Page 1 of 3 123 LastLast
Results 1 to 25 of 60
  1. #1
    Join Date
    Aug 2004
    Posts
    371

    Sempron 3100+ vs Xeon 2.8GHZ

    I previously posted a message speculating that a Sempron 3100 Socket 754 platform should outpeform Xeon 2.8GHZ due to tis architectural superiority. The socket 754 semprons has

    1) Direct Connect Architecture with FSB running at CPU frequency 1.6GHZ
    2) Embedded memory controller that reduce memory latency
    3) Data Execution prevention with NX bit

    The sempron share the traits of Opteron, but has less cache. However, it uses faster unbuffered ram(though it's single channel).

    I don't have a sempron 3100+ machine. Could anyone with these cuties run some tests and post some scores?

    Two tests

    1) OpenSSL
    you run it by typing

    openssl speed rsa

    at the command prompt

    2) mysql becnchmark
    You may have to download it from
    http://dev.mysql.com/doc/mysql/en/mysql-benchmarks.html

  2. #2
    Join Date
    Jan 2004
    Location
    North Yorkshire, UK
    Posts
    4,164
    This under an x32 or x64 environment?

    Dan
    █ 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).

  3. #3
    Join Date
    Jun 2000
    Location
    Washington, USA
    Posts
    5,990
    The Xeon is likely to be faster, as the Sempron is rated in comparison to a Celeron. However, the performance difference isn't likely to be very much.

  4. #4
    Join Date
    Nov 2004
    Location
    Switzerland
    Posts
    1,307
    The Sempron sucks. I am using a sempron 2800+ to post this message. I prefer a PIII instead!
    Enterprise Consultant
    CCNP Enterprise - CCNP Security
    .:. Travels From West to East .:.

  5. #5
    Join Date
    Jun 2003
    Location
    Indiana, US
    Posts
    1,355
    Originally posted by edelweisshosting
    The Sempron sucks. I am using a sempron 2800+ to post this message. I prefer a PIII instead!
    The 2800+ and the 3100+ are completely different sockets. How your 2800+ performs is irrelevant to this thread.

    This under an x32 or x64 environment?

    Dan
    I wasn't aware either of the processors were 64-bit capable?

  6. #6
    I'm sure many people are aware of this but just in case. There will be certain applications where the Xeon outperforms the Sempron or vice versa. Generally, when choosing a large server I vary between AMD and Intel depending on my application needs. Benchmarks involving openssl and mysql are going to be biased as that isn't a complete overall comparision of either processor architecture.
    Last edited by Snipz; 05-04-2005 at 07:21 PM.

  7. #7
    Join Date
    Jan 2004
    Location
    North Yorkshire, UK
    Posts
    4,164
    Originally posted by HstCreations
    The 2800+ and the 3100+ are completely different sockets. How your 2800+ performs is irrelevant to this thread.



    I wasn't aware either of the processors were 64-bit capable?
    Doh! Half asleep as usual, was thinking he was on about an Athlon 64.

    A Sempron is actually comparable to a P4, especially in the lower numbers (not HT etc), but I wouldn't go head to head with a Xeon. In this case much as I love my AMD's, I think the Xeon will probably win
    █ 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).

  8. #8
    Join Date
    Nov 2004
    Location
    Switzerland
    Posts
    1,307
    The 2800+ and the 3100+ are completely different sockets. How your 2800+ performs is irrelevant to this thread.
    If this is true, there is a big communication error from AMD. If these 2 processors are differents may be they deserve different names.

    In my office, I have a P4 2.8 Ghz, a sempron 2800+ and PIII 500 Mhz, this last with Windows 98. I order these machines [all with 512 MB RAM] from the faster to the slowest:

    The P4 2.8 Ghz
    The PIII 500 Mhz
    The Sempron 2800+
    Enterprise Consultant
    CCNP Enterprise - CCNP Security
    .:. Travels From West to East .:.

  9. #9
    Join Date
    Jun 2000
    Location
    Washington, USA
    Posts
    5,990
    Something must be messed up with your Sempron box then. I've got a 2600+ and it smokes a P4 2.0Ghz with ease.

    As for the Semprons, they make both socket 462, and 754 models. Those models do overlap to some extent at various rated speeds.

  10. #10
    Join Date
    Jun 2002
    Location
    sydney.au
    Posts
    248
    The socket 754 Semprons are based on the Opteron architecture, but with reduced cached and with 64-bit mode disabled. Socket A semprons are Athlons; I don't remember which exact version of the core.

    The Sempron 3100+ is a 1.8GHz chip. While I'd definitely take a 1.8GHz Opteron over a 2.8GHz Xeon; the Sempron with it's much smaller cache might be a wash.

    Anyway, Anandtech reports a Sempron 3100+ doing 364.9 1024-bit RSA signs per second, so if you can find someone with a Xeon 2.8...

    edelweisshosting: If your Sempron 2800+ really is running slower than a PIII 500, it's not the chip's fault, since even an Athlon 500 will beat a PIII 500. There's something screwed up with that machine.

  11. #11
    Join Date
    Jul 2003
    Posts
    332
    The Sempron would get wasted by the Xeon.

    You cannot compare AMD's budget chip to Intel's top of the line server chip. The Xeon has like what... 4 times the amount of cache? That would kill in database environments.

  12. #12
    Join Date
    Aug 2004
    Posts
    371
    Ok, I found some benchmarks here
    http://www.anandtech.com/printarticle.aspx?i=2170

    Sempron 3100+ does mysql select in 378 seconds
    Sempron 3100+ does 364 1024 bit RSA signs per second


    Now, how about Xeon 2.8 GHZ?

    From http://www.webhostingtalk.com/showth...4&pagenumber=4

    Dual Xeon 2.8GHZ produced 480 1024 bit RSA signs/second.
    I can deduce that one 2.8GHZ Xeon will do about 240 signs/second.

    from

    http://www6.tomshardware.com/cpu/200...#database_test

    Xeon 2.8GHZ finishes the MySQL select test in 595 seconds.

    Conclusion:

    Sempron is faster than xeon 2.8GHZ in openssl RSA tests

    Sempron 3100+ (socket 754) is decisively faster than Xeon 2.8GHZ in MySQL tests. Sempron took 378 seconds to finish, Xeon 2.8 took 595 seconds.
    (Note although different mysql version is used, however, the tests themselves are the same)

    To compete with Sempron 3100+ (32 bit, 1.8GHZ, socket 754, 1.8GHZ FSB), you will need at least a XEON 3.6GHZ EM64T running 64 bit Linux.

    Guys, numbers don't lie!

  13. #13
    Join Date
    Aug 2004
    Posts
    371
    I can explain why Sempron 3100+ trashes Xeon 2.8GHZ in all meaningful server benchmarks,
    1) Sempron has Direct Connect Architecture
    2) Sempron has hypertransport running at 800MHZ
    3) Sempron FSB 1.8GHZ
    4) Sempron has only 12 stage pipeline

    Xeon is slow
    1) shared bus of 533MHZ to 800MHZ
    2) 31 stages of pipe line, one branch misprediction, you waste 31 clock cycles...

  14. #14
    Join Date
    Jun 2002
    Location
    sydney.au
    Posts
    248
    Originally posted by wm2100
    Xeon is slow
    1) shared bus of 533MHZ to 800MHZ
    2) 31 stages of pipe line, one branch misprediction, you waste 31 clock cycles...
    Only the Prescott-core Xeons. The older Northwood core Xeons had a 20-stage pipeline.

  15. #15
    You could've just looked at your other post. Your conclusions are incorrect. Rethink you experiment and try again and try to base your conclusions on REAL data. People may believe what you are stating as fact.


    http://www.webhostingtalk.com/showth...5&pagenumber=1

    Dual Xeon

    sign verify sign/s verify/s
    rsa 512 bits 0.0003s 0.0000s 2981.2 35243.0
    rsa 1024 bits 0.0017s 0.0001s 581.9 11931.4
    rsa 2048 bits 0.0100s 0.0003s 99.5 3577.8
    rsa 4096 bits 0.0661s 0.0010s 15.1 1008.8

    Standard 2.8 Ghz w. HT

    Multi-Threaded Results


    sign verify sign/s verify/s
    rsa 512 bits 0.0007s 0.0001s 1506.0 17543.9
    rsa 1024 bits 0.0034s 0.0002s 291.5 5970.1
    rsa 2048 bits 0.0201s 0.0006s 49.7 1787.3
    rsa 4096 bits 0.1323s 0.0020s 7.6 504.6

  16. #16
    Join Date
    Aug 2000
    Location
    Sheffield, South Yorks
    Posts
    3,627
    You can't go round comparing machines from different benchmarks, it's just not accurate at all, the machines need to be as identical as possible. i.e. Same speed RAM, same disk subsystem, same graphics card, same addon cards etc. etc. The only thing that should change, is the motherboard and the CPU, nothing else, it should all be absolutely identical. As for deducing that a Single Xeon will get half the signs as a Dual Xeon, that is just plain wrong, everyone and his dog knows you don't get a 100% increase in performance in moving from Uni to Dual Processor, if you're lucky it'll be around 60-70% but it all depends on the application in question.

    Whilst you may think you're doing AMD some good by singing their praises, by using a poor methodology, you're giving everyone the impression you're just an AMD fan-boy without much clue (I'm not saying you are, it's just the impression you give with poor methodology).
    Karl Austin :: KDAWS.com
    The Agency Hosting Specialist :: 0800 5429 764
    Partner with us and free-up more time for income generating tasks

  17. #17
    Despite the fact that I'm also consider myself as an AMD fan I was quite surprised by the comparison and the way it was performed.

    You need to compare Xeon vs. Opteron, as the Sempron is a value processor built to compete against Celeron.

  18. #18
    Join Date
    Jun 2002
    Location
    sydney.au
    Posts
    248
    Originally posted by secureserver
    Despite the fact that I'm also consider myself as an AMD fan I was quite surprised by the comparison and the way it was performed.

    You need to compare Xeon vs. Opteron, as the Sempron is a value processor built to compete against Celeron.
    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.

  19. #19
    Join Date
    Aug 2004
    Posts
    371
    Numbers don't lie!

    The same MySQL selects, Sempron 3100 finish it in 300+ seconds, Xeon 2.8 does it in 500+ seconds.

    Sempron does 300+ rsa signs in one second, Xeon does 200+.

    Do we blindly believe Xeon is better simply because it's priced higher and labeled as a server chip by INTEL?

    Those Xeons needs bigger cache, why?

    Because its architecture is oudated, its FSB and memory interface so slow, and its pipeline so long, it has to have big cache, otherwise the core will be stalled.

    Xeon has 31 stage pipe, meaning one instruction takes 31 clock cycles to finish.

    Sempron and Opteron has 12 stage pipe, one instruction takes 12 clock cycles to finish.

    Pentium-M has 10 stage pipe, ans we see dothan 2GHZ faster than 3.6GHZ P4.

    Itanium has 8 stage pipe, so 1.6GHZ Itanium is faster than 3.8 GHZ xeon.

    But Itanium is slower than Opteron, because Opteron has direct connect architecture.

    INTEL can no longer play the frequency trick, because it has hit a ceiling....

  20. #20
    Join Date
    Aug 2000
    Location
    Sheffield, South Yorks
    Posts
    3,627
    Question for you, what memory did the Xeon have? The Sempron? Disk? If they aren't the same, then it isn't a valid comparisson.
    Karl Austin :: KDAWS.com
    The Agency Hosting Specialist :: 0800 5429 764
    Partner with us and free-up more time for income generating tasks

  21. #21
    Join Date
    Aug 2004
    Posts
    371
    In Anandtech's Sempron benchmark, Sempron 3100+ has 1GB of ram (2 x 512MB).

    In Toms test, Xeon 2.8 also has 1GB of ram, 4x256MB

  22. #22
    Join Date
    Aug 2000
    Location
    Sheffield, South Yorks
    Posts
    3,627
    Type of memory? Disk?

    Also the SQL Select is only one specific type of benchmark, you can't conclusively say one CPU is better than the other in general terms without having a comprehensive suite of tests on machines with as similar specification as possible. The only time you can ignore the rest of the spec is with the SpecInt tests etc. which measure $/transaction.
    Karl Austin :: KDAWS.com
    The Agency Hosting Specialist :: 0800 5429 764
    Partner with us and free-up more time for income generating tasks

  23. #23
    Join Date
    Jun 2000
    Location
    Washington, USA
    Posts
    5,990
    Karl is right. The benchmarks need to be have comparable hardware.

    Also, the Xeon and Sempron will go back/forth in benchmarks. AMD vs. Intel has always done that, each CPU has it's strong points.

  24. #24
    Join Date
    Jun 2002
    Location
    sydney.au
    Posts
    248
    Originally posted by KDAWebServices
    Type of memory? Disk?

    Also the SQL Select is only one specific type of benchmark, you can't conclusively say one CPU is better than the other in general terms without having a comprehensive suite of tests on machines with as similar specification as possible. The only time you can ignore the rest of the spec is with the SpecInt tests etc. which measure $/transaction.
    You mean TPC - and those always specify the hardware in excruciating detail.

  25. #25
    Join Date
    Aug 2004
    Posts
    371
    Originally posted by ikeo
    You could've just looked at your other post. Your conclusions are incorrect. Rethink you experiment and try again and try to base your conclusions on REAL data. People may believe what you are stating as fact.


    http://www.webhostingtalk.com/showth...5&pagenumber=1

    Dual Xeon

    sign verify sign/s verify/s
    rsa 512 bits 0.0003s 0.0000s 2981.2 35243.0
    rsa 1024 bits 0.0017s 0.0001s 581.9 11931.4
    rsa 2048 bits 0.0100s 0.0003s 99.5 3577.8
    rsa 4096 bits 0.0661s 0.0010s 15.1 1008.8

    Standard 2.8 Ghz w. HT

    Multi-Threaded Results


    sign verify sign/s verify/s
    rsa 512 bits 0.0007s 0.0001s 1506.0 17543.9
    rsa 1024 bits 0.0034s 0.0002s 291.5 5970.1
    rsa 2048 bits 0.0201s 0.0006s 49.7 1787.3
    rsa 4096 bits 0.1323s 0.0020s 7.6 504.6
    Ikeo, the anandtech benchmark shows sempron 3100+ does 364 RSA 1024bit signs per second. Your data shows that Dual (Two) Xeon 2.8GHZ does 581 signs per second. In this openssl test, two CPUs are twice as fast. Therefore, one Xeon should be in the ballpark of 581/2 = 291 signs/second. xeon is slower.

Page 1 of 3 123 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
  •