    2,082
    How many threads are you going to start about this subject?

    Buy what you want and leave the rest of us to buy what we want.

  18. #43
    Join Date
    Apr 2003
    Location
    Los Angeles, CA
    Posts
    820
    I won't dwell on the subject too much, but I can't resist and will contribute my pointless results (sarcasm mode off):

    Code:
    Pentium II, 400 MHz (801 bogomips), 256 MB RAM:
    
                      sign    verify    sign/s verify/s
    rsa  512 bits   0.0035s   0.0003s    282.0   3397.2
    rsa 1024 bits   0.0174s   0.0009s     57.5   1151.8
    rsa 2048 bits   0.1028s   0.0029s      9.7    349.1
    rsa 4096 bits   0.6880s   0.0103s      1.5     97.1
    
    Sempron 2800+, 2 GHz (3940 bogomips), 256 MB RAM:
    
                      sign    verify    sign/s verify/s
    rsa  512 bits   0.0006s   0.0001s   1586.8  18209.9
    rsa 1024 bits   0.0029s   0.0001s    344.7   6717.0
    rsa 2048 bits   0.0166s   0.0005s     60.4   2107.4
    rsa 4096 bits   0.1077s   0.0017s      9.3    600.9
    
    Athlon XP 2500+, 1.83 Ghz (3620 bogomips), 256 MB RAM:
    
                      sign    verify    sign/s verify/s
    rsa  512 bits   0.0008s   0.0001s   1241.0  16967.6
    rsa 1024 bits   0.0035s   0.0002s    283.4   5943.1
    rsa 2048 bits   0.0200s   0.0006s     49.9   1765.8
    rsa 4096 bits   0.1326s   0.0021s      7.5    477.5
    
    Dual Opteron 242, 1.6 GHz (3178 bogomips each), 1 GB RAM:
    
                      sign    verify    sign/s verify/s
    rsa  512 bits   0.0004s   0.0000s   2688.7  30911.1
    rsa 1024 bits   0.0014s   0.0001s    723.2  12002.6
    rsa 2048 bits   0.0080s   0.0003s    124.3   3982.1
    rsa 4096 bits   0.0529s   0.0009s     18.9   1175.5
    
    Dual Xeon, 2.8 GHz (5570 bogomips each), 2 GB RAM:
                      sign    verify    sign/s verify/s
    rsa  512 bits   0.0008s   0.0001s   1322.3  13993.6
    rsa 1024 bits   0.0039s   0.0002s    253.7   4669.4
    rsa 2048 bits   0.0245s   0.0007s     40.8   1378.5
    rsa 4096 bits   0.1673s   0.0026s      6.0    384.3
    No, it's not the same hardware by no means... hence pointless, but fun nevertheless.
    Last edited by luki; 05-06-2005 at 03:02 AM.
    Pings <1 ms, Unlimited Transfer, Lowest Price: http://localhost/

  19. #44
    Join Date
    Apr 2003
    Location
    Los Angeles, CA
    Posts
    820
    Sorry, one more...

    Code:
    Intel Celeron, 2.4 GHz (4836 bogomips), 1 GB RAM:
    
                      sign    verify    sign/s verify/s
    rsa  512 bits   0.0009s   0.0001s   1171.7  12459.5
    rsa 1024 bits   0.0047s   0.0003s    214.8   3950.3
    rsa 2048 bits   0.0294s   0.0009s     34.0   1142.9
    rsa 4096 bits   0.2020s   0.0031s      5.0    317.6
    And finally one thing to note: on a completely idle dual-CPU machine, the openssl test only uses one CPU, the other CPU remains 99.9% idle. This is so on the Xeon and Opteron systems above. All systems were either devel system or currently idle cluster nodes.

    I'll shut up now.
    Pings <1 ms, Unlimited Transfer, Lowest Price: http://localhost/

  20. #45
    Join Date
    Apr 2003
    Posts
    279
    Originally posted by keepr
    How many threads are you going to start about this subject?

    Buy what you want and leave the rest of us to buy what we want.
    Until everyone joins the wm2001 Benchmark Warrior legion. I agree with Ikeo, you're only showing the slower results for the Xeon, instead of the results where the Xeon beats the Sempron or Opteron.

    I told you this in your last insane thread about this, different CPUs and hardware configurations serve and execute different applications and so forth DIFFERENTLY. Look past your propoganda of constantly being 'religious' towards one, and reflecting on Intel's lack in a 64bit market that is not mainstream OUTSIDE of webhosting and a few industries that require 64bit technology anyways, and Intel's major focus is in the industries of Cinema and 3D rendering for companies like Pixar and so forth.

    So really, by the time Intel catches up to the '64bit' direct connect architecture, which you believe to be 2007, which isn't confirmed, and could be earlier, the rest of the market will have started shifting that way also.

    Facts are not only determined by benchmarks.
    Shawn R. Lockheart

  21. #46
    Join Date
    Aug 2004
    Posts
    371
    There were people who found the idea of comparing Sempron 3100+ to Xeon 2.8GHZ unthinkable, one is AMD's cheapish desktop dog, one is INTEL's lower end server chip. But once you open your mind and look at the real performance instead of brands, the comparison makes sense. The sempron 3100+ is a scaled down opteron...

    So far, not a single test showed up with xeon 2.8GHZ outperforming sempron 3100+.

    If you read INQ's recent report on INTEL's roadmap, the next generation of INTEL chips will be based on Pentium III and clocked at around 2.5GHZ.

  22. #47
    Join Date
    Apr 2005
    Location
    Atlanta, Georgia
    Posts
    521
    If anyone wants a sempron 3100+ to benchmark on we'll have another shipment comming in on monday if you want me to test anything on it PM me what test you would like ran with what parameters and I'll post up the benchmarks from those systems.

    Sempron3100+ (Socket 754) 1024 PC3200

    We've actually been super impressed with what we've seen out of these things.

  23. #48
    Join Date
    Aug 2004
    Posts
    371
    The openssl test is the easiest one, it's built in. Just type openssl speed

    To do the mysql test, you need to install the mysql benchmark suite.

    But, Anandtech has done these two tests on Sempron 3100+ already. what we are missing is these tests on a single xeon.

  24. #49
    Join Date
    Aug 2003
    Location
    East Coast
    Posts
    2,082
    TulipSystems

    Why dont you go ahead and post your scores. Lets see if you get the same results anandtech got.

  25. #50
    Join Date
    Apr 2005
    Location
    Atlanta, Georgia
    Posts
    521
    ah, ok I didn't see the post about the 3100+ already.

    Still, when the equipment comes in monday I'll do a quick test on it and post the scores anyway just as a verification.

Page 2 of 3 FirstFirst 123 LastLast

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •