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
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
$2999 for the mac 2x2ghz
creation science book
Microsoft Windows XP Pro Upgrapde: $199
http://shop.microsoft.com/Referral/Productinfo.as
MacOS X 10.3/2/1 Full price: $129
http://www.apple.com/macosx/
Microsoft Windows XP Pro (5 Users): $1315.60
MacOS X 10.3/2/1 (5 Users): $199
If you bought Windows XP ($299), and then can upgrapde to Longhorn for $199, you paid $498. If you bought MacOS X 10.1, 10.2, 10.3, and 10.4, you paid $516. Pretty similar, and that's assuming you only have to pay $199 for Longhorn. In the meantime, Apple users enjoy continued advance, while Windows stagnates for 4+ years.
Do the same with a family licence of 5. Buy Windows XP for $1315.60, then upgrade for $875.60: $2191.20 (over 4 years, for 5 people: $109.56/user/year).
Buy MacOS 10.1, 10.2, 10.3, 10.4 (5 User Licence): $796 (over 4 years, for 5 people: $39.80/user/year).
Using http://shopper.cnet.com I found a copy of Windows XP Pro for $207, and an upgrade for Windows XP Pro for $177. I found a copy of MacOS X 10.2 for $98.
If these prices hold over to the newer Operating Systems these companies release, then Windows would cost $384 (23% savings), and MacOS X would cost $196 (24% savings). If you bought every point upgrade Apple released it would cost $392.
Dak
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
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;
Mac is cheaper than PCs, even in laptops. It all depends on what you'll use it for. I'm trying to convince a pal that the iBook would be a better purchase for his mother than a Fujitsu Siemens PC. The iBook is rougly 9500 NOK and the Siemens 11500 NOK.
The difference in speed mHz, RAM etc. is irrelevant for a person just getting used to computers and the net. When she is having weekly problems with the Siemens / Windows machine, it will be lost money in time. While the Mac is cheaper in usage, because of less "frustration time" and less hassle.
This is an argument I would strongly disagree with, if you asked me two years ago. But since then, I have come to the conclusion that the Mac simply work better for the lay people. It does the work, and faster because there are less frustrations and less hassle.
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
>>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.
It was a hardware simulator running on a soundstage.
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'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.
I'll make that argument any day of the week if you want to consider TCO. My family got a Powermac G4 in 1999, and it is still the daily use computer for them. (I have my own Cube, which is basically the same for performance comparisons.) That thing still does everything that they can ask of it and then some. Hell, it can still play all the games that I want to play, save UT2k3. The great part is that it is still humming along perfectly, and I don't see any reason why it won't last two or three more years. Find me a PC that you will still be using daily 6 or 7 years later.
This doesn't even take into account all the time and headaches that have been saved from using a Mac. Taking out the "Did you accidentally kick the power cord out?" type phone calls I've gotten to help them, I can think of maybe twice that they have had to call me and troubleshoot. There is no pricetag on this peace of mind.
"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.
Hyperthreading isn't a magic double-the-speed-of-your-processor feature. In fact, ti can slow a computer down. What it is nice for is for running multiple threads or programs more efficiently.
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.
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...
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
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.
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
actually you only need 3 db to double the volume (which, btw, has little to do with loudness). and 1000 dBa is, I hope, impossible.
--[Nothing important]--
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.
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 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.
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.
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