P4 2.8 vs. XP 2600
 

http://www.digit-life.com/articles2/...nxp-26-mx.html

Is that article correct? As I have seen other benchmark tests and such, and the results were different ... does it depend on the system architecture?

Tests can definitely vary depending on system setup, though in general you can't do the impossible. I'd agree with the conclusion that the P4 2,8GHz has a small upper hand over the Athlon XP 2600+. However, there are a few things in this test which should be duly noted:

They're using a VIA KT400 mobo for the Athlon processor with DDR400 RAM, which in fact hampers the Athlon's performance. There is afaik no official standard for DDR400 yet, and it performs worse than DDR333 does on a KT333 motherboard.

Also, they appear to be running something called Intel Application Accelerator 2.2.2, which sounds like it's somewhat biased towards the Pentium processor.

I dont get why they are comparing an AMD 2600XP to and Intel 2.8

The intel is like 500MHz faster than the AMD......of course the intel will win most things.

Because both are the current top-of-the-line products for Intel and AMD?
The P4 2,0GHz was almost 500MHz faster than the Athlon XP1800+, still the AMD beat it in the benchmarks.

Intel's processors may have a higher clockspeed, the AMD processors have a much higher ipc (instructions per clockcycle), which means they need a lower amount of MHz to attain the same performance.

yes but more recent benchmarks I have seen have shown Intels to beat their equivalent AMDs i.e. Intel 2.6 beats an AMD 2600XP.

Yup. The newer P4 Northwood performs a lot better than the old core, and of course having a 533MHz FSB doesn't hurt it either.

What would you like your benchmark to say? I can whip you something up that will say whatever you like. You want AMD to come out on top? Sure, no problem :) You prefer Intel? Just as easy.

Unless I'm very much mistaken the Whetstone benchmark is the best way of comparing pure CPU performance. Anything else is too system dependant.

heres a set of benchmarks that seem fairer :)

http://www6.tomshardware.com/cpu/02q4/021001/index.html

they seem pretty equal in these ones

I think you're looking at the AMD Athlon XP2800+ there. The 2600+ is definitely the lesser one when compared to the P4 2800MHz in these benchmarks as well.

i know its the XP2800+, thats why its fairer

I presume the first review was written before the XP2800+ was released. As such, the XP2600+ was AMD's fastest processor, whereas Intel's was the 2800MHz version.

Comparing the top models of the two market leaders is fair - if one is considerably slower than the other, the other company is simply doing better :)

Please everyone buy these chips, so that when I want to buy one in 8 months or so I don't have to pay nearly as much.

* a XP 2600 isn't the same as a P4 2600 *

Btw - would (example) 2 XP 1300 chips run just as fast as a single XP 2600 chip ?

You'd need MP not XP. And the answer is no, not under any normal conditions.

Generally a multiprocessor system will be slower in a straight line than a faster single processor system, but won't slow down so much under load.

If you take combustion engines as an analogy - a sportscar is your single processor and a lorry is your multiprocessor system. Generally you wouldn't expect a lorry to do 200mph, just as you wouldn't tow 10 tonnes up a hill with a Ferrari.

great analogy Mis :) - I understand it now :)

what would you rather have - speed or load ?

depends on what u need

if its for gaming, go for speed

if its a server of some kind, who has connections all the time go for load

yep, its a great analogy :D

further more, using your example 2x XP1300+ will never be as fast as a single XP2600+ in tasks that take advantage of multiprocessing. This is because its not always possible to evenly split the work between the two processors.

2x usually yields around 75% of the 1x equivalent

Miserableman is teh win.

Also, if you take an excellent SMP OS like Solaris 8/9, typically you can expect to get between 125% and 150% from a dual proc box, compared to a single proc box (being, obviously, 100%).
Most applications (and a heck of a lot of server applications) are written with no SMP in mind. 99.99% of games take no advantage of SMP. The reason why SMP is sovery popular on servers is that a lot of servers run a unix variant. Unix has the philosophy of "lots of small programmesbounce cleanup error flush lmtp local master nqmgr pickup pipe qmgr qmqpd showq smtp smtpd spawn trivial-rewrite virtual, all very good at one thing only", so a single application, say an email server, will have one process that accepts incomming mail, one process that checks whether its valid, one process that sends it to the MDA... etc

A counter example would be MS exchange server, which is one application that handles the sending, receiving, storing, filtering, sorting of mail, all functions that are taken care of by seperate processes under *nix. The hint is in the names, for email you need an MDA, an MTA and a MUA. Typically on unix, each of these tasks (possibly with the exception of the MUA, which is a users perogative anyways) are several different processes. For example, the popular MTA Postfix consists of 18 applications*, most of which are not daemons at all. Therefore, postfix will run exceedingly well on multi proc boxes (lots of small processes will get assigned and spread evenly over the processors available)

GUESS WHO HAS BEEN READING TANENBAUN'S MODERN OPERATING SYSTEMS?

* For those geeky enough to care, bounce, cleanup, error, flush, lmtp, local, master, nqmgr, pickup, pipe, qmgr, qmqpd, showq, smtp, smtpd, spawn, trivial-rewrite and virtual

there's a time and place for everything MT, but for once, you took it TOO FAR - k.