View Poll Results: Reliable?

Voters
23. You may not vote on this poll
  • Dual Opteron Quad Core 2356

    6 26.09%
  • Dual Xeon E5405

    7 30.43%
  • Intel Core i7 920

    8 34.78%
  • Intel Core i7 860

    4 17.39%
Multiple Choice Poll.
Page 1 of 2 12 LastLast
Results 1 to 25 of 28

Hybrid View

  1. #1

    Dual Opteron Quad Core 2356 vs Dual Xeon E5405 vs Intel Core i7 920

    poll:
    what You prefer? Reliable and high powerfull

    Dual Opteron Quad Core 2356 2.3hz
    Dual Xeon E5405 2.0hz
    Intel Core i7 920 2.66ghz
    Intel Core i7 860 2.8ghz

  2. #2
    Join Date
    Oct 2007
    Location
    United States
    Posts
    1,182
    Core i7's have been amazing when it comes to virtualization.
    www.opticip.com - Optic IP LLC

  3. #3
    Join Date
    Jul 2007
    Location
    Tech Belt
    Posts
    8,160
    Core i7 920 I'd have to say would be one of my picks, although the Dual Quad Opteron is pretty tempting. I'd say it all comes down to what you're gonna do, if its Virtualization the Core i7 might do a better job than the Dual opterons.
    Nothing here right now.

  4. #4
    Join Date
    Nov 2007
    Location
    India
    Posts
    843
    i will go with Dual Xeon E5405 it will be good
    HostNotch Hosting Services 99.9% uptime Shared Hosting, Reseller Hosting
    yajur | Sales Team
    CPanel Hosting • R1 Soft • Offsite-Backup • Great Uptime
    http://hostnotch.com sales @ hostnotch.com

  5. #5
    Join Date
    Jan 2003
    Location
    Chicago, IL
    Posts
    6,957
    For most server uses the i7 options would be the least powerful, as you'd only have four processing cores, while you have 8 with the dual CPU options. Out of the dual CPU options the AMD should give you better performance as they're clock-for-clock roughly the same as the Intel's, but that one has a higher clock speed.
    Karl Zimmerman - Founder & CEO of Steadfast
    VMware Virtual Data Center Platform

    karl @ steadfast.net - Sales/Support: 312-602-2689
    Cloud Hosting, Managed Dedicated Servers, Chicago Colocation, and New Jersey Colocation

  6. #6
    Quote Originally Posted by KarlZimmer View Post
    For most server uses the i7 options would be the least powerful, as you'd only have four processing cores, while you have 8 with the dual CPU options. Out of the dual CPU options the AMD should give you better performance as they're clock-for-clock roughly the same as the Intel's, but that one has a higher clock speed.
    yes but i7 show 8 CPU

  7. #7
    Join Date
    Jan 2003
    Location
    Chicago, IL
    Posts
    6,957
    Quote Originally Posted by Up2vps View Post
    yes but i7 show 8 CPU
    Yes, with Hyperthreading you see 8, it physically has 4 cores. The physical cores is what will get you actual processing power.
    Karl Zimmerman - Founder & CEO of Steadfast
    VMware Virtual Data Center Platform

    karl @ steadfast.net - Sales/Support: 312-602-2689
    Cloud Hosting, Managed Dedicated Servers, Chicago Colocation, and New Jersey Colocation

  8. #8
    Join Date
    Oct 2002
    Location
    Vancouver, B.C.
    Posts
    2,699
    Quote Originally Posted by KarlZimmer View Post
    Yes, with Hyperthreading you see 8, it physically has 4 cores. The physical cores is what will get you actual processing power.
    At the CPU level, hyperthreading allows two threads running concurrently on the same core at different stages in the pipeline to be presented as two separate processor units to the OS. However, without hyperthreading, you are able to run multiple instructions on the same core at different stages of the pipeline anyway for any x86 cpu since 486. As not all instructions can be pipelined in this manner, hyperthreading will allow for better utilization of each core but nowhere near double.

    With both hyperthreading and architectural improvements, one nehalem core averages around 25% more performance clock per clock in comparison to Core 2 in benchmarks. Real-world performance would depend on your application, but just using benchmarks to formulate a rough estimate using the clockspeeds of the listed CPU's, you would get more processing power out of the dual CPU configurations.

    i.e. some very rough calculations
    Dual Opteron Quad Core 2356 2.3hz - 8 x 2.3GHz = 18.4GHz
    Dual Xeon E5405 2.0hz - 8 x 2.0GHz = 16.0GHz
    Intel Core i7 920 2.66ghz - 1.25 x 4 x 2.66GHz = 13.3GHz
    Intel Core i7 860 2.8ghz - 1.25 x 4 x 2.8GHz = 14.0GHz

    So as Karl said, the Dual Opteron Quad Core 2356 configuration would be expected to have the best performance for most applications.

    Now, if your application sees 31% or more performance increase with Nehalem, you would see better performance with the new single CPU configurations. (8/5 / 2.8/2.3 = 1.31)
    Last edited by hhw; 09-18-2009 at 06:32 PM.
    ASTUTE INTERNET: Advanced, customized, and scalable solutions with AS54527 Premium Performance and Canadian Optimized Network (Level3, Shaw, CogecoPeer1, GTT/Tinet),
    AS63213 Cost Effective High Performance Network (Cogent, HE, GTT/Tinet)
    Dedicated Hosting, Colo, Bandwidth, and Fiber out of Vancouver, Seattle, LA, Toronto, NYC, and Miami

  9. #9
    Join Date
    May 2009
    Posts
    217
    Quote Originally Posted by hhw View Post
    At the CPU level, hyperthreading allows two threads running concurrently on the same core at different stages in the pipeline to be presented as two separate processor units to the OS. However, without hyperthreading, you are able to run multiple instructions on the same core at different stages of the pipeline anyway for any x86 cpu since 486. As not all instructions can be pipelined in this manner, hyperthreading will allow for better utilization of each core but nowhere near double.

    With both hyperthreading and architectural improvements, one nehalem core averages around 25% more performance clock per clock in comparison to Core 2 in benchmarks. Real-world performance would depend on your application, but just using benchmarks to formulate a rough estimate using the clockspeeds of the listed CPU's, you would get more processing power out of the dual CPU configurations.

    i.e. some very rough calculations
    Dual Opteron Quad Core 2356 2.3hz - 8 x 2.3GHz = 18.4GHz
    Dual Xeon E5405 2.0hz - 8 x 2.0GHz = 16.0GHz
    Intel Core i7 920 2.66ghz - 1.25 x 4 x 2.66GHz = 13.3GHz
    Intel Core i7 860 2.8ghz - 1.25 x 4 x 2.8GHz = 14.0GHz

    So as Karl said, the Dual Opteron Quad Core 2356 configuration would be expected to have the best performance.
    Actually the i7 920 would beat the dual Quad 2356 barcelonas by roughly 20% or so. The only reason you should get the Dual quad Opterons is if you need more than 12GB of ram that the i7 920 will give you.

  10. #10
    i already Have Dual Quad Opteron with 32gb Ram I plan to get new another server and I want the option,

  11. #11
    Join Date
    May 2009
    Posts
    217
    Quote Originally Posted by Up2vps View Post
    i already Have Dual Quad Opteron with 32gb Ram I plan to get new another server and I want the option,
    i7 920 performance:

    http://www.spec.org/cpu2006/results/...81024-05720.ps

    Dual 2356 Opteron performance

    http://www.spec.org/cpu2006/results/...0623-04654.pdf

    So i7 gets 109/102 SpecInt vs Opteron 102/87. 7-15% difference.

    If you need 32GB ram, get Dual E5504s which gives you up to 72GB ram. Benchs around 137 SpecInt.

  12. #12
    Join Date
    Oct 2002
    Location
    Vancouver, B.C.
    Posts
    2,699
    Quote Originally Posted by tshen83 View Post
    i7 920 performance:

    http://www.spec.org/cpu2006/results/...81024-05720.ps

    Dual 2356 Opteron performance

    http://www.spec.org/cpu2006/results/...0623-04654.pdf

    So i7 gets 109/102 SpecInt vs Opteron 102/87. 7-15% difference.

    If you need 32GB ram, get Dual E5504s which gives you up to 72GB ram. Benchs around 137 SpecInt.
    SPECint's CINT2006 and benchmarks are more oriented towards scientific computing, which is a much different application than your typical hosting environment. Unfortunately, there aren't any SPECweb results for these CPU's yet.

    That's not to say the benchmarks I looked at to derive my very arbitrary 25% were the best reference either (Anandtech if anybody cares), but are somewhat more balanced workloads than raw integer and floating point performance.
    ASTUTE INTERNET: Advanced, customized, and scalable solutions with AS54527 Premium Performance and Canadian Optimized Network (Level3, Shaw, CogecoPeer1, GTT/Tinet),
    AS63213 Cost Effective High Performance Network (Cogent, HE, GTT/Tinet)
    Dedicated Hosting, Colo, Bandwidth, and Fiber out of Vancouver, Seattle, LA, Toronto, NYC, and Miami

  13. #13
    Join Date
    May 2009
    Posts
    217
    SpecWeb is heavily IO dependent, and favors max memory configs. So dual Socket systems with obscene 8GB DIMM ram win all the time.

    SpecPower2008 is the newest benchmark people look at, which is aggregate performance per watt at different load levels. So I base my decisions on SpecPower2008 ratings per dollar now.

  14. #14
    Join Date
    Apr 2003
    Location
    San Jose, CA.
    Posts
    1,624
    As a server builder... I'm inclined to say the Dual Opteron boxes...
    However, they real ly would be much better as Shanghai not Barcelona cores.
    Daved @ Lightwave Networking, LLC.
    AS1426 https:/www.lightwave.net
    Primary Bandwidth: EGIHosting (NLayer, NTT, HE, Cogent)
    Xen PV VPS Hosting

  15. #15
    I used both i7 920(HT) and dual e5502(**2.2G), I7 920 with HT tech is obvious faster. about Dual Opteron Quad Core 2356 2.3hz i haven't used amd, however, i don't think it has better performance than top2

  16. #16
    Join Date
    Jan 2003
    Location
    Chicago, IL
    Posts
    6,957
    Quote Originally Posted by tshen83 View Post
    Actually the i7 920 would beat the dual Quad 2356 barcelonas by roughly 20% or so. The only reason you should get the Dual quad Opterons is if you need more than 12GB of ram that the i7 920 will give you.
    Show me ANY benchmark in a web hosting environment where the i7 is 20% faster. Show anything that can truly take advantage of all 8 cores and you'll see the AMD is clearly faster.
    Last edited by KarlZimmer; 09-21-2009 at 01:20 PM.
    Karl Zimmerman - Founder & CEO of Steadfast
    VMware Virtual Data Center Platform

    karl @ steadfast.net - Sales/Support: 312-602-2689
    Cloud Hosting, Managed Dedicated Servers, Chicago Colocation, and New Jersey Colocation

  17. #17
    Join Date
    May 2009
    Posts
    217
    Quote Originally Posted by KarlZimmer View Post
    Show me ANY benchmark in a web hosting environment where the i7 is 20% faster. Show anything that can truly take advantage of all 8 cores and you'll see the AMD is clearly faster.
    I already showed you that with Spec CPU benchmark, the i7 920 beats the Dual Quad Opteron 2356s. I can pretty much throw the same argument back at you:

    Show me ANY benchmark that the dual Barcelona wins over the i7 920. Please don't show me high memory configs, because that would be unfair. If you want to bench 32GB-64GB configs, then the Nehalem-EP kicks the **** out of anything AMD has, including the Istanbuls, at performance, performance per watt, per watt per dollar, however you want to skin the cat.

    A lot of things take advantage of all 8 cores. For software that doesn't, you can always virtualize 8-16 VMs and run parallel instances of the same software and load balance. Even with software that scales to 8 cores and beyond, i7 920 wins because of the hyperthreading. OS sees 8 faster threads out of the i7 vs 8 Barcelona cores.

  18. #18
    Join Date
    Jan 2003
    Location
    Chicago, IL
    Posts
    6,957
    Quote Originally Posted by tshen83 View Post
    I already showed you that with Spec CPU benchmark, the i7 920 beats the Dual Quad Opteron 2356s. I can pretty much throw the same argument back at you:

    Show me ANY benchmark that the dual Barcelona wins over the i7 920. Please don't show me high memory configs, because that would be unfair. If you want to bench 32GB-64GB configs, then the Nehalem-EP kicks the **** out of anything AMD has, including the Istanbuls, at performance, performance per watt, per watt per dollar, however you want to skin the cat.

    A lot of things take advantage of all 8 cores. For software that doesn't, you can always virtualize 8-16 VMs and run parallel instances of the same software and load balance. Even with software that scales to 8 cores and beyond, i7 920 wins because of the hyperthreading. OS sees 8 faster threads out of the i7 vs 8 Barcelona cores.
    1) That test did not show any correlation with them running a web hosting environment.

    2) Neither of those showed the 20% improvement you claimed.

    I can't seem to find anything directly comparing the i7 920 to a dual Opteron system, but I did find some other interesting benchmarks comparing them against Phenoms, which are roughly equivalent to the Opteron 2300's.

    http://ixbtlabs.com/articles3/cpu/am...x4-940-p2.html - Phenom X4 wins the web server benchmark, and yes, the clock speed is higher, but has half the cores of the setup we're comparing here.

    http://www.phoronix.com/scan.php?pag...ynnfield&num=9 - The single Phenom X3 beats the i7 in PostgreSQL and Apache benchmarks.

    I'm not cherry picking either, those were the first two web server related benchmarks I found that were at all relevant.
    Karl Zimmerman - Founder & CEO of Steadfast
    VMware Virtual Data Center Platform

    karl @ steadfast.net - Sales/Support: 312-602-2689
    Cloud Hosting, Managed Dedicated Servers, Chicago Colocation, and New Jersey Colocation

  19. #19
    Join Date
    Oct 2007
    Location
    United States
    Posts
    1,182
    I have tested a Dual Xeon 5410 (8 Cores) and a Core i7 920....unfortunately we have been building nothing but Core i7 920 for our VM series. The CPU loads are a lot smaller on the Core i7 as oppose to the high loads the Dual Xeon 5410's were giving me.
    www.opticip.com - Optic IP LLC

  20. #20
    Join Date
    Feb 2004
    Posts
    768
    Have you considered the new 6core Istanbul chips from AMD? You could also even look at the Shanghai 4 cores.
    linveo.com | Shared Hosting | AMD and Intel KVM VPS | Dedicated Servers
    Direct Admin | VirtFusion | TenantOS | 24/7/365 Support

  21. #21
    Join Date
    Oct 2007
    Location
    United States
    Posts
    1,182
    Dual Xeon 5410 (8 Cores)
    32GB Ram
    4x1.5TB Raid 10

    (NOTHING ON THE SERVER, NO PROCESSES RUNNING OR CLIENTS)

    WHT Benchmark:
    INDEX VALUES
    TEST BASELINE RESULT INDEX

    Dhrystone 2 using register variables 376783.7 20122163.7 534.1
    Double-Precision Whetstone 83.1 1672.5 201.3
    Execl Throughput 188.3 6454.8 342.8
    File Copy 1024 bufsize 2000 maxblocks 2672.0 129034.0 482.9
    File Copy 256 bufsize 500 maxblocks 1077.0 34354.0 319.0
    File Read 4096 bufsize 8000 maxblocks 15382.0 1387319.0 901.9
    Pipe-based Context Switching 15448.6 312665.8 202.4
    Pipe Throughput 111814.6 3486970.3 311.9
    Process Creation 569.3 17171.9 301.6
    Shell Scripts (8 concurrent) 44.8 1150.8 256.9
    System Call Overhead 114433.5 2943254.6 257.2
    =========
    FINAL SCORE 337.3

    (Server is completely empty, still a horrible benchmark with NOTHING on it)


    2nd Server:
    Core i7 920 (8 virtual cpus)
    12GB Ram (Triple Chan)
    2x1TB Raid 1

    (SERVER FULLY LOADED WITH VPS's)

    INDEX VALUES
    TEST BASELINE RESULT INDEX

    Dhrystone 2 using register variables 376783.7 24686392.7 655.2
    Double-Precision Whetstone 83.1 1693.4 203.8
    Execl Throughput 188.3 14975.7 795.3
    File Copy 1024 bufsize 2000 maxblocks 2672.0 489130.0 1830.6
    File Copy 256 bufsize 500 maxblocks 1077.0 137528.0 1277.0
    File Read 4096 bufsize 8000 maxblocks 15382.0 3024574.0 1966.3
    Pipe-based Context Switching 15448.6 423001.5 273.8
    Pipe Throughput 111814.6 3948717.2 353.1
    Process Creation 569.3 37823.9 664.4
    Shell Scripts (8 concurrent) 44.8 4037.0 901.1
    System Call Overhead 114433.5 3673292.6 321.0
    =========
    FINAL SCORE 652.1




    If the Core i7 had 0 clients on it, I think it would be close to the 1000 benchmark. The dual xeon 5410 is completely empty and the benchmark score is absolutely horrid!
    www.opticip.com - Optic IP LLC

  22. #22
    Join Date
    Jan 2003
    Location
    Chicago, IL
    Posts
    6,957
    There is something wrong with that 5410 system... There is no way the RAID 1 array should have 2-4 times the performance of the RAID 10 configuration... That seems to be where you're losing it, as the top couple CPU based benchmarks seem to be about the same, which is about what I'd expect in that comparison. That is definitely an IO issue, not a CPU issue there.
    Karl Zimmerman - Founder & CEO of Steadfast
    VMware Virtual Data Center Platform

    karl @ steadfast.net - Sales/Support: 312-602-2689
    Cloud Hosting, Managed Dedicated Servers, Chicago Colocation, and New Jersey Colocation

  23. #23
    Join Date
    May 2009
    Posts
    217
    Karl:

    You showed a benchmark where the Phenom II 965 edges out the i7 920 in the web benchmark using PHP+MySQL, and you are trying to correlate that Barcelona has twice the cores so, must be faster. It is false.

    1. PHP is known to not scale well over multicores. MySQL also has vertical scaling issues unless you use Percona builds or beta 5.4. I am sure that review site used the default 5.0 MySQL which is only scalable to 4 threads, which means if you use a CPU with 8 threads, it is actually faster to disable HyperThreading for MySQL(in 5.0 and 5.1 standard builds in Linux). It is the same reason why facebook CTO said that he didn't find the newest Nehalem-EP server give him more performance.

    2. Phenom II 965 can't be correlated to Barcelonas because it uses DDR3 memory. Also it is high clocked 3.2Ghz? Shanghai cores that MySQL likes.

    As to the Phenom x3 beating the i7, something has to be wrong or IO constrained for Phenom x3 to beat 8 threaded i7 920.

    Review sites can swing performance benchmark results +/- 20% for a processor depending on who is "paying the bill". If you are to pump the AMD CPUs, you use benchmarks that aren't well threaded so Nehalem advantage won't be felt, ie games so that it is GPU bound, or MySQL or PHP. If you are to pump the Intel CPUs, you use well threaded applications such as Microsoft SQL Server, VMware, Oracle, SAP. You get the point.
    Last edited by tshen83; 09-21-2009 at 11:24 PM.

  24. #24
    Join Date
    Jan 2003
    Location
    Chicago, IL
    Posts
    6,957
    Quote Originally Posted by tshen83 View Post
    Karl:

    You showed a benchmark where the Phenom II 965 edges out the i7 920 in the web benchmark using PHP+MySQL, and you are trying to correlate that Barcelona has twice the cores so, must be faster. It is false.

    1. PHP is known to not scale well over multicores. MySQL also has vertical scaling issues unless you use Percona builds or beta 5.4. I am sure that review site used the default 5.0 MySQL which is only scalable to 4 threads, which means if you use a CPU with 8 threads, it is actually faster to disable HyperThreading for MySQL(in 5.0 and 5.1 standard builds in Linux). It is the same reason why facebook CTO said that he didn't find the newest Nehalem-EP server give him more performance.

    2. Phenom II 965 can't be correlated to Barcelonas because it uses DDR3 memory. Also it is high clocked 3.2Ghz? Shanghai cores that MySQL likes.

    As to the Phenom x3 beating the i7, something has to be wrong or IO constrained for Phenom x3 to beat 8 threaded i7 920.

    Review sites can swing performance benchmark results +/- 20% for a processor depending on who is "paying the bill". If you are to pump the AMD CPUs, you use benchmarks that aren't well threaded so Nehalem advantage won't be felt, ie games so that it is GPU bound, or MySQL or PHP. If you are to pump the Intel CPUs, you use well threaded applications such as Microsoft SQL Server, VMware, Oracle, SAP. You get the point.
    I agree, the benchmarks I gave aren't perfect, but those were the most relevant benchmarks I could find. If you find something better, say so, but those benchmarks are just as good, if not better than yours, as it is showing the applications used in a hosting environment, and this is Web Hosting Talk. Personally, I have really never seen a synthetic benchmark that has any true correlation to real world performance, as you said, people will hand pick benchmarks or optimize for specific benchmarks, which is the reason I prefer MySQL and Apache benchmarks, etc.
    Karl Zimmerman - Founder & CEO of Steadfast
    VMware Virtual Data Center Platform

    karl @ steadfast.net - Sales/Support: 312-602-2689
    Cloud Hosting, Managed Dedicated Servers, Chicago Colocation, and New Jersey Colocation

  25. #25
    Join Date
    May 2009
    Posts
    217
    Quote Originally Posted by KarlZimmer View Post
    I agree, the benchmarks I gave aren't perfect, but those were the most relevant benchmarks I could find. If you find something better, say so, but those benchmarks are just as good, if not better than yours, as it is showing the applications used in a hosting environment, and this is Web Hosting Talk. Personally, I have really never seen a synthetic benchmark that has any true correlation to real world performance, as you said, people will hand pick benchmarks or optimize for specific benchmarks, which is the reason I prefer MySQL and Apache benchmarks, etc.
    Let's just call it a day. I personally don't see how you disregard SpecInt results since it is the defacto industry standard benchmark for CPU integer performance, which is relevant in pretty much every way. On the contrary, you quote some non-industry standard benchmarks from Phoronix using Phenom x3s when you were asked to quote barcelona results. Look at the Phoronix results where they showed that Phenom x3 beats the i7, the i7 920 and i7 860 both were beat by the i5 750. So the benchmark was screwed up from the beginning. Seriously, I don't see how Phoronix test suite and Phenom can be more relevant than SpecInt.

    If you really want to look at some industry standard "web benchmark", even though it is heavily IO oriented,

    http://www.spec.org/web2005/results/web2005.html

    Look at those results especially those by HP. Clock for clock, Nehalem based cores beat Shanghai based cores by 80%+.(It is not the only benchmark that shows this) Since Shanghai is about 15% better than barcleonas clock for clock, it is not hard to see why i7 920 is better than dual quad barcelonas.

Page 1 of 2 12 LastLast

Similar Threads

  1. Intel P4 $69,Core2Duos starting $95,Quad Core $139,2x Dual Core AMD Opteron $159
    By B2Netsolutions|Adull in forum Dedicated Hosting Offers
    Replies: 0
    Last Post: 09-18-2009, 11:19 AM
  2. Replies: 0
    Last Post: 08-03-2009, 06:32 PM
  3. Replies: 0
    Last Post: 07-22-2009, 05:33 PM
  4. Replies: 0
    Last Post: 07-15-2009, 12:08 PM
  5. Replies: 0
    Last Post: 07-07-2009, 12:26 PM

Posting Permissions

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