Home

News

Forums

Hardware

CPUs

Mainboards

Video

Guides

CPU Prices

Memory Prices

Shop



Sharky Extreme :


Latest News


- Outdoor Life: Panasonic Puts 3G Wireless Into Rugged Notebooks
- Averatec Launches Lightweight Turion 64 X2 Laptop
- Acer Fires Up Two New Ferrari Notebooks
- Belkin Debuts Docking Station for ExpressCard-Equipped Notebooks
- Logitech 5.1 Speaker System Puts Your Ears At Eye Level
News Archives

Features

- SharkyExtreme.com: Interview with ATI's Terry Makedon
- SharkyExtreme.com: Interview with Seagate's Joni Clark
- Half-Life 2 Review
- DOOM 3 Review
- Unreal Tournament 2004 Review

Buyer's Guides

- September High-end Gaming PC Buyer's Guide
- September Value Gaming PC Buyer's Guide
- October Extreme Gaming PC Buyer's Guide

HARDWARE

  • CPUs


  • Motherboards

    - Gigabyte GA-965P-DS3 Motherboard Review
    - DFI LANPARTY UT nF4 Ultra-D Motherboard Review

  • Video Cards

    - Gigabyte GeForce 7600 GT 256MB Review
    - ASUS EN7900GT TOP 256MB Review
    - ASUS EN7600GT Silent 256MB Review
    - Biostar GeForce 7900 GT 256MB Review





  • SharkyForums.Com - Print: Itanium performance item

    Itanium performance item
    By ua549 April 10, 2001, 12:11 PM

    In a document published by Cordiant, the 800 MHz Itanium processor showed a greater than 10 fold advantage in processing secure transactions over existing, RISC-based processors. For example, a server configured with four Itanium processors performed more than 1,376 security transactions without hardware acceleration, compared to 114 transactions achieved on a comparably configured Sun Enterprise 420R system.

    "Itanium processor-based systems achieve the best 4-processor security performance results we have ever seen in our labs," said Eric Packman, chief technology officer, Cordiant.

    Does anyone know what a Sun enterprise 420R system is? Is it really comparable to a 4x800 MHz Itanium system?

    By Arcadian April 10, 2001, 12:51 PM

    Do you have a link for that info? Thanks.

    By Moridin April 10, 2001, 01:03 PM

    There is an entry on the SPEC page for this machine (1 of 4 processors enabled) the results are not impressive. (205 SPEC Int, compared to 438 for a 1 GHz PIII).
    http://www.spec.org/osg/cpu2000/results/cpu2000.html

    The processor in this is a 450 MHz USII. Not Sun's new USIII which would be considerably faster. The 750 MHz USIII may be almost twice as fast. (as a 450 MHz USIII)
    http://www.sun.com/servers/workgroup/420r/

    Most Sun boxes are still likely shipping 450 or 500 MHz USII's. The USIII is a little newer then the P4 and still is in short supply.

    This aside though, I seem to recall speculation that Itanium would be particularly good (or was that bad) at encryption/decryption. I can't recall why this was supposed to be so but I will think about it a little.


    By Arcadian April 10, 2001, 01:36 PM

    Well I found the link on the Encryption performance.
    http://biz.yahoo.com/bw/010410/0544.html

    I would assume that the very large register file would contribute to this. I'm sure Itanium can make a killing against any of the RISC processors out there in terms of security processing.

    As for the SPEC scores, Moridin, you'll have to provide a reliable link. I find it hard to believe Itanium would get a SPECint score of 205. I was expecting more along the lines of 500-550, with a SPECfp of about 600.

    Keep in mind, also, that the liklihood of uniprocessor Itanium systems will be very small. I imagine DP will be the smallest config available, and multiprocessor performance on Itanium is supposed to scale far better than x86.

    By Moridin April 10, 2001, 01:47 PM

    quote:Originally posted by Arcadian:
    Well I found the link on the Encryption performance.
    http://biz.yahoo.com/bw/010410/0544.html

    I would assume that the very large register file would contribute to this. I'm sure Itanium can make a killing against any of the RISC processors out there in terms of security processing.

    As for the SPEC scores, Moridin, you'll have to provide a reliable link. I find it hard to believe Itanium would get a SPECint score of 205. I was expecting more along the lines of 500-550, with a SPECfp of about 600.

    Keep in mind, also, that the liklihood of uniprocessor Itanium systems will be very small. I imagine DP will be the smallest config available, and multiprocessor performance on Itanium is supposed to scale far better than x86.

    I guess I got distracted, I thought I put the link in. Itanium should easily outperform the USIII for most tasks IMO. Despite being a new chip, much of the USIII's fundamental technology is outdated.

    The only real advantage it has over Itanium is it's on board memory controller and hardware prefetch, (and the hardware prefetch needs to be disabled because it is causing data corruption.)


    BTW it doesn't look like the SPEC submission was prepared by Sun so this score may be low relative to other scores.


    By ua549 April 10, 2001, 03:21 PM

    quote:Originally posted by Arcadian:
    Do you have a link for that info? Thanks.

    http://news.moneycentral.msn.com/ticker/article.asp?Symbol=US:INTC&Feed=BW&Date=20010410&ID=683854

    By Marsolin April 11, 2001, 05:45 PM

    I have seen a demo comparing an Itanium 667MHz system with a Sun 450 USII system. I think both had four processors. The Itanium won easily for IPSEC calculations even when a dedicated hardware decryption card was added to the Sun box.

    An USIII 750 should be slightly faster in SpecINT, but about 50% of Itanium's SpecFP scores.

    By Marsolin April 11, 2001, 05:49 PM

    quote:Originally posted by Moridin:
    [BDespite being a new chip, much of the USIII's fundamental technology is outdated.
    [/B]

    Something both chips have in common. I'm not sure which takes the title for most delayed.

    By Hattig April 13, 2001, 09:11 PM

    quote:For example, a server configured with four Itanium processors performed more than 1,376 security transactions without hardware acceleration, compared to 114 transactions achieved on a comparably configured Sun Enterprise 420R system. [/b]

    Or for the cost of a server with 4 Itanium servers inside it, you could buy a much cheaper solution and stick 4 nCipher nForce 400 SCSI cards on it for 1600 transactions per second.

    Why use the CPU to brute force something when some elegantly designed hardware can do it better - that is why we have GPUs, SCSI adapters, etc...

    Also, what was the cost differential between the Sun system and the Itanium system? Also bear in mind that a favourite of the benchmarker is to compare your up-and-coming with the competitions old-and-dying.

    By Arcadian April 13, 2001, 10:07 PM

    quote:Originally posted by Hattig:
    Or for the cost of a server with 4 Itanium servers inside it, you could buy a much cheaper solution and stick 4 nCipher nForce 400 SCSI cards on it for 1600 transactions per second.

    Why use the CPU to brute force something when some elegantly designed hardware can do it better - that is why we have GPUs, SCSI adapters, etc...

    Also, what was the cost differential between the Sun system and the Itanium system? Also bear in mind that a favourite of the benchmarker is to compare your up-and-coming with the competitions old-and-dying.

    Then again, I believe that there are other applications that can take advantage of Itanium's large register file, like encryption does. This is just one comparison, and probably not the only one where Itanium will accel. Of course with all marketing, take it with a grain of salt. You wouldn't have read about how good George Bush is from his own web page, would you?

    By Hattig April 14, 2001, 12:00 AM

    quote:Originally posted by Arcadian:
    Then again, I believe that there are other applications that can take advantage of Itanium's large register file, like encryption does.

    I agree. However, when you are creating a server for a specific purpose, you should see what the most cost effective may of achieving that purpose (over the servers lifetime, of course) is. For an e-commerce secure server under very heavy load, you would want a good hardware accelleration card, a piece of hardware designed to do the task in hard very quickly and efficiently, without burning a hole in the servers case.

    quote:This is just one comparison, and probably not the only one where Itanium will accel. Of course with all marketing, take it with a grain of salt. You wouldn't have read about how good George Bush is from his own web page, would you?

    No, I don't live in the land of nearly-freedom. I live in the land of foot-and-mouth and stupid annoying farmers with too much political lobbying power who should all be made to get animal health insurance instead of getting our taxes to pay for their inept tax dodging disease spreading behaviour.

    Anyway, I am waiting for McKinley benchmarks. They will be interesting and more relevant. Not that I could ever afford one of these chips anyway, so it is a moot point I suppose.

    Graham

    By Conrad Song April 14, 2001, 12:29 PM

    quote:Originally posted by Arcadian:
    Then again, I believe that there are other applications that can take advantage of Itanium's large register file, like encryption does. This is just one comparison, and probably not the only one where Itanium will accel.

    One of the more interesting advantages is that Itanium does not have a hardware floating point divider. Instead, it uses a software algorithm that allows it to spit out divisions at a higher throughput than even the Pentium 4 with SSE2. From my calculations it would take a P4 over 2GHz to get the same rate.

    By Arcadian April 14, 2001, 03:08 PM

    quote:Originally posted by Conrad Song:
    One of the more interesting advantages is that Itanium does not have a hardware floating point divider. Instead, it uses a software algorithm that allows it to spit out divisions at a higher throughput than even the Pentium 4 with SSE2. From my calculations it would take a P4 over 2GHz to get the same rate.

    Yeah, EPIC works a lot differently than x86. Most CISC architectures support a large variety of calculations in hardware, but these actually don't pipeline very well, and the logic required to implement these complicated instructions sometimes compromise other instruction latencies.

    Intel has come out with many case studies to aid developers in writing Itanium friendly code. They have the advantage to start from the beginning to build good habits from developers.
    http://developer.intel.com/software/products/itc/architec/itanium/itan_wpdg.htm

    Also, check out these links.
    http://developer.intel.com/software/products/mkl/mklspecs_new.htm
    http://developer.intel.com/software/products/opensource/libraries/numdown1.htm
    http://developer.intel.com/software/products/itc/architec/itanium/itan.htm
    http://developer.intel.com/design/ia-64/downloads/24547401.pdf
    http://developer.intel.com/software/products/itc/architec/itanium/optapp_mod/tutorial/index.htm

    It's a lot of links, but it gives you an idea of how Intel has prepared the industry for Itanium support. There are many developers that have had a head start in optimizing their applications, and it is my opinion that we will be pleasantly surprised with the content that gets released at the Itanium launch.


    Contact Us | www.SharkyForums.com

    Copyright 1999, 2000 internet.com Corporation. All Rights Reserved.


    Ultimate Bulletin Board 5.46

    previous page
    next page





    Copyright 2002 INT Media Group, Incorporated. All Rights Reserved. About INT Media Group | Press Releases | Privacy Policy | Career Opportunities