Home

News

Forums

Hardware

CPUs

Mainboards

Video

Guides

CPU Prices

Memory Prices

Shop



Sharky Extreme :


Latest News


- Dell Joins the Netbook Movement with its Inspiron Mini 9
- Kingston is the First to Offer Ultra Low-Latency DDR2 Notebook Memory
- Logitech Unleashes Three New Keyboards
- NEC Adds New Servers to Product Line
- Lian-Li Launches New Power Supply Line, Rack Mount Kit and Fan Blower
News Archives

Features

- SharkyExtreme.com: Interview with Microsoft's Dan Odell
- SharkyExtreme.com: Interview with ATI's Terry Makedon
- SharkyExtreme.com: Interview with Seagate's Joni Clark
- Half-Life 2 Review
- DOOM 3 Review

Buyer's Guides

- July High-end Gaming PC Buyer's Guide
- May Value Gaming PC Buyer's Guide
- March Extreme Gaming PC Buyer's Guide

HARDWARE

  • CPUs

    - AMD Phenom X4 9950 BE & 9350e Review

  • Motherboards

    - AMD 790GX Chipset Review
    - Gigabyte GA-MA790FX-DS5 Motherboard Review
    - AMD 780G Chipset Review

  • Video Cards

    - PNY XLR8 GeForce 9800 GX2 1GB Review





  • SharkyForums.Com - Print: Questions on AthlonXP

    Questions on AthlonXP
    By elimc October 09, 2001, 10:42 PM

    As many of you know, the AthlonXP has come out and, looking at most benchmarks, is the best performing processor that you can buy right now using today's applications. However, I made a quick trip to Ace's Hardware and they used the SPEC test in which Intel pretty much dominated the competition. Why is this???

    Also, is XP the Palomino core. I'm assuming it is because of the 20% heat reduction (around what SOI would give them).

    By jtshaw October 10, 2001, 01:47 PM

    While I can't tell you why the AthlonXP didn't do well on the SPEC benchmarks I can tell you that it is in fact using the Palomino core that the AthlonMP and Athlon4 use. Tomshardware has a good article that talks about the differences between the thunderbird core and the palomino core and the packaging differences.

    quote:Originally posted by elimc:
    As many of you know, the AthlonXP has come out and, looking at most benchmarks, is the best performing processor that you can buy right now using today's applications. However, I made a quick trip to Ace's Hardware and they used the SPEC test in which Intel pretty much dominated the competition. Why is this???

    Also, is XP the Palomino core. I'm assuming it is because of the 20% heat reduction (around what SOI would give them).

    By pm October 10, 2001, 05:41 PM

    I can't answer your question regarding SPEC scores, but I will say that the best benchmark for a processor is comparing the relative performance of the applications that you intend to use. I realize that your sig is (probably) sarcastic, but if, for example, all you intended to play is Quake3 with your future system, then Quake3 benchmark scores are the only thing that you should be looking at. SPEC is intended to show the relative performance of processors of different architecture by levelling the playing field as much as possible, but it is still essentially a synthetic benchmark that attempts to replicate real-world performance

    The AlthonXP is using the core codenamed Palomino but the AthlonXP currently is not being fabricated using SOI technology. The power reduction of the AthlonXP design was probably acheived by a combination of fabrication advances on their 0.18um process and design tweaks.

    By elimc October 10, 2001, 05:46 PM

    Ok, so was the SPEC test a quirk? Why should we use the SPEC test if no applications are based on it?

    By pm October 10, 2001, 06:30 PM

    quote:Originally posted by elimc:
    Ok, so was the SPEC test a quirk? Why should we use the SPEC test if no applications are based on it?
    I can guess why the SPEC scores are the way they are, but since I work for Intel I'd rather not try to guess.

    Why SPEC? I personally think it's the best cross-platform benchmark available. It attempts to solve a lot of the problems of traditional benchmarks - such as executable optimization for a certain platform, microarchitectural optimization for a certain code sequence, etc. The benchmark itself contains plenty of relevant code that implements commonly used code sections such as compression, games and data processing. Here's the application list for SPEC CPU2000: http://www.spec.org/osg/cpu2000/CINT2000/

    My point is not to say that SPEC is irrelevant or a bad benchmark, but to say that the best benchmark is the application that you intend to run on the system - not a general cross-platform benchmark no matter how good it is. I think SPEC is about the best that we can hope to achieve in a general purpose cross-platform benchmark, but I still think you should make your system selection based on applications you are likely to run if possible.

    By Conrad Song October 10, 2001, 11:40 PM

    quote:Originally posted by elimc:
    As many of you know, the AthlonXP has come out and, looking at most benchmarks, is the best performing processor that you can buy right now using today's applications. However, I made a quick trip to Ace's Hardware and they used the SPEC test in which Intel pretty much dominated the competition. Why is this???

    Also, is XP the Palomino core. I'm assuming it is because of the 20% heat reduction (around what SOI would give them).

    SPEC is one of the few benchmarks that allows targeted optimizations, since code is compiled and run using the compiler and platform of your choice. Other benchmarks are legacy precompiled code and tend to fair badly on new architectures.

    In the Pentium 4's case, this is exasperated by the fact that the larger cache line and memory ordering restrictions can give it fits on legacy code. With SPEC, code is truly compiled for the Pentium 4, and the results are obvious.

    It must also be mentioned that the latest SPEC results were made with the 010525 (2001, May, 25) build of the Intel 5.0.1 compiler, which happens to be newer than the October 2000 scores. To contrast, SpecInt scores did not change much but SpecFp improved by around 10% base.

    By elimc October 10, 2001, 11:54 PM

    Thanks guys, that cleared things up. So when applications begin to be truly optomized for the P4 we will see the SPEC scores represent real world scores?

    By duron-burger-man October 11, 2001, 01:29 AM

    quote:Originally posted by elimc:
    Thanks guys, that cleared things up. So when applications begin to be truly optomized for the P4 we will see the SPEC scores represent real world scores?

    Not exactly, as pm mentioned, it depends on what your running. Certainly everything being optimized for SSE2 won't hurt P4 hehe, but that does not mean it will tear up Athlon or Hammer. That doesn't mean it won't though either, just wait and see.

    SPEC only mimics real world performance, its synthetic, perhaps the most fair synthetic benchmark, but by no means is it Quake 3, Photoshop, or whatever you choose to run.

    By Conrad Song October 11, 2001, 01:41 AM

    quote:Originally posted by elimc:
    Thanks guys, that cleared things up. So when applications begin to be truly optomized for the P4 we will see the SPEC scores represent real world scores?

    I don't want to mislead you in any way. It may be that it will never happen.

    Part of this is to do with the fact that the set of benchmarks may never be applicable to the personal user. For example: specfpu has a couple of benchmarks dealing with meteorology, but the image recognition and neural network benchmarks may become common in the future.

    I suggest you take a look at SPEC's CPU2000 benchmarks individually. There's also plenty of data on memory usage.
    http://www.spec.org/osg/cpu2000

    By Moridin October 11, 2001, 11:49 AM

    quote:Originally posted by pm:

    SPEC is intended to show the relative performance of processors of different architecture by levelling the playing field as much as possible, but it is still essentially a synthetic benchmark that attempts to replicate real-world performance


    I think synthetic is the wrong term here, since it implies that real application code isn't used. SPEC does use real (though not commercial) applications.


    By Moridin October 11, 2001, 11:52 AM

    SPEC represents the performance of the P4 on a set of (generally) scientific applications running code generated to Intel's recommended standards.

    First, current code is usually generated to P5 or P6 (I'm talking about the core here) standards. AMD designers put great emphasis on running older code well, but this is (probably) done at the expense of how well the processor performs on more optimized code. This limits the upside potential for the processor, but allows it to do very well on existing code.

    Intel (IMO) puts more emphasis on the upside potential of the processor at the expense of mediocre performance on existing code. This makes sense for them because they have the muscle to eventually get most code to fit the newer (higher performing) standards.

    In this sense, SPEC is a good predictor of how these processors will perform down the road. If you put away an model 1800+ Athlon XP and a 2Ghz P4 for 4 years and run "real application" benchmarks on them then you are likely to the results much closer to today's SPEC scores.

    Like pm said though, there is no substitute for benchmarking the applications you intend to run. It will always give you far more accurate results than any general benchmark like SPEC. It is difficult to do this when the application does not exist yet, and this is where SPEC can fit in. It should be a decent predictor of this as long as the software vendor uses Intel's preferred coding standards.

    Second, the type of work done in other benchmarks does not necessarily correspond to the work being tested in SPEC. The natural variation in how processors perform in different tasks can skew the results considerably.

    This is not to say that the stuff tested in SPEC is irrelevant, in fact can be and it may be even more relevant in the future as games start to work more on the underling physical models of the world.


    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