Slashdot Mirror
NASA Benchmarks the New G5 Powermac
sockit2me9000 writes "Well NASA's Langley Research Center recently benchmarked the new G5 dual 2ghz Powermac against a dual 1ghz Xserve, a dual 1.25 ghz Powermac, a Pentium4 2 ghz, and a Pentium4 2.66 ghz. To make things fair, the second processor in the G5 was switched off, as well as the other dual sysytems. Then, they all ran Jet3d. Even with un-optimized code and one processor, the G5 performance is impressive."
Because I have a strong feeling this is going to be asked:
;)
For those of you who were wondering, you too can switch off one of your Mac's dual CPU's with the Apple CHUD Tools. Look near the bottom of the page. It'll make you appreciate your second processor
Personally though, I want to see how well it runs Seti@Home.
Vonal Declosion
By adding a second processor, the MFLOPS/Mhz output only dropped from 0.127 to 0.125 MFLOPS/Mhz. This chip can definitely perform in a multi-processor environment. The P4 scored 0.096 MFLOPS/MHz with a single processor.
Apple's benchmarks which were highly criticized by some, gave the Dual 2GHz Power Mac G5 a 194.5% performance advantage over a 3GHz P4 in SPECfp_rate_base2000. The G5 getting a score of 15.7, and the P4 getting an 8.07.
NASA's study found the Dual 2GHz Power Mac G5 to score 498 MFLOPS for their Jet3D performance. A P4 running at 2.66GHz scored 255 MFLOPS: a 195.3% performance advantage for the G5 in this test. If we assume a direct correlation between MHz and MFLOPS for the P4 (which would actually overstate the performance of the P4) and increase the P4's score by 12.782% this would give the 3GHz P4 a score of 287.594 MFLOPS. This is still a 173.16% performance advantage for the G5, and NASA states that a 20% increase in performance for the G5 would be reasonable "when G5-aware compiler tools become available."
So it would seem NASA's benchmarks go a long way in validating the benchmarks for the G5 that Apple released last month at the WWDC. In fact, NASA appears to be giving the G5 even better scores than Apple and Veritest did.
The vector tests that NASA performed to test the G5's AltiVec instruction set produce some even more impressive results, and would be a good indication for why the G5 outpaced the Xeon and P4 by such dramatic amounts on real world tests (at times more than 700% faster than a 3GHz P4). "The vector version of Jet3D runs an order of magnitude faster than the scalar version (speedups of 10X-13X are typical)." The dual 2GHz G5 was benchmarked at 5177 MFLOPS (a 1040% increase over the scalar test) and 1.29 MFLOPS/MHz. This also seems accurate considering Ars Technica's claim that the AltiVec engine wasn't as well integrated into the G5 as it was in the G4. The 2GHz G5 (single cpu) scored 2755 MFLOPS, or 1.378 MFLOPS/MHz, which shows a slightly larger performance hit for vector operations than floating point operations when moving to a dual G5.
Dak
I love the wording:
"Benchmarks from the scalar version of Jet3D are shown in Figure 1 (MFLOPS) and Figure 2 (MFLOPS normalized by MHz). In terms of raw MFLOPS, the 2GHz G5 is about 32% faster than the 2GHz P4, 97% faster than the 1.25GHz G4, 142% faster than the 1GHz G4, and within 1 MFLOP of the 2.66GHz P4."
Translation: Slower than the P4 for anyone who didn't look at the grid. And M stands for million. Not one.
I wonder why none of the NASA boxes were running Windows? I mean, if it's an all sinigng all dancing solution to everything...
(It's ok, you can mod me -1 Troll now. I'm just bitter about an edict on a project I'm working on.)
Beep beep.
What the heck is that a benchmark of? Divide by the Mhz to create a higher number on their nice little graph for the G5?
Why a Pentium 4 2.66mhz?
Why no Athlon?
Why no Opteron?
Why an old old version of RedHat 7.1?
and so on....
I have very bad experience with Portland compiler; I wonder what the results would have been if they had used Intel Fortran Compiler....
It is interesting to note that they used the Portland group compiler instead of the intel compiler. For the CFD code that I work on (which is mostly Fortran), the Intel compiler produces much faster code than the Portland group compiler (as much as 50%).
If there budget is such that dualie 2Ghz G5's are a possibility, then it's somewhat surprising that they A) used such low powered P4's B) that they didn't include Itanic2 systems. Seems that their report really just pointed out stuff that we already knew, the PPC is typcially faster per mhz than the P4 (hell, just about anything is better per mhz than the P4). Interesting to note that on the vector test, the G5 outperformed the G4 is a fashion that is almost purely based on the increase in Mhz (i.e. other system improvements didn't really seem to help much). Compiler perhaps, though some of the architectural improvements would seem to be not dependant on that?
Apple has just created a new market for itself among the hardcore engineers who use workstations for numerical simulations like HSPICE, etc. Steve Jobs lucked out -- again.
By the way, the bell tolls. It tolls ominously for Sun Microsystems.
The only benchmarks that matter is my impression of the system while using the apps I use. Everything else is opinion.
People who bite the hand that feeds them usually lick the boot that kicks them
Something which seems to get lost in the Mac/PC debates *sometimes* is the cost factor. I looked those graphs and thought "Wow - mac is faster at this benchmark". Then I looked up pricing - minimum I can get that mac for would be $1999. An equivalent PC system with the P4 2.66ghz is probably under $900 (didn't spec it out entirely, just did a rough lookup on Dell). Great - Mac is faster. But I can apparently get within reasonable range on PC hardware for probably 50% less cost.
I'd read some thread a while back on another board saying that "Macs are cheaper than PCs". I still can't believe anyone would make that argument. Doesn't being really good in a few areas satisfy the mac people? Do they have to try to spin higher costs as 'lower' (craziest thing I'd ever heard...)
creation science book
The G5 really is a powerful machine, and i wouldn't mind owning one... if only the price for maintaining a mac wasn't so high...
Their scheme for OS X is the equivalent of Microsoft charging $100+ for a service pack, I just don't understand it.
I've used OS X, and it blows everything else out of the water in elequency and it seems the perfect balance between productive and 'cool factor'
But until I win the lottery, I'll stick with my cheap x86 machines
Excuse me, I don't mean to impose, but I am the ocean
$2999 for the mac 2x2ghz
creation science book
Why no Athlon? They probably didn't have one and with the P4 at 3ghz and climbing, the old althon is becomming less and less significant for these pure number crunching apps. Plus maybe they've done previous tests that said that P4 they uesd was faster than an equivelent athlon, so it didn't need to be tested.
Why no Opteron? They probably didn't have one. This is the most valid question you ask.
Why RH 7.1? That's probably what they use. They are benchmarking PURE CPU so the OS doesn't matter too much for this kind of thing.
Comment forecast: Bits of genius surrounded by a sea of mediocrity.
The major OS version updates are when new features are added, etc. That is the equivelant of upgrading 98 to XP. The cost of buying a Mac is high. The cost of maint. is probably less than a Windows box.
CAn'T CompreHend SARcaSm?
Wait, let me get this straight, your assertion is that the benchmark is invalid because they "probably compared the i386 generated code against the one fot he G5". "Probably" IOW, you have no freakin' clue if this is the case and are just spouting off. Brilliant! Worse yet, you got a +1 Insightful...
It was interesting to see this paper devote so much effort to the completely useless metric of MFLOPS/MHz. This measurement has absolutely nothing to do with performance, but rather, with the approach taken by the chip manufacturer. You can do more in one clock cycle, as AMD historitcally has done, or less, but optimize for faster clock speeds, as Intel has in recent flavors of the Pentium.
One might be tempted to design a chip that does more in one cycle and then clock it as fast as a chip that does less in one cycle. Unfortunately, while reality is a little more complex than this, the basic reason is that the more a chip does per cycle, the more heat it generates per cycle. If you try to squeeze too many cycles through it in a second it will fry.
So showing that the G5 has better performance per clock cycles is no more useful than showing that an AMD chip has better performance per clock cycle than an Intel chip. All that matters is how much performance you can get from a chip before it cannot be clocked any faster without requiring unreasonable cooling methods.
All this paper shows is that, while the G5 is designed to do more in a clock cycle than a P4 is, the chip tested is ultimately not any faster than the P4 they benchmarked it against. It remains to be seen how the G5 will do at higher clock speeds. With this in mind, it would be *far* more useful to see heat dissipation stats on the G5 since that might give us some idea how close to it's design limits. If it is cranking out high-end P4 performance and running cool *then* I will be impressed.
Maybe because that was the hardware and software available to the testers. Contrary to popular belief, government employees do not have unlimited budgets to buy stuff. The last time I worked in a government office, some of the furniture was older than I was and my PC was built from scrounged parts.
Mea navis aericumbens anguillis abundat
I guess this makes me the 10th fan...oh wait...
but deep down inside . . everything is linux right?=)
How Now Brown Cow
Linux is almost as popular as Windows 95! Seriously, 1% of that pie is probably tens of millions of people.
Apple has been making better software for years, everyone agrees; They just never had the hardware to back them up. Then every time they do crop up with better hardware, everyone criticizes them and says that it's just not possible, PC hardware is always better they say. But now they've proven you wrong... TWICE, and some trolls STILL don't believe them. It's a sad world. I just wish Apple would open up at least their motherboards a little more, make Macs more customizable, more like PC's so they can start dominating again.
There's also one benchmark I'de love to see. Power Mac G5 vs Sun UltraSPARC III. It's fair: they're both 64-bit procs, and it would really make people look at it in businesses that only look at supercomputers as viable. Then maybe people would start giving Apple and IBM some credit.
My 2 cents (Canadian). Thanks.
"Victory means exit strategy, and it's important for the President to explain to us what the exit strategy is." G.W.Bush
I hope they didn't use gcc (the yet-another free and hopeless compiler).
It should be noted that Apple uses gcc to compile Mac OS X and most of their applications, so it would be appropriate to use gcc on the G5. Intel's compiler might be a more appropriate choice for the Xeon.
$x='S24;r)>63/* h@<5+oZ)32"5cz';$me='phroggy'x$];
$x=~y+ -xz+\0-Tx+;print$_^chop$me for split'',$x;
Um...no.
BSD and Linux are two different code bases.
-twb
> Tell me this is a troll. Please!
Yes, your posting is a troll.
Do you feel any better now?
k2r
Why not show an mflops/$ chart? Related to my 'cost' post as well, but I felt it deserved its own post. :)
This seems to confirm my belief that most mac people don't buy their own hardware, but get it through work or school.
creation science book
This article says 9 fans. Say what you like about the source. :)
"The new PowerMacs feature an all aluminum enclosure featuring 9 fans but with be up to twice as quiet as current Power Mac G4's. The new Power Macs will be shipping in August."
That's nice that they're twice as quiet. But 9 speicalised fans can be expensive to replace over time. They are not immortal.
RTFA.
$2999 for the mac 2x2ghz
How much for the dual Xeon system they were comparing that too? Yeah, you can build a P4 for $900, but not a dual Xeon.
However, yes, the $1999 low-end G5 is definitely more expensive than a P4 of similar performance, if the only consideration is raw speed. Macs are cheaper than PCs at the high end, not the low end.
$x='S24;r)>63/* h@<5+oZ)32"5cz';$me='phroggy'x$];
$x=~y+ -xz+\0-Tx+;print$_^chop$me for split'',$x;
>>Vector performance of the G5 remains excellent, and is inline with current G4 systems on a per clock cycle >>basis. As a result, raw vector performance of the G5 will be boosted simply by its higher clock speeds relative >>to current G4 systems.
This would seem to be one of the more interesting points made, actually. Prior to the announcement of the G5s, speculation on the PPC 970 suggested that it would be stellar with FP & so-so with integer; the real question surrounded how well IBM would implement SIMD. Many were pessimistic. Given that it seems like they've managed to add it efficiently a scaled-down POWER4 core, future refinements could make this series of chips (PPC 9X0s) real monsters.
But the future viability of that roadmap (given how ruthless the company as a whole tends to be when faced with departmental money losses) depends as much upon the success of IBM's Linux strategy as it does on its success in the PowerMac line.
[With apologies to BadAndy of the Ars Technica boards; thanks for sharing your insights.]
Habit is the ballast that chains the dog to his vomit - Samuel Beckett, "Proust"
It's not Apple's G5, it's IBM's 970 and it's the shizzle.
Karma: The shiznight, mostly because I am the Drizzle.
Yes, I see the article talks a lot about performance per MHz, but that's a useless measurement. That would be like comparing cars by looking at speed/RPM. If you want to get somewhere, car A that goes 60 MPH and runs at 6000 RPM will get you there faster than car B at 50 MPH and 2500 RPM, even though car B has a better speed/RPM ratio.
G5 compared to Itanium and Opteron as soon as someone actually writes a G5 compiler, news at 11:00.
Benchmarking a system thats not out yet with unoptimized code with compilers written for systems from 3 years ago, old news.
Good thing we have some more bogus benchmarks, it's a slow nerdy news day.
- Adam L. Beberg - The Cosm Project - http://www.mithral.com/
i just wanna see a dual 2ghz G5 vs a dual 1.8ghz Opteron
If one thing is 80 dB and one is 90 dB, the second object is twice as "loud." Each 10 dB jump either doubles of halves "loudness." ie: If you're at 1000 dB vs 1010 dB, the 2nd object is twice as loud.
So, based on what was said at the keynote (and my interpretation), the G5s are 10dB quieter. Twice as quiet sounds more impressive. Note that saying "half as loud" still implies "loud" so psychologically it's not as impressive.
If I'm wrong, I'm sure someone will jump on me soon enough. I'm holding on tight.
It was a hardware simulator running on a soundstage.
...if this recent Ask Slashdot post was related?
SCO (noun.)- A Slimy Corporate Ogre. Often seeks free money.
I think it's pretty obvious why they tested the G5: their Altivec program is 13X faster than their scalar program. They don't mention the SSE2 so I assume they have an investment in Altivec programs. Therefore they would naturally be interested in comparing the G5 versus the XServe and G4. Until Intel releases the 34.5GHz P4 (13X 2.66GHz), there doesn't seem to be any reason to run out and buy a latest P4 just for this comparison.
And surely the version of RH Linux hardly matters. Maybe they benchmarked using this OS because (shock, horror) it is what they use daily.
I was RTFA, and this caught my eye:
...
Additional Notes: The G5 system was running Mac OS X 10.2.7 and
I'm only running 10.2.6, and Software Update says nothing new is available. What's up with that?
Tuus crepidae innexilis sunt.
I'll be more impressed if the folks at the Langley Research Center compared the Apple Power Macintosh G5 with the 2.2 GHz PowerPC 970 CPU against a system running the Pentium 4 3.2 GHz CPU (which has Hyperthreading instruction registers to have almost dual-CPU performance).
My guess is under Jet3D the P4 3.2 GHz machine will likely outpace the 2.2 GHz G5 PowerMac handily in single CPU mode.
"The vector version of Jet3D runs an order of magnitude faster than the scalar version (speedups of 10X-13X are typical)." The dual 2GHz G5 was benchmarked at 5177 MFLOPS (a 1040% increase over the scalar test) and 1.29 MFLOPS/MHz."
5177 MFLOPS when running a Velocity Engine optimized version of Jet3D.
Now, how much does an P4 extrapolated to 3.2 GHz get? Like 288 MFLOPS?
Someone please explain to me how 5177 MFLOPS and ~300 MFLOPS are even comparable.
As the Mathematica guy said, the competition is no longer high-end PCs, it's now $10,000 UNIX workstations...and the G5 is still faster than any of them.
No wonder the G5s smoke the dual Xeon in the Photoshop, Mathematica, Logic, and Luxology app bake-off. All these apps would have been optimized to use the Velocity Engine.
If I were a scientist doing lots of image processing and vector calculations, I'd need a cluster of about 18 or so 3.2 GHz P4 machines to keep up with the dual 2 GHz G5 PowerMac running a typical Velocity Engine optimized app.
That's a sweet 5177 MFLOPS for you - evidence the G5s rock as hard as Apple has been indicating.
Um, yeah, sure is lucky Apple found the G5. I'm sure they had nothing to do with its development. It's not like Apple has been involved with development of the whole PowerPC architecture since the early 90s.
But Jobs wasn't at Apple during that time. What's the timeline:
1983: Scully joins Apple.
1985: Scully fires Jobs. Jobs, no longer at Apple , begins Next.
1991: PowerPC alliance between Apple and IBM.
1993: Scully leaves Apple, Spindler becomes CEO.
1996: Gil Amelio becomes CEO. Jobs approaches Amelio with idea for Apple to buy Next. Apple buys Next.
1997: Amelio fired from Apple.
Besides, at the $3000 launch point for the basic twin proc G5, it's not a lot differnet to a dual processor P4 3GHz.
And no - I'm not a Mac fan. ;P
oh brave new world, that has such people in it!
actually what you'll find is that when you buy that shiney new G5, it actually has this little piece of paper inside that offers you future OS upgrades for free, or $30 per. (i don't remember offhand).
also, last i checked the number of a release was rather arbitrary relative to what was added or not added. the differences between 10.1, 10.2 and 10.3 are pretty hefty. well worth $30 per.
but i suppose it's much better to assume you have to pay full price for each.
This serves to only debunk the usefulness of benchmarking...
Tom
Someday, I'll have a real sig.
i think they used redhat 7.x as it is the last in the redhat range supporting the ppc chipset. isos are not available for 8 or 9; anyone know any different?
Any benchmark that's not against another real 64bit chip with a flat memory space is completely and totally irrelivant. Comparing optimized (no gcc need apply) code against optimized code like in the real world.
Has anyone seen any such relivant benchmarks?
- Adam L. Beberg - The Cosm Project - http://www.mithral.com/
Hahah, so the P4 does, on average 'less work' per cycle then the g5, so much for the g5 being superior "RISC" and the p4 being inferior "CISC"
Okay, I'm joking. But seriously, that information is a little useful. We can uses it as a baseline to compare new CPUs. So we'll know that a 3ghz g5 is still slower then a 5ghz p4.
This also tels you how many floating point ops per CYCLE the chip can do, which is always intresting.
autopr0n is like, down and stuff.
They were only using one of the G5 processors, RTFA
You mean this wasn't a Tom's Hardware review?
I especiallly love the way G4s used to "smoke a Pentium" in Apple magazines and advertisements. You know what else "smokes" a Pentium? A PII.
Probably but it won't run half life 2, because its not being ported to the mac (last I heard) :)
Personally though, I want to see how well it runs Seti@Home
My bet is you still won't find any signs of Alien life. So it won't be any better than my old crappy ass P1 166.
But good luck to ya.
Redhat 7.1 and 7.2 are the standard versions used by industry and research. They are concidered more stable then newer releases. Most GNU/Linux hosting in the US uses RH7.2. It works and does not have some of the issues of OSs that use the kernel of pain. I do all of my development to target RH7.1. After the 2.6 kernel gains some maturity, I will probably change this. BTW, the 2.4 kernel realy has little to offer in the way of improving performance for the type of application NASA is intersted in. The same will not be true of the 2.6 series as can be seen in the current 2.5 tree.
The apple scored 0.39% less then a 2.6ghz p4. Given that there are 3.2ghz p4s out there, as well as dual athlons, I doubt this is the 'fastest ever'
autopr0n is like, down and stuff.
The Pentium comparison was just a fun side show.
Moderators should have to take a reading comprehension test.
3.06GHz with 512k cache? Dell has been selling these for a while.
What about Dual Opteron?
Why do people tests Macs that aren't even out yet against CPUs from the competition that are several speed increments below the top?
The 498 MFLOPS figure was WITH 2 G5s!!!!
With a single G5, the 2ghz got a 254, and the 2.66ghz P4 got 255 MFLOPs...
Please read the article more clearly, this DOES NOT IN ANY WAY validate apple's earlier claims... here's the quote that was misread
"Though dual processor benchmarks are not presented in detail here, it is worth noting that the G5 system benchmarked at 498 MFLOPS and 0.125 MFLOPS/MHz for scalar Jet3D performance when two processors were used."
Followed by a chart showing the P4 2.66ghz with 255MFLOPS at the top and a G5 2ghz with 254MFLOPS at the bottom...
So you could guess that a dual 2.66ghz would get about 499-500MFLOPS which would be a 0% performance advantage to the G5, and the P4 3.2ghz would be even faster...
If this was done by NASA, why isn't the report on a NASA web server? (as opposed to some guy's personal homepage) kinda limits the creditability of this whole thing...
What you're not factoring in is that the G5 competes against the Xeon rather than the P4.
Go spec a Dell with a Xeon processor, a DVD burner, Digital Optical Audio I/O and a Huge hard drive (doesn't _have_ to be SATA) and a good graphics card and you will have a more realistic comparison.
Factor in the amazing case design (yes, as a tech it's so much easier to work on Apple machines than anything else) and things turn the other way...
Try adding 8GB of RAM to the PC and... oh, wait, you can't!
As for the benchmarks with HyperThreading tutned OFF they did that to speed up the PC, it was slower in those particular benchmarks with it turned off than it was with it enabled. HyperThreading isn't two real CPUs, it's not magic that speeds up _everything_
- k
Specialist Mac support for creative pros, Melbourne
They used RedHat 7.1 only on the faster P4, the slower P4 was with RedHat 7.3.
You can easly build a dual Athlon system that could trounce an equivilant g5 for less cost. Xeon's arnt the only MP capable x86 chips out there...
autopr0n is like, down and stuff.
Found this from last Jan:
Date: Mon, 13 Jan 2003 23:29:38 -0500
From: Craig Hunter
Subject: G4 vs. P4 performance
I have been following the discussion of Rob Galbraith's benchmarks with much interest, as I have spent a good deal of time testing, optimizing, and benchmarking software for the G4 (OS X) and P4 (Linux).
The first thing to realize is that there are numerous benchmarks that show the P4 is faster, and there are numerous benchmarks that show the G4 is faster. What matters? Well, probably the benchmarks that apply to the kind of work you do. For people doing photo processing with the software Rob tested, his results are extremely relevant. But, someone working with a program optimized for AltiVec and dual processors might have a completely opposite experience.
Just to give an example of a benchmark that goes the other way, see this chart.
(You're welcome to mirror this benchmark image, since my web site may not handle a lot of traffic). These real-world results come from the Jet3D computational fluid dynamics noise prediction software, which I developed for my doctoral thesis and currently use in my work at NASA. Jet3D is written in a combination of FORTRAN 77, FORTRAN 90, and C, and is optimized for AltiVec and dual processors on G4 hardware. When compiled on Linux using Intel's ifc compiler tools, Jet3D also becomes optimized for the P4 (using the various SIMD extensions available on the P4).
As you can see, the G4 does quite well here. A dual processor 1.25GHz G4 system is more than 3.5X faster than a single processor 2GHz P4 system. Though it's not shown on the chart, a single 1.25GHz G4 processor benchmarks at about 1589 MFLOPS, 1.9X faster than the P4. If you look at MFLOPS per MHz for a single processor, the G4 comes in at 1.27 MFLOPS/MHz, while the P4 comes in at 0.42 MFLOPS/MHz. If you want a good example of the MHz myth, look at the Cray, which comes in at 1.78 MFLOPS/MHz with only a 500MHz processor, beating both the G4 and P4.
Without AltiVec, the Jet3D benchmark would be about 794 MFLOPS on the dual-1.25GHz G4, which erases the performance lead over the P4. And then, using only a single processor, the 1.25GHz G4 benchmarks at about 418 MFLOPS, which is about half as fast as the P4. And all of a sudden, the G4 doesn't look very compelling. For the Jet3D benchmark, AltiVec and dual processors are key (AltiVec more so than dual procs). This is true for most benchmarks I have looked at; thus numerically intensive applications that can't use AltiVec and/or dual processors are likely to suffer on the G4.
In the case of Jet3D, it was easy to optimize for AltiVec. I was able to hand-vectorize about 10 lines of code within the guts of the FORTRAN algorithm and convert the computations to C for easy access to AltiVec hardware instructions. It had a huge effect for not a lot of work. For other more complicated cases, it may be possible to use the VAST compiler tools to automatically vectorize and tie in with AltiVec (VAST has parallel tools also). But in some cases, vectorization is not possible or feasible. In those instances, you're stuck with the processor's scalar performance, and the P4 generally has better scalar performance than the G4 in my experience. One final note: these are my personal views, and do not represent the views of NASA Langley Research Center, NASA, or the United States Government, nor do they constitute an endorsement by NASA Langley Research Center, NASA, or the United States Government
Oh boy.
/just works/.
Those systems aren't *exactly* what I would call comparable. A HD that is 4x bigger, a superdrive, and thats just the stats you posted (I'm sure I could draw it out with things like the Airport Antenna).
"I still haven't seen reliable benchmarks with the dual 2.0GHz facing a P4/3.2GHz with Hyperthreading on"
Veritest disabled HT for tests where the system would be slower, left it on where it was faster. They also enabled SSE2. You can check all of that in their report off of their website.
"if they are three times the cost of a PC, buyers will have a hard time justifiying it."
Apple doesn't sell in the low-range (exempting iBooks), they sell mid-range and up. For those of us who purchase Apple systems, we don't want the cheapest system we can get, we want a system that
Integrate Keynote and LaTeX
If, as widely reported, the PPC 970 goes from 2GHz to 3GHz in the next 12 months it will definitely be more than competative.
It has been along time since I've seen performance increase by near 50% in a year. Takes me back to the 486DX2-66 days.
"It's the height of ridiculousness to say for those 9 lines you get hundreds of millions."
The point, however, is that it gives you a basis to compare any G5 system to any P4 system, and get a rough idea of how they will perform relative to each other MHz for MHz.
Sinced the stated goal of the benchmarking was to be able to gauge G5 performance relative to G4 and P4 performance in fluid-dynamic-number-crunching, this seems a reasonable unit to use. One can easily use it to do cost/performance analyses on current or future configurations of Macs and PCs.
Apple will NOT be hard-pressed to sell their systems. People will be clamoring more for these systems than they did for the G4. You can't put a price on OS X's unprecedented stability, unprecedented integration with UNIX, and unprecedented UI. In addition, compare the oldest Intel-based computer you've seen working with the oldest PPC-based computer you've seen working. By far, Power Macs give you your money's worth, if only because maintenance costs and other long-term costs are so low. Macs have been well-known to last users more than 5 years...can you say that about Intel boxes?
Ever notice how fast Windows runs? Neither do I - get Mac OS
Did they just use whatever they had lying around? Seems to me that if you were gonna benchmark apples fastest, you'd pit it against intel (or amd)'s fastest. I would have expected to see some 3ghz processors (which are still cheaper than G5 2 ghz processors) in the test.
It's nice to see that Apple will finally have some decent power to back them up, but I have STRONG doubts about their "Fastest PC in the world" claim. If a single 2 ghz g5 more or less keeps pace with a 2.66 ghz intel chip, I fail to see how dual 2 ghz system would outperform a dual 3 ghz (or faster)intel-based system.
Nevertheless, compared to the G4, it does seem that apple is making strides. Congratulations on that.
Though dual processor benchmarks are not presented in detail here, it is worth noting that the G5 system benchmarked at 498 MFLOPS and 0.125 MFLOPS/MHz for scalar Jet3D performance when two processors were used.
That was the above poster's point. Mkay?
</karma burn>
StrategyTalk.com, PC Game Forums
You're comparing apples to oranges. Do you feel that putting Windows on a fast computer is better? Would you consider a Windows XP vs Mac OS X a equivalent trade off? Are you out of your mind?
Even at twice the price Mac OS-X beats Windows. (and I'm not even a Mac person!) These are the same type of arguments Mac people must have to put up with every day. I would buy a Mac. Just like I would buy a Corvette instead of a Chrysler.
Linux is only free if your time has no value. Windows is only free if you threaten to use Linux.
I guess the NASA won't care about the price difference, and will care a lot about performance/herz (you know, mission critical CPUs that would rather run at slower clocks due to radiation issues).
I don't know about you, but if a P4 at 2.66Ghz is cheaper and better (even marginally) than a G5 at 2Ghz, I won't give a rat's ass about the clock speed.
Better scalability? maybe. Just show me the whole computer's performance and price tag.
Talk about performance/price and stop the rhetoric.
Someone please explain to me how 5177 MFLOPS and ~300 MFLOPS are even comparable.
They're not, which is what makes this whole benchmark so entirely useless.
Look at it: The conclusion, basically, is that there's no point in running CFD code using scalar FP. So why didn't they port their code to SSE2? P4's, and particularly the new 800MHz FSB P4 get data through SSE2 code like there's no tomorrow.
Nah, I'll listen when someone compares SSE2 and AltiVec properly. Until then it's just more blah. Don't get me wrong, I'm rapidly turning into the biggest Mac fanboy you've ever seen (Cocoa, since you ask) but the G5's are not the quantum leap Apple are making them out to be. Back in contention? Sure, but I promise you a dual Opteron 2GHz will blow the doors off a dual G5.
Dave
I write a blog now, you should be afraid.
Yeah, yeah. Add 120GB hard drive, firewire, firewire 800, optical audio in and out, a fricking Pioneer DVRA05, PCI-X slots, AGP 8x, dual-head video card (dvi, dvi), and a full movie editing and dvd-authoring solution...and adobe's albulm software (to match iPhoto). Oh, and also try to stick 8GB of RAM in the Dell. And to be fair, also put XP Professional on the machine.
For crying out loud, if you can't afford a Mac, keep it to yourself and buy a Dell. And don't give me any more of this "in this economy" B.S.. People are still buying brand new gas-guzzling SUV's "in this economy."
"The code used was developed on a G4, and the vector tests use Altivec, so these results are what you can expect for an optmized application."
They give both the optimized and the unoptimized results.
Integrate Keynote and LaTeX
you should change the theme pic from the old g4 to the new g5
What are you talking about? Are you looking at the same graphs I am? A midrange p4 beat the highest end g5 available. How can you say that the g5 'won'?
autopr0n is like, down and stuff.
Actually they did mention SSE2, and weirdly enough, said it made things slower, and so they didn't use it
Well this story really makes me nervous. After taking in the comments from fellow /. readers, I am now convinced that Apple has conspired with the NASA Langley Research Center to make the G5 CPU look fast. The real clever bit was how they only used one CPU, and and unoptimized compiler, just to make it look "authentic" I suppose. Clever bastards. But they'll never fool me. They have my crappy ass Wintel box when they pry it from my numb, pale fingers.
Apple claims 15.7 for the Dual 2GHz G5, and the 3GHz P4 getting an 8.07. NASA gives the Dual 2GHz G5 498MFLOPS and the 2.66GHz P4 255MFLOPS.
If you use your math skills: 15.7 / 8.07 about equals 498 / 255. So therefore we can draw the conclusion that they have similar results.
Now, NASA only used a 2.66MHz P4 while Apple used a 3GHz P4. Although remember NASA's figure that the P4 had 0.096 MFLOPS/MHz? Give the P4 333 more MHz, and you find it has about 286.968 MFLOPS. NASA also suggests a 20% performance increase can be expected with compilers that take advantage of the G5.
Although, even without this increase Apple's benchmark and NASA's benchmarks are very close. Which would lead one to draw the conclusion that Apple's benchmarks were in fact valid.
I should also note that a P4 would not perform as well in a dual system as the G5 does. So your 500 MFLOPS number is a little rediculous. The G5 which is an amazing dual proc chip saw it's 254 MFLOPS for a single processor (508 when doubled) drop to 498 MFLOPS in a dual system. And the P4 isn't designed for a dual system, doesn't support HyperTransport, etc.
Dak
Who on earth would spend thousands and thousands of dolars on high-end hardware and then not even bother to use the compiler's optimization flags? Sheesh.
autopr0n is like, down and stuff.
- The Mac has a DVD-R/CDR drive
- 120 Gig HD is just a little bit bigger than the 40 Gig
- The "free" monitor is only worth $100 to Dell
- The Mac includes video editing software
- The Mac's video card is dual output with ADV and VGA connectors, dual-monitor capable
An "equivalent" Dell system is nearly $1400, according to their website, if I start with your example as a basis.Sorry, but it's worth $1000 to me to have a computer with a better ROI and no Windows.
NetInfo connection failed for server 127.0.0.1/local
You are aware of this system where people make comments, and then other people can comment on their comments. My reply was to the comment above, not the article.
autopr0n is like, down and stuff.
NASA's study found the Dual 2GHz Power Mac G5 to score 498 MFLOPS for their Jet3D performance. A P4 running at 2.66GHz scored 255 MFLOPS: a 195.3% performance advantage for the G5 in this test. If we assume a direct correlation between MHz and MFLOPS for the P4 (which would actually overstate the performance of the P4) and increase the P4's score by 12.782% this would give the 3GHz P4 a score of 287.594 MFLOPS. This is still a 173.16% performance advantage for the G5, and NASA states that a 20% increase in performance for the G5 would be reasonable "when G5-aware compiler tools become available."
etc. Please try not to jump into discussions without knowing what people are talking about.
autopr0n is like, down and stuff.
Real Translation: 0.4% slower, at 75% of the clock speed.
Real translation fastest G5 0.4% then mid-range p4. Don't forget it was you Mac zealots who screamed 'clock speed doesn't matter'! You could probably make custom hardware that would chew through this application at 100mhz if you really wanted to.
autopr0n is like, down and stuff.
what kind of RAM is on the Dell? Not the high end RAM that's on the Mac, I'm sure.
A DVD-R on the Cac is far superior to the CD on the PC
I've seen those $50.00 monitors... I'd rather buy mine separate
Don't forget to add FireWire to the PC
And the Dell comes in a case that's a bitch to open and work on, so figure another $100-300 on a REAL case, not some plastic $60 PoS.
Sure, if you want the cheapest piece of shit you can buy, you can do better with the PC. But if you want something that you can admire, be productive on and work on with ease, the Mac is a better choice against comparable high end PCs.
I just love it when PC myopics try to compare a bottom of the line PC vs a mid range or high end Mac. They always forget that you GET MORE WITH A MAC. The SuperDrive, the case, the RAM, the larger Drive, built-in FireWire and on some models, wireless and bluetooth.
You can't compare Apples and Oranges. You have to re-spec the PC to match what comes stock on the Mac to even be in the ballpark for a valid comparison.
Tell you what, let's split up Intel's market share 2:1 Athlon:PPC, how's that for fair?
Some applications are going to do better with the G5, no doubt others will be speedier on the Athlon. From what I can gather, over the next 12 months, the G5 will go from 2Ghz->3Ghz, Intel will go from 3.2Ghz->3.6Ghz but I don't know what the Athlon's are roadmapped for. If the 2Ghz is competitive with today's Athlon and the roadmap (IBM's roadmap this time, not Motorola's) has the G5 ramping up speed significantly faster than the Athlon, it makes sense for people in a certain part of their upgrade cycle to go to G5 even though today's Athlon might be faster.
How about a more fair comparison? Namely, between similarly configured high-end single-processor systems:
Apple PowerMac G5:
1.8GHz PowerPC G5
250GB Serial ATA - 7200rpm
SuperDrive (DVD-R/CD-RW)
512MB DDR400 SDRAM (PC3200)
Mac OS X
AppleWorks
ATI Radeon 9800 Pro
56k V.92 internal modem
No Monitor
$2874
Dell Dimension XPS:
3.2GHz Pentium 4
200GB Ultra ATA - 7200rpm
DVD+RW/DVD+R/CD-RW
512MB DDR400 SDRAM
Microsoft® Windows® XP Professional w/ Microsoft® Plus!
Microsoft® Works Suite 2003
ATI Radeon 9800 pro
No Monitor
$3062
And if you are to believe the benchmarks, it seems that Apple is selling the faster system for a lesser price than a similarly configured Dell.
Apple has never competed at the low end. It is not starting now.
-Mike
Schrödinger's cat is not amused—maybe.
The unit MFLOPS/MHz is a little weird. Let's simplify.
/ /
/ MFLOP \
| ----- |
\ S
-----------
/ MCYCLE \
| ------ |
\ S
Multiply by the reciprocal...
MFLOP S
----- * ------
S MCYCLE
Millions cancel, seconds cancel...
FLOP
-----
CYCLE
So it seems that this unit is equal to 1 floating point operation per CPU cycle. That makes a little bit more sense as a unit.
I think the thing that most people on /. seem to keep missing is this: MacOS X and Linux both use GCC as their primary compiler. The Linux kernel is compiled with GCC, as is Darwin. Most software for each platform is compiled with GCC.
Now, with all these Linux-heads around here insisting that Linux is faster than Windows on x86, you'd think GCC for x86 might be a good compiler. Certainly the SPEC tests Apple (and Veritest) did with GCC on the G5 with OS X and the dual Xeon Dell with Red Hat had to have been a valid comparison between those two situations.
I also keep seeing all these comparisons to Dell computers without full specs of the Dell. The base configurations for the PowerMac G5 is positively loaded. How many $500 Dells come with Gigabit Ethernet? How many have the same level of engineering into the thermal managment?
Only time will tell for sure. In the mean time, remember that IBM will be producing blade systems with the 970. We'll get a chance to compare those as well eventually.
--
The internet is the greatest source of biased information in the history of mankind.
I found it interesting that the G4 and P4 have similar MFLOP/MHz. This argues that (at least for this kind of test) the G4 is not more efficient per clock cycle than the P4, as some have tried to argue. On the other hand, it looks like the G5 is.
This is also a useful metric if you want to extrapolate performance of anticipated systems with higher clock speeds.
I recently got the chance to do a testrun, doing some airflow simulation on a G5 1.8GHz demo machine, and with altivec optimizations it clocked in at roughly 2100MFLOPS average for 5 runs(I could probably get better results with a better compiler though), while the dual Opteron 1.8(which the place where I did the testrun has bought 10 boxes of for their renderfarm), running Suse Linux, and my program re-compiled for x86-64 and SSE2 performed at about 2960MFLOPS average, but that could probably be improved with a better compiler too, but I had to use GCC at this time. Both machines had 4GB RAM btw.
what i'd like to see is some G5 versus AMD XP+ 1.5GHz to 3.2GHz reviews, since P4 seems to be a dog compared to P3.
No, let's make your "summary" a little more accurate. The 2Ghz G5 is *very* slightly slower than the older 2.66Ghz P4 -- when running unoptimized code on an unoptimized version of OS X.
As far as I'm concerned, all of this benchmarking is meaningless at this point in time. The G5 is simply too immature of a CPU to benchmark until the software catches up with it.
This same problem came up when Intel first released the Pentium 75Mhz CPUs. Nearly everyone said "This thing isn't even quite as fast as my 486 that costs a lot less!" This also happened when Intel first added "MMX" support to their chips, and when the PIII's first came out as successor to their PII.
Any time you believe the early benchmark results on a new line of CPU, all you find out is that you're not going to get immediate gratification as an "early adopter". So what? This is common knowledge.
Apple fans have every reason to celebrate with the G5's release. It gives Apple a whole new starting point on CPUs they can "ramp up" from, instead of struggling to squeeze some meager improvement from the maxxed-out G4 chip. If you want to use Apple's OS and Apple compatible software - your only option is an Apple computer. Ultimately, that's the key, and explains why some people (myself included!) shelled out thousands for a system (dual CPU G4) that doesn't outperform Intel's best offerings. I want to do things I can't do with anything Intel sells, such as run OS X.
...because it gets more miles to the gallon.
160GB Serial ATA, SuperDrive (CD-RW, DVD-R), Firewire 800, USB 2, ATI Radeon 9600 Pro...
Plus all the standard things that apple don't actually mention in their specs any more of course. 16 bit sound, gigabit ethernet, and so on.
And that's without the free software. iMovie, iCal, iPhoto, iDvd, etc., (which you may or may not want, so adjust value of system accordingly).
The thing with your comparision, is that you are comparing the G5 processor with the P4 processor. Now try comparing the G5 system to a PC system with a P4 with equivilant specs. Hard drive, DVD writer, latest firewire, gigabit ethernet, etc, and you soon see the PC cost more.
And that's without touting the advantages of Mac OS X and all the other points that have already been made in this thread...
This is Apples advantage and disadvantage. Because they won't sell you anything less than top quality in their pro line... And that costs!
People seem to miss the big picture some times!
Eddie, are you kidding?
a single 970@1.8Ghz scored 2100MFLOPS and a DUAL opteron@1.8Ghz scored 2960 !? I assume your code was multithreaded/multiprocess? Your's sounds like a much more interesting test than nasa's - could you give any more info?
Except the Dell would beat the Mac. If we generously give the 2x2Ghz Mac 1.5 times the performance of a similarly spec-ed out 3.06 Ghz P4, (based on the fact that Intel benchmarks show 20% better than Mac and Mac benchmakrs show 30% better than Intel). assuming an extra processor increases performance by 80% (also being generous to Mac, all their benchmarks show higher than that, which makes the single processor systems look flimsy), that makes the Dell 40% faster.
Plus, I configured the machine you're talking about, and it was $2,439, which makes it 14% cheaper than the Mac as well.
Actually, if you would read the link I posted, and the sentence I mentioned below that where I added a CD-RW and a 120GB hard drive, I did configure a similar system for $919.
So you what the whole enchilada, eh?
Dell Dimension 4600
Pentium 4 2.66GHz
512MB RAM
200GB hard drive
4x DVD+RW/CD-RW
16x DVD-ROM
128MB ATI Radeon 9800 (which is higher-end than the video card in the $3000 Mac, and MUCH nicer than the one in this Mac)
XP Home
$1448 (And that's still including the free 15" flat-panel monitor. To those who said this monitor is worth "$50", please tell me where I can buy a 15" flat-panel that is even half as nice as this one for $50.)
Now, if you want to say that Macs are worth twice the price, that's fine. I don't necessarily agree, but that's a value judgement (opinion). But to argue that the systems aren't comparable -- not only are they comparable, but the PC is 2-3x cheaper with a monitor. This particular one also has a larger hard drive and a better video card, and it's still averaging half the cost of a Mac.
On a purely cost basis, the PC wins, hands down. Is the Mac worth twice as much? That's up to individual buyers to decide. I'm just pointing out the facts.
Simpli - Your source for San Jose dedicated servers and colocation!
"...a testament to the quality of the code several lines are still in use ..."
What percentage of the original lines?
You're forgetting the G5 is dual CPU, these test are on single CPU mode.
(\_/)
(O.o) This is Bunny. (> <)
I think that NASA has requirements for computers that cost $10k sometimes. For a lot of things, the p series is too pricey.
You can get dual-proc Xeon and Athlon PCs if you want.
autopr0n is like, down and stuff.
So this proves the Mac is an :-)
excellent computer/Mhz
I would like to know the price per Mhz, and
then the price of a Mflop in all systems.
Z.
By a whopping 0.4%, and with one of the G5's processors disabled. You can spin it any way you want, but the clear fact is that with the G5 Macs are competitive in CPU performance again. I don't see why this disturbs you so; competition is good.
How to solve most of our problems: 1.Lots of nuclear plants. 2.Cure aging.
If Apple had release at least a dozen G5's to various internet review sites that specialized on hardware, and specified which benchmarks they MUST include, none of this anger would be directed at Apple.
Instead, well see the true prejudice these so called "unbiased" hardware review sites have towards all things NOT within their interests.
I would have given the first box to amdzone.com. Then everyone could blast them for their subjective view on all things not AMD, and leave Apple alone, or even console them for being the victim.
They couldn't get a hold of a super-cheap Athlon XP or even a sub $1500 Opteron server, but they got a hold of an Apple G5 which isn't even available for sale???
Puh-lease... The gripe with the SPEC benchmarks was that Apples numbers for the competition were WAY below the OFFICIAL numbers, ot that Apples numbers for their own equipment was crap.
Jeeze...let's at least wait till these things are SHIPPING.
[RIAA] says its concern is artists. That's true, in just the sense that a cattle rancher is concerned about its cattle.
The Jet 3d website was /.'d ....
It's Christmas everyday with BitTorrent.
The g5 is 64 bit and can address a shitload of memory. Can we please see comparison against 64 bit apps ? Compare to t5he Itanium or Opteron please.
Apps were not optimized for the g5, in real world of 3d graphics, engineering, and science, optimized apps would be used.
It's unfair to apple to have people benchmark it without utilizing the full capabilites and feature set of the processor.
"...what you'll find is that when you buy that shiney new G5, it actually has this little piece of paper inside..."
Which is how it REALLY smokes the P4.
(tig)
"We do not inherit the land from our ancestors"
"We borrow it from our children"
Ignorance and prejudice and fear
Walk hand in hand
The Portland compiler traditionally underperforms the Intel compiler, and often gcc as well.
Does it hurt to hear them lying? Was this the only world you had?
Uhhh, looks like someone didn't read the article...
Actually, you're both correct, and you're both missing something.
.18u, .15u, .13u, etc) means there's less time necessary to charge the capacitors, and thus speeds are increased. (This is only one aspect, and I'm no expert here).
Originally, the pipelining was segmented based on the I-Fetch, D-Fetch (register/etc), Exec, Reg-Write-Back, with expensive floating point doing with different timing considerations (externalized or delay-locking multi-stage execution). Then they started sub-dividing each of those stages (especially in CISC archetectures). Now its common to see 15 integer execution pipeline stages - either with shared resources, such that you can only have one divide occuring at any given time (early P-I, P-II, P-III), or with fully independent/concurrent resources (AMD's Athlon).
The addition of the pipelinable-stages between I-Fetch, D-Fetch, exec, and WB was somewhat trivial, because prior to pipelining, there were still seperate events on seperate clock-ticks with inter-stage latching. However, in CPU's with exec-stages that are pipelined, you are introducing additional latches that cause additional undesirable propagation delays.
So a 15 stage integer multiply unit (excluding fetch/WB) has 15 x [guestimating] 4 propagations of additional latency over a single-stage I-unit. If there are resource-based stage-interlocks, or worse data-dependencies, then the pipelining is useless and you're totally hit by the excess propagation delays.
Still, marketing being what it is these days, adding more stages means less propagations per stage, thus less worst-case propagation time, and thus higher clockability (all else being equal; temperature, etc).
The P4, however, compensated by double-clocking the core integer stages, so the number of advertised stages is somewhat misleading.
On a side note, due to the latching in pipelining, you're definately doing more total work for a given instruction. And more importantly, the designers have to think of totally different logic-algorithms to efficiently pipeline than to single-stage. My guess is that the pipelined versions will always be less efficient (especially considering that not all stages will fully utilize their allotted clock-time), and thus there's an additional loss.
Ok, so this supports your post, but here's the part about power/heat.
There are two types of transistors used in modern CPUs (everything past the Pentium). BiPolar and Field-Effect. The CMOS-FET refers to Complementary Metal-Oxide-Semiconductor Field-Effect-Transistor. This acts similarly to a capacitor in that there is a charge and discharge time with little waste current, and power dessipation is typical V=IR, Pwr = I^2 * R. The gate capacitor charge-time is the killer, and what limits switching speed (and thereby clock-speed). Shrinking the area of the capacitor (related to the micron-size stated,
There's another way of reducing switching speed.. Increasing the amount of current running through the wires that ultimately charge/discharge the gate-capacitors. FETs are poor amplifiers, but BiPolar (while more complex and harder to make small) are phenominal. In addition to their complexity, Bipolar also are power hogs. While a FET only consumes power while turning on or off, BiPolars are always on, consuming power (there is current bleeding from the switch). So what often happens is that designers sprinkle BJT's here and there to amplify the current (at the expense of cost/complexity and power-dessipation), and continue using FETs everywhere else.
The bigger and greater number of BJTs that are used, the faster some heavily loaded FET gates will charge and the quicker their switching time will be.
If you up the voltage on a CPU, you're enhancing the amplifier's ability to charge the capacitors and thus gives you more safety-room to up the external clock-speed.
Again, this deviates somewhat from my knowledge domain, but you can often merely co
-Michael
Plus or minus 3dB represents a doubling or halving of the power, respectively. However, quietness or loudness is a subjective quality. Most statistically normal humans seem to agree that plus or minus 10dB doubles or halves the apparrent loudness. Psychoaccoustics bears no relation to math or physics.
3dB is only a factor of 2 for power. It takes 10dB to get a factor of 2 on loudness/quietness.
16 bit sound
Considering my G4 PowerMac has 24bit audio, I think it is safe to say that the G5 have at least that too.
sin(6cos(r)+5A)
I think the operative questions are:
Is it easy to optimize for Altivec?
Is it easy to optimize for SSE2?
Suppose you have a year to complete a series of simulation runs. After spending a few months writing code, you discover the intel machine, using code compiled by icc is 20% faster than a G5 coupled with gcc.
Now, you could start your runs. Or you could spend another month or so hand vectorizing code for a tenfold increase in speed (at least on the altivec platform). Is it as easy to produce such speed gains by hand optimizing SSE2 code?
We'll only stop running benchmarks until Intel wins.
We'll only stop counting until Bush wins.
Buncha damn liberal hippies anyway...
What are you talking about?
Single CPU Jet3D Vector Benchmarks - MFLOPS/MHz
G4 1.25 -> 1.29
Single CPU Jet3D Scalar Benchmarks - MFLOPS/MHz
P4 2.60 -> 0.96
Sure, the G4 using the scalar benchmark is just a bit faster, but as soon as you use Altivec the G4 beats the P4 on a per clock basis soundly. That is what all those arguments are refering to. Thanks to Altivec the processor is more efficient.
sin(6cos(r)+5A)
The G5s uses dual processors more efficiently than pentium systems use dual processors. The load balancing tends to be more even, from what I can tell. In fact, the architecture was, to some extent, optimized FOR dual processors. Additionally, the G5 is designed to be a good floating point/vector processor.
====
Crudely Drawn Games
I could not find these benchmarks anywhere on NASA's site. The benchmarks are only posted on this persons personal home page.
Or could someone reply with a direct link to the NASA page instead of the guys ISP homepage.
Otherwise I can't take these benchmarks seriously at all.
http://www.kubuntu.org/
you could spend another month or so hand vectorizing code for a tenfold increase in speed
You could do, or you could spend a month being scared of it, then actually get it done in a couple of days (when push comes to shove), which is what I did.
I recently managed to get MMX code into some software I'm wirting using GCC instrinsics and, y'know, it wasn't that hard. The hard part is architecting the application as such that the data just happens to be hanging around in exactly the right structures in the first place.
Dave
I write a blog now, you should be afraid.
How is a dual Athalon going to "trounce" a G5 when the 970's trounce a dual Xeon, which will trounce a dual Athalon? I wont argue that you can build a cheaper dual Athalon than buy a dual 970 from Apple, or that you might get more price/performance by going AMD, but trouncing the new Macs? ...bullshit.
As far as your unprecidented overuse of unprecidented, I don't think so. As far as stability goes, there are still systems out there that are far more stable than OS X. Pit OS X against a system running VMS and OS X looks buggy. As far as UNIX integration, take a look at a Sun, or an SGI workstation for real UNIX integration; while you're looking at that SGI workstation, examine it for real case design that looks good, yet isn't over-flashy. As far as UI goes, the Mac UI has always bugged me; it's intangible, I know, but simply put, the interface has always left a bad taste in my mouth.
Of course, I may be a bit biased here. I'm clamoring for a nice Athlon64 system, which will be coming out at about the same time as the G5, and will actually be affordable. I'd be willing to place money on a good Athlon64 system being 1/2 the cost of the cheapest G5 at introduction. I just see AMD's 64 bit line giving the better bang for the buck than Apple.
Marxism is the opiate of dumbasses
I thought that P4s didn't do SMP which is why you see single P4 and dual Xeon comparisons.
The big deal with the Athalon will be the upcoming Athalon-64 which will likely compete more with the G5. From everything I've seen the Opteron is already much faster than the G5. Intel, I don't know what they are going to do. Certainly the P4 will continue to improve. I've heard some exciting things about upcoming chips but the Itanium clearly isn't aiming at the same market as Apple or AMD. (Well, some overlap, but I think Apple has more work to do to get the workstation/server market)
You have got to cut down on that bad crack. Altivec is generally acknowledged at the one shining star extracted from the entire Motorola mess. It's a much better implementation when you have code optimized for it (competing with code optimixed for Intel/AMD's version of SIMD).
Is the Intel compiler available under Linux? Will it make Linux-compatible object files or binaries?
The free version of Intel Fortran comes with this caveat
Sometimes, timely response to compiler issues is invaluable to a
scientific programmer. If a compiler produces a slight inaccurate result
in a game, the programmer might not care very much. The same inaccuracy
in a scientific simulation could mean that a year's worth of runs will
have to be thrown out.
Damn liberal editors caving int those artsy-fartsy mac users again...
How dare you print a story that derides the paragon of the IT hardware industry, Intel, as being second best to a processor installed in a machine maniufactured by those long-haired elitist at Apple.
And to think that you would accept the word from those commie "scientist" down at NASA, who sit around doing nothing all day except "thinking" and playing with thier toys at taxpayer expense.
Shame, shame, next thing youl be tryin' to convince me that Linux OS is more secure than the hallowed work that our hero in capitolism, Bill Gates, does at microsoft. It's bad for the economy, I tell you. All these long haired, smelly, weirdos keep messing with our good old American way of life. What, you say you want a choice? You have a choice! You'll choose from what choices we here at the Central Office tell you to choose from!
Take your Open Source, MacIntosh, OS-X (You don't think I know whaty OS-X really means, do you, but I do, smarty pants, I do indeed..), and whatever else it is you keep trying to give us and go back to Russia, or wherever it is that you came from. Who invited you to our red-blooded American industral party anyway? Nobody! That's who!
Why you pnko, beatnik, hippy, givin' it all away, don't think you need the establishment faggo...
and so on,
and so on,
and so...
Read, L
What is the difference between a PC and a workstation? I think its pretty clear the term workstation as a distinguishing label died long ago, some time after "PC" performance on the desktop destroyed "workstation" performance.
I they wanted they could *easily* have got one.
How? (you might ask)
If someone from NASA had sent an email to marketing@amd.com; basically asking*:
Howdy!
I work at NASA and do some testing of some new hardware, among it the new G5 Powermac.
Could i borrow a Opteron for some testing?
AMD would have sent it on the day.
*Or a phonecall to any of these numbers.
Disclaimer: I don't work for AMD or any other related companies
Melius mori in libertate quam vivere in servitute.
Wrong Conclusion!
According to this pdf (page 13) G5 @ 1.8GHz has 1051 SpecFP.
At the same time Power4 @ 1.7GHz has 1598 SpecFP !!!
It is very clear that Power 970 (G5) is much-much slower in floating point than it's Power4.
Also if you had a previous version of windows then you simply purchase the upgrade which is around $200 and if you find some decent dealers it can be had for a lot less.
:)
OS X costs around $100 for almost each upgrade. I started out with 10.0 which was extremely buggy. It came with my G4 which also had os 9. At the time i just used os 9. 10.1 was free i think and it should be as it's just some bug fixes. 10.2 costs another $120 something but it was the first really useable versions. Almost crash free for the first time. I haven't bothered with 10.3, yet... is it free
But the software cost really isn't big because if you go amd then you'll get huge savings over apple no matter what. XP would have to cost at least $500 for macs to be a better price value.
Hmmm... Pie...
I'm sorry if i seem a bit agitated but i'm getting tired of all this misinterpretation. If you read that article it basically says that the p4 2.66 ghz (old gen as the 3.0 ghz 800 mhz fsb stuff came out some time ago) is still beating a new chip that will come from apple. I think there were some tests in which apple won but most were when the dual cpus were turned on.
EVERYONE should know that by now that intel makes processors that do less per mhz but to reach a higher mhz level. Infact even p3's do more per mhz then p4's not to mention all of amd's cpu's since the original athlon. This benchmarks clearly shows that apple doesn't have a chance again the current 3.2 ghz p4 cpu's with 800 mhz buses. If you take overclocking into the picture (can't oc macs) then apples are just blown out of the water.
Besides in the next few months the biggest cpu launches from both amd and intel since a long time ago will happen.
Hmmm... Pie...
If you payed attention to any hardware comparisons of the latest macs (g4's verses p4's) then you'd notice that it was always getting trounced.
Macs are great but hardware performance isn't the reason.
Hmmm... Pie...
Do not forget, the G5 has 64 bit processing while everything else has 32 bit processing. You cannot draw the same comparisons. Wait until there is a 64 bit processing for the P4 or AMD processor, then compare.
Look at what nasa and what apples uses. Low end dells with slow pc2100 ram for comparison. NASA compared that with a product not even out yet. The PC you built for $3000 will beat the mac that won't be out for months. The 3.2 ghz p4 has the new 800 mhz fsb that the 2.66 ghz one compared in the article didn't. In reality by sept. we'll have the athlon64 and possibly intel's prescott. So no matter how you look at it, this benchmark is irrelevant unless and pc's are still cheaper.
Hmmm... Pie...
Although I know it was asked, but why aren't there AMD processors? What about (DEC, Compaq, HP, or whatever the hell it is now) Alphas? Sparcs? Also, why rate in MFLOPS/MHz? CpF is a lot more easier to understand. CpF is Cycles per Floating Point Opteration. As far as I know, the Itaniums are the most efficient, while running in native mode. Whatever the G5 has to offer, doesn't even compare.
Now, getting back to AMD processors, I think that they should have been included in there. Although, since the Athlons ARE more efficient in pretty much everythan that the Pentiums, they could have easly gotten high scores. Perhaps Langley doesn't want to show how equal, or worse even, the G5 is compared to a sub-100 dollar USD processor. No processor in existance compares in performance for price to Athlons.
Well, even though there is another, perhaps even credible to some, source for benchmarks, but I still say I don't trust them one bit. Call me old fashioned, or a stubborn 16 year old, but I don't trust any benchmark done by anyone. I like to test things before I buy them. Why doesn't anyone else? Why trust people that are funded by the companies that make the product they benchmark? Well, anyway...
This great article discusses the general specs of the CPU and it's architecture.
A really well written (but highly technical) article. As you can see, IBM designed it to beat Sun and Alpha (still alive at the time), not Intel. As far as Apple is "involved" in the development, it's in the latter stages of implementation rather than design itself.
Please note,
gcc + g++ for MacOS X on G5 platform = $0
Intel's C++ compiler for pentium based platforms = $399
prices
Now factor in Operating Systems prices, and general software.
MacOS X 10.2 $129 Windows XP Professional $143
iTunes $0 MusicMatch Jukebox Plus $19.99
AppleWorks $0 WindowsXP Office $297
Software Total: MacOS X: $129 WinXP: $459.99
(prices taken from www.newegg.com)
Error 407 - No creative sig found
I showed a photo of the new G5 to my dual 1.25 GHz G4, and said ...on your mark....get set....GO!!!, and neither of them moved an inch.
I had a stopwatch, start/finish line, champagne girls on the podium and everything.
I have to admit, after all the talk recently, I was pretty surprised at the results.
The G5 has requested another match, later in the summer, so we'll see if things come out differently then.
Seriously.
They included the MFLOPS/Mhz graph that makes the G5 look great, but they forgot the MOST IMPORTANT graph of all: MFLOPS/Dollar.
What you get inside the box are these three little coupons that can later be redeemed if and when Apple decides to have a discount for previous OS owners. If Apple decides to charge full for it, the coupons are useless (on rare occasions they're also used for certain rebates). I have some of them from my new iBook purchase, and they're essentially scrap because they're charging for Panther. (Not that I necessarily have a problem with that; I'm more than willing to pay for an OS upgrade of this magnitude [not to mention I get educational pricing])
Want Slashdot headlines on your site? Try SlashHead
And a 2.66GHz P4 barely edged out the 2GHz G5 in scalar floating-point operations, a task it isn't even particularly good at per-clock. If you read the great-grandparent, I think you'll notice that he mentioned the dual CPU score of the G5. So while you're reading the fucking article, you can consider reading the fucking comments, too.
Apple zealots, I have all the time in the world.
You have exactly 314 seconds to come up with a less retarded plot.
The reason why there are no RH 8 or 9 ISOs for PPC is that basically they are leaving development of the PPC branch of Red Hat to the makers of Yellow Dog Linux. Which, right now, is G5 ready and is basically RH9 for PPC. http://www.yellowdoglinux.com/
Knowledge is power. Knowledge shared is power multiplied.
Just want to share a story, There is once a father and his son who owns a donkey. They were walking towards a village with the son riding the donkey and the father towing it. When they entered the village, the people starts critizing the son for being a useless son. They decided to go into the village again later but this time the father is riding the donkey and the son tow them. In the village, the villagers start commenting that the father is a cruel father for letting the son tow the donkey while he rides it. They thought for a while and when into the village again; but this time both tow the donkey behind them. Guess what? The villagers call them stupid. Donkeys are to be ride. In the end, both of them carried the donkey. Think about it..
Are these 1 or 2 LSB errors or are the answers flat-out wrong? References please.
Contribute to civilization: ari.aynrand.org/donate
Yes, I know. I was trying to be humorous. As in, when the little piece o' paper inside the G5 catches fire etc etc. Guess I failed. Again. *sob sniff sniff sob*
(tig)
"We do not inherit the land from our ancestors"
"We borrow it from our children"
Ignorance and prejudice and fear
Walk hand in hand
Does anyone remember the dual-ppc that the BeOS people were originally selling? Anyway, you could turn off one of the processors but if you wanted to, you could turn them both off. I think they evenrtually removed this feature. =)
-
Insightful posts by people like you. I also came to the same conclusion. I'm Glad to see that some people can figure things out without it being presented to them on a silver platter.
I used to think that my brain was my favorite organ, but then I remebered who was telling me this. --Jack Handey.
Well.. maybe. Or Maybe not. But Definitely not sort of.
I'm talking about the non-vectorized benchmarks:
Numbers were 0.096 P4 vs 0.105 G4 MFLOP/MHz for the non-vector comparison
As stated in NASA study,
In other words, for the many uses that don't involve Altivec, the G4 and P4 actually are comparable on a MHz basis, despite the substantially different designsIs it just me, or do other people notice the growing trend of Intel really sucking hind tit? Sure, they are over 3Ghz, but AMD and now even Apple are sporting processors that are more than one thousand mhz slower, yet, over all, perform the same or better. I figure by the time Intel hits 4 or 5 Ghz, Via's Cyrix processor will be just about 1.8Ghz, but will have been worked to where it's performance will be on par with the Intel offering. Ol' Intel had better double their advertising budget, or even Dell will have to start offering non-intel systems.
By the way, no, I don't benchmark systems I use, as that, in my mind is like putting a car on a dynometer. I mean, who realy cares how much horsepower a machine has, if it's gearing is completely mis-calculated? Based upon actually using systems, I find that G3s seem very slow to use, G4s not bad, P4s a little better, and AMD Athlons wicked quick. Yes, all systems were slightly different clock speeds, but all had 256MB of RAM, except the P4, it has 512MB of RAM. In just normal usage, nothing I've come across can touch the Athlons for performance. However, I also do NOT do video editing, sound editing, etc. I just play a few games, do programming for school, the internet, and such. So no, I do NOT care about how fast something can open Photoshop, or if these systems can do real time video editing. I didn't build/purchase them for that.
There's my two cents, could I use that for a down payment on a new dual G5?
For those who describe their systems as 'boxen', do you order multiple 'boxen' of corn flakes also?
you can remove the $100 or so (CAN$129) you put on for Appleworks and use ThinkFree, OpenOffice.org or StarOffice (free?).. all work seamlessly with Word and Excel files, but Thinkfree Office is by far the best at the job, with AppleWorks 6.2.4 not far behind.
Really.. no Power Mac (desktop) owner is going to buy AppleWorks.. and we really don't want Word.
Will this give me porn a million times faster?
RTFM; please, I beg you.
One must remember that the Apple compiler still isn't programmed to take full advantage of the PPC970 instruction set, including 64-bit modes.
I really think this article is taken out of context. It's not about NASA pimping Apple hardware. I think they were just looking at the G5 processor itself to see what kind of legs it has for scientific computation.
-------- -------- Support Wesley Clark for president!!!
I wish even half of the people on this site posted as well as this. Really, really nice explanation; thanks for taking the time to explain it to us laymen.
I don't understand why the new G5 processor is constantly benchmarked against the P4 and Xeon processors. The G5 is a 64-bit processor and it damn well better be faster than any 32-bit processor. I would be interested to see benchmarks comparing it to an Opteron or Itaniam2. With the already mediocre results against the 32-bit processors I think that Apple will fall short this winter when Intel breaks 4GHz.
That's why I'd never buy a computer from Apple (or a high-end computer from Dell, apparently); I could buy or build more computer than I'd ever want for approximately $2,000 and it would be comparable to those systems (better in some ways and worse in others, most likely).
However, I've never spent more than $1,100 on a computer; in fact, lately I've spent less. And I've still been quite happy with what I've gotten for my money. For me, it's all about getting a computer that fits my needs, and maxing out that magical price/performance ratio. Therefore, my computer is invariably a non-Intel x86 built from parts from (or by) my local generic computer shop.
I actually considered getting a Mac when there were PowerPC clone vendors; I thought it was very cool that Apple was finally opening up its hardware. But even the clones were pricey, and lo and behold Apple squashed them so fast that it didn't even matter. Also, Rhapsody didn't deliver on its promises at the time, so I'm glad I didn't get stuck with an old, overpriced G3, and I'm certainly not going to end up with a new, overpriced G4 or G5.
But hey, no hard feelings; I'm just not in their target market of suckers with too much cash lying around. If I had that much cash and I wanted a PowerPC of some sort, I'd probably look into getting something from IBM instead.
pb Reply or e-mail; don't vaguely moderate.
These prices apply if you're doing your own development; the issue most people are raising has to do with the operating system and off-the-shelf applications which are compiled with whatever the developer happened to be using. For Mac OS X, that's often gcc. For Windows, that may be Intel's compiler. How much money it costs the developer really isn't at issue here.
By the way, AppleWorks is not included with the G5 as far as I know (it's only bundled with consumer Macs, not pro Macs); it retails for $79. In my opinion, AppleWorks is really not that good, compared to MSOffice, so if you're comparing MS Office to MS Office, you should notice that MSOffice for Mac is $499.95, or Word and Excel are $369.95 each.
$x='S24;r)>63/* h@<5+oZ)32"5cz';$me='phroggy'x$];
$x=~y+ -xz+\0-Tx+;print$_^chop$me for split'',$x;
These benchmarks prove once again that while high-end RISC CPUs outperform commodity CISC CPUs per clock cycle, it really doesn't matter, because the CISC CPU runs so many more clock cycles that the CISC CPUs can blow away the RISC CPUs.
;)
Some people will point out that the Intel chip barely beat out the G5. But the intel chip is available *now*, while the G5 won't actually be shipping in quantity until late August, at which point the intel CPU will have dropped in price and been replaced by faster models.
Of course, I would still rather have a new Apple over any more Lintel boxes
True the G5 Macs have 9 fans but given that they have that number in order to have it quieter they need to have them strategically placed and smaller so that the total cooling may not be comparable to what you may think at first when you think of 9 fans which would mean that a P4 cooled with an equivalent cooling power may not be overclockable that much.
"The obvious mathematical breakthrough would be development of an easy way to factor large prime numbers." Bill Gates,
yes, but the article tells us that the G5 does more work per cycle than an equivalent P4. However, considering that the 2GHz (top of the line) G5 does less work than the 2.66GHz (yesterday's news) P4, in terms of raw MFLOPS (according to the article) your point is completely invalid. The fastest P4 is over 3GHz. The fastest G5 costs more and doesn't do as much work per second. Obviously the P4 is the faster chip, and the greater efficiency of the G5 has nothing to do with this argument.
So, in this case, megahertz actually *do* matter, since the only difference between the 2.66-P4 and the 3.2-P4 is clock speed. The factor that kills the G5 in the real world competition actually *is* clock speed.
Oh, and RTFA.
the article tells us that the G5 does more work per cycle than an equivalent P4. However, considering that the 2GHz (top of the line) G5 does less work than the 2.66GHz (yesterday's news) P4, in terms of raw MFLOPS, (according to the article) your argument is invalid and misleading. The fastest P4 is over 3GHz. The fastest G5 doesn't do as much work per second. The G5 does more work per cycle, which is completely different. The P4 is the faster chip, and the greater efficiency of the G5 has nothing to do with this argument because the P4 ends up doing MORE WORK per second.
So, in this case, megahertz actually *do* matter, since the only difference between the 2.66-P4 and the 3.2-P4 is clock speed. The factor that kills the G5 in the real world competition actually *is* clock speed. The analogy would be better stated by saying that increasing the MHz is like boring out the engine; it increases max output up to a point.
Oh, and RTFA.
G5 is a small step for a computing but quantum leap for Apple.
Emacs is good operating system, but it has one flaw: Its text editor could be better.
Actually it's Apple's G5 too. G5 stands for generation five of Apple's Power PC architecture. The machine they tested was built by Apple. These test test not only the processors but the various busses and IO chips at least two of which were designed totally by Apple (The Memory controller & PCI-X bridge).
IBM plays a huge part but you can't run Nasa's software on a 970 alone. You need the whole platform. (Imagine if the 970 was still tied to a 180Mhz bus)
The real test I expect is MFLOPS/$ where $ is a retail price of the desktop being tested.
Less is more !
I remember how Apple was ridiculed when they said their G4s were "super computers". But in fact they were right. There was a government standard for those and they met that. Compared to the ENIAC any old 386 box would pass as a supercomputer. And compared to my iMac DV the G5 very well goes as workstation. But we have to see that standards shift. And for a workstation it just misses features. I mean, come on: No second 8X AGP slot, you must be kiddin' me... ;-P
I'd rather have a choice.
I'd rather have as little problems with my hardware and software as possible. My job is challenging enough for me to avoid all unecessary challenges.
MIPS and MFLOPS are not the only important factor when choosing a computer, just as maximum velocity is not the only factor when choosing a car. A complex machine is worth as much as its weakest link. So I don't think a MFLOPS/$ chart would say anything informative - I'd rather see a 3D chart with "How many problems will I most likely have with this machine" axis versus MIPS versus $.
::Smacks head off table:: Doh. Must have missed that bit. My bad.
I missed the bit in the article where it talks about dual processors. That's what I get for doing everything in a hurry. My bad. Sorry. Guess I was a little overzealous in trying to stamp out the usual FUD against Apple which was a little unjustified on my part in this case.
Yes, my program is multi-threaded.
The dataset I ran the test on was about 500MB in size, thus easily fitting into the RAM.
The simulation does two major things: It gives a simple visualization that it outputs to high-resolution but graphically simple pictures, for easy analysis. The second is that it dumps all the stuff to a log-file, saving coordinates, motion curves, dynamics states etc, that can be imported into Maya for prettier visualization.
Thus it's fairly dependant upon HD too, which is one place where the speed problems could be, since they are using MSI motherboards(bleurgh).
Another thing is the compiler, GCC. I compiled for x86-64 with the switches -m64 -O2 -msse2 -mpreferred-stack-boundary=4(Which sets it to 16 bytes which SSE/SSE2 needs) -mno-push-args -maccumulate-outgoing-args -mtthreads, but I don't think GCC properly read the stack boundary size. The Opterons should be faster than the result I got.
For the G5, the most important switches I used were -mpowerpc -O2 -maltivec -mnew-mnemonics -mabi=altivec.
Each increase of 3 dB is a doubling in volume. Every 10 dB increase means the sound is 10x louder.
;-)
That means the G5s are 1/10 as loud, not half as loud.
Back to Physics 101 for you.
You're quite wrong. MFLOPS/MHz (the "H" is always captalized in honor of Mr. Hertz) can be used to project performance.
For instance, NASA tested a 2.66 GHz P4 system. By using MFLOPS/MHz, they can estimate the performance of a 3.06 or 3.2 GHz P4 system.
Likewise, IBM has promised a 3 GHz G5 within 12 months. The MFLOPS/MHz metric allows you to project the potential performance of such a system as well.
That is why the NASA test doesn't simply look at raw performance. They want some idea where things are going as the G5 gets faster and Intel tries to eke another 0.167 GHz out for a "next generation" Pentium4. (How long did it take to jump from 3.06 to 3.2 GHz?)
I respectfully disagree with ciroknight. There have been times when Apple has had the hardware to take on the Intel world.
The Mac 512K of 1984 at 8 MHz held its own against the 6 MHz IBM AT. Too bad it didn't have a decent hard drive.
The 40 MHz Mac IIfx outperformed the 33 MHz 386 and 20 MHz 486 systems of its era (1990), especially with Apple's accelerated video card.
MHz for MHz, the better Quadras ran circles around Windows PCs (back in the Windows 3.1 era).
In 1994 Apple went PowerPC, which didn't result in an major immediate improvements. Then again, they didn't have the Pentium math bug, which was probably responsible for the biggest CPU recall in computer history.
When the Power Mac G3 was introduced, it was the most powerful CPU on any desktop system. Remember the snail ads? The bunny suits? The BYTEmark results -- and BYTE magazine getting gobbled up by a pro-Windows company and the magazine being discontinued?
The G4 was a breakthrough as well, although the folks at Motorola seemed to believe that Moore's Law was an unatainable goal.
And now the G5 offers more power. About 25% (ballpark) more than the G4, MHz for MHz. About 25-50% more than the P4, MHz for MHz (we'll know a whole lot more when production units ship with a full release version of OS X 10.2.7 -- the NASA and other tests are all using prerelease hardware and a prerelease OS).
Not only has Apple has Intel-beating performance several times in its history, but the original Mac OS was a far more efficient OS than Microsoft's bloated consumer Windows releases (3.1, 95, 98, Me).
Unfortunately, OS X trades a lot of that efficiency for stability and eye candy, making it harder for modern Macs to feel perky compared with the old Mac OS or 2-3 GHz Windows systems. The fast G4s helped, but the G5s are going to take the OS Xperience to the next level.
On top of that, Macs are simply more productive than Unix workstations. You can run high end software, but also AppleWorks, Microsoft Office, SimCity 4, Mathematica, and a host of other apps not available for Solaris, Linux, BSD, and the like.
Sure, along with a 10 page NDA requring not to publish unfavorable benchamrk results.
US-UK-Israel: The real Axis of Evil
Project Builder and Interface Builder and all Apple programming tools = $0. MS Visual Studio .NET Enterprise > $3000.
It's not fair to compare the 64-bit G5 with even the much more expensive 32-bit Xeon, and simply out of order to the crappy P4 which is not even capable of dual CPU configuration.
The dual 2GHz G5 Mac costs $3000, and a dual 3 GHz Dell is $3000, but the Mac is still 20% faster in SPECfp 2000 and 200% in real app tests.
I asked this myself earlier today and hit pricewatch. I came up with $1800 for an E7505 based dual Xeon box in an aluminum case that would be the x86 equivalent of Apple's flagship G5 at $3,300 (I had to bump the stock video card up as I couldn't find 64MB 9600s on pricewatch). Run the numbers yourself. What you believe about the lower expense of high end Macs just isn't true.
chip except a larger memory address space.
Really? It seems unlikely to me that cpus without any floating point processors at all would have higher MFLOPS/MHz than a P4 or G4. Can you direct me to some benchmarks that show this?
This is totally OT, since the FA was about benchmarks done by NASA, not by Apple and not by IBM and not by Intel.
You want to know what a Mac is really worth? Buy any Mac Laptop or Desktop and buy any Brand PC Lapptop or desktop. Use your machine for three years. Then put it up for sale on ebay. See which one fetches a higher percentage of the origional price.
It's called an investment. Case closed.
I can't believe some people. First they say Apple need better hardware... then Apple comes out and delieves on that... and now they're yelling "it can't be true - it just can't". Give me a break.
I don't think it's fair to compare the system performance of the Mac with one of its processors off to the P4. The highest end Mac has two processors for a reason.
"No!" you cry. "That's not a fair comparison! The P4 only has one processor, so to make the benchmark equal, we have to make sure the systems are on equal footing!"
Well, too bad. It's a system limitation that Intel put in. If I buy a TOP END system, I want to see top end performance. The fact that the P4 was so (badly) designed isn't my problem, or Apple's. Their top end machine spanks ANY P4 system you can put to the test because they decided that a good way to make the system faster would be to make sure it supported dual processors.
Now, if you're stricly comparing PPC970s to P4s, and you want strict PROCESSOR benchmarks, then fine, just test the processor. However, even results like this benefit from the architecture of the system. A fast, data starved processor is useless.
I'm uninterested in raw CPU results. I can't do anything with JUST a CPU on my desk. I need a whole system.
This is the highest performing, lowest cost Dell that anyone posted to this discussion:
Dell Dimension 8300: (same as XPS except not blue)
3.2GHz Pentium 4
200GB Ultra ATA - 7200rpm
DVD+RW/DVD+R/CD-RW
512MB DDR400 SDRAM
Microsoft® Windows® XP Professional
Wordperfect
ATI Radeon 9800 pro
No Monitor
$2239
This is the base 2GHz Apple. So, the cheapest, highest performing Apple.
$2,999.00
Dual 2GHz PowerPC G5
1GHz frontside bus
512K L2 cache/processor
512MB DDR400 128-bit SDRAM
Expandable to 8GB SDRAM
160GB Serial ATA
SuperDrive
Three PCI-X Slots
ATI Radeon 9600 Pro
64MB DDR video memory
56K internal modem
And that doesn't list the computer controlled cooling or expensive case, the fact that the Apple is quiet, the gigabit ethernet, the optical digital audio and analogue audio in and out, or any of the engineering that went into making all of those things fit together. For only $700 more, you get an extra processor, superior performance, and an operating system that's actually worth using.
Remember, you get what you pay for...even with Apple. Dell puts togher fine PCs, and they're certainly cheaper. However, (for the first time in a while) the Mac is BETTER.
your skipping the $250 for the 4 year service agreement dell offers. Now I'm not here saying that there isn't a single institution on the face of this, or any other planet that would be more likely to suddenly become cost-conscience then to pass up a service-agreement, what I'm saying is that your dreaming if you think that an institution is going to skip the service agreement, also microsoft works? office'll be another $250. and finally, the P4... how SMP is that? you're going to need to go for the gusto. a P4 SMP, i.e. "XEON" which'll give you a larger price, only 3.06ghz in the pants, and a 533 FSB, though the place for a front side bus in strictly mathmatical benchmarks like these *as I understand it* won't matter much
Apple might have shot itself in the proverbial foot with the nomenclature they chose for OS X.
People equate the "point one", "point two" names as small increases (0.1 of an increase in fact), not the names of the operating system. The small increase comes with the second point - the point releases.
10.1 is to Windows 95 as 10.2 is to Windows 98 (well, if we're being fair, 10.2 = Windows 3257, but there we go, I'm biased).
There were probably as many changes between Windows 95 and 98 as there were between 10.1 and 10.2, and people weren't up in arms about having to pay for 98 were they?
Well spotted, it's an FP IPC (or CPI) test.
If you want high IPC (instructions per clock) then look at Alphas. DEC were the first company to hit 1.0 IPC mean on SPEC, and even had 2.0 IPC before anyone else. Having said that, now Intel has bought vast chunks of Alpha tech, the Itanic is pretty darn high IPC too.
YAW.
Your head of state is a corrupt weasel, I hope you're happy.
610 posts, and only one AC mentioned the cache size difference in the machines? Here are some interesting numbers:
Program use 1024 KB of memory -- that's unusally low.
G5 data cache:
hw.l1dcachesize = 32 KB
hw.l2cachesize = 512 KB
G4 cache:
hw.l1dcachesize = 32 KB
hw.l2cachesize = 256 KB
hw.l3cachesize = 2048 KB
P4:
cache size: 512 KB
It looks like the G4 should walk all over these other processors- the whole dataset fits in cache. One really interesting thing about the dual G5 is that each processor can access data in the other's cache... Since the 2nd processor was still installed, I wonder if its cache was still operating! If so, then this might boost the G5's cache to roughly the size of the data set.
HIV Crosses Species Barrier... into Muppets
Was that the 3.06GHz Xeon that Apple demonstrated, or a slower dual-Xeon system?
Does that include most of the hardware features of the G5? Gigabit Ethernet, FireWire, optical 5.1 audio, CD-RW/DVD-R drive?
A friend of mine ran the numbers on Pricewatch last week and didn't have the same results you did. I don't recall what he found, but for similar specs it wasn't cheaper than the G5.
$x='S24;r)>63/* h@<5+oZ)32"5cz';$me='phroggy'x$];
$x=~y+ -xz+\0-Tx+;print$_^chop$me for split'',$x;
I see a lot of people arguing, so here's my summary/explaination:
- A single G5/970 top of the line is probably 20% slower than P4 top of the line, on unoptimized software. By that I mean regular integer and floating point (no AltiVec), and with unoptimized compilers. The compilters may or may not improve significantly over time, so that shouldn't be afactor on release date, but may pan out later.
- The G5/970 is not designed/optimized to ship as a single CPU, instead the bus, etc. are designed for SMP of between 2 and 8 CPUs (might technically work with more, I don't know). So, you'll see that, in a dual CPU comparison that the G5/970 will scale better than the P4 (xeon) duals, narrowing the performance gap, and in some applications, putting the G5/970 on top.
- With software that is recoded to work with AltiVec, compared to SSE(2) optimized software, the AltiVec performs noticably better, but that depends on the application. When comparing AltiVec to regular (non-vector optimized code), there can be as much as an order of magnitude increase in speed, again depending on t6he application.
So, expect benchmarks to vary by a tremendous amount, depending on the number of CPUs, and how optimized for AltiVec the software is, and how mature the compiler that was used. Since a lot of software can be vectorised, I would expect software on the G5 to initially lag P4 performance, but then to dramatically speed up in some areas.
Also, expect a lot of people who don't grasp this to be getting their panties in a knot!
...will the G5 cruch SETI@home faster than an x86 based solution? Because you know I want to be the first to establish formal contact with the Zeta Riticulans, the Nordic Aliens, or the Time Lords... :)
"Right now, somewhere in this world, Scott Baio is plowing a woman he doesn't love," - Peter Griffin, *Family Guy*
Maybe like, here. Or here. Someday indeed.
I don't consider my self being a Terrier walking yuppie, though i have a mac.... I wouldn't buy a Toyota to drive in, if i could get an BMW... Given both cars would take me from PointA to PointB, one would do it with comfort and style.... I'll leave you to figure out which of the cars im talking about here... However its an incredible view you have, even for a AnonCoward, i must say...
I'll speak some about having actually used SSE and Altivec. I've used Apple's gcc altivec flags, and SSE (and SSE2, but since Altivec doesn't have 2x64bit FP vectors, it's unimportant) assembly.
Having written some code, and hacking it up to do 4x4 matrix multiplication, I set it up to multiply about a million times. I tested standard unvectorized floating point (on both a G4, and P4), altivec on a G4, and SSE on a P4.
My results? Altivec provided a nearly 4x speed boost (even though I was writing results out to memory nearly every single operation), and SSE performed horribly, running almost 2x as slow. I figured out that this was because I was using "movups" instead of "movaps" (unaligned, vs aligned) but I couldn't get aligned moves to work, because GCC wouldn't align my vector to a 128-bit boundary. (you can aligned the stack to a 128-bit boundary, but then the %ebp register goes on, and it craps out everything else)
After trying numerous things, I just outputted the C code to assembly, and hacked it so my vector was alligned, and then SSE showed minimal performance gains at all.
I'm working on rewriting my library to make it so that I don't have to write out all the vectors every single add/mul/madd. But this thread will be old by then, and no one would probably read it.
I am unamerican, and proud of it!
You can switch the second processor off permanently by simply taking a hammer to it.
10.2.7 is only available to slashdot subscribers.
Tyan Thunder i7505 motherboard with PCI-X, 8X AGP, Gigabit Ether, Firewire, USB 2.0, and 6-channel Dolby audio (including SPDIF out) = $415.
Dual Xeon 2.66 GHz @ $315 X 2 = $630.
512MB Kingston CL2 ECC DDR = $80.
ATi Radeon 9800 Pro with 128MB = $305
Seagate Barracuda 7200.7 Plus - 160GB/8MB Cache = $165
Pioneer DVR-A06 DVD burner (Supports DVD-R, DVD+R, DVD-RW, DVD+RW) = $250
Lian-Li Aluminum Case = $115
TruePower 550 Watt power supply (nice and quiet!) = $110
Add more cash if you want to skip stock cooling to lower noise or overclock by buying a pair of Zalmans or Swiftechs = $100.
Oh yeah, throw more cash at a modem, keyboard, and mouse = $100.
Now that totals $2,270 right now for the x86 equivalent of a $3,300 PowerMac available in September. One could stick with the stock 64MB Radeon 9600 that will come in the top PowerMac and drop the extra $300 off it's price. That would allow you to drop $160 off the PC and it would still leave you with a 128MB card, a fast, quiet drive (8MB cache which does add a boost -- see storagereview.com), CAS 2 ECC memory, HyperThreading for that 4-way feel, an optical drive compatible with all formats, and a cool grand to spend as you please. If you don't care about PCI-X or integrated Firewire you could get the Tiger i7505 instead and save another $150 to boot. If you want this PC to be even cheaper you can wait until the G5 PowerMacs are actually available before buying this dual Xeon box.
I work with os x standard and server almost daily. I meant the server edition rather then addition... OS X has several version is all I was pointing out in my previous post. Yes it costs tons more... but thats not related to anything as does windows server.
Hmmm... Pie...
it's been a while but when i was trying out os x installs and version i remember entering serial numbers. Might have been the server edition...
In any case buy xp pro cds and it will work on all pc's that can run xp as long as you have the correct number of licenses.
Hmmm... Pie...
>> So why didn't they port their code to SSE2?
Probably because it's not worth it. While Altivec speeds up Jet3D by a factor of 10 to 13, there is no evidence that SSE2 can even double the speed.
I'll take the word of the NASA guy over the slashdot rants.
Ok, state your arguments, and I'll state mine. After all, I seriously doubt you can conduct a civil discussion in the first place, based on your preference for ad-hominem attacks instead.
:)
Apple and Microsoft both make piles of money; that doesn't mean I like how they conduct their businesses.
As for the Xerox bit, I admit I never looked into it, but after a little googling, it sounds like Apple did have some sort of agreement with Xerox, or at least "In the end, Xerox got a large block of Apple stock for sharing the technology." (according to Woz himself). That doesn't make it any more innovative, however.
pb Reply or e-mail; don't vaguely moderate.
Apple's benchmarks... gave the Dual 2GHz Power Mac G5 a 194.5% performance advantage over a 3GHz P4 in SPECfp_rate_base2000. The G5 getting a score of 15.7, and the P4 getting an 8.07.
15.7 is 94.5% greater than 8.07 -- not 194.5% greater.
That that is is that that that that is not is not.
You said:
Altivec is nice, for what it is meant for (mainly media type calculations, signal processing, etc.) But scientists will prefer SSE2.
Well I was under the impression, that if I need double-precision floating point arithmetic I use the FPU (which is quite fast with the G5 as you can see in Fig. 1 of the Article). ;-)
Concerning fast vector operations - some problems can be "dumbed down" to take advantage of the faster single precision units and lastly you can do an additional Newton-Raphson refinement step where double precision is needed.
Your claim "scientists will prefer SSE2" was proven wrong by the article, where the Jet3D test suite was only ported to take advantage of the Altivec SIMD unit, not the SSE2 SIMD unit - unless of course you argue that NASA engineers are not scientists
IMHO real scientists use whatever they think is best suited to solve their specific problem.
Code is Speech. No to Censorship.
Well, first of all, I work at NASA as a researcher, doing numerical work (although I'm not a civil servant, and don't speak for them).
:)
I agree that AltiVec is superior to SSE (ie. single precision), but you compared it to to SSE2, which is a bit apple-to-oranges (no pun intended, btw). If the G5 FPU is faster than current SSE2 at double precision, it just proves the well thought out design of the PowerPC architecture (and the unfortunate legacy of Intel's FPU instruction set, which is still a handicap even with SSE/SSE2, due to the need to mode switch).
But SSE2 is still immature, and I expect compilers to improve, as well as chip implementations. Once they do, a more meaningful comparision can be made.
The Intel chips NEED stuff like SSE/SSE2 to achieve faster floating point speeds, whereas the PowerPC can get by without it, thanks to a much better FPU design, and thus, PowerPC makers will probably not spend the silicon to make a double precision SIMD instruction set anytime soon.
I stand by my claim that while most consumer and media software can get by with single precision, scientific computing (ie. large matrix calculations, to be blunt) quite often needs double precision (hell, you can get libraries that use 128 bit long doubles, these days), and will ultimately prefer SSE2. Scientists fuss with single precision SIMD simply because many of their applications can benefit so much from SIMD that it is worth the pain to use single precision (with proper conditioning and verification, etc.) Now that double precision SIMD is available, I can only predict they will want to jump to it, once tools for using it are there.
Granted, if Intel can't make a double precision SIMD unit that outperforms a double precision general FPU like the G5's, for matrix problems, then they don't deserve to design chips for scientific computing.
"It's overkill, of course. But you can never have too much overkill." - Anonymous Slashdot Coward
MFLOPS = (million floating point operations) / second
.
MHz = (million cycles) / second
MFLOPS / MHz = (million floating point operations) / (million cycles)
=> (floating point operations) / cycle
So this unit is deliberately chosen to disfavor the P4 since it measures only number of operations per cycle and and is not a measure of speed or how fast. To measure speed it has to be per second
We all know that the P4 has more cycles per second than the G5, this benchmark is the other extreme of the MHz myth. More floating point operations per cycle does not mean much without considering the number of cycles per second.
A computer that does 10 ops per cycle, and 1 cycle in 10 seconds only does 1 op per second.
Another computer that does 1 op per cycle and does 10 cycles in 1 second, would do 10 ops per second.
Which is faster? I'm sure you don't need a research institution to tell you that.
Nasa compares them only on the ops/cycle metric. And some G5 zealots here tend to agree. What a disgrace to common sense.
"Fighting terrorists with millitary might is like killing a mosquitor on your Dad's forehead with a rifle."
I find it deeply depressing that this is the second week in a row I have had moderator points... only to find that everything is moderated at 5, and, what is even worse, that is exactly how I would have moderated!!! "Use 'em or lose 'em" eh? I think it's the /. Bayesian filter that has a loose screw! I mean, how many people are moderating, really? And how come we are all in agreement??? Like Senator Joe "MAC" Carthy, brilliantly immitated by the previous poster, this is, no doubt, a communist plot to overthrow /.
I'm sure you'd agree that ops/sec is also not a very useful metric either, since two CPU's will have vastly different efficiencies of work accomplished for an average instruction.
A CISC "alu [dstMem], [srcMem], [srcMem]" instruction that (when efficiently using a cache) can complete in fewer cycles then the explicitly defined RISC instrs:
"ld [src1]",
"ld [src2]",
"alu",
"st [dst]"
due to a reduced load on the register set is an example.
More critically, there are the vector operations, cache-sensative prefetching instructions, etc.
Ultimately, all that you can do is have comparable top-tier compilers render the same application on different platforms and compare the aggregate outputs. Individual enhancements don't generally make or break a CPU; it's the whole shibang; platform, compiler, OS, language-choice, etc.
-Michael
Bull. It is still the best metric. Compilers don't matter. User the best available for each platform and compare systems of equivalent price. The one with the better ops/second wins -- simple.
If there is a vector unit present, it should enable the proc to do more ops/sec so it is included in the metric. No of procs don't matter either, we're talking systems here. They should learn from the TPC benchmarks. What matters should be how fast and how chep, not what compiler or how many CPUs or how many MHz etc.
"Fighting terrorists with millitary might is like killing a mosquitor on your Dad's forehead with a rifle."
>> So why didn't they port their code to SSE2?
Probably because it's not worth it. While Altivec speeds up Jet3D by a factor of 10 to 13, there is no evidence that SSE2 can even double the speed.
Yeah, especially if no one tries to port something to SSE2 to see.
you guys are hilarious. lol. fighting like cats and dogs. when it boils down to it, computer geeks use pcs and artsy types use macs.
I love it I quote a troll and I get modded down as a troll because it was a STUPID troll?
;)
WTF...
Oh well the moderation is all screwed up anyhow. They gave ME mod point.
As you can see I don't care about my karma.
That looks pretty similar to what you already wrote up; I'm not convinced, but maybe you're right. In any case, I obviously won't be buying a Mac, and at least for those reasons. Their additions--much like Microsoft's--generally annoy me more than anything.
However, you should probably collect those comments together and distill them into a stock reply on the subject; the topic does come up often enough, after all.
pb Reply or e-mail; don't vaguely moderate.
Now, with all these Linux-heads around here insisting that Linux is faster than Windows on x86, you'd think GCC for x86 might be a good compiler.
I believe the kernel design is a more important factor than how fast code the compiler generates.
Do you care about the security of your wireless mouse?
Octuple-precision floating point on Apple G4.
Lars T.
To the guy who modded me down from perfect to terrible Karma - Apple haters still suck
I suppose people for whom life is about pool better stay with games & virus boxes indeed. Meanwhile we Macophiles get on with our business and prosper ...
SETI@Home is not vectorized. Even on a G4, they could get a factor of almost 10 on throughput, were they to vectorize the client.
When this was pointed out to them by Apple employees, the SETI@Home people did nothing about it, preferring to keep the client code the same on all platforms, rather than achieve their workload numbers in a fraction of the time (apparently one Apple employee by himself is sitting at about 15th place on number of workunits processed).
It's actually really annoying, but from what I hear, SETI@Home is wrapping up, and they are working on new clients for different problem sets (also not vectorized).
-- Terry