AMD Duron vs. Intel Celeron

DeadBugs writes: "With all the hype surrounding the new Athlon XP and P4 2.2 GHz, the more affordable processors have been ignored. Tech-Report has a great article comparing the new AMD Duron and Intel Celeron. Both are now running at 1.2 GHz and have upgraded cache. The new Duron contains XP technology, while the Celeron is a PIII Tulatin with a 100MHz bus and built on the .13 micron process."

Comments

Talk about extending the timetable

[Wells2k]

From what I remember of the Intel roadmap around 1998, Intel was supposed to have come out with a 100Mhz Celeron chipset sometime in the year 2000. Obviously this didn't happen on time...

Re:Talk about extending the timetable

[Anonymous Coward]

You're right, the first Celeron with 100 MHz FSB was the Celeron2 800 which was announced January 2001, missing 2000 by a couple days. That was only a year ago so the news probably hasn't trickled down to /. yet.

Re:Talk about extending the timetable

[questionlp]

Part of the problem is that Intel is being plagued by supply issues and they don't want their "crippled" or previous generation processors beating out their newer, more expensive processors. Remember that in some cases, the Tualatin Pentium III beats the Pentium 4 processor while having a lower clockspeed and lower heat dissipation.

I wish Intel wouldn't have cut off the Tualatin P3 so quickly, as it would make a decent dual processor system... but now I'll be getting a dual Athlon instead :)

Re:Talk about extending the timetable

[mjed]

In 1998 (or maybe in 1997, not positive), Intel alrady released a 200Mhz Pentium chip. It's big selling point was MMX. Intel promised that MMX was "the future of computing". Remember those days? I sure do miss them.

Re:Talk about extending the timetable

[GigsVT]

MMX was 1996 or 1995 man. I know because I bought a pentium 200 non-MMX in late 1996, and it was very cheap because non-MMX chips were "old technology" by then, and 166 and 200MMX were the standard midrange in late 1996.

In early 1997 the Pentium II 233 and 266 came out. The celeron didn't exist yet.

Re:Talk about extending the timetable

[questionlp]

And when the Celery^W Celery did come out, it was a castrated bugger... no L2 cache. I have no clue what Intel was thinking when they released that thing. At least second revision was better, mainly since the cache was running full speed and was on-die (128KB though, but even then it still beat the Pentium II in some cases).

Re:Talk about extending the timetable

[b0r0din]

As has been said numerous times, Intel got hurt by failing to go with the DDR standard, the benchmark still proves the Duron to be a bit better but considering they are running at the exact same clock speed, you would almost expect the Duron to outperform. Then again, Intel is doing little more than taking PIII cores and calling them Celerons. Smart move to cut manufacturing and R&D costs, but it kind of reminds me of this old cartoon.

Personally, I prefer AMD to Intel. Their K2 and Athlons have helped stave off a monopoly and they drive costs low. Works for me.

Re:Talk about extending the timetable

[Yottabyte84]

Celerons ran 100MHz starting with the 800, I think that WAS released in Dec 2000, but i may be wrong.

Re:Talk about extending the timetable

[steveha]

I really want to make a dual Athlon system, but I am waiting for the 0.13 micron process Athlons. I'd like to have a system that doesn't need roaring loud fans, and two CPU chips that dissipate about 60 Watts each is a bit much heat!

I was wavering, almost considering breaking my "no Intel CPUs" rule, because the 0.13 micron version of the Pentium III is sweet. But Intel reminded me why I have that rule -- the 0.13 micron Pentium III was deliberately made incompatible with Socket 370 motherboards. I hate it when companies play those sort of stupid games.

When the 0.13 micron chips come out, they will crush the Pentium 4. Right now the best Athlon is neck and neck with the best Pentium 4, and the Pentium 4 has the benefit of a 0.13 micron process (i.e. a much higher clock rate).

steveha

Re:Talk about extending the timetable

[Toraz Chryx]

the P166 MMX and P200 MMX were released on Jan 7th, 1997

IIRC

Re:Talk about extending the timetable

[jsoderba]

...the 0.13 micron Pentium III was deliberately made incompatible with Socket 370 motherboards.

Kind of how most older KT133A boards can't run Athlon XPs? New manufacturing processes almost always mean lower voltages.

Re:Talk about extending the timetable

[Toraz Chryx]

the Tualatin doesn't just run at a lower voltage, intel changed the pinout too..

Re:Talk about extending the timetable

[GigsVT]

I'm thinking the OEMs had the chips before the "official" release.

Tualatin "Celery"

[questionlp]

The Celeron is also crippled by the poor FPU that hasn't really changed since the Pentium II came out. The only reason why I would buy a Celeron-based computer is if heat and noise are not tolerated, beyond that, even a slower Athlon or the Duron would be the processor of choice (both for people on a budget or for people who crave speed).

Re:Tualatin "Celery"

[flight666]

This is incorrect. The latest Celerons are based on the PIII core and have the exact same FPU as the PIII.

Re:Tualatin "Celery"

[questionlp]

But the P3 still uses the same x87 FPU units as the Pentium II processor. The P3 adds SSE (yeah, like that is used a lot) and brings the cache on-die with the Copper[less]mine processors.

The Celeron reviewed actually uses the Tualatin P3 core, not the older P3 core.

Re:Tualatin "Celery"

[Sivar]

Intel chips' FPUs (Floating-Point units for those not acronymnically inclined) have been inferior since the release of the first Athlon. Then the P4 core comes around and its SLOWER than that of the P3.
Of course, its clockspeed is higher so I guess that's all that matters to Joe Sixpack.

(0, Troll)

[Anonymous Coward]

Tom's Hardware did a review on this a couple of weeks ago: http://www.tomshardware.com/cpu/02q1/020103/index. html

As you can see, the Celeron is actually at 1300 MHz, not 1200. Funny thing is the Duron still beats it by a good deal.

hmm

[RainbowSix]

Unfortunately, AMD apparently isn't ready to move the Duron to a 266MHz bus just yet. That's really a pity, but AMD wants to differentiate between the Athlon and Duron

They're not ready because to put the Duron and Athlon at the same bus speed would make their performance levels nearly equal. With the hardware prefetch and SSE we've already seen the 1 gig duron keeping up with the 200mhz fsb 1 gig Athlons. To put the cheaper Duron at 266 would give little incentive to buy an Athlon of the same grade (save for the cache).

Re:hmm

[Spazntwich]

"(save for the cache)"

I'm sacrificing my ability to mod on this thread to reply to you, so I hope you're happy. You are doing a great misdeed to discount the importance of cache so greatly.

Duron: L1 Cache: 128kbytes L2 Cache: 64kbytes

Athlon: L1 Cache: 128kbytes L2 Cache: 256kbytes

A difference in cache sizes of this magnitude will ensure (insure?) the Athlon a victory if it is running at a clock speed the same as a Duron. Don't just assume that they're both as fast as one another simply based on bus speed.

Re:hmm

[mgblst]

Yes, but how much by? Not that much. I think that when Intel released the Celerons, they relised that the Cache doesn't make that big a deal. So they deliberatly kept it behind in clock speed. AMD know this as well.

Re:hmm

[mountain_penguin]

I'm also sacrificing my ability to mod on this thread to reply to you, so I hope you're happy
as well
I sit here on my dual duron 1.1 gig machine. Yes the L2 cache is smaller but this machine is faster than a 4 way ultrasparc 450 which has 8 Meg of cache per processors. Cache coherency and pre fetchm make more of a difference than you think. It is certanly faster than the 1.1 gig athlon i had cos i was sent the wrong chips. Put in SMP form the duron kicks arss esp considering the price difference between a duron and MP or XP has to be doubled if u a re using a dual proc machine

Re:hmm

[MonkeyBoy]

I think that when Intel released the Celerons, they relised that the Cache doesn't make that big a deal.

Really? The original Celerons had zero L2 cache. It took the "A" series, released a few months later, to add L2 cache to the package, running at full processor speed no less. There were less than a handful of name-brand systems released that were based on that original Celeron, due to horrible performance.

Cache size matters, as does the speed of the cache. A Coppermine P3 has no problem keeping up with an older P3, despite having half the cache size. So size matters, but speed (and I know I'm oversimplifying and ignoring bus lines, latency, etc) does too.

This is why taking a Celeron 300A and overclocking it to 450Mhz (changing the 66Mhz FSB to 100Mhz FSB) was just a big thing. Sure, you only had 128K of L2 cache, but it was running so damn fast that the difference 90% of the time was negligible. Just needed to keep that sucker cool...

I think the idea is that AMD needs to keep a nice wide margin between the cut-throat (Duron) and the profitable (Athlon) chips. The slimmer the margin, the less money they make, and in this economy...

Re:hmm

[software_god]

All these sacrifices... So many people giving, kinda brings a tear to my eye.

leading zeros

[klund]

Celeron is a PIII Tulatin with a 100MHz bus and built on the .13 micron process

By the way, that's "0.13 microns."

As my Nobel-Laureate physics lab professor used to say, "ALWAYS use leading zeros with decimal points; that way your readers can tell the difference between a fraction and fly shit."

Go ahead and mod me down, but I'm not a grammar nazi, I'm a math nazi!

Re:leading zeros

[Wavicle]

I assume that was important when the only way to get stuff was on printed paper and there could be ambiguity between a decimal point and an artifact (or an excrement). If you have a problem with fly shit on your monitor, you should clean it. I regularly clean my monitor with a soft lint-free cloth and a 10M solution of HCl.

Re:leading zeros

[xah]

Again, the word "nazi" here is used in a positive sense. I find that regrettable. The Nazis did not become best known for their strictness, but for their acts of overt evil.

For all numerical values greater than or equal to zero, and less than or equal to one, the numerical value is used in the singular sense. Thus, "0.13 micron" is the proper English usage.

One could list a few values as: "Zero micron, 0.5 micron, 1 micron, 1.5 microns, 2 microns," etc.

Re:leading zeros

[Alsee]

&GT the .13 micron process
By the way, that's "0.13 microns."
Go ahead and mod me down, but I'm not a grammar nazi, I'm a math nazi!

No, it's 0.13 micron.

Go ahead and mod me down, but I'm not a grammar nazi or a math nazi, I'm a "that just sounds stupid" nazi!

Go ahead, just *try* saying it out loud: "0.13 microns process". That 's' rolls off the tongue like an anvil.

-

Re:leading zeros

[nikoftime]

I assume he meant molar, not million. Don't assume it's a different quantification just because you've never taken basic chemistry courses.

Re:leading zeros

[Saeger]

You know, using toilet paper and H2O works just as well if you're willing to put a tiny bit more effort into the cleaning (and as a bonus, you'll appear less anal retentive at the same time)! :)

--

Re:leading zeros

[ackthpt]

As my Nobel-Laureate physics lab professor used to say, "ALWAYS use leading zeros with decimal points; that way your readers can tell the difference between a fraction and fly shit."

Marginal on-topicness, but I have this coupon in my pocket for .55c off something. Like you say, is that 55/100 of a Cent or do they mean 55 cents? :)

Re:leading zeros

[DeadBugs]

That's Ok my Slashdot profile now says "Deadbugs has submitted 1 stories"

Although I believe the spellchecker allies defeated the grammar nazis several years ago

Re:leading zeros

[PurpleFloyd]

10M HCl? Dear lord, you must have some serious pitting in the plastic parts of your monitor! I might suggest Windex or a similar ammonia-based glass cleaner.

Re:leading zeros

[Wavicle]

Dear lord, you must have some serious pitting in the plastic parts of your monitor!

Yeah, but there's no fly shit on it. :)

Re:leading zeros

[sam_handelman]

I regularly clean my monitor with a soft lint-free cloth and a 10M solution of HCl.

I believe you can find enlightenment by embracing the tangential.

Chlorine gas (which HCl releases in small amounts) is bad for your eyes. 10M HCl also burns cheap plastics, and isn't especially good at disolving oily residue. If you have metal deposits on your moniter, HCl is the way to go, otherwise, no.

The secret moniter cleaning solution only lab chemists new about (until now)- 50% water, 50% acetonitrile. Unlike cheap malt liquor (which I used to clean my moniter with - seriously) it doesn't leave a stink or a funny residue.

Two things to remember - only use it in a well ventilated area (which you should have for your computer anyway, in case it starts putting off Ozone) and try not to spill it on yourself, it permeabilises your skin (although not as much as DMSO.)

Re:leading zeros

[ShoeHead]

A Nobel Prize winner for a lab instructor? Where are you going? I get the bottom of the barrel for my TA's...

Re:leading zeros

[elmegil]

And if you pronounce 0.13 correctly as "13 hundredths" it's obvious that it's "13 hundredths of a micron" not "of a microns".

Re:leading zeros

[shokk]

*sigh*

Like most people, you are attacking the symptoms and not the problem.

I see a curtain of fly paper surrounding every computer as a good fix. No more fly shit on monitors. Off to ETrade to find a fly paper manufacturer to invest in!

Re:leading zeros

[EnderWiggnz]

well.. at least you arent using HF...

that would just be silly...

Re:leading zeros

[EnderWiggnz]

the real question there is - COO-pon or Que-pon.

Re:leading zeros

[Kalabajoui]

I like your sig, but you should give Pravin Lal the proper credit for it.

Re:leading zeros

[tmarzolf]

I assume that was important when the only way to get stuff was on printed paper and there could be ambiguity between a decimal point and an artifact (or an excrement). If you have a problem with fly shit on your monitor, you should clean it. I regularly clean my monitor with a soft lint-free cloth and a 10M solution of HCl.

Scrolling works well too!

Re:leading zeros

[Detritus]

I don't think he is going to complain.

Re:leading zeros

[mrfiddlehead]

More than silly, without the proper gear you'd be dead before the day were out.

Nasty nasty nasty stuff that HF (hydrogen flouride, in case you were wondering).

Re:leading zeros

[DrSkwid]

ahhhhh

yes you can't beat vinegar & newsprint

Re:leading zeros

[fredrik70]

...it permeabilises your skin
um, not sure what it means but it sounds a bit nasty, really....

Re:leading zeros

[Kalabajoui]

That's what you think! You should have heard him
going on and on about being plagerized. I finaly
had to post on the topic to get him to shut up.

Re:leading zeros

[tanpiover2]

And the little dude with the hearts and arrows is "COO-pid"?

Re:leading zeros

[sam_handelman]

...it permeabilises your skin
um, not sure what it means but it sounds a bit nasty, really....

There are some substances that cannot pass through your skin and into your blood (in appreciable amounts), but if you dissolve them in acetonitrile they pass through your skin in much greater amounts. The effect is even more striking with DMSO (dimethyl sulfoxide) - if you dissolve methanol (wood alchohol) in DMSO, and poor it on rats, the rats go blind.

It has been suggested that you might use a stream of DMSO as a delivery system for chemical agents (a death squirt gun), but it's only barely feasible (DMSO is volatile, and fairly expensive, and if the gun leaks on you you die...) so, really, dart guns are a clearly superior technology.

Re:leading zeros

[bugg]

Hydrochloric acid releases small abounts of chlorine gas? Under what mechanism? I would imagine that any measurable chlorine gas that you'd find in a bottle of HCl was in solution from when the HCl was manufactured.

I'm not a Chemistry whiz, but my old textbook says that the energy of reduction for 2 chlorine ions to form chlorine gas (Cl2) and 2 electrons is -1.36 volts- so I don't see why this reaction would occur spontaneously.

I won't argue with you that HCl vapor is bad for you, but it's certainly nowhere near as deadly as chlorine gas.

So, am I forgetting any of my chemistry?

Re:leading zeros

[sam_handelman]

I'm not a Chemistry whiz, but my old textbook says that the energy of reduction for 2 chlorine ions to form chlorine gas (Cl2) and 2 electrons is -1.36 volts- so I don't see why this reaction would occur spontaneously.

I won't argue with you that HCl vapor is bad for you, but it's certainly nowhere near as deadly as chlorine gas.

The question of whether or not a reaction proceeds spontaneously also depends on the concentrations of the reactants/products. Even with 10M HCl, the equilibrium concentration of Cl2 is really small - but it isn't zero. Actually, the Cl2 is fairly likely to break down into two Cl-s when it hits your screen (cathode ray tube = electron gun, after all.)

Anyway, I've been told by other people in the lab's I've worked in that it's the tiny amounts of chlorine gas from the HCl that irritate your eyes, and not the HCl gas, itself. For one thing, the HCl is driven into Cl2 because I don't think HCl is very volatile, but Cl2 sure is (so it floats away and can't turn back into HCl.) I don't know what would accept the electrons (probably water, to make 2 OH- and 1 H2) if you look up the voltage for that half cell and subtract it from 1.36 volts, you can calculate the equilibrium concentrations. Keep in mind that 10M HCl is about 50% water, and that the "starting" OH-, Cl2 and H2 concentrations (the products) are all three of them basically zero.

Re:leading zeros

[shokk]

Pravin Lal is from the future. Where do you think he got it from?

Just so you know...

[Daniel Wood]

Tom's Hardware already did a review of the Celeron 1300 vs the Duron 1200 (benchmarks included the Celeron 1200) where the Duron simply spanks the Celeron.

SIS has restored my faith in AMD. The ECS K75SA motherboard is only $64 after shipping and works with any Socketed Athlon/Duron cpu. It is fast and stable, accepts DDR and SDR, built in networking and sound(ok, AC'97 isn't that great), a real winner. You can build a 1GHz system and only pay $120 for the cpu, heatsink/fan, and mobo.

Re:Just so you know...

[MtViewGuy]

Tom's Hardware already did a review [tomshardware.com] of the Celeron 1300 vs the Duron 1200 (benchmarks included the Celeron 1200) where the Duron simply spanks the Celeron.

The test shows why the Celeron is inferior to the Duron: the Duron's vastly superior FPU unit allows it to substantially outrun the Celeron on FPU-intensive tasks. That is the reason why the Duron has become the choice for many do-it-yourself computer builders.

Re:Just so you know...

[GigsVT]

Speaking of SiS and ECS....

I just got in an ECS k7SEM with onboard everything, and a duron 950. All the onboard stuff is well supported in Linux. You had better have a recent Xserver to handle the onboard video though, the one that ships with Red Hat 7.2 works well (4.1.0). The sound card gets a little flaky under heavy processor load (sometimes XMMS won't be able to open the sound if it changes tracks while the processor is loaded heavily), but it sounds great.

So I got this setup working well. I ordered 6 more of them to build a MOSIX cluster with.
from www.mwave.com:
6-ECS K7SEM motherboards
6-950 Mhz Durons
1-16 port 10/100 switch

Total w/shipping: about $880

from www.sofistic.com:
6-128 meg Micron DIMMs PC133
6-el cheapo cases with 300 watt powersupplies

Total w/shipping: $337 (watch out they rape you on shipping, but their prices are so low it offsets it)

Anyway, so I am building a 6 node supercomputer for $1200. This is what a low end PC used to cost. Boy we have come a long way.

There will be some other costs, like there will need to be a hard disk somewhere for these things to boot from, but no other major costs.

Re:Just so you know...

[Daniel Wood]

Speaking of cheap systems, does anyone know how overclockable the new Duron is? I'm almost tempted to buy one and another K7S5A and bump the FSB up to 133. A 1600MHz DuronXP(might as well call it that) should compete nicely with an Athlon XP 1900+, and at little more than 1/3 the price.

Re:Just so you know...

[debrain]

"works with any Socketed Athlon/Duron cpu"
You never tried the Athlon 1.4 Ghz T-Bird with the K7S5a.On a message board with, on average, 44 posts per topic, there were 14,000 posts on the Athlon 1.4 + K7S5a. Someone did solve the problem, that being total system instability, by putting a 200 ohm resistor in parallel with something underneath the chip (soldered onto the motherboard), but I wasn't brave enough for this and settled with upgrading to an Athlon XP which works fine. Strangely enough, this issue really only reared itself en masse with revision 4 of the board, which constituted the most shipped by far. Revisions 1-3 were flakey, and oddly enough revision 0 was rock solid, from what I read (so this is hearsay), and I stopped paying attention by the time revision 5 was out.

Re:Just so you know...

[Geek In Training]

The ECS K75SA motherboard is only $64

I must also chime in as a fan of this board. I run it in my gaming rig with an Athlon XP 1600, and 512 megs of DDR. It whomps ass! Onboard 100baseT and ATA100, 4x AGP, and the SIS 735 chipset requires no fan. I got mine for $57 at newegg.com, whom I highly recommend for parts (this is an unsolicited testimonial for an independent party :).

Also, if you look at chipset reviews, the SIS735 comes in JUST behind the high-end Via chipsets, at many $$ less.

Yes, I put an Audigy in and disabled the onboard sound, but the AC97 is very workable if you're running a single pair of speakers or headphones.

Just my $0.02. (Note the leading zero.)

Re:Just so you know...

[DeadBugs]

Yeah I read the article on Tom's like 5 minutes after I posted this article. Tech-Report gave more detail and I liked the pure 1200 vs. 1200. But the main topic was to point out that there are other new processors out there that don't cost Between $350 & $600

Re:Just so you know...

[keefebert]

I have also had very good results with newegg.com, as did my place of employment, if any one cares.

Re:Just so you know...

[larsl]

The K7S5A is a basket case in Linux. Lots of problems with DMA on the IDE controller that lkml still hasn't figured out yet. trouble more problems

Re:Just so you know...

[Tim Doran]

An earlier post might be of use here...

Benchmark woes

[Cerebris]

A decent review, I suppose.

I think it was a tad unfair to compare a Duron using DDR to a Celeron on PC100/133 (depends on the motherboard and how they set it up). They did acknowledge it directly when discussing the memory bandwidth (which showed the expected numbers, Duron was around 2x Celeron), but I think it shows only part of the picture (especially with DDR prices back up in the stratosphere compared to say, a month or two ago). This is one reason I take benchmarks with a grain of salt...it's very difficult to objectively compare AMD and Intel CPU's now due to the drastically different architectures.

The article also mentioned the Intel headspreaders...these should be reflexive on all processors. I can't count how many "Cracked core" thread's I've read on the [H]ardOCP forums...and a reasonable number of these guys are shall we say slightly above your average user.

My $0.02...

-Colin

Re:Benchmark woes

[Jay Bratcher]

it's very difficult to objectively compare AMD and Intel CPU's now due to the drastically different architectures.

The primary difference is actually the chipset, and not the chip itself (at least in the case of DDR vs. SDR memory). That being said, look back at comparisons between any DDR and SDR system - memory bandwidth is dramatically increased, but real world performance rarely improves more than marginally. What I am trying to say is, it's fair to compare the 2 chips, especailly since they run the same software.

Re:Benchmark woes

[SoftwareJanitor]

DDR prices back up in the stratosphere compared to say, a month or two ago

When did DDR prices go up? I just looked at the price list of one of the local parts stores (www.laboratorycomputers.com), and a 128M DDR is $38 vs. $31 for a 128M PC133. That is only a $7 difference. The difference is percentage wise the same on 256M modules, which are $69 and $55 respectively. The difference is percentage wise less on 512M modules, which are $138 and $122 respectively for a difference of only $16.

Probably the biggest difference in price if you go with DDR is in motherboards which are about $35 or $40 more for boards with DDR support.

Re:Benchmark woes

[Spazntwich]

Difficult to objectively compare the processors? You're just an intel whore talking out his ass!

In order to compare the processors objectively, all you need is to run the processors on testbeds as close to one another as possible, and when you can't have a certain piece of hardware crossover (i.e. motherboard), you insert the best piece of hardware available for that platform.

Perhaps you haven't noticed, but running a Celeron on a DDR system doesn't give it any benefit. Wonder why? It's max speed for it's front side bus is 133mhz SDR.
Don't whine about non-objective reviews if they ran each CPU on the best test-bed they could.

Re:Benchmark woes

[gordon_schumway]

On 29.October, I ordered 256MB DDR DIMMs from Crucial for $35.09 each.

Around Christmas they were $70-75.

Today they are $97.19.

Re:Benchmark woes

[dbarclay10]

I would agree with you, if the review was actually a comparison of technologies.

Of course, if that were the case, then it wouldn't be a review - it would be a comparison of technologies :)

However, since it *is* a consumer-oriented review, the focus is obviously on performance vs. price and a number of other factors; all easily summed up in the term "value".

Since both the Duron and the Celeron have similar prices and are both targetted at the same market(at least in retail), then it's totally fair to compare them, despite the fact that they have some relatively different technologies.

Now, I would say it's unfair to compare, say, an Athlon XP 2000+ to a 386 used in "embedded" markets. This review, however, is more than fair :) Saying it's otherwise would be like saying it's unfair to compair the first- and second-place winners in the Olympic men's triathlon; yes, obviously one is faster than the other. Maybe they've got more endurance(greater memory bandwidth), maybe their muscles are bigger(stronger FP units), but if you're not going to compare those two, what else are you going to compare? The winning triathlon athlete vs. the winning 100m swimmer? :)

Thought so ;)

Re:Benchmark woes

[talonyx]

This review, however, is more than fair :) Saying it's otherwise would be like saying it's unfair to compair the first- and second-place winners in the Olympic men's triathlon; yes, obviously one is faster than the other. Maybe they've got more endurance(greater memory bandwidth), maybe their muscles are bigger(stronger FP units), but if you're not going to compare those two, what else are you going to compare? The winning triathlon athlete vs. the winning 100m swimmer?

Thanks. The sports analogy helped me understand this computer-related subject a great deal better.

not.

Re:Benchmark woes

[cdipierr]

Yes, but PC133 is up as well, so it's not as if DDR has increased while PC133 stayed low, it's just that memory in general is higher.

duh, duron will win hands down!

[Anonymous Coward]

faster bus speed, faster instruction execution, faster floating point, there is now way it can't.

Celerons are a better choice over PIII

[AnimeFreak]

(This starts off as a bit offtopic, but I felt like posting this)
Price.

Yes, you get that 1.2 second advantage in Photoshop if you have a Pentium III 900, but I could care less if I have a Celeron over Pentium III.

It also goes for games too, like I care if Quake III Arena runs 20 frames faster than on a Celeron, all the computer has to do is meet 30 frames per second. The human eye is not fast enough to see 100 frames per second. (Yes I know video cards play a factor, but the processor does do a fair bit of work).

All I care about is if the processor can run applications at a decent speed.

Also, is it just me or does the name "Duron" not sound very catchy? "Pentium," "Celeron," and "Athlon" all have catchy names. Yet for some reason, "Duron" doesn't sound all that great and sounds like a processor from the waste bucket.

I know of a few people who own them and actually don't have anything negative to say about them.

Re:Celerons are a better choice over PIII

[nexex]

As I see it, if your game is runnning at only 30 fps, when ever the cpu has to allocate more time to other things be it AI, networking, cron jobs, etc, your fps will drop below and you will see the difference. Granted more ram will help eliminate this, but it's still something to consider I think. Also, perhaps it is just me, but I can see a difference between 30 fps and say 60-70 fps. It just looks smoother. You can also feel it as you play, it stays smooth during the whole game; perhaps that is what I think looks smoother...

PS - Everytime I hear Celeron, I think of celery, when I hear duron, I think of durable. :)

Re:Celerons are a better choice over PIII

[bunhed]

Also, is it just me or does the name "Duron" not sound very catchy?

The Duron is actually the middle processor of the "Moron" line of precessors, designed specifically for XP. If you look in the start menu on XP and you're running a Duron you'll see the My Moron icon. The "Freon" is the fasted at 3.2 GHz but requires a gallon of coolant every 100 web pages. The "Peon" is the entry level processor. You can read the whole article here

Re:Celerons are a better choice over PIII

[questionlp]

I usually call the Celeron, "Celery"... but when I see Duron, I think of Durian. Don't know if it's because I eat durian or some other sub-concious thing :)

Re:Celerons are a better choice over PIII

[Wavicle]

Out of curiousity, how do you know that the human eye cannot see 100fps? When you say that, what do you mean? Cannot make out detail? The flash from a camera usually lasts less than 10ms, but we see it.

I ask this because I once believed as you did and created a simple experiment to prove I was right. I took an old manual camera, removed the lens, set the shutter to 1/60 sec (its flash sync), opened up the back, put my eye up to the shutter curtain, pointed the camera at my wife, asked her to hold up some fingers but not to tell me how many, and snapped the shutter.

To my somewhat surprise, I had no trouble telling how many fingers she was holding up. So I started clicking the shutter up to higher speeds. Only at 1/1000 sec did I start having trouble.

Thus I had proved myself, and conventional wisdom, wrong. If you are getting 100fps, and you monitor is drawing 100ftps, if something important happens even for only a frame or two there is no reason your brain won't register it.

Re:Celerons are a better choice over PIII

[Simon Garlick]

> all the computer has to do is meet 30 frames per second. The human eye is not fast enough to see 100 frames per second.

Tell me, can you discern a difference between your display at a refresh rate of 60Hz and a refresh rate of 100Hz? I thought so.

Re:Celerons are a better choice over PIII

[root2]

Hmm, to a non-physicist, that seems to make sense.

Think about it. A still image on the TV is actually 30 (29.97 for NTSC, actually) identical images being shown per second. When one says that the human eye can only see 30 FPS, what you're saying is that if the one of the images was different, we'd be able to see that.

Now, if that's the limit, then we can only detect an image if it lasts longer than 1/30 of a second. So if an image lasts for only 1/100000 of a second, we wouldn't be able to see it.

I'm sure this is wrong but would appreciate someone EXPLAINING why.

Re:Celerons are a better choice over PIII

[jelle]

I suppose what you're seeing in the short-duration flash is similar to an echo in sound.

Re:Celerons are a better choice over PIII

[CmdrSanity]

Your experiment seems biased. One of the eye's primary functions is to detect differences (motion) from one scene to the next. Motion detection is one of the lowest level functions subconsciously performed by the eye. So if you have an experiment where one scene is flat black and the next is an image, you aren't testing in a situation that is on par with reality. Instead, you'll need to run a test \w fluid images, then your wife's fingers, then more fluid images. So try editing a movie reel and inserting images randomly into the photo cells. If you run the film through a projector faster than 24fps, the average human eye will not detect inserted images. Instead, your eye will try to blend the oddball image into the two surrounding images.

Pardon me if I got any of the neuroscience wrong on that description, but I think the general idea is correct.

Cheers

Re:Celerons are a better choice over PIII

[aspeer]

What I find rarely mentioned is that the new Celerons (1.2Ghz and above) should generate less heat than their AMD counterparts because they are based on a 0.13 micron process.

As far as I am aware the AMD products are based on a 0.18 micron process.

Heat was an important consideration for myself, as my emphasis was on having a *quiet* PC, so I did not want to have a million fans running at X,000 RPM just to keep the CPU from becoming a red blob on the motherboard.

Re:Celerons are a better choice over PIII

[TheOnlyCoolTim]

30 fps is not the greatest the human eye can percieve.

Try this: Set your monitor to a 60 hz vertical refresh rate. Now your monitor is making 60 "frames" a second. (That is how fast it will redraw the screen.) Then stand back from your computer and look at a point off to the side of the monitor. You should notice it is flickering madly. Some people get eyestrain if they use a monitor that is running at 60hz or from cheap flourescent lights that flicker at 60hz.

Personally, I find I can no longer notice the flicker once I set it to 85 Hz, so I try to aim for 60-85 fps in my games as a tradeoff between speed and quality. Some people, however, find they can notice a flicker in 85hz and have to go up to around 120 hz or so before they feel comfortable.

Re:Celerons are a better choice over PIII

[jamp]

There IS a difference between a monitor "flickering" at 60Hz, and a game running at 25-30fps.

At 60Hz, your eye effectively sees the scanning of the electron gun on the CRT - thus the flashing. Running at higher refresh rates makes this more difficult to notice.

However, when a game is running at 30fps, the images on the screen are only changing 30 times per second, but you'd not notice them changing. The images appear on the screen more consistently at 85Hz (i.e. no flicker), but you can't tell if there's a higher frame rate than about 30fps.

Jamie.

relative cost

[Anonymous Coward]

Do these low cost cpu's matter anymore? Celerons do not stand for value.

In the mid 90's sure it was a huge cost difference $100 for that celery 300a@450mhz vs p2-p3 450 at about $500.

As of right now celeron ghz is about $58
( http://www.pricewatch.com )
AMD XP 1500 $107

Thats the battle, now I'll give you 3 guesses which is a better value.

Re:relative cost

[SoftwareJanitor]

I don't know how a Celeron 1.2GHz can be discounted that much, as it is just over 1/2 the street price locally (Austin, TX) while the AMX XP 1.5GHz price you quote is not that much lower than the local street price. The pricing at a local place (www.laboratorycomputers.com) looks like this:

Celeron 1.2GHz $115

AMD XP 1.5GHz $133

For comparison:

Duron 1.2GHz $99

TBird 1.2GHz $112

If Intel is trying to compete with AMD, it sure looks like a no-brainer choice to go with AMD. The only question is which AMD is the best value.

On the other hand:

P3 Tualitan 1.2GHz $273

So if you have to have "Intel Inside" and you want a "1.2GHz computer", then the Celeron looks like a good deal in comparison to the P3.

Re:relative cost

[(H)elix1]

Another factor is the motherboard.... you want to use the 1.2G Celeron? You need to have a board that supports it. A slocket adapter, even a fancy powerleap adapter, will not save my lovely SuperMicro SBU (slot 1, BX, SCSI) motherboard. I hate that!

Price vs Performance, its really hard to beat the Durons. My mom is getting a new box - found a 1G duron from newegg.com for $54USD, and have it tracking here as we chat from a vendor who has been good to me in the past. Looking at an intel solution, it was an extra $20 to get something that performed like a $30 AMD chip. No brainer - though my mom is fine now with my old 486dx2/50.... Still, my cash, a guy has to have standards...

The price jump to the latest greatest is way too steep right now to look at the 1.2 duron, but when the 1.3 comes out this month I suspect this might drop into that great deal catagory too.

Re:relative cost

[CtrlPhreak]

Parent Post: AMD XP 1500 $107

Above reply: AMD XP 1.5GHz $133

I'm very glad to see that AMD's marketing strategy is working just as planned.

Re:relative cost

[shokk]

They're not cheap enough for you *now*?

Athlon XP is about $140 for something decent.

Soyo mobo to support it is about the same.

Memory: picked up 768MB SDRAM from Crucial for about $100.

Hard drive: 30GB for $100 months ago.

Case, CDROM, floppy and the rest are practically free.

Even a 19" .24mm KDS monitor is about $199.

They're all commodities and we should be able to pick up a new fast game machine for $400 next year. No wonder the computer industry is having trouble making money. They're still following the same price plan from the dot-bomb days. If they slow down on the innovation for just a second, they might be able to swing two quarters with the same pricing and product instead of killing themselves with this single-quarter pricing schedule.

Re:relative cost

[wallsg]

Somebody mod this up!

I wish I could but I used up my moderator points a couple of days ago.

I think this AMD XP 1500 thing shows very well the intellectual dishonesty of AMD. In the very name of their processor they intentionally attempt to deceive twice: first by trying to link themselves to Windows XP, and second by misleading the consumer on the actual clock rating of the processor (the same tactic that failed the losers at Cyrix, BTW).

I use both Intel and AMD processors but I doubt I'll buy another AMD while they continue the attempts to deceive like this.

Re:just short the L1 bridges for 266mhz bus

[Magila]

Unfortunatly the pencil trick will not work for XP processors (including the new Duron).

http://www.vr-zone.com/guides/AMD/AthlonXPUnlock/

XP Technology?

[nexex]

If the duron comes with XP technology, does that mean there is a complementary holographic image of Bill Gates on the top of the cpu? Or does it mean you have to "activate" it and if you switch mobo's you have to call AMD to have it unlocked?

Re:XP Technology?

[Spazntwich]

I think the XP technology in it just means it's extra easy to break the darn things. Ever shattered an AMD cpu putting on a heatsink? It's not pretty.

Contains XP Technology!

[Proud Geek]

That really impresses me. I mean, they could shrink the die size, or use SOI, or even borrow a cue from the Alpha and use SMT. But no, they went way beyond all the rumours and used XP Technology. I wonder if they had to license it from Microsoft?

Re:Contains XP Technology!

[b0rken]

The "XP" technology refers at least to the incorporation of the SSE instruction set---my 1GHz duron has that. I think there are other non-marketing differences between "XP" and older K7s, but find someone more technical than me to explain them.

Celeron

[ehiris]

Intel is not advertising XP comaptibility of their Celeron.

Celerons do work really good with XP from my previous experience but they didn't post it on their compatibility list.

Celeron Compatibility: Fully compatible with an entire library of PC software based on operating systems such as MS-DOS*, Windows* 3.1, Windows for Workgroups* 3.11, Windows 98, Windows 95, OS/2*, UnixWare*, SCO UNIX*, Windows NT, Windows 2000, OPENSTEP*, and Sun Solaris*.

Re:Celeron

[Anonymous Pancake]

it says 'such as' these operating systems, not 'ONLY' these operating systems. Celerons work fine with linux too which isn't listed there.

Re:just short the L1 bridges for 266mhz bus

[strictnein]

Wonder if that vr-zone.com site could have more pop-ups.

Blah...

Why bother?

[jerkychew]

A Duron 1.2GHZ costs $79, while a 1.2GHZ Athlon costs $72... someone explain to me why the Duron is the 'budget' CPU.

Re:Why bother?

[vipw]

i just wonder which one of those performs faster in most tasks. the duron has prefetching, but the tbird has more cache. more on topic though, the duron consumes less power and outputs much less heat than the 1.2GHz thunderbird, lowering the price of heatsink/fan unit and decreasing the need for high wattage power supply and case cooling.

Re:Why bother?

[fireant]

vipw mentioned it, but to be explicit...

The new Duron is based on the Morgan core (think Athlon XP), while the 1.2 GHz Athlon is a Thunderbird, which came out at the same time as the old Spitfire core for the Duron (about a year and a half ago). I suspect that the T-bird would still beat the new Duron due to a bus speed advantage as well as the extra cache, so your point is still valid.

Why get either?

[Sloppy]

For the price of a 1200 MHz Duron, you can get a 1000 to 1100 MHz Athlon that will run rings around it. These "value" processors don't really look like much of a value to me.

yes... very cheap!

[moosesocks]

When you go down 200mhz to a still-respectable 800mhz, the price plumets down to $40ish... from a decent vendor. Then you get potential. You can build a $250 countertop pc. Which is exactly what I did...

Add $50ish for the suprisingly nice ECS motherboard (for $50, it's a GREAT deal. it's nice on quality and features. i would gladly pay twice that price for the same board) and $10 for a heatsink (amd-approved, of course).
Add another $50 for a Geforce2MX. The MX isn't a horrible card, it is sufficent for running most games...
Add (whatever) for RAM... it's SOOO cheap now. I just dug up 256mb from my closet full-o-parts. You can get that for about $30 from crucial
Add about $30 for an AMD approved psu (i can't stress the approval more. it really makes a diffrence!)
Then i added a keyboard and mouse (i got a refurbished wireless keyboard and mouse for $25. most people can find a cheapo keyboard and mouse that is more than sufficent).
I used a hard drive from an existing pc. It's fine for my purposes.

Volia! For about $250, i have a NICE countertop pc, which is hooked to the hi-fi as a jukebox(hey, that's what it was advertised as when it was purchased... oh 20 years ago :) ).

Now, granted, I used some parts i already had, which isn't fair to the price factor. Take those off, and you go up to about $400 (case+ram+hdd). There are many people who can't afford a pc due to the astounding prices. To dell and friends, i say poo!

This says a lot about the beuracocracy behind dell and all the big pc makers. Once must wonder the profit margin dell is enjoying (remember their little deal with intel... surely they are sold chips at dirt cheap prices)?

Re:yes... very cheap!

[Spazntwich]

"Add (whatever) for RAM... it's SOOO cheap now. I just dug up 256mb from my closet full-o-parts. You can get that for about $30 from crucial"

Checked the crucial website recently? 256 megs of pc2100 is up to ~72$. Check Here

Duron Price

[56ksucks]

You know, if you look on pricewatch.com, the Duron is only $5-$10 cheaper than an Athlon thunderbird of the same clock speed. Some $20. Might be a good reason why they get ignored sometimes.

The celeron might be slower...

[cosmicg]

The celeron might be slower, but it beats the PII 400 I've used it to replace. I just upgraded to the celeron on my 3.5 year old Dell. $170 buys the chip and slotkit.
Because Intel is still producing inferior chips with slow bus speeds, I can play Black & White. Part of the fun of tech advances, is the way they pull up the rear, while dropping prices.

Software Optimization

[DarkEdgeX]

The P3 adds SSE (yeah, like that is used a lot) and brings the cache on-die with the Copper[less]mine processors.

Actually, this brings up an important issue-- compiler technology, and the run-time libraries (RTL's) they use (in the case of C/C++, the standard libraries, in the case of Pascal/Delphi, the RTL and possibly parts of Borland's VCL/CLX). The problem, it seems to me, is that compiler authors don't seem to take advantage of architecture specific improvements like they used to (and as they should). Sure, some libraries/RTL's take advantage of it (and the compiler may have switches to emit optimized code), but if the standard libraries/RTL's are re-compiled (or even re-written) to take advantage of it, then it's all for nothing.

It seems to me that Intel has the right idea (the FPU is really useless if you know HOW to use SSE and SSE2 properly), and that if anything, it's poor software authors and poor compiler writers that are to blame for the lackluster performance of code on Intel's CPUs. It's saddening to me to see the optimization skills software engineers *used* to have back in the day diminishing year by year as the ability to right crappy code is justified by ever-faster CPU's. (Why spend the weeks or months needed to engineer everything to run properly now, when Intel/AMD will have a 'fix' for our sloppy code out in a few months?)

I wish authors such as Michael Abrash still released optimization guides for assembly language (or even just updated versions for C/C++ and assembler).. his 'Zen of Code Optimization' (ISBN: 1-883577-03-9 *or* FatBrain.com's description (out of print)) was probably the best investment *I* ever made.

Re:Software Optimization

[Grishnakh]

it's poor software authors and poor compiler writers that are to blame for the lackluster performance of code on Intel's CPUs.

I think this one can be blamed on Intel too. If Intel really wanted to sell its processors, they'd invest a little money in helping push compiler improvements to take advantage of their processors--such as contributing to GCC. Instead, they do invest money in compiler technology, but only in their own proprietary compiler, and then try to sell that as competition for the other two mainsteam, more popular compilers that everyone uses (GCC and MS VC++). Then they wonder why software isn't optimized for their processors.

The compiler authors don't have time to make processor-specific optimizations for every single flavor of x86 architecture out there; they already have to deal with P5, MMX, P6, K6, 3D-now, etc. Why is Intel's newest fad so special that they should get extra attention? It's not. Compiler authors are going to write their compilers to perform the best on the majority of processors out there, instead of concentrating too much on one specific technology.

Re:Software Optimization

[NonSequor]

If you are designing a processor that is backwards compatable to a 10yr-old line of chips, removing something like the FPU and expecting software to change is a bit hipocritical, if you ask me.

Actually, it's even worse than that. Here's a quote from the Intel Architecture Software Developer's Manual:

The developments leading to the Intel Architecture can be traced back through the 8085 and 8080 microprocessors to the 4004 microprocessor (the first microprocessor, designed by Intel in 1969). However, the first actual processor in the Intel Architecture family is the 8086, quickly followed by a more cost effective version for smaller systems, the 8088. The object code programs created for these processors starting in 1978 will still execute on the latest members of the Intel Architecture family.

So the Pentium 4 is a descendent of the first microprocessor ever created and the core of the x86 instruction set is over 20 years old.

Re:Software Optimization

[egeorge]

Sure, some libraries/RTL's take advantage of it (and the compiler may have switches to emit optimized code), but if the standard libraries/RTL's are re-compiled (or even re-written) to take advantage of it, then it's all for nothing.

I think this is one argument in favor of the dynamic compilation in the latest generation of Java VMs (Oracle 9i and Sun's HotSpot).

With this technology, you could swap out the optimization logic in the VM without recompiling the libraries (they are in bytecode). In theory (although I haven't heard about anyone offering this), you could load a processor-specific optimization package when you installed the VM and get all of the advantages of your processor without having to worry about what version of the libraries you were using.

These dynamic compilers are not without tradeoffs, but I do think it solves this type of hardware variability fairly well.

Re:Software Optimization

[Joe U]

Much to my suprise, Microsoft came out with a processor pack upgrade to VC6 to support the newer processor extensions.

Good luck finding it, it's almost never mentioned and it's buried in the VC website which is currently too busy hyping Visual Studio.net

Re:Software Optimization

[LarsWestergren]

It's saddening to me to see the optimization skills software engineers *used* to have back in the day diminishing year by year as the ability to right crappy code is justified by ever-faster CPU's.

Well, its just a matter of economics. In the beginning of computing, you could get maybe 50 programmers for the price of one computer. So the time of the computer was valued more than the time of the programmers. The programmers had to spend a lot of time optimizing.

Now, the price of computers has fallen, until you get a lot of computing power for the price of one programmer. The time of programmers is valued a *lot* more than the time of computers. So the rational economic choice is to buy more (or more powerful computers) to make life easier for programmers.

When you look at how some of the old time programmers react to this change, I think it is insteresting to look at the medieval guilds. The programmers are angry at the newcomers and try to put them down, and make it as hard as possible for them to advance ("RTFM!"). The guilds on the other hand made it forbidden by law to practice the craft in question outside the guild. Both guilds and programmers occasionally justify their behaviour by the need to preserve the fine traditions of the art, and distrust new techniques and technologies that make things "too easy".

But what it really is, is a fear of competition. Instead of trying to improve themselves and keep up with the times they try to stomp out the competition.

This may cost me some karma...

Re:Software Optimization

[DarkEdgeX]

Uh, actually it's copyrighted 1994-- that makes it six or seven years old. Strangely, except for suggesting that people use REP MOVSxxx, it's fairly timely considering that the Pentium processor is covered (and later iterations of the Pentium line used almost the exact same core). Hence why I also suggested that it might be nice if the author put out an updated book, taking into consideration current processor technology (the Pentium 4 being the first big change since the Pentium Pro).

As for optimizing skills coming naturally, that's a bad assumption to make. You'd be surprised how many VB coders couldn't optimize their way out of a paper bag. (And really, I've seen some badly written C/C++ code as well...)

Re:Software Optimization

[DarkEdgeX]

Why is Intel's newest fad so special that they should get extra attention? It's not.

Oh, but I think it is. Any new technology that can give you a 10:1 performance improvement is worth ANY persons time to investigate. That can make the difference between an MPEG2 video stream compiling in 10 hours or 4 hours.

You seem to forget the whole point of optimizing the compiler is to make the compiled code perform faster and better (and possibly with fewer instructions, thus decreasing bloat). Now sure, SSE and SSE2 (and really, even MMX) require more thought from the actual developer as well, but implementing these enhancements in standard libraries and in commonly used functions would surely help people realize some of these benefits up-front with a simple re-compile of their code.

There's also your issue of considering it a "fad". This technology is here to stay, and Intel has pretty much said that using the FPU is the wrong way to go about things in their newer CPU's. By this very virtue, one could pretty much surmise out of the gate that if a test or benchmark uses *only* the FPU, Intel will likely fail, and fail miserably. Intel has really only introduced three new technologies since the release of the Pentium--

MMX (MultiMedia eXtensions) - SIMD (single instruction multiple data; basically performing the same operation repeatedly on a set of data, by only executing a single instruction) integer routines, useful in certain video functions, and in some string functions. Introduced in some later Pentium processors (not in the Pentium Pro, I believe, not sure though).

SSE (Streaming SIMD Extensions) - SIMD again, but applied to floating point work. Introduced in the Pentium III.

SSE2 (Streaming SIMD Extensions 2) - More SIMD, again applied to floating point (I believe, I've only begun to read the info on this from Intel's developer website (developer.intel.com has tons of manuals in PDF format you can read on the subject though). Introduced in the Pentium 4.

Only three "fads", as you call them. Other than this, very few instructions have been added that a compiler team need worry about; the only ones I can even think of worth dealing with in a compiler are the CMOVcc instructions (and with these, it depends on the amount of time you save using these instructions; I assume they're an improvement though since it saves you the trouble of branching in your code).

BTW: The compiler issue is really more of a Wintel based problem (afterall, the benchmarks being run are usually on Windows based systems).. take a look at Borland's compilers for example; Borland Delphi offers little optimization configurations. Borland CBuilder offers more options, but as far as I recall, no ability to compile with MMX/SSE enhancements. As someone else pointed out, you can in fact purchase Intel's compiler technology as a plugin for Visual Studio, but Microsoft should *really* be adding these features directly to their compiler.

And the latest optimization issue (which I don't believe can be addressed through compiler changes, but who knows) will likely be when Intel releases SMT-capable processors.. suddenly everyone running Windows XP Professional (or if MS decides to allow dual processors on WXP Home Edition) will be able to run multiple threads at the same time-- the problem is, not a lot of software utilizes multiple threads, even though the technology has been around since Windows '95 (obviously in a single processor environment, you don't get a speed increase, but it does make your app more responsive (the UI doesn't freeze during long operations, for example)). Benchmarking an AMD processor without any form of SMT against an SMT capable Intel processor would be unfair at that point.

Re:Software Optimization

[Lurks]

The compiler authors don't have time to make processor-specific optimizations for every single flavor of x86 architecture out there;

Some compiler authors do not. My company treats processor-specific optimizations as a core feature especially when it comes to SIMD implementations. Intel C/C++ 5.0 does some vectorization for Intel CPU features. Our compiler, VectorC, is designed to vectorize real-world media operations for all CPUs on the market today.

CPU specific binaries come out of the compiler essentially for free. One can either use libaries or seperate executables for the equipped CPU. In our view these still far too much generic compilation of performance sensitive code.

AMD's recent domination of performance to price ratio has also made some unlikely new friends. Namely that of distributed Linux-based high performance and scientific computing. Formerly Intel based platforms with the Intel compiler, now there's another option to consider - AMD based machines with our compiler. Especially when we have C++ at any rate.

Re:Software Optimization

[TRoLLaXoR]

Nice post.

FYI the Pentium Pro was the first 686 core, while the 586 (Pentium) core was the first to include MMX. The Pro lacked MMX.

The Pentium II, in regards to your post, was basically the 686 core with MMX.

The Pentium III was a 686 core with MMX and SSE (also known as KNI, or "Katmai New Instructions" when it was first released).

ANd now supposedly the Pentium 4 is a 786-class core, the first true microarchitecural change to the Intel x86 line since the Pentium Pro.

Re:Software Optimization

[delta407]

SMT -- "simultaneous multi-threading" (or more commonly SMP, symmetric multi-processing) -- is achieved by having two CPUs executing different instruction streams. This is great for things like database servers (possibly handle multiple requests simultaneously), but for anyone running XP Home Edition, there would be no increase in speed.

"will likely be when Intel releases SMT-capable processors" -- yeah, like the Xeon line. In addition, most P3s and even some earlier Celerons can be placed into a dual-processor environment.

As far as "the technology has been around since Windows '95", if you're talking about SMP, no, it hasn't. SMP has been around since NT 4 (possibly 3.5, I don't know) and above (in the NT line) -- while you can have two CPUs running under '95, the kernel doesn't know what do, so one will run a whole lot cooler than the other :)

"but it does make your app more responsive" -- assuming it's written to take advantage of multithreaded operation, in which case it would [often] be more responsive than a single-threaded version. "the UI doesn't freeze during long operations", which again assumes that the long operation takes place in a separate thread, which would keep the main message pump running just as it would under a single-CPU system.

In addition, introducing another processor means that both of them have to keep from stepping on each other's toes -- the two shouldn't try to write to the same region of RAM simultaneously, for example. This extra locking makes SMP-enabled kernels slower for a single thread than non-SMP kernels.

The bottom line is that adding an extra processor won't help the average Windows XP user.

Re:Software Optimization

[moogla]

...considering that the Pentium processor is covered (and later iterations of the Pentium line used almost the exact same core).

No. The transition from P5 to P6 (Pentium Pro) is the introduction of an entirely new core.

The P5 core had two instruction pipes (thank god), the U and V pipes. I remember he discussed how you can optimize assembly for this in one of the chapters. Basically, you would schedule instructions that didn't step on each other's feet one after the other. However, the P6 core (Pentium Pro-Pentium III) is quite different. It has a RISC core and a variable (chip dependant) number of ALUs and FPUs. It is nowhere nearly as predictable as the P5, thus optimizing it is a tricky, black art. None of the old optimization tricks work, instead you basically follow guidelines for emitting code that won't break the longish pipeline.

Re:Software Optimization

[DarkEdgeX]

SMT -- "simultaneous multi-threading" (or more commonly SMP, symmetric multi-processing) -- is achieved by having two CPUs executing different instruction streams. This is great for things like database servers (possibly handle multiple requests simultaneously), but for anyone running XP Home Edition, there would be no increase in speed.

Which is why most likely, and I thought I made this clear in my first post, that Microsoft would need to release a service pack to enable dual processor configs in Windows XP Home (and likely allow quad processor configs in WXP Pro, and move the formula up for their other releases of XP/.NET Server).

As far as "the technology has been around since Windows '95", if you're talking about SMP, no, it hasn't. SMP has been around since NT 4 (possibly 3.5, I don't know) and above (in the NT line) -- while you can have two CPUs running under '95, the kernel doesn't know what do, so one will run a whole lot cooler than the other :)

I wasn't referring to SMP.. I was referring to multi-threading. Windows '95 was (AFAIK, unless Win32S had this under Windows 3.1) the first MS OS to have a function such as CreateThread (and the other thread-related support routines needed to manage threads). Multi-threading *HAS* been possible, and with the advent of SMT technology multithreaded applications will suddenly perform better without any effort on the part of the developer (as far as the OS is concerned, it's as if you have two CPU's installed).. this was the angle I was trying to present. =)

"but it does make your app more responsive" -- assuming it's written to take advantage of multithreaded operation, in which case it would [often] be more responsive than a single-threaded version. "the UI doesn't freeze during long operations", which again assumes that the long operation takes place in a separate thread, which would keep the main message pump running just as it would under a single-CPU system.

That's what I was assuming. =) As for the last bit about the message pump-- you mis-state how Windows handles multiple threads, it doesn't use messages to do this. The kernel is responsible for allocating slices of time to each active thread, and thread priorities are taken into account during said allocation (hence the 'idle' thread priority, for things working in the background, and the 'real-time' priority for things that MUST respond immediatly to input, for example).

In addition, introducing another processor means that both of them have to keep from stepping on each other's toes -- the two shouldn't try to write to the same region of RAM simultaneously, for example. This extra locking makes SMP-enabled kernels slower for a single thread than non-SMP kernels.

This is true, but only in complex situations where you need one thread to talk with or cooperate with another thread.. thankfully the Win32 API has this covered through Semaphores and Critical Sections. (Lookup CreateSemaphore or CreateCriticalSection in the Win32 API.) I've never done multi-threaded programming under Linux, so I'm not sure what facilities are available in an SMP/SMT environment, but I'd imagine similar functions are there as well (and if not, if you think about it, critical sections are *really* easy to implement in user-code).

The bottom line is that adding an extra processor won't help the average Windows XP user.

No, the bottom line is that with properly written applications, an extra processor WOULD help the average Windows XP user by being 1) more responsive and 2) unable to lock the system in the event of an endless loop being entered (or during a long operation involving one thread). (This I've experienced first-hand on my dual processor Pentium III system-- sometimes I've found bugs in shareware/freeware that involves the app entering a seemingly endless loop-- this endless loop consumes all the CPU resources; but only for one of the two processors. =) The other one is still unused and allows me to kill the errant application. In a single processor system, using the whole CPU could lead to the mouse being unresponsive and the keyboard not working as expected (if the errant code were running in the context of a 'real-time' priority thread/application). Specifically however, SMT *WOULD* help the average Windows XP user while being cheaper than buying two processors outright. As far as I understand it, SMT isn't the exact same as having two processors, but it DOES allow two threads (two execution streams) to execute at the same time when certain parts of the CPU are going unused (which, during typical UI refreshing, is probably quite likely). That's my take on it, again this assumes that Microsoft realizes people will be livid that their SMT enabled CPU's aren't being used in WXP Home, and releases a service pack to fix the issue.

Re:Software Optimization

[delta407]

> you mis-state how Windows handles multiple
> threads, it doesn't use messages to do this.

Not ususally, though separate threads may have message pumps (if you *really* wanted to, but I think thats silly). As far as my statement about the long process and the UI, here is what I was trying to convey:

1. Main thread -- contains message pump and handles the UI
2. Worker thread(s) -- executes long processes, separate from main thread

Scheduling prioritites aside, if a long operation (say, blocking I/O) happens in the thread that has the message pump, the UI will become unresponsive.

> This is true, but only in complex situations
> where you need one thread to talk with or
> cooperate with another thread.

The kernel has to ensure it does not break itself (hence, the locking), which creates overhead, regardless if two user threads are being silly or not.

> more responsive

Do you think that CPU power is the major bottleneck for the average XP user?

> unable to lock the system in the event
> of an endless loop

No, it can still do this; just have the endless loop running in two threads :) Seriously, though -- one program cannot (normally) lock up the computer by placing lengthy operations in a single thread; and there really is not a point to making a thread realtime for anything other than device drivers. (It doesnt go faster, it will just receive the CPU more often; if most of it is idle [which for the average desktop user it is] no performance is gained). Sure, CPU time may be 100%, but the scheduler *does* preempt that thread and give others their fair share. Granted, that does nothing to help poorly-written programs that blow up parts of the kernel on occasion, but...

In any case, as standard desktop applications (such as word processing or e-mail) do not contain much, if any, parallel code, and while having an extra processor to redraw all those icons in the system tray sounds nifty, it gives little increase to performance. Again, how much CPU usage do you think writing a letter (at 20 WPM) produces?

Re:Software Optimization

[delta407]

Oh, one other thing:

> Which is why most likely, and I thought I made
> this clear in my first post, that Microsoft
> would need to release a service pack to enable
> dual processor configs in Windows XP Home (and
> likely allow quad processor configs in WXP
> Pro, and move the formula up for their other
> releases of XP/.NET Server).

NT4 workstation and Win2k Pro both support 2 processors. Server (IIRC) supports up to 4, Advanced Server supports up to 8, and Datacenter supports up to 32. No fuss, no muss. Out-of-the-box SMP support.

Also, it is not just a matter of enabling dual-processor configurations; it is a matter of modifying single-processor-specific sections of the kernel, as well as (again) adding locking in the appropriate places.

I can think of no reason that Microsoft would change their plan now instead of before the release date, and I find the idea of Microsoft releasing a service pack to add functionality that was taken out for the release of XP Home unlikely. if you want dual-processor support, buy Pro. (They get more money that way, you see.)

Re:Software Optimization

[DarkEdgeX]

Scheduling prioritites aside, if a long operation (say, blocking I/O) happens in the thread that has the message pump, the UI will become unresponsive.

Which is as good a reason as any to design future and current applications as multi-threaded; breaking it up so applications can remain responsive even during blocking-events. (By blocking, one instantly thinks of Winsock API calls, though this could easily apply to anything using Semaphores or Critical Sections.)

The kernel has to ensure it does not break itself (hence, the locking), which creates overhead, regardless if two user threads are being silly or not.

I might be wrong, but Windows XP (and 2000/NT) have a pluggable kernel model-- ie: different USER32.DLL's and so forth are loaded depending on if you're running an SMP (and likely SMT) or single processor environment. I assume Microsoft takes care to avoid using bus-locking instructions in the single-processor environment simply due to the lack of any need.

Do you think that CPU power is the major bottleneck for the average XP user?

I think the lack of an ability to execute multiple threads is a bottle-neck, yes-- CPU power isn't the issue, for a word processor a 200 MHz processor is really just about all you need (or ideally, to make everything respond quickly, a 450 MHz CPU should suffice). However, the issue I'm discussing (and have been trying to explain) is RESPONSIVENESS-- an SMT enabled processor running at 200 MHz would be more responsive than a 450 MHz processor that didn't have SMT (with properly multi-threaded applications). The end user cares about one thing-- how fast the computer responds to their commands. When the UI doesn't refresh (or worse, is overwritten by other UI elements and doesn't refresh it's own), the user believes (rightly) that the application isn't responding. (Hell, even Task Manager uses this as a method of telling if an app is hung-- when an app isn't responding to messages, it's considered dead to the OS.)

In any case, as standard desktop applications (such as word processing or e-mail) do not contain much, if any, parallel code, and while having an extra processor to redraw all those icons in the system tray sounds nifty, it gives little increase to performance. Again, how much CPU usage do you think writing a letter (at 20 WPM) produces?

Microsoft Word is multi-threaded, actually. When you print, you can immediatly go back to working on your document, instead of waiting for it to finish the print job (by finish, I mean the act of spooling it into the print queue-- not the entire act of printing through the print driver). In a small 1 or 2 page document, you'll never notice or appreciate this, but when you're printing a 50-100 page document, you will. Word also uses multiple threads to spell-check and grammar check as you go. There are also certain graphics applications (NOT Adobe Photoshop-- Adobe has lackluster support for multiple threads) which take advantage of threading, applying filters to one document while allowing you to continue to work on other open documents. I'm not saying end users need 5 GHz processors, I'm saying they need a cheap way to have dual processor systems (and perhaps as SMT is rolled out, Intel will manage to allow more threads to execute at the same time (think 4, then 8, etc) all on one CPU) to make the system both more responsive and hence, more user friendly. And back to my original point, we need more software writers to realize the advantages of multiple threads and take advantage of it NOW, so when these processors arive we aren't waiting for software to be written to take advantage of them.

Re:Software Optimization

[DarkEdgeX]

NT4 workstation and Win2k Pro both support 2 processors. Server (IIRC) supports up to 4, Advanced Server supports up to 8, and Datacenter supports up to 32. No fuss, no muss. Out-of-the-box SMP support.

You're telling me things I already know. =)

Also, it is not just a matter of enabling dual-processor configurations; it is a matter of modifying single-processor-specific sections of the kernel, as well as (again) adding locking in the appropriate places.

Which Microsoft has already done, hence the different drivers in Device Manager depending on if you're running a dual CPU system or a single CPU system (or a quad, eight-way and so forth).

I can think of no reason that Microsoft would change their plan now instead of before the release date, and I find the idea of Microsoft releasing a service pack to add functionality that was taken out for the release of XP Home unlikely. if you want dual-processor support, buy Pro. (They get more money that way, you see.)

You can't think of one because I already GAVE YOU one. WXP (and really, 2000/NT) were released prior to Intel releasing CPU's with SMT capabilities. As SMT CPU's come into the consumer level (likely in the next year or so) people will wonder why their system isn't benefitting from these new features-- Microsoft will likely bump the Home Edition to dual processors, the Pro version to 4 processors and so on and so forth through Server and Advanced Server to Datacenter Server. WXP Home still won't have the ability to connect to an NT domain or Active Directory, which is part of the reason businesses avoid WXP Home Edition.

Re:Duron vs Celeron: the utlimate test

[stressky]

Yeah, but you didn't see what happened later on in the night outside the bar, where Duron called on his big pal SledgeHammer to dish out some "rough justice" :-)

Source?

[hendridm]

Where are you getting your prices from? Pricewatch shows a different story...

After much debate, Apple has released

[bryan1945]

a new processor licensed from Motorolla. In the vein of the Duron and Celeron, Apple has dubbed its chip the Dodecadon. It will feature a spherical chip package with one wing.

After a lukewarm reception, Apple changed the name to the "Celery", and sales went up 1400%.

(If you don't get the above [bad] joke, please "move along, nothing to see here!")

Re:After much debate, Apple has released

[analyst99]

ROFL! Just what I needed after a long day!

DW

Re:Athlons at the same speed cost 5 dollars more!!

[fmaxwell]

Athlons may be only $5 more for the same speed, but Athlon XPs are not. The older T-Bird Athlons are slower than the new Durons. And the new Athlons run MUCH cooler. So you can pay $5 more for a 1.2ghz T-bird and get a slower CPU that runs a lot hotter.

DDR prices

[Cerebris]

I'm referring to the trends I observe on Pricewatch (for bottom of the barrel) and Crucial.com (for "premium.").

-Colin

News???

[SquierStrat]

How is this news? The Duron has be kicking the Celeron's tail end since the Duron's release. The Duron can usually keep up with a P3 of the same speed, or at least trail very closely behind. The fact that they are both over 1ghz is not new either, not to mention that the so-called advances for the Celeron trully haven't helped it much, as it's an aging architecture. Please people, save money, get a better product from the AMD processors. (I'd still prefer a XP over a P4 personally, I'm yet to see a P4 that didn't give me this feeling that everything for some reason loads slower, though benchmarks seem to say it can be faster in some cases...yet it's only slightly faster than a much lower clocked Athlon XP, see a problem here?)

The problem

[stressky]

The problem with Athlon-esque CPUs is that the motherboards are still rather expensive (at least over here in Oz), thus although AMD CPUs are cheaper to purchse, you end up spending more on the motherboard anyway and the whole price advantage that the Duron has over the Celeron is negated. The price difference between the Athlon and Duron, last time I checked, was so small that the Duron doesn't really compete.

For this fact alone, if I had a choice between building a Duron box or a Celeron box, I'd choose to build an Athlon box instead :-)

Re:The problem

[stressky]

Just had a look at my favorite online computer gear store, and it seems that I'm wrong, there are a few well-priced Athlon mobo's out there...but I'd still rather go for a decent board that will set me back about 300 or so bucks...

re: the problem

[stressky]

Also turns out that the price difference between the Duron and Athlon CPUs has increased since I last cared to look (which isn't often, I'd rather go with the Athlon anyway...) and duron is now about 130 dollars cheaper...

Re:The problem

[netwerk]

heh, AUD or USD.

...tard.

Re:The problem

[titaniafq]

That would be 300 Australian Dollars or about 80 UK pounds.

Think then post please. Or at least read the htread!!! He says he is from Oz!

(tired)

Re:just short the L1 bridges for 266mhz bus

[roguerez]

No, the new Duron is still unlockable with the penciltrick, unlike the XP.

At least the 1 GHz Duron with Morgan core is: I have one and unlocked it this way.

Re:just short the L1 bridges for 266mhz bus

[shokk]

Adshield.org on WinXX. Lets you pick which sites are filtered so that you can still answer surveys at sites that do them through popups (CNN comes to mind) and still see pages at other sites without the bazillion popups. Sites like CNET tend to do very bad when you block because they seem to have tied everything to 3rd party cookies, but there are other sites to see besides theirs.

Intel has gone totally mad

[Anonymous Coward]

The new Celeron 1200mhz eludes me more than any product I've seen Intel release. Just recently I read www.tomshardware.com 's artice on celeron vs duron and although he doesn't touch on this subject at all... I can CLEARLY see in many of the benchmarks (sisoft cpu bench, as well as mp3 encoding speed) that the Celeron 1200 is indeed OUTPERFORMING the pentium 4 1400 and 1500mhz. Now is that silly? yes I think so. Do I find this terribly disturbing? yes. Why? Because enginners at intel seem to think higher MHZ is better than good cpu design. And Intels own marketing strategy is going to bite itself in the ass. Oops too late.

Re:Intel has gone totally mad

[RadioheadKid]

I know that there seems to be a lot of bashing of the Celeron and Intel's marketing, but in some ways I see it as a response to the market.

It used to be when you talked about a PC, you gave the specs of your hard drive, RAM, graphics adapter, whether or not it had a soundcard, and what number came before the 86 in the processor name 2,3, or 4.

Now having over 256 MB of RAM is not unreasonable. Hard drives size is mostly irrelevant, sound cards are standard, and except for the gamers, a graphics card is where you plug your monitor in and it works. So what's left to spec? MHz! It's a number, it sounds technical and the Wintel PC marketing machine has jumped right on it. So much to the point that AMD now puts 4 digit numbers in their processor model name that don't necessary represent the clock speed of the processor, but keep up with Intel's current MHz release.

Re:Intel has gone totally mad

[Lord of Caustic Soda]

Actually the Pentium 60 was quite a bit faster than the 486DX2/66, and I'd imagine you'd need some truly bad hardware to get the 486 to outperform the P60. But then the P60 is running the bus at 30MHz instead of the 33MHz of the DX2

I remember playing around with the alpha release of Quake and it ran okayish on my P60 and chokes on my friend's 486DX2/66, even though he got a video card that is a bit faster than mine.

I suppose the good thing about Duron is that you can upgrade the CPU later, whereas the Celeron means a new motherboard.

Re:Intel has gone totally mad

[f00zbll]

Good points on the cycles of product development, but I find it really annoying that rather than build systems with say 512mb of RAM and 1ghz processor as the basic system, manufacturers and reviewers buy into the whole speed BS. Frankly, how many people besides hardcore gamers need a 2ghz cpu? It's not like programmers will benefit a whole lot. How much faster do you need your code to compile?

Humans can't out type a 300mhz P3 with windows let alone a 2ghz system. Sure office runs slower with each new release and requires faster CPU, but more than anything double or triple the ram will make the system more usable. Now that graphic cards manufacturers are increasing the processing power, there's no point in a 2ghz cpu for a gaming system. You're better off getting a video card with 128mb of ram.

I like fast processors as much as the next geek, but it's stupid to push a CPU that is meant for the server market to joe smo. I would rather Intel and AMD focus on lower power consumption (which they do occasionally), heat dissapation and faster bus. Or how about shrinking P3 mobile even further to make laptops lighter and more efficient so that it doesn't need a heat sink.

Re:Intel has gone totally mad

[Kuad]

I remember playing around with the alpha release of Quake and it ran okayish on my P60 and chokes on my friend's 486DX2/66, even though he got a video card that is a bit faster than mine.

Integer-wise, the late 486-120s from AMD performed about on par with a Pentium-75. When you considered price, 486s still looked pretty good compared to Pentiums. Quake was the first true-3D game, and as such used heavy floating point math. The Pentium's FPU is an order of magnitude more powerful than that found in a 486. My old 486-160 wasn't a match for a Pentium 60 at heavy FP math.

Re:Intel has gone totally mad

[biohazard99]

Problem is, we are quickly coming up on the 4GB memory barrier for ia32, well before *hammer or itanium is ready for the desktop, that is one train wreck I do not want to see.

Re:Intel has gone totally mad

[f00zbll]

You're absolutely right we are coming up to the barrier, but do I really need 4GB of RAM to run windows and office? Oh I forgot I'm supposed to do what ever microsoft tells me. 4GB here I come :)

Re:Intel has gone totally mad

[Junta]

Actually, the P-60 runs at 60 MHz bus, no clock multiplier, or at least it does in m system, which is still running. It was the P-90 and P-75 models that used a multiplier, IIRc.

Re:Intel has gone totally mad

[Junta]

Actually, if AMD weren't around, I'd bet the P4 would not have such crappy performance and higher clock speeds.. One of the major factors in the P4 design was that the Athlon core was able to scale to 1.2 GHz and beyond, while the PIII core started to choke at 1.13 GHz, so to compete with AMD, they lengthened the pipeline and did other not-good stuff to try to pass AMD..

Re:Intel has gone totally mad

[Lord of Caustic Soda]

Sorry, I meant the PCI and VL bus...... I assume those would be run at 60/2 = 30MHz...

Dual Proc Systems

[Glonk]

You wanna see a sweet dual processor system? Check out the World's First Prestonia benchmarks. 2 CPUs (1.8GHz) with 512KB of full-speed L2 cache, and functioning SMT. The BIOS reports it as 4 CPUs.

The benchmarks speak for themselves. Zoom zoom. :)

Re:Dual Proc Systems

[Glonk]

I did submit, they rejected it.

It's far more interesting to read about Durons and Celerons than the first mainstream implementation of SMT, I guess.

ECS == PCCHIPS

[Anonymous Coward]

You do realize who ECS is, right? You can get started with some info from these scheisters here.

They are a distributor of PCChips, the same people who:

- Fake the speed of their motherboards
- Pirate their BIOSes
- Rewrote bioses to display fake cache amounts and glued black plastic squares to their board with bits of metal sticking out from them and etched SRAM looking part numbers onto em
- Create deceptively similar but not at all related to their brandname counterpart chipsets. VXpro, VX+, BXpro, etc. are all famous PCChips parts.
- Relabel chipsets that look/are too cheap with the above names
- Seem to be gaining a reputation on newsgroups for selling complete garbage motherboards (just look for "PCchips shit/junk/suck" on deja)
- Have an absolutely amazing number of aliases. Eurone, Houston Tech, ECS, Amptron, Protac, Aristo, Minstaple, Matsonic, Fugutech -- the list goes on and on.
- Don't label their boards properly
- Use huge jumper blocks to set memory type (no, not just 5 or 6, but jumper blocks the length of a DIMM slot!)

BTW: As a tech at a computer store I build systems with ECS motherboards (its what we sell -- no I won't say where I work :). They are unreliable garbage, IMHO.

But AMD approved power supplies are a good idea. I don't know how many times I've seen cheap power supplies with the ratings simply stuck on them with an extra sticker.

I'd really reccomend swapping the board, though. Just a reccomendation. But if it is performing alright for you right now, well, stick with it. Occasionally some are actually built ok! :)

BTW: If you are using the board I am thinking of, its also "made" by amptron in red. Or blue, I forget which colour was for which company sometimes. With their most popular boards, by selecting the manufacturer you can select the colour. F-U-N

When Tom's Hardware Guide gave a solid rating to a PCChips board (I think its the K7S5A) I decided I'd never trust one of his reccomendations again. If you are going to rate boards you should at least do some homework on the company first!

Re:ECS == PCCHIPS

[uebernewby]

I'm not proud of it, but I used to work for PCChips (or PC Shit as we used to call them), and I've heard it whispered a lot of times that most of what you say sounds familiar. In fact, I even have it on good authority a few senior staffers in their European distribution center were prosecuted for filing away serial numbers on CPU's, so they could sell them as higher end models than they really were.

The chipset naming issue (my absolute favorite example has to be the slogan "rember, it's BXtoo!") you bring up, however, isn't PC Chips' fault per se: a lot of these get made by shady Chinese plants and slapped on PCShit mobo's just because they're cheap.

reading THG?

[EnderWiggnz]

geezus ... anyone else blink at that acronym and remember (T)he (H)umble (G)uys?

wow. i'm old.

Speed differences unknown to consumer market

[orgnine]

The average consumer would assume that the Pentium chips are much faster as Intel has branded a 'fancy' new 2.2 GHz chip. Even that AMD chips *model names* only reach to +1900. (about 1.63 GHz).

Almost hilariously, AMD doesn't have to get their chips running at a 2.2 GHz frequency to get nearly the same performance.

The same speed differences per frequency show up in the lower bus-speed chips (Duron / Celeron).

The average consumer is completely unaware of the closeness between speed of the chips of each company.

AMD chips are much better priced, and carry more value for their money. Stability is excellent, speed is unmatchable in identical frequency ranges. It has been this way for a couple years now.

Aside, AMD has likely changed their naming system to make their chips 'sound' competitive compared to Intel chips? (i.e. Athlon XP 1600+ sure sounds like 1600 MHz doesn't it!).

orgnine

Re:Speed differences unknown to consumer market

[Junta]

This is only partially correct. True, the P4 architecture has poor performance per clock, enabling AMD offerings clocked at 1.67 GHz to be quite competitive with 2.2 GHz P4 offerings.

As far for the conservative model numbers, AMD wanted to play it safe and avoid heavy criticism by claiming more than they could back up completely. Better to underestimate yourself than overestimate, as reviewers tend to be much more friendly to companies that do this.

This does not, however, apply to the Duron vs. Celeron scenario. The Celeron is based on the PIII core, a core that acheives much more competitive performance per clock, but will not scale to the higher clock. The P4 architecture move intentionally sacrificed performance per clock to allow higher, thus more marketable, clock speeds. That is not to say I would go with Celeron, Durons are typically cheaper, but the XP vs. P4 discrepencies are not the same thing as Duron vs. Celeron.

Re:Speed differences unknown to consumer market

[Kanasta]

what I don't understand is why AMD decides to limit their chips to the performance of the top Intel chip.

If the tech exists to produce a 2.2GHz chip, then shouldn't AMD be able to produce a chip at that speed?

Re:Speed differences unknown to consumer market

[PeteABastard]

>what I don't understand is why AMD decides to limit their chips to the performance of the top Intel chip.

At the moment the Intel P4 chips are on a 0.13 micron core, while the AMD XP's are on a 0.18 I think. AMD probably want to maximise their profits from the current line before replacing it with a 0.13 line, which they are working on for the future.

Since AMD currently have the fastest chip they have no reason to rush to the 0.13 core, until they can no longer keep up with the P4's. This will eventually happen as the 0.13 micron P4 cores can potentially be used at much higher clock speeds, as they dissipates much less heat, and use much less power.

Comment removed

[account_deleted]

Comment removed based on user account deletion

I thought..

[tomblackwell]

K7S5A's have a reputation for being cheap, fast and generally hard to overclock.

Hey, learn a little chemistry.

[TheOnlyCoolTim]

10 M means 10 molar. It is used to measure the strength of the acid. If I remember right 10 molar means that there are 10 moles or 60.2 x 10^23 (a real big number) of HCl for in every liter. But I could have forgotten something somewhere...

It is also entirely possible to get HCl without taking it from your stomach. I don't remember how you get it though (as in what kind of chemistry to use, you can always buy it).

Re:Hey, learn a little chemistry.

[SmittyTheBold]

Or 6.02 x 10^23, if you want to be proper about it...

Scyld Beowulf

[lucidvein]

There will be some other costs, like there will need to be a hard disk somewhere for these things to boot from, but no other major costs.

Look into the Scyld Beowulf. All slave nodes can run diskless, booting from a floppy or CD. They pull the base system from the master. So you can walk around with the install CD (which contains the node booter) and bring up your slaves one by one. Nice setup and easy to run.

That is unless you're doing distributed storage... Then you'll need the hard disks, unless you have a lot of RAM, heh.

One thing about the K7s5a's

[TheOnlyCoolTim]

They are indeed excellent motherboards. There is however two warnings I will make:

1.) If you have a crappy case it might not fit. I am rather sure that it is the largest size under the ATX standard. Cheap cases, especially those made by OEMs like Dell and HP, might not be able to fit it. Hell, I've seen cheap OEM cases that can't even fit a mid sized ATX motherboard.

2.) This isn't really specific to that motherboard but from what I hear people who are trying to make an inexpensive computer (or going for Athlon with many gigahertz and a Geforce 3 but trying to cut corners elsewhere) will buy this inexpensive motherboard and then a cheap crappy power supply that can't handle having . The result is that it doesn't work in various ways. I will admit that this happened to me. I had a cheap 250 W from my old computer and I figured I might as well at least try it with ECS k7s5a, Athlon 1.4, and Geforce 3 before I went and spent money on a new power supply. It actually worked for about a week and a half with no stability problems or anything, but after that I came home to find the computer that I had left on was now off with the smell of burnt insulation coming from it. I opened the power supply and at least 4 components inside had burnt up. So I purchased a nice powerful 350 Watt and now everything is good.

Pricing

[Seehund]

What's really behind the price difference between the Athlon/Pentium and the Duron/Celeron? What is so much more expensive in the fabrication process of the "real" CPUs compared to their "value" counterparts? Actually, you'd think that the Durons/Celerons would be more expensive than the Athlons/Pentiums, since I presume that the first are made in smaller volumes. I'd also guess that the redesign of existing "real" CPUs into "value" versions would mean higher R&D costs than the common cranking-up of clock frequencies in the "real" CPUs.

Perhaps you can see that IANAHP (hardware person). I simply don't see a raison d'être for the "value" versions, other than for use in laptops.

Re:Pricing

[Peter McC]

The key manufacturing difference is the reduced amount of cache. This cuts down piles of space from the die, saving a decent amount of money. You can see this idea most clearly in the original Celerons, which had no cache at all. However, with no cache performance sucked ass, so they later rejigged the Celeron line to have some cache (Celeron 300A and later).

Peter.

the difference

[darweidu]

SO what's the big difference between the two companies' products? Although I read the occasional article on arstechnica, I have to admit I don't know much about processors.

Which would you recommend for gaming? I've always heard people say AMD is better price/performance for games, but it seems like intel is finally making progress with their new products.

what to do!? what do to...

http://www.alexchiu.com/affiliates/clickthru.cgi ?i d=dar

Re:just short the L1 bridges for 266mhz bus

[kilrogg]

It's in the preferences now "Advanced"->"Scripts & Windows"

Battle of the CHIPS

[System_level]

I feel there is lot of Hype on core clock speeds. With the Busses slowing down throughput. I see comparable results between 2 systems working on clock speed differences of 500 MHz.

Interesting final analysis

[Enrico Pulatzo]

The author points out that at 91 vs 86 dollars, the duron is better bang-for-your-buck. However, the duron uses DDR-SDRAM, while the Celeron uses good old cheap PC-100's. Cost vs Performance really needs to look at the entire package, not just the chips themselves.

Re:Interesting final analysis

[Mastagunna]

You can still use SDRAM, even with XP's you just have a faster option. Since when is giving the consumer choices something that should be frowned apon.

Ummmmmm

[Sycraft-fu]

The Intel Compiler isn't competition for MS VC++, it's a plugin. It just replaces the compiler and linker of VC++ with Intel's optimised one. It is well worth your money if you're oging to be doing serious development as it is just all around mroe efficient, even for Athlon chips.

micron? micrometer!

[vrt3]

Actually, it is 0.13 m (micrometer). Micron has since long been deprecated in favour of micrometer, part of our beloved SI system of units.

Voltage != Power!

[Erich]

Does anyone besides me find this type of thing annoying:

The Duron 1.2GHz requires 1.75 volts of power.

You would think that someone who wanted to be respected for writing a hardware review would at least try to remember the Voltage, Current, Resistance, and Power relationships that he or she learned in high school.

Your processor may only use a supply voltage of 1-2 V, but if you still need a heatsink and fan it's obviously using tens of watts of power.

Actually, one of the problems with modern processors is that power is going up in processors while supply voltage is going down. That means much more current. Around half of the pins on modern processors are for power and ground. You start to see wierd problems with metal migration on the power supply wires, ground bounce problems from wire inductance, and all sorts of other strange problems.

Wrong: XP 2000+

[Penguinoflight]

which costs $326 on pricwatch right now, given the fact that the athlon xp 1800+ whoops the p4 2ghz, the XP 2000 will beat the p4 2.2ghz, and the p4 2.2ghz costs $629 for the processor alone, plus double the ram cost, and more for a motherboard.

I'd rather have a XP 2000+ anyway.

the use than benchmarks

[john_uy]

i do not believe in benchmarks. they are not fair and inaccurate. systems being compared are not in equal specifications.

my other point is when you are using a celeron or a duron, do you feel any speed difference in using word processing, spreadsheet, presentation, e-mail, surfing? when running games, can you notice a 100fps from 95fps?

i am sure most will buy the faster cpu but there is not much difference with all the cpu unless you have specialized applications such as encoding, graphics rendering, and other professional applications.

so for a home user, a celeron 500 may do very well than a celeron 1.2ghz or a duron. just count the times the actual usage of your cpu, most of the time, it is idle.

Who cares about clock rate?

[Acy James Stapp]

"When the 0.13 micron chips come out, they will crush the Pentium 4. Right now the best Athlon is neck and neck with the best Pentium 4, and the Pentium 4 has the benefit of a 0.13 micron process (i.e. a much higher clock rate)."

Does the process really matter if they are "neck and neck"? If the best Athlon is as fast as the best P4, buy the damn Athlon.

OT: Another kick for the dead horse

[lostboy2]

>>> the .13 micron process

>> By the way, that's "0.13 microns."

> No, it's 0.13 micron.


One might argue that it's really "0.13-micron", but according to webster.commnet.edu, we may all be right.

I'm not a nazi, I'm just anal-retentive.

-- D.

Re:OT: Another kick for the dead horse

[Alsee]

I just read your link, and OUCH. Painful. That's why I can program in several computers languages, pretty much learn a new one in a few days, but I only speak one language*. (Yes, I am an IgnorantAmericanTM. God bless America. Because we need it. Just not quite as bad as everyone else)

I'll have to remember the pain of that webpage the next time I feel like grumbling about stupid compiler errors.

*I took a year of Spanish, but only speak it for vanishingly small values of speak.

-

Durian (OT)

[timothy]

Durian is one of the weirdest foods in the entire world. It's like an urban legend for a lot of people ("Yeah, there's this fruit which smells like garbage, and tastes sort of like a garlic pudding ..."). Bad name for a child, just because of the smell factor. If not for that, would sound good for a son or daughter.

I've had it, once. I enjoyed the *experience* but don't exactly have a hankering for it as a result.

timothy

Re:Hey, learn a little chemistry u 2.

[SmittyTheBold]

No, wrong. 6.02 x 10^23 is the number of molecules in one mole.

Anyhow, my comment was a jab about the incorrect usage of SI. You only you use a number between 1 and 10, then your exponent.

So, in trying to correct me, you were wrong in your formatting, and you were wrong mathematically by an order of magnitude. I'm sure you are glad you posted AC so you look less stupid...

Re:just short the L1 bridges for 266mhz bus

[amunter]

Nice. Thanks for pointing that out. I did not check the new options that were added with the new Mozilla build I downloaded, but this makes me like the browser even more.

Celeron, Duron, Moron,,,

[tmjva]

When is some Far Eastern manufacturer with
poor English (and marketing) skills going
to come out with a chip called the 'Moron'?

Don't you just love these intelligent comments?

[Zymurgy]

You know, I really honestly believe that Slashdot should stop allowing Anonymous Coward posts. Now, granted, all I needed to do was change my threshold, but I keep it at zero because I do like to see the AC posts.

OK, now it's looking like I've contradicted myself, but here's the deal: if we disallowed AC posts, the ones who have anything intelligent to say would just get a username. That's good anyway, right?

Now the question may be: what about the "kewl d00d" that posted the message I'm applying too; do we really want these morons all being official "Slashdotters?" Well, we need some kind of system (perhaps working off e-mail address) that would forever ban you from Slashdot if you ever try to create an account with a username like 1337pImPdAdDyOWNZj00!!! or pHaT_gAmEr_H4X0R as I'm sure our friend just couldn't resist doing.

Anyway, here's the reason I posted these thoughts in the first place. I don't think Slashdot should allow posts from people who regularly say things like:

• "My daddy just bought me, uh, I mean, I just got this pimp-daddy-yo new 'puter with and Athlon XP 2666+ and 2GB DDR so I can beat everyone on my online Quake 3 tournaments."
• "Oh, I mean, not Quake 3! That's just so last week, I mean, like, GAWD! I only play modern games!!!!!"
• "Ha ha! I'm better than you are because my computer's faster than yours is! Ha ha! And I'm also more of a 31337 H4X0R than you are! Ha ha!"