Slashdot Mirror

It beats the itanium2 on int_base (845 vs 807) and gets 1016 vs 1356 on fp_base. That would place it in third place on fp, and second on int, in the graphs they have.

So it's kind of strange that it's missing. Inter-department rivalry?

I think it's great that they actually use the spec benchmarks too, and not some kind of homebrewed BS crafted to give themselves an edge, like some other manufacturers do. Signs of a quality corporation.

/August

Okay, tell me what processor is faster than the current high end AMD or Intel processors?

A Pentium IV 2.8Ghz offers a peak SPECint2000 score of 1010. Meanwhile an Alpha running at just over 1/3 of the same clock speed offers 2/3 of the performance. That same 1 GHz Alpha has only 80% of the SPECint2000 performance of the AMD Athlon 2600+. These numbers were grabbed from the full list of published SPECint2000 results.

Don't be ridiculous. Of course Mhz != performance. But the PPC loses in clockspeed by something like 2.5x. At that disparity, even if the PPC could keep its pipelines full 100% of the time due to its great architecture, it STILL wouldn't be as fast as the fastest P4. This is getting old. Go to ArsTechnica's forum, and look up G4 vs P4. Read the posts. A lot of people a whole lot smarter than you or me have put up some good info there. As for SPEC, go here. Yep, GCC and PERL sure sound like artificial benchmarks to me!

There are some benchmars out actually. Make of them what you like.
This is the sheet for itanic2 1GHz.
And you can compare it to the others

Its still too early to see what effect the itanics will have but they look quite respectable, if they havent priced them selves out of the parket

There are some benchmars out actually. Make of them what you like.
This is the sheet for itanic2 1GHz.
And you can compare it to the others

Its still too early to see what effect the itanics will have but they look quite respectable, if they havent priced them selves out of the parket

In any case, the about 400 games to be played should suffice to find the true chess champion among both dual-processor setups.

Because of benchmarks like this? (Note how, ignoring the hardware cost for a moment, the top-of-the-line 16-processor IBM pSeries machine running zeus supports 2.5x more users than the best-available 8-processor IIS server.) Also, zeus (and, may be, netscape enterprise, etc.) is known to have better single-machine scalability because of serveral interesting I/O techniques it tends to use---these benefits are more pronounced when run on operating systems like solaris that support fine-grained user-level threads to kernel-level thread mappings. On top of the raw performance, many also support application-level clustering and redundancy (may be important for some portal sites that demand underlying data consistency, and, which, therefore, require more app-level work to scale-up/failover than just adding more server instances). However, for the vast majority of the sites out there that serve out mostly static and simple dynamic traffic, I think apache is more than sufficient (these sites tend to be bottlenecked by the n/w, not by the server), and I would pick apache anyday over IIS for simplicity, stability, and security reasons (even the humble tux server almost matches the best-available IIS5.0 on the same hardware in the benchmark above in terms of performance; there is no need to go into security/stability comparisons).

SPEC

see the 1st quarter 2002 results for CPU2000

This was my final year project thesis. Just remember the golden rule unstructured 2 structured == convert 2 XML I wrote a [very bad] program in C++/Perl/tcsh IPC=pipes to add XML tags to English, and then index them into a search engine which would use the lingual data stored in the XML tags to help the search.

NIST does a MASSIVE competition on this annually. I don't want to be an XML-buzzword whore <Arnold Schwarzenegger accent> (XML commando eats Green berets, C++, Java, Perl, COBOL for breakfast)</Arnold Schwarzenegger accent> but you can't beat XML for easily converting anything that you can make sense out of into computer readable format. Real h3cKoRs use SGML, but us underlings have to stick with things we can understand like XML. As for expandability, if we want to encode something else into the document, then just tag-it-and-go

It took me 200 hours to fish out all these links (before the Google days), I don't want anyone to have to waste as much time as I did feeding the search engines exotic foods. It's a year old so pardon me for the odd broken link, armed with these you could probably turn jello into XML ;-)

My favourite bookmarx
PROJect[21 links]
Beginners' Guide[13 links]
Berkeley Linguistics Dept. Course Summaries, general stuffzzzzzzzzzzzzzzCryptic IR Vocabulary defined
Explanations of weird words like hypernym zzzzzzzzzzzzzzHow do we produce and understand speech
How Inverted Files are Created - Univeristy of Berkeley zzzzzzzzzzzzzzNLP Univ. of Indiana, very good basics e.g. word sense d
Simple langauge - useful.... zzzzzzzzzzzzzzWhat is Natural Language Processing, links
What is POS tagging........ zzzzzzzzzzzzzzWord Sense Disambiguation defined
Word Sense Disambiguation in detail, scroll down far zzzzzzzzzzzzzzWord Sense Disambiguator - LOLITA (tested at MUC-7 and SENSEVAL competition as best)
XML for the absolute beginner

HTML, XML stuff + parsers[19 links]
Apache plug-in that uhhh does stuff with XML zzzzzzzzzzzzzzConvert COM to XML
convert XML, HTML to Unix pipeable formats zzzzzzzzzzzzzzconverters to and from HTML
expat XML parser zzzzzzzzzzzzzzHTML Tidy - converts HTML 2 XML + source code!!
Parse DB (RDBMS, whatever) to XML zzzzzzzzzzzzzzPerl-XML Module List
PHP Manual XML parser functions - what the hell are they talking about, PHP Virtual M... zzzzzzzzzzzzzzPublic SGML-XML Software
Pyxie - XML Processor for Python, Perl, etc. zzzzzzzzzzzzzzSGML+XML tools.org
The XML Resource Centre - massive number of links zzzzzzzzzzzzzzW4F wrapper - wrapper converts XML to HTML
XFlat - convert flat file into XML zzzzzzzzzzzzzzXML Parsers and other XML stuff
XML.com - Parsers, etc. zzzzzzzzzzzzzzXML-Data Catalog System - uhhhh looks close
XTAL's general converter - convert anything 2 XML

other Background[8 links]
Is Linux ready for the Enterprise, scalable... zzzzzzzzzzzzzzLinux reliability
Linux Versus Windows NT, Mark(sysinternals bloke) zzzzzzzzzzzzzzPC reliability (pcworld)
SPEC - Standard Performance Evaluation Corp. zzzzzzzzzzzzzzSystems benchmarks
TPC - Transaction Processing Performance Council zzzzzzzzzzzzzzUnix Beats Back NT In EDA Workstation Arena
Proper TREC(-8) QA systems[2 links]

pg. 387 LIMSI-CNRS pretty deep parsing[2 links]
More links....
NLP, IR links - lots to corpii, etc.

pg. 575 U. of Ottawa and NRL (shit system, got 0%)[1 links]
LAKE Lab
pg. 607! University of Sheffield (crap system, but OPEN SOURCE!)[2 links]
GATE - FREE IE app w`source code
LaSIE - ER, coreference, template (cv)

pg. 617 Univ of Surrey (inconclusive matches)[2 links]
System Quirk - Or is this their search system..... Hmmmmmm
Univ of Surrey - pointers (hopefully this is their WILDER search system...)

SMU - Pg. 65[1 links]
Natural Language Processing Laboratory at SMU

Textract[2 links]
Cymfony - Technology
Textract - State of the Art Information Extraction

Xerox uhhhhh maybe[1 links]
Xerox Palo Alto Research Center
(OVERVIEW) 1999 TREC-8 Q&A Track Home Page
NLP bloke, Univ Sussex


Tcl-Tk[4 links] Tcl tutorial
Tcl-Tk Contributed Programs Index
Tcl-Tk Resources, sources
TclXML - manipulating XML using Tcl-Tk
Artificial Natural Language - Is this what I'm trying to parse into...
Comparison of Indexers - Prise vs. Inquery vs. MG, etc.
Eagles - Language Engineering Standards
Language Technology Group - lots of modules!
LDC - Linguistic Data Consortium, lots of corpora
Lexical Resources
Links 2 resources, indexers.....
Lots of IR stuff, University of uhhh
Managing Gigabytes Indexer
Managing Gigabytes Manuals and stuff
Htdig search system
NLP & IR (NLPIR, NIST) Group
OVERVIEW OF MUC-7-MET-2
Perl XML Indexing - XML search engine type thing
Phrasys Language Processing Software Components (money)
QA HCI bullshit
SIGIR - TREC-type thing, resources
SMART indexer system documentation
Text REtrieval Conference (TREC) Home Page
The Natural Language Software Registry
Thunderstone IE and IR products
WordNet - FREE DOWNLOADABLE lexical English database

Page created with URL+, nice utility for working with internet shortcuts

One of the big problems here though is that, relative to PCs, Macs have very slow processors. Yeah yeah, I know Steve Jobs likes to try and convince everyone that a 2MHz Mac will outperform a 20GHz PC, but in reality, the G4 is only marginally faster, clock for clock, then a PIII, about on-par with an AMD AthlonXP. The downside? The fastest G4's are running at 867MHz, the slowest AthlonXP's are running at 1.33GHz.

The end result is that the Macs have reasonably good components but slow CPUs as compared to most PCs. The Dell's and Compaq's of the world, on the other hand, have very fast processors and generally piss-poor other components.

All in all, you definitely DO spend a lot of money on a Mac, considering what you get. None of the components in even the high-end Macs are truly top-notch components, at best they're above-average stuff, unless you customize the system yourself (where you get REALLY expensive systems, because the markup on high-end components from OEMs is HUGE, regardless of whether it's from Dell or Apple).

PS. Anyone who actually believes Steve Job's comments about the speed of G4's should try and find ANY industry standard benchmark. ANY of them! Every major CPU out there today has SpecCPU 2000 scores EXCEPT for the Macs. Reason? Third-party SpecCPU scores haven't been too hot for Macs. There's a very good reason why Apple ONLY uses Photoshop, because it's the absolute easiest benchmark in the world to "prove" that any given architecture is faster then another. Apple's benchmarks show that the G4 is faster then the P4 in Photoshop, while Intel's benchmarks show the exact opposite. Neither of them is lieing, just using different benchmarks.

They should just use Spec numbers...
"The new Athlon 701i 634f!!!! Get it now!"
(Epox 8KHA+ Motherboard, Athlon (TM) XP 1900+ )

"Get rid of that old Pentium 4 609i 628f!"
(Intel D850GB motherboard(1.8 GHz P4))

Of course, that gives you the German car kind of luster (lustre U.K.), too... "Dude! My new Athlon 740i protection compiles a kernel superfast and corners like it is on rails!"

Or maybe not....

Check out specbench.org its pretty much the industry standard for performance benchmarking. They have many types of programs that they test with, dbs, compilers ...etc. The two main categories are floating point and integer tests.

You can see this if you compare various benchmarks for different systems. The Sun Fire 280R has a CINT2000 of 375, and a CFP2000 of 324, the Fujitsu PRIMEPOWER400 (600MHz)has 390 and 314, so the Fujitsu has faster Integer performance, while the Sun has faster FP performance. Which one is faster in 'real life' depends on the mix found in your application.

What a novel idea!

Sarcasm aside -- the SPEC benchmarks have been around for a long time and are well respected. You can see some SPEC CPU 2000 results here.

Your altivec optimezed CPU are playing in a different leage. UltraSparcs are for servers and are top of the line when it comes to 16 cpus. Something I have never seen on a mac.

Why don't you present some fact to back your statement up. Of course it's hard to find because Apples only bench Photoshop. Acording to specbench a 900mhz US-III is about as fast as an Athlon Tbird 1.2 on a shitty mobo with 133mhz mem. Of course mac/G4 benches are nowhere to find on that site.

I can not understand why you should compare compare one US-III to two G4. seems odd.

Simple, you define your "target mix" (see here) for the test to include no dynamic content. Or you choose "dumb" dynamic code that really isn't dynamic (the benchmarker gets to pick his own dynamic code).
Also note that this article does not say that the specweb99 benchmark was used this time, it only says that eWEEK worked closely with Dell, who did some specweb99 benchmarks of Tux before.
So basically we have 1) no indication that the specweb99 benchmark was used this time, and 2) no requirement that dynamic content handling is needed to complete the specweb99 benchmark.
And while it is entirely possible that Tux does indeed do dynamic content and is very fast at it, this article offers no proof of it. It would help a lot if they actually mentioned what benchmark they used, and gave some more results than just the few numbers given. It would be even better if benchmark results applied to real-life situations as well.

First off do not equate clock speed to clock speed. Second, even the numbers I quote are
pretty much meaningless. The only TRUE comparison would be to actualy try it.
Also there is no such thing as a P3-1000. You meant a P6 (PIII chip, probably
Coppermine).

Being as such it has 32K of L1 cache and 256K of L2 cache.
Its SpecINT and SpecFP is aproximatly 410 and 407 respectivly.
Opposed to this is the Ultra3 with 100K L1 cache and 2M of L2 cache.
Its SpecINT and SpecFP is 467 and 438.
However do not be mislead by these numbers. Real tests of performance
would be to put the systems side by side and run whatever application you have to test with.

http://www.specbench.org
SPECint is a set of benchmarks that tell you something about integer performance for certain
sizes of code, often with ultra-heroic compiler tuning that sometimes doesn't happen on real
applications, some fo which includes pattern recognition of the specific source code that
breaks real applications. SPECfp is a set of benchmarks that tell you something about floating
point code of certain sizes.
Lastly rember SpecINT != Integer performance and SpecFP != floating point performance

In limited cases where AltiVec acceleration was available and tested under FORTRAN, the G4 showed a clear advantage with 4-7X greater performance and a 5-8X greater cost effectiveness than all other workstation systems evaluated.

I found the link at the bottom of this page. slashcode isn't rendering my html

Let's take something from a IBM's site, just for fun:

IBM PPC Benchmarks

SpecINT 95 is the geometric mean 10 different benchmarks, including a GCC compilation benchmark, a GO-playing program, a LISP interpreter, and database core, some PERL code...in other words, it's a pretty wide spectrum of performance, and is really *THE* number to talk about in terms of integer performance. It's not perfect, but it's a benchmark from an independent source, and measures a wide variety of metrics. SpecFP is a similarly structured suite stressing floating point performance.

The chart is a bit old, topping out with the 750 @ 500 Mhz. The numbers given are 23.8Int/14.5Fp.

From SPEC's own site, you can see a 550Mhz Athlon scores about 25Int/20.6FP

So the cores are roughly on par in a clock-for-clock basis. Since I can't find any numbers for the new 766Mhz PPC, I'll assume it scales linearly (This is generous), and say it goes to roughly 32Int/22Fp. a 1 Ghz Athlon (which isn't even the top of the line anymore) rates 42.9/29.4.

This isn't, of course, complete system performance. It says nothing about IO speed, hard storage speed, etc. but I'll be a monkey's uncle if Photoshop is anything other than CPU or memory bound.

Apple is being deceptive. Plain and simple. They've chosen a nonrepresentative benchmark to portray the PPC as faster despite the clock disadvantage, and that's just not true.

Measuring the performance of machines is way to complicated of an issue to use something like a MIPS rating. Because MIPS factors out the instruction count needed to get something done, you can inflate your rating by doing a large load of useless instructions really quickly.

MIPS = (InstructionCount) / (ExecutionTime *10^6)

ExecutionTime = (InstructionCount * AvgClockCyclesPerInstr * CycleTime)

The InstructionCount's cancel out, leaving

MIPS = 1 / (AvgClockCyclesPerInstr * CycleTime *10^6)

So if another computer can do the same amount of work with ten times less instructions, it doesn't show up in a MIPS rating.

Measuring performance just isn't as simple as looking at a single numeric rating. Sometimes you are interested in measuring responsiveness, somtimes throughput, and a lot of it depends on the specific applications you want to run. Just asking "how fast" is a processor is almost meaningless.

To my knowledge, the spec benchmarks, while not free, are the best standardized benchmarks out there. For integer performance alone, it tests data compression (gzip and bzip I think), FPGA circuit placement, compiling c code, chess, running perl, ray tracing, database stuff, etc; I can't even remember all of the stuff it tests for floating point performance. Obviously, because it isn't free you probably won't be using it to test your home linux box, but if you are doing serious bench marks, the money would be worth it.

However, I'd be interested in seeing another round of benchmarks, especially between Linux 2.4.0 (preferably whenever it's officially released) and Windows 2000.


On identical hardware, linux 2.4 with Tux web server beats the crap out of Windows 2000 with IIS Click here for result and don't forget to click here also for comparisons with different numbers of CPUs/NICs.

For file system spec test, it isn't pretty. Reiserfs home pages list comparisons against ext2 - they are quite comparable. NTFS hardly wins ANY performance test against any file systems.

However, I'd be interested in seeing another round of benchmarks, especially between Linux 2.4.0 (preferably whenever it's officially released) and Windows 2000.


On identical hardware, linux 2.4 with Tux web server beats the crap out of Windows 2000 with IIS Click here for result and don't forget to click here also for comparisons with different numbers of CPUs/NICs.

For file system spec test, it isn't pretty. Reiserfs home pages list comparisons against ext2 - they are quite comparable. NTFS hardly wins ANY performance test against any file systems.

Then you'd be oh so wrong. The Spec2000 figures are up at specint2000 and specfp2000 and Mhz for Mhz the US3 beats any x86 by 10% integer and 60%+ Floating Point. Only the Alpha touches it, and there's still no useful software for them.

Then you'd be oh so wrong. The Spec2000 figures are up at specint2000 and specfp2000 and Mhz for Mhz the US3 beats any x86 by 10% integer and 60%+ Floating Point. Only the Alpha touches it, and there's still no useful software for them.

low end I2's had 100, 150, 200 or 250 mhz R4400's while high end I2's had 175 or 195 R10000's. Some also has 150 or 180 mhz R5000's. There was also a R8000 variant which used a 75-80 mhz chip. Sgi hrware faq is here.

My experience with I2's shows that an R4400 is about comperable to 1.25 mhz on a Pentium 2, while an R8k or R10k is about 2-3 mhz P2. I've never used an R5k; they seem to be aimed at the graphics types, while the 44k, 8k, and 10k were aimed at the numerical simulation folks. Your results may vary, of course; Spec 95 CINT results for an R10k 195 were very disappointing- comparable to a PII 266. Some similar POV-ray render speed benchmarks can be found here as well.

Rev Neh

With regard to testing, please see the project I've been working on recently, glean. I can certainly use more contributions.

And Alpha's have been running Linux almost from
day one.

HP should spend some of that money on alpha instead.

just check out this site.

www.dcginc.com

They have a 700Mhz Alpha 21264 system using the
AMD Irongate chipset for just $5000.


700Mhz EV6 system

Or screaming Dual 750Mhz Alpha 21264 system with
all the frills for just $15000
mind you thats with 8MB of cache per processor.

DUAL 750 Mhz Alpha 21264 System with 16MBytes of cache

looking up the spec site we see that a 700Mhz EV6 gets 39.1 specINT and 68.1 specFP.

scaling that these 750Mhz 21264's should have
42 specINT and 72 specFP


On a side note does HP think IA64 will not
be as hot as once predicted.
why are they now pushing linux for PA Risk.
if they are going to abandon PA Risk in a year or two
why start now to port linux?

Or is HP going to pull an SGI :)

Oops. PA Risk=PA-RISC.

The best fair metric for chip performance seems to remain SPECint95 and SPECfp95 (www.specbench.org).

By that metric, the fastest recorded RISC and CISC speeds (1 per architecture) are the following:

700 MHz 21264A Compaq AlphaServer GS60E:
39.1 SPECint95 / 68.10 SPECfp95 (or w/ minimum optimizations, 34.7 SPECint_base95 / 54.5 SPECfp_base95)

733 MHz Pentium III (i840) Intel:
35.6 SPECint95 / 30.4 SPECfp95 (no _base figures available, figures from Intel, not yet on SPEC website)

440 MHz PA-8500 HP N4000:
34.0 SPECint95 / 51.4 SPECfp95 (30.8 int_base / 48.7 fp_base)

450 MHz UltraSPARC-II Sun Ultra 60 Model 1450:
19.7 SPECint95 / 27.0 SPECfp95 (16.2 int_base / 23.90 fp_base)

300 MHz MIPS R12000 Origin 2000 2-way:
18.4 SPECint95 / 34.4 SPECfp95 (18.1 int_base / 30.1 fp_base)

340 MHz PowerPC RS64-II IBM H70:
16.0 SPECint95 / 21.2 SPECfp95 (13.7 int_base / 20.2 fp_base)

Thus the answer to your question is "Yes, Alpha remains the fastest", with the important caveat that the 10% performance advantage over Intel comes at a significantly higher price. All other RISCs are slower than the fastest Intel systems, at least in terms of uniprocessor integer performance, the best single predictor for most CPU-limited applications.

Note that Apple G4 performance, and performance of IBM's latest S80 (450 MHz Power RS-III) aren't discussed by their respective vendors. If you extrapolated the G4 performance from the mildly similar 340 MHz Power RS-II, performance of a 500 MHz part would be around 23.53 SPECint with SPECfp at 31.8. IBM's Power PC 604e parts have slightly lower integer performance and much lower floating point performance at the same clock rates as the RS-II (375 MHz 604e runs 15.1 SPECint, 10.1 SPECfp,) so even if there are some other G4 improvements, I doubt the 500 MHz G4 will be beating a 733 MHz Pentium III.

Note that these benchmarks don't measure performance of vector-processing chip features like MMX used by a few apps like Photoshop.

--LP

P.S. (Sidebar: The minimal performance value-add of RISC over Intel is is why I think Linux's highly touted multi-architecture support for RISCs is exactly a glowing scalability feature as some apparently make it. I guess 64-bits and floating point comes in handy for a few apps. Other than that, it's a nice plus for legacy hardware that grows less relevant by the day. )

And VA Research has SPEC numbers on their site last I checked.

But there aren't any SPECWeb results on the SPEC Web site (that's "SPEC" "Web site", not "SPECWeb" site; i.e., it's the Web site for the Standard Performance Evaluation Corporation).

A search for "SPECweb" on VA Research's Web site turned up nothing; where did you find their SPECweb numbers? (SPECCPU numbers, say, aren't SPECweb numbers; the only "SPEC numbers" that count as a response to Microsoft's claim are SPECweb numbers....)

(There are NT+IIS numbers on the SPEC site.)

What you have requested (a level playing field benchmark agreed upon by vendors which any party can submit their own results doing their own tuning) already exists! It's called SPECweb96 (see www.specbench.org). The major conventional system vendors have agreed upon the benchmark, running procedure, and reporting procedure for webserver performance.

There are already 4-way Xeon Windows NT-based results. There are 4-way Xeon Solaris for x86 results. So where are the 4-way Xeon Linux results?!?!

Now unlike what you requested, the results can be submitted on any hardware setup; that doesn't stay constant, making it somewhat tricky to compare results at times. But one-upsmanship on the same hardware is fair game.

Some Linux vendor (Red Hat, Caldera, VA Research, Compaq, etc.) or ambitious user needs to wake up and do these. There would seem to be a substantial payoff in terms of third-party objective SMP credibility for Linux.

--G

SPECint95 SPECfp95
Celery 400 + 14.9 10.6
PII 400 + 16.9 12.8
PII 450 ! 17.2 12.9
PII 450 + 18.5 13.3
PIII 450 + 18.7 13.7
PIII 500 + 20.6 14.7
21164 500 ! 15.7 19.5
21164A 533 * 16.1 18.8
21164A 667 * 23.0 30.0
21264 500 * 26.0 49.0
21264 500 ! 27.7 58.7