AMD Athlon MP 1800+ Processor Review

Lars Olsen writes: "Amdmb.com has posted a review of the new AMD Athlon MP 1800+ processor -- a big speed jump for the dual Athlon processor family with the new processor running at 1.53GHz. There are also 1600+ and 1500+ Athlon MPs available as well right away at stores around the World. Dual AMD Goodness is now running just as fast as its desktop counterpart ! Here's a quote: 'Those of you who want to jump into the dual processing Athlon world will finally be able to do so with the knowledge that your processors are the top speed that the Athlon family has to offer. And for anyone who already has a Tyan Thunder or Tiger MP board and a pair of Athlon MP processors, you may just want to pop a couple of these new Athlon MP 1800+ CPUs in your system to boost performance.'" Some of the comments following yesterday's "dream system" article addressed dual-Athlon complications, so make sure you read before you buy.Update: 10/15 15:14 GMT by T : Check below for LinuxHardware.org's take on this chip, and Athlon MP systems in general as well.

Augustus writes "LinuxHardware.org takes a look at the Athlon MP platform under Linux and the newly released Athlon MP 1800+ is included. Covered in this article is not only the technology and performance of the AMD-760 MP chipset and the Tyan Thunder K7 motherboard but we also look at why anyone would consider a multi-processor system."

Comments

Why bother ? its an excuse to write bad code

[MosesJones]

I'm running at home on a PII 400Mhz and it runs everything I possibly need. My Mother and Father in Law are on a P166 and wondered if they should upgrade. I said no as it really doesn't need it, they just do basic database, spreadsheet and word.

All too often developers use the increased memory and processor speed to write worse implementations, or to create pointless bloatware. I know this will continue no matter what I say but at the end of the day who really needs this much power, QuakeIV players ? QuakeV ? QuakeIII runs fine with my upgraded graphics card, and top of the line sound card, the processor does bugger all.

Moore's law is great, it means computers can do more and more, but for the home market its just silly, 90% of people would be fine not changing their machine for 4 years, but they are forced to upgrade by market perception.

Faster this, faster that.... but never ever actually "better", "more reliable" or "stable".

Hardware is the excuse for bloatware, its not H/W engineers fault but it isn't an excuse to use....

(and yes this is partly a dig at the huge swap requirements on the 2.4 kernel)

Re:Why bother ? its an excuse to write bad code

[Anonymous Coward]

That doesn't change the facts, though. If you do have enough memory, you will benefit from a faster CPU. And I don't know what type of video/audio editing you've done, maybe it's just assembling clips. But when you do much more than that, the processor needs to render new output froms from the inputs for stuff like transitions, color adjustments, and overlays and combinations. Ideally, you would want to be able to do this in faster than real-time, and be able to have the full power of digital computerized video editing for use on-the-air at reasonable costs. Similarly, the sound card doesn't do much other than just playing back the digitally encoded streams you have. If you want to change their contents, it's again a CPU job.

I don't see why people disparage using faster processors for legitimate applications. I've done video editing, and no matter what CPU I do it on, I wish I had more. And no, it wasn't disk I/O bound, because I had no trouble playing the input and output videos at full-speed.

Re:Why bother ? its an excuse to write bad code

[rabidcow]

I want to use C++ and lotsa nice OOD that's easier to write, easier to read, easier to expand, easier to debug, and easier to maintain.

In theory, you should be able to write such classes so you can define one flag and the debug stuff will compile away to nothing. (or just a few extra pointers)

So the developers need good machines, everyone else doesn't.

Except some companies are shipping debug builds as their final product. I'm not sure why. (Black & White, for example, includes the debug mfc & msvcrt dlls.)

Re:Why bother ? its an excuse to write bad code

[Grab]

In short, no, nor will 99.99% of the world's serious computer users, ie. those for whom a computer is a tool to be used rather than an end in itself. The 0.01% who do will likely not be doing it on a regular basis (eg. a couple of semesters of maths lectures). The 0.01% of those who do it on a regular basis (eg. maths professors/postgrads) can go out and get a multi-processor mobo and umpty-GHz CPUs - or more likely, will get their uni department or company to buy it.

In other words, no-one needs this unless they (a) need to compile mega-programs or (b) do heavy maths work. So no home user and most business users have no need.

I speak as someone who switched from a P233 to a Duron 800 only bcos the mobo broke - I refused to spend £80 on a new Pentium mobo when £200 would get a complete new system!

Grab.

Damn! Slashdotted!

[Arethan]

Note to web programmers, MySQL doesn't like it when it runs out of connections. Try increasing the connection pool size. Also, instead of having the page try to open the connection just once, and fall all over itself if the connection fails, try putting the connection request in a timed loop with a timeout of around 5 minutes, and a sleep(5) in the middle to help throttle a little. Your MySQL server will thank you, and your web page viewers will thank you.

Interesting, this.

[dave-fu]

I like how the little guys are going to benchmarks to indicate how their product actually performs while the big boys (Oracle, I'm looking at you) are recusing themselves from it.
Too bad that IT managers go with what they know (everyone else is using) and what's worked for them in the past.
It may be confusing for Jane Consumer, but it's nice to see that AMD's finally gotten a marketroid with a clue as to what works. Now if only their stock would start working, too...

About the naming

[TheMMaster]

I've been an AMD fan ever since, erm well always actually even my 486 was an AMD ;-)
I really think AMD will have te expect some problems with this. Back in the good old days (r) of the pentium and the cyrix 6x86 I worked in a computer store and we also sold cyrix computers to customers that didn't want too spend too much money (so sue me)
Very often people came back because they saw that their Cyrix PR200+ wasn't actually running on 200Mhz and demanded a refund (which they didn't get ofcourse) we had to explain the whole thing and it costed us a lot of time
That's why we stopped selling them back then
Another thing is that the semi-geeks (the dudes that THINK they are geek but basically know nothing) won't buy them because "they are already overclocked"

What about Dual Durons?

[Brento]

Thresh's Firing Squad has a review of the Tyan Tiger with dual AMD Duron MPs, which is probably of equal or more interest to us geeks. For those of you who weren't aware, AMD Durons work in multiprocessor mode as well, and they're very, very close to Athlons in terms of performance (and obviously cheaper.)

Big speed boost?

[wholesomegrits]

I didn't realize that people took press releases as gospel. What's all this crap about a "big speed boost?"

We're talking about mhz increase of less than 10% -- in how many months? It's been over a year since the T-birds were introduced.

Yeah, they have a new core. Whoopee. It's not a dramatically new improvement, and apparently AMD has decided that if its chips, in name, are as fast as P4s, they should cost as much too.

I like AMD stuff, but the Mhz Myth shit hasn't worked for apple, ever, and it won't work for AMD. Apple tried the Mhz Myth stuff back when the ppc601 came out, and despite 6 or 7 years of PR bunko, it's not caught on.

Anandtech's review

[acm]

Anandtech has a good review that compares all the latest p4's and athlon xp's. Check it out here.

Re:So what is good code?

[Grab]

This is a problem unrelated to language. It's surely possible to write good code in C, or C++, or Java, or C#. The problem is that ppl don't, and won't regardless of the language.

Read any software QA textbook, and you'll find they all agree (and experience tells you the same). How do you learn to code? It's not by being taught, it's by hacking away in a dark room somewhere. Individual coders/engineers may being incredibly skilled, but the experience doesn't get passed on, so the next generation of engineers make the same mistakes as the last one! Personally, I split up software developers into "hackers" and "engineers".

The "hacker", when given a vague problem to solve, sits down on his own and bashes out a piece of code without reference to requirements clarification, design documents, etc. It may even work - but it will be an unmaintainable nightmare, and if it doesn't work first time (or if it works sporadically) then it's over to printf and the debugger for months. Documentation, where it exists, will be written post-facto, and you'll be lucky if it explains the code properly. No-one else will be able to rework the code, and the hacker himself may not remember how it worked 6 months later!

An "engineer", OTOH, spends most of their time working in Word and a CASE package working out what they want to do and how they're going to achieve it, and runs his ideas past someone else to see whether a fresh pair of eyes can spot anything wrong. By the time the engineer goes for his favourite text editor, the problem's most of the way solved, and any bugs can be found by comparing design against code (ie. peer review). Any future changes are simple to include, as the design explains how everything works in sufficient clarity that anyone can pick it up and rework it.

A really good engineer (and I'm not one, yet :-) can distance himself from his own work enough to review it himself to make sure that a new reader can follow it easily. This differs from a hacker in the same way that a solo round-the-world sailor differs from the nutter who sets off across the Atlantic on a boat he bought at a garage sale: the former starts off knowing that there are risks, but has the experience to avoid or minimise them; the latter sets off not knowing that there are any risks, and only finds out when he hits the rocks. :-) An engineer doing RAD may well have a few trial hacks at the problem to see what works - but the difference is that the final result will not be constrained by these, ie. the experiments will likely be thrown away so that the final version is not cluttered with legacy crap from when the problem wasn't understood properly.

I've not run Netscape 6 for more than a few hours total, and it's already crashed on me more than once. Java is no magic bullet. Sure, there's some ways C will let you kill things which Java doesn't let you do. But coding standards such as MISRA define "safe" subsets of C, and by following them you will minimise the risks. Is it better to be coding in C, knowing how to avoid the problems, or coding in Java without knowing about any pitfalls? And as for timescales, Netscape are hardly a shining example, are they? :-) I'm not saying that Java is bad and C is the one true way, I'm just saying that more layers of indirection and "slower" code do not necessarily make it more reliable. What makes it more reliable is good design, and that is something you have to learn, not something you're born with.

For a typical user running typical productivity software, a 300MHz CPU and 128MB of RAM is all they'll ever need. More power will only be required for a new "breed" of programmes - maybe the Metaverse, maybe not. But your typical home computer user will not require any more processing power until a new killer app comes along. OfficeXP is not that killer app.

Grab.

Re:Athlon MP restricted by AMD760 mobo

[Ice_Hole]

Actually, this is somthing that is slowly but surely changeing. We are seeing new mobo chipsts which actually have 2 seperate busses for each processor. But only thing that is combined would the the bus that the RAM runs on. But this also is somthing that is in the works, and may not be a bottle neck for too much longer.

The problem is, their will ALWAY's be a bottle neck. No matter what you are dealing with. Wether it be the internet, the computer memory sub system, or the traffic on the way to work. Once we make one thing faster, it shows that another isn't quite up to par. So, that is th next thing that needs to be worked on. Wether it be the mobo manufactures, processor manufacturers, the wonderful people that lay that precious fiber optic cable, or the road crews that interupt my morning commute to work.

Things like the nVidia nForce shipset are (At least IMHO) going to advance computer technology even more than a newer, slightly faster processor. Why? Because of battle necks such as this memory issue we are seeing with the AthlonMP SMP systems.

Granted we have to give them some credit. When the Via chipsets were first released their memeory bandwith was HORRIBLE. Even to the point that it was better to stay with the BX chipset over upgradeing to the newer Via133 chipset. But, that has been fixed for the most part through things as simple as BIOS updates.

Their is a lot to a computer system, and their is a lot that makes it function properly. And if I had time I would get into the bandwith limitations between the northbridge, and the southbridge, the interactions with a SMP systems and the different cache's available to that processor, and their bandwith/ latencies, etc.

/pointless blabbering

- Ice_Hole