Slashdot Mirror

Client side, that is true. Server side, its just as fast or sometimes faster. See http://kano.net/javabench/ and http://www.aceshardware.com/Spades/read.php?articl e_id=153

http://www.aceshardware.com/read.jsp?id=50000347

PHP just can't cut it.

Send a big cheer to Rufus at Aces Hardware for grabbing all of the presentation slides.

Worth reading IMO.

http://www.aceshardware.com/forums/read_post.jsp?i d=120057079&forumid=1

http://rufus.hackish.org/~rufus/amd/big.html

Well, for one thing, it is slower than native code.

Patently false. It has been false for years now. Ever since Chris Rijk published his earth shattering benchmarks. (More recent benchmarks here.)

It's now down to the skill of the programmer. A good programmer will write speedy code, and a bad programmer will write garbage. Who'da'thunk?

For another, its garbage collection has a tendency to result in really bad performance stalls

When was the last time you used Java? 1.1? The modern hotspot JVM uses a generational collector which should NEVER stall during runtime unless it begins running into memory pressure. Go try this game and tell us how many stalls you see. If you think that's too "simple", try this one.

For another, its portability has been hampered by not fully supporting interesting OS features, which means that there are all these OS-specific extensions to add things like audio support,

Is there something wrong with the javax.sound packages? I'm REALLY thinking that you haven't tried Java since 1.1.

They don't integrate well with other apps, don't do a good job of supporting OS services, etc.

Psst!

Finally, Java makes it hard to add debug functionality into your code without a performance hit.

That's just a weak argument. Debugging info can really screw up a codebase and should be removed after debugging. But if you're wedded to the idea, get one of the three billion preprocessors that are available.

The bottom line is that pretty much any compiled language has great advantages over Java.

The bottom line is that you haven't used Java since the days of 1.1, but you feel that you're fully qualified to make statements about a platform you know nothing about. Whether you intend to or not, you are trolling, sir. So I would ask you to stop spreading FUD by not commenting on Java until you are again familiar with the platform.

I may buy/build a workstation soon that will need the capability of holding a large amount of RAM (say 32GB - calculations requiring a lot of memory but not a huge amount of CPU) at some point down the road (I can probably put off fully populating the memory for a while). I came across this post, which seems to say that motherboards for DDR2 will allow more DIMMs (16 2GB sticks is a lot cheaper than 8 4GB sticks right now, at least for DDR). It is talking about DDR2 with Opterons. Is there a launch date for DDR2 on Opteron? Is the capacity actually greater with DDR2? Is DDR likely to become scarce down the road, causing DDR2 to be a cheaper option for future expansion? Any opinions are appreciated (I haven't had an excuse to buy hardware in a long time, so I haven't kept up on such things).

Sidenote: Yes, I am aware of the iWill DK88 (16 DIMMs DDR) - anybody have any experience with it (especially with Linux)?

Early 1999.
http://www.aceshardware.com/read_news.jsp?id=55000 489

You mean it wouldn't perform well on real-world floating point applications then? Like SPEC fp 2000?

Spec fp 2000 results

Oh, I dunno, I think those AMD results look pretty good...

Actually, I don't think it would help much. Most games now don't benefit from 2 way SMP, so the benefit from 64 way is debateable to say the least. Still for servers, this thing might help. I suspect that most server applications/os's will have servere scaleability problems once you go this far SMP though.

BTW, Has anyone heard of the MLX1. Makes you wonder what would happen if you put a bunch of these on a chip with some clever caching and the mother of all memory controllers. x86 Niagra anyone.

>PowerPC is a much better design than x86.

3 2260 Pentium M 1839 1812 HTML
14 2200 PowerPC 970 1040 986 HTML

The PowerPC is an inferior processor to the PM, both raw speed and speed/watt. Futhermore, both the PM and PowerPC are RISC, only the PM has a translation unit to internal micro-ops that is becoming more negligable as die real estate is used for cache and more functional units or dual core. As for some "aesthetic" appeal of RISC vs CISC assembly, asm is asm, and since you can use SSEs on Pentiums in C, whereas Altivec *requires* you to use asm, it would seem that even the ISA is of a poor design.

Intel has made a processor that is both faster, better designed, and more work/watt, as well as being able to supply them in quantity reliably. Go take a graduate course in Computer Architecture or at the least read and understand H & P before you spout some kneejerk emotional reaction to your beloved Apple using the same processor as us unwashed masses.

>PowerPC is a much better design than x86.

3 2260 Pentium M 1839 1812 HTML
14 2200 PowerPC 970 1040 986 HTML

The PowerPC is an inferior processor to the PM, both raw speed and speed/watt. Futhermore, both the PM and PowerPC are RISC, only the PM has a translation unit to internal micro-ops that is becoming more negligable as die real estate is used for cache and more functional units or dual core. As for some "aesthetic" appeal of RISC vs CISC assembly, asm is asm, and since you can use SSEs on Pentiums in C, whereas Altivec *requires* you to use asm, it would seem that even the ISA is of a poor design.

Intel has made a processor that is both faster, better designed, and more work/watt, as well as being able to supply them in quantity reliably. Go take a graduate course in Computer Architecture or at the least read and understand H & P before you spout some kneejerk emotional reaction to your beloved Apple using the same processor as us unwashed masses.

I believe its called register renaming, but I probably know less on the topic than you.

The limitations of this are described in more depth here http://64.233.179.104/search?q=cache:5mZte35ICdQJ: www.answers.com/topic/register-renaming+limitation s+of+register+renaming&hl=en&client=safari
And here http://arstechnica.com/cpu/03q1/x86-64/x86-64-3.ht ml
And here
http://www.aceshardware.com/Spades/read.php?articl e_id=53
One significant problem is that this process is not transparent to the coder; it's done on the fly, by the processor, with hints from the compiler.

Working on the assumption that you weren't trolling, here's how things looked about two years ago.

Conclusion then: "the Opteron simply destroys the competition."

Today AMD has dual-core Opterons and Intel have... well, they have the same old Xeons with a miniscule FSB-bump?

Your post is utter, utter bullshit. Congratulations on the 4, Insightful - it's certainly not informative. Almost everything you say is incorrect, pure speculation, or irrelevant:

slashdot would still be bashing itanium if it were from amd.

this is speculation.

few people like paying $1000+ for a cpu alone, for example.

this is irrelevant. "few people" could mean 20,000 per year buying 10 CPUs each (200,000 is roughly how many Itanium 2 CPUs have been purchased in the last 12 months, and most people buying Itanium buy 8-way systems or higher. 2-way Itanium2s are quite rare and nobody _makes_ a 1-way Itanium system any more) for around $400million in annual sales, not too shabby for a "nice product" (the average Itanium CPU sale price is closer to $2,000

itanium is a niche processor filling a tiny tiny tiny market.

This is bullshit. $400million in annual CPU sales corresponds to way, way over $1billion annually in system sales. That is a niche market, yes. "tiny, tiny, tiny" it is not.

and it is already hitting scaling issues.

I suppose you're referring to Itanium 2, which is now a 4 year old product? That will be refreshed in ~5 months, where the next 4 years of scaling will be set up.

itanium also has yet to deliver on most of its performance promises.

This is total bullshit. Itanium's big performance promise was that by offloading complexity from the chip itself to software such as the compiler or virtual machine, you can gain improved efficiency as more of the CPU core can be dedicated to logic that does real, computational work. They delivered.

it's quite telling when a lot of the intel engineers and scientists involved with itanium are calling it a huge mistake. the p4 guys aren't impressed either :-)

Can you name even one Intel engineer who called it a huge mistake? You're just making up bullshit now. I suppose you might have been thinking of Bob Colwell but he certainly didn't say it was a huge mistake.

From this point onwards actually, the rest of your post is just bullshit you appear to have made up because it sits well with whatever the hell prejudice you have - I guess that's why this whole article got posted in the first place.

Anyway, you might want to try the "FreeBSD is dying" crowd, they are a bit better established than the "Itanium is dying" mob which is a bit younger (though over the years I'm sure it will grow to be every bit as respectable a trolling subcommunity! ;) )

Forgot the link

P4 spec performance.
The only consumer processor comparable to the P4 is AMD64, and it has a lower base int score, with higher int peak, and lower base/peak fp scores.

>To this day, a Dual 2.5GHz G5 still pounds a 3+GHz P4/AMD whatev into the ground
See here
Top 20 2-way SPEC systems

Top 20 SPECint_rate2000
2 2600 Opteron 40.5 36.1
6 3200 Pentium 4 Xeon 34.3 32.9
10 2200 PowerPC 970 21.5 20.2

Top 20 SPECfp_rate2000
2600 Opteron 45.8 42.3
3600 Pentium 4 Xeon 28.6 28.2
2200 PowerPC 970 20 19.2

Extrapolating linearly results for a 2.5GHz, x86 is still about 1.5x to 1.75x on ints, and 1.4x to 2x on floats. From this I must conclude that you are as the subject says, or that "pounds into the ground" has aquired the slang usage meaning "is pwnd".
For some reason, IBM PPC processors seem to have aquired Jobs' RDF, from the G5 to Cell.

>the popular consensus WAS that PPC's WERE better than anything in the x86 camp. That is, during the G4 era.
And, as usual, popular concensus turns out to be wrong.
The x86 still dominates any other processor available today in raw speed, though not on throughput. If you want to program in assembly, it's easier for a beginner to pickup risc because of its regularity, but I highly doubt most that the people complianing about the ISA actually used it and were just complaining that it seemed to them to be unelegant, and today it's pretty much a non-issue, as evidenced by the fact that special instructions like SSE2 are automatically used by compilers, making it unnecessary and actually more harmful sometimes to code assembly due to the way modern CPUs schedule instructions.

>it pushed more numbers with far less power
I'll give you that it required less power per flop, but check the results about to see how the mythical power of the g4 was all apple hype.

>AltiVec showed a ton of promise
According to wiki, Altivec has been around since the late-1990s, while SSE2, comparable in power to it, debuted in 2001. Any chance that Altivec would improve performance would've happened, and IBM and Motorola even had a falling out over whether to include it, according to the register.

The PPC has not been comparable in performance to anything offered by x86 for years. It added no competition to the market. I bid it good riddence so Apple might actually do some innovation that is actually about usability, instead of blatent lies about performance.

As for petty whining about Microsoft having a monopoly - guess what, Windows and x86 have succeeded where Apple failed because they understood what the market, and engineered it to be good enough, as opposed to being control freaks about the purity of their products.

You're not the first person to ask this, and Johan responded to the same thing over at Ace's Hardware (where he was before heading over to Anand). Basically you can only do so much while having a reasonable article:

Johan's response

SPEC FP results.

Look, a Athlon FX-55 is nearly TWICE as fast at SPEC FP - when the clock speed is only 30% higher. SpecINT will be about equal looking at it.

The Pentium M is good, but outright performance isn't something it has.

AMD single processors can't handle multiple threads at once?

No, they cannot. They quickly switch between threads, but cannot process them simultaneously. Intel chips with Hyperthreading process the threads simulataneously.

What planet do you live on?

Earth. Welcome, visitor.

So you're saying I can't run multiple threads of the same application on an athlon and do it effectively?

And he sets up the strawman, and he misses. No, I didn't say that and nowhere from reading my post could a reasonable person draw that conclusion. I only said that the single core Athlons couldn't process multiple threads simulataneously, but now with dual cores they can, giving AMD's dual core chips a bigger boost in benchmarks than Intel's dual core chips received.

While this may not be as effective as Intel's hyperthreading technology, I'd take an athlon64 3200+ 939 pin or athlon xp 3200+ over a pentium4 3.2ghz any day, simply because of the fact that I haven't noticed any difference with hyperthreading on or off on the intel systems.

Me too, and I did choose an Athlon 64 3200 over a P4.

All it does is make a fake virtual cpu out of a single cpu and reduce the processing time given to a single thread.

It does a lot more than that. It actually improves performance quite a bit, and anyone who has analysed benchmarks would see its effect in multi-threaded tasks.

Educate yourself on the subject before making a post flaming me.

http://www.aceshardware.com/read.jsp?id=50000319

But if you're naive about the net and you go online maybe once a month...then you're a raw piece of meat in a pool full of sharks.

I am one of those naive that really don't have a clue what hardware review sites to trust. My comfort is that I am probably far from alone, in this matter.

To assist me and other naives(sp?), please join this silly poll and review the following sites (regarding credibility) with a scale ranging from 1 to 10, where 1 is "No credibility at all" and 10 is "Perfect credibility, these guys wouldn't post a biased review for world domination":

About PC Hardware Reviews
Ace's Hardware
Anandtech
Ars Technica
Beyond 3D
Cnet Reviews
Dan's Data
Dev Hardware
Extremetech
Firingsquad
[H]ard|OCP
Hardware Analysis
Hardwarecentral
Hardwarezone
IT Reviews
OcPrices
Overclockers.com
ProCooling.com
The Tech Report
The Tech Zone
Tom's Hardware
TrustedReviews
Viperlair
Xtreme Resources

If you know only a few of them, give your opinion on those.

Maybe someone with the right facilities could set up an independent poll?

Uhm, nevermind Itanium then.. http://www.aceshardware.com/SPECmine/index.jsp?z=2 &s=1&v=1&if=0&ncf=1&nct=256&cpcf=1&cpct=2&mf=250&m t=3800&o=0&o=1&start=20 (ugly URL)

so it might not be as good as an equally clocked Power5 based processor

Man, you do realize the Power5 is F#$%ing FAST. The "PowerPC Processing Element" looks like it won't have much in common with the Power5 at all. I doubt you will get 1/100th the speed of the current Power5s for generic PowerPC programs on the Cell.

Not really. Intel has been playing catchup all this time, first with 64-bit and now with dual core. Opteron was built from the ground up to support more than one core, which is the beauty of it.

Here's a long discussion on the current dual core situation on Ace's Hardware. (They use a lot of codenames. "Smithfield" I think is what this /. story refers to. "Yonah" is somewhere in the future.)

Intel chips are x86-64 too, and they run up to 3.8Ghz.

Not that it helps them much on Spec benchmarks -

http://www.aceshardware.com/SPECmine/top.jsp

But for most of the last couple of years, the best performing x86 managed to run at a higher frequency than the best performing Risc, or turn in better SpecInt benchmarks - at least for single CPU, or sometimes both.

PC hardware tends to suck for multi CPU stuff though - even Opterons are a bit disappointing to be honest, considering that decent SMP performance was supposed to be one of the design goals. On the other hand, I don't think switching instruction sets would help here, it's something that will be fixed with multicore chips (because they make SMP more mainstream and thus more important) and better bus protocols - maybe AMD will license the Power5 one like they did for Alpha.

FP performance is a bit disappointing too - especially as the Athlon FX-55 result is with a decent 64 bit compiler - it's still a fair bit behind Itaniums and POWER5. On the other hand, how much mainstream computing is limited by FP performance?

Actually, Power5 does better than I'd expected - much better than PPC970, but it isn't enough to get people to switch from x86. Back in the Alpha days, Risc chips had a advantage in virtually all benchmarks, and conventional wisdom said that it would increase, but it still wasn't enough of an incentive to get significant numbers of people to switch over.

Which is my point really. It's not that x86 has an particularly good, it is that PPC and other Risc chips no longer better enough to make people switch.

You can see it at
Ace's Hardware

If I'm not mistaken, the 386 doesn't have an internal FP unit. It might be a better idea to do this with 486s, and double the performance at the same time.

I've seen this cobble-together-old-processors idea around before, but there's got to be a reason nobody has bothered to try it. I'm sure that "controlling logic" has got to be hideous.

You might find Sun's Niagara of interest.

Ugh, forgot the .com: Ace's Hardware

POWER5 is already shipping, and it beats the dual-core Opteron in floating-point while the Opteron wins in integer code. Of course YMMV.

(i.e., a 1.4 GHz dual core CPU will outperform a 2.4GHz dual processor machine)

That would be nice if it was possible, but it's not. Actually, a 1.4 GHz dual-core Opteron would be slower than two 1.4 GHz single-core Opterons.

You mean the "rumors" aren't officially "news" until they appear on /.? Forget what we've been reading since Febuary on http://www.anandtech.com, http://www.tomshardware.com, http://www.theinquirer.net, http://www.arstechnica.com, http://www.hardocp.com, http://www.aceshardware.com, and of course http://www.intel.com, it's not true until it appears on /. ...

PSSST!!! I've heard the rumor that Apple is planning on ditching Motorola's chips for IBM processors in their upcoming Macintoshes. Has anyone elseo heard about something called a "G5"? Some say it might also be 64 bit? Heavens-to-Betsy, let's post it to /.'s FP.

Intel's really falling behind the curve if that's the case. If I were a stockholder, I'd be pretty annoyed after AMD64^H^H^H^H^H EM64T, and the Itanic debacle. Now they're cloning a Sun processor design? Heh.

</flamebait>

Monday Forbes reports Intel told software companies they should license a multi-core chip as one processor. Also on Monday, Intel compared their new Itanium to the "best published RISC" machine. Their graph indicates a 64-processor Itanium is about the same SpecIntRate as a 64-processor RISC machine. Now the funny part is for the RISC result they used the 32 chip Power5 SpecIntRate as 64-processors. So 64 Itanium-2 chips are really about the same as 32 Power-5 chips. So while Intel advocates per-chip licensing, they use per-core benchmarking. It is also interesting to note that this new Itanium-2 SpecIntBase of 1590 is just a bit faster than a 2 Ghz Pentium-M and much slower than a 2.6 Ghz Athlon-64-FX.

Usefull site for getting usable spec numbers is ace's SPECmine. Try SPECint2000 or SPECfp2000 and check "CPU MHz". The opteron's fare quite nicely.

I checked, and the your definition for new does not exist. "New" does not mean never sold. It means "fresh" or "current" or "latest." Taking your logic, a car made in 1944, put in crate & never sold is "new." No, it's just a 60 year old car that was never sold. It's still quite old. Pentium IIs were released between May 1997 & August 1998. In PC years, those are ancient.

Ace's Hardware is a pretty good (though low-volume) cross-architecture news site. Still covers x86 mostly, but does go into other architectures from time to time, and the people there seem to really know their stuff.

UltraSparc IV: 66 million transistors
Pentium IV Prescott: 125 million transistors
Power4: 170 million transistors

So how many transistors are in the TR-1?

4

For everything else, there's vacuum tubes. (Or diodes, depending on your radio set.) :-)

32 bit intel binaries will run on an Opteron and will probably give some gain as seen here at Aces.
Aces article

However, there was a desktop "engineering prototype" that was kind of quiet. No picture was shown. It sounds like they made a multichip module from 2 die to test things. If they had a dual-core Xeon, they would have said so clearly and not just mentioned after a demo that the machine was using an engineering prototype dual-core.

What makes you say "certainly not"? And what makes linking that phrase to marketing propaganda insightful? Ultrasparc is running out of gas, folks. It's not scaling up and instead of finishing and releasing their new core they actually had to scrap that effort and release a multi-core processor instead because the ultrasparc is getting left behind badly by POWER5. Even Opteron seems to be faster; from what we know about its processor interconnect technology it should scale well, and the 4-way Opteron in the above-linked benchmark looks like it would beat the UltraSparc III with half as many cores. (It's only compared with 1/4 as many cores as the sample USIII system.) USPARCIV is basically a dual-core USPARCIII since they couldn't manage to bring their actual new core out. Put another way, an 8-processor (16 core) USPARCIV should be no faster for CPU-intensive tasks than a 4-processor (8 core) Opteron when such a beast becomes available - which will be soon.

Hence, unless Sun comes up with a new UltraSPARC soon, which seems unlikely, SPARC is done. It's over. There's no reason for Sun to keep flogging this particular deceased equine when it can just buy Opteron processors and build systems around that.

Of course, there's no reason to buy such a system from Sun, either, once PCs start getting onboard peripherals that lie along a PCI-E bus. Right now you're lucky if your onboard peripherals that need more than 133MB/sec of bus bandwidth are even on 64 bit or 66MHz PCI buses internally. I'm not sitting at my workstation just now but some of its hardware is on one or the other type of PCI bus, but not both...

This isn't made for rendering, or for stuff that requires floating point power.

Since the chip has 8 cores, each core is quite simple. This kind of chip is more suited for database and web servers, where there are lot of simultaneous requests, but fullfilling a single request is quite simple task.

You can find more information about Niagara here.

Why? You can get a 1U Sun system for under 1k$. A dual AMD 64 system comes in under 5K$ as well. I'm not saying that the IBM system isn't worth the money, just that it isn't "low cost".

It appears that POWER5 wipes the floor with UltraSPARC IV. An 8-core POWER5 system has better performance (according to this benchmark) than a 16-core USIV system in every benchmark in which they are compared.

... double the FLOPS of an Opteron ...

A couple months after AMD said they had taped out their dual core Opteron, Cray and others said they would be upgrading to that in 2005. Putting two Opterons next to each other with their hypertransports talking to each other is so easy that I suspect AMD's first silicon worked. I suspect dual-core Opterons will be in production way ahead of dual-core Itanics, since they are so much smaller. And we will see 8-core Opterons before we see 4-core Itanics. A number of people have working 8-way Opteron motherboards and I have not heard of anyone getting more than 2 Itanics on a motherboard. All the while AMD production volumes will be far higher than Itanic volumes.

If SGI wants to stop loosing money, they should come out with an Opteron CPU-brick fast.

"Won't be safe for long" means Itanic is sinking. Really.

... double the FLOPS of an Opteron ...

A couple months after AMD said they had taped out their dual core Opteron, Cray and others said they would be upgrading to that in 2005. Putting two Opterons next to each other with their hypertransports talking to each other is so easy that I suspect AMD's first silicon worked. I suspect dual-core Opterons will be in production way ahead of dual-core Itanics, since they are so much smaller. And we will see 8-core Opterons before we see 4-core Itanics. A number of people have working 8-way Opteron motherboards and I have not heard of anyone getting more than 2 Itanics on a motherboard. All the while AMD production volumes will be far higher than Itanic volumes.

If SGI wants to stop loosing money, they should come out with an Opteron CPU-brick fast.

"Won't be safe for long" means Itanic is sinking. Really.

Perhaps this is what you are looking for

There are older dual and quad Opteron vs Xeon reviews around.

When it comes to (Java) webservers and/or MySQL, the Opteron definitely has the advantage. In some cases, the Opteron simply annihilates the Xeon, but luckily for Intel the latter offers some resistance in our GZIP dominated benchmarks.

Humorously, the also say this:

The Opteron will probably remain the fastest CPU for the server tasks tested here until Intel introduces Nocona, the Xeon Prescott at 3.4-3.6 GHz (1 MB L2, 800 MHz FSB) at the end of the 2nd quarter of 2004.

Now we know that the Nocona is here, and it's getting slaughtered at the Altar of The Opteron.

http://www.aceshardware.com/forum?read=115093869 Sorry, forgot tags...

I though that these benchmarks looked a little strange when you're using Jack the Ripper as one of your major comparisons. There's a nice thread going on over at Ace's bashing the benchmarks, including a post from the author of the chess benchmark stating:
this test they did was flawed in all respects.

I though that these benchmarks looked a little strange when you're using Jack the Ripper as one of your major comparisons. There's a nice thread going on over at Ace's bashing the benchmarks, including a post from the author of the chess benchmark stating:
this test they did was flawed in all respects.

and I predict that Sun will replace their entire product line, except for the extreme high-end, with Opteron, in the next 5 years.

Where does Niagara http://www.aceshardware.com/read.jsp?id=65000292
fit in this?