Slashdot Mirror

More informative might be the CPU2000 benchmark figures at SPEC. Lots of CPUs tested for both integer and FP performance and throughput. LINPACK is a linear algebra library. I think the SPEC FP benchmark includes codes that do lots of linear algebra (among other things).

High [sic] Definition Porn
YES!
oh, and Star Trek will look nice as well.

There are 2 unusual lessons from the development of digital video. The first lesson is that pornography is the driver of improvements in video equipment -- from VCRs to HDTVs. Adult video is something that you simply must see -- with the eyes. If the resolution improves by 50%, then the porn video becomes 50% better. Adult video is a multibillion industry if you measure all the videos produced in the world.

The second lesson is that there is simply no justification, whatsoever, for allowing H-1B workers to work in the United States of America (USA). Allow me to explain.

We are all familiar with the comparison between the UltraSPARC III and the SPARC64-V. The management of Sun Microsystems says that it absolutely must have H-1B employees in order to build the "best" microprocessor, the UltraSPARC III. By comparision, Fujitsu hires almost exclusively native employees (i. e. Japanese citizens) and used native employees (not foreign workers) to build the SPARC64-V. It implements the same insruction set that the UltraSPARC III implements but significantly outperforms the UltraSPARC III on 3 key benchmarks: SPECint2000, SPECfp2000, and TPC-C. Please verify the performance characteristics of the SPARC64-V at the web site for SPEC and the web site for TPC-C.

Another interesting example is high-definition television (HDTV) Please read "The History and Politics of DTV".

The typical foreigner claims that H-1B workers help to give American companies a technological edge to create whole new industries. The typical foreigner claims that foreign workers helped the Americans to leapfrog the Japanese in the area of HDTV. The Japanese had established an HDTV standard prior to 1989 and had begun broadcasting HDTV programs by 1989, but this standard was based on analog techniques. In the 1990s, the Americans developed an HDTV standard based on digital techniques. The foreigners claim that foreign brainpower helped the USA to leap ahead of Japan.

Not really. Once the HDTV standard based on digital techniques was established, the Japanese (and the Koreans) commercialized the technology. Most of the HDTV products that you see in the USA are manufactured in Japan (and Korea).

The lesson here is that, "even if" terminating H-1B employment may diminish American innovation and simultaneously increase innovation in India (for example), the Indians may not be able to effectively commercialize the technology. Given the superiority (e.g., low rate of software piracy) of Western society, the Americans would be able to out-commerialze any innovation that appears in India before the Indians achieve commercialization.

The phrase "even if" is used because there is simply no evidence to suggest that terminating H-1B employement will diminish American innovation. Innovation is a hard thing to predict -- for the very reason that creativity is hard to predict. For example, Japan was initially ahead in HDTV by using analog techniques in the 1980s; at the time, many "experts" predicted that the USA would fall permanently behind in HDTV technology. Nonetheless, the USA raced ahead in HDTV by using digital techniques. Now, Japan is ahead again by pushing the envelope of HDTV technology to create ultra HDTV.

... from the desk of the reporter

I can't speak for SCSI, Firewire, SIDE, or any other drive techs 'cause I'm a cheap S.O.B. and won't pay the big bucks for them.

We moved an application from 2 UltraSPARC III 750 MHz CPU's to 6 UltraSPARC III Cu 900 MHz CPU's and saw very little improvement in performance. Then we moved the disk for the application from 9 internal drives to 20 external SCSI over FC drives, and voila our IO wait dropped from 60% or so to 10% +/-. Our query response times dropped by a factor of three or more. Faster, and even more, CPU's are not the answer to data intensive problems, I/O is. Slower (clock speed wise) 64bit CPU's, with better efficiency, more memory addressing, etc. are the norm in the data center for just this reason. IF you can take advantage of your L1/L2 cache then faster clock speed on the CPU will improve performance. The reason most Intel PC's benchmark better than an older box is because the disk, memory and video sub-systems have improved, not because the CPU is making a huge difference.

As proof, search SPEC's benchmark results using Dell and then Sun as your search criteria. Notice the following:

• A Dell PE2550 with a PIII 1.13 GHz CPU has a CINT baseline of 561.
• A Dell PE2650 with a Xeon 3.06 GHz CPU has a CINT baseline of 1014
• A Sun 280R with an US III 1.2 Ghz CPU has a CINT baseline of 637

Things get even worse when you start looking at the SMP capabilities and scalability. In a truly linearly scalable SMP system you should be able to go from 1 CPU to 2 CPU's and have the benchmark double. Even the best SMP systems (Sun UltraSPARC and IBM Power) can't quite achieve that. But Itanium really has trouble. Search on Dell and look at the CINT and CFP rates benchmarks. Look at 1, 2 and 4 CPU scores for the Dell 7150.

Bottom line? If you are doing heavy lifting on a server, go SMP with 64bit RISC, or, in some cases, use a cluster of 2 CPU x86 servers. If you are a PC user, you are unlikely to see a significant performance increase with new Intel CPU's unless you upgrade the whole system, not just the CPU.

This whole thing of adding clock cycles and deepening the pipeline is not working out well.

But don't listen to me. I'm not an IBM fanboy and I've been bitchslapped.

Yeah, it's not like you can just look them up. Oh wait, Apple doesn't like to send them scores. Wonder why that is....

As for benchmarks, it's silly to omit the G5s from the comparison for religious reasons, and people want to see how they stack up. The only way to counter Apple's marketing drivel is to do actual real-world benchmarking using cross platform apps and benches.

Apple has put out documents (the technical documents at the G5 site) which claim that their SPEC values are greater than the P4. If you want industry standards, you just have to go to SPEC and verify that their claims are bogus. The P4 at 3 GHz has a spec_int value of 1200 and Apple claims that it is 700. This is an outright lie. Combine this with their usual "supercomputer" claims, and you will see why people have zero respect for Apple. And by real world apps, did you mean PhotoShop?

AMD created the 3DNow! extensions to the 80x86 instruction set architecture (ISA), also known as IA-32. They were a significant improvement over the original set of MMX extensions. However, later, Intel created the SSE (and SSE2) extensions. Guess what? AMD was forced to incorporate them into its future chips in addition to the 3DNow! extensions. Ignoring the SSE extensions would have cost AMD dearly in terms of marketshare. The fact of the matter is that Intel sets the global standard for the IA-32 ISA.

Now, AMD has created its own x86-64 extensions to the IA-32. You can be sure that Intel has created a different set of 64-bit extensions (which we shall call "INTEL-64") to the IA-32. After all, why would Intel support AMD in any way? Once Intel activates the INTEL-64 extensions in the upcoming Prescott, AMD will be forced to go back to the drawing board to incorporate the INTEL-64 into all future chips. The current Athlon64 will be like the K-5 -- interesting but without a future.

AMD will probably take an additional 2 years to produce an INTEL-64-compatiable chip. By that time, Intel would have locked 90% of the 64-bit desktop market with Prescott.

The worst news is for Sun. With Prescott, Intel has a 64 bit chip that will be significantly faster than the UltraSPARC III/IV. Right now, the Pentium 4 crushes the UltraSPARC III in performance. Please review the performance characteristics of the Pentium 4 at the SPEC web site. Since Prescott (successor to the Pentium 4) will be faster than its predecessor, Prescott will clean UltraSPARC's clock. Moreover, the number of applications that will run on Prescott -- the heir to the software empire of the x86 -- far exceeds the number of applications that run on UltraSPARC III/IV. On the key TPC-C benchmark, Prescott will clearly deliver outstanding performance, compared to the UltraSPARC III/IV.

In short, when Intel activates the INTEL-64 extensions in Prescott, Intel will force (1) AMD back to its usual state of borderline bankruptcy and (2) Sun into being a software company.

... from the desk of the reporter

This "extreme" version of the chip has to be aimed at a very niche market, at least for the next couple of years until more processor intensive software catches up.

The processor-intensive software is already here. It is called HSpice, Verilog, fluid-dynamics simulation, etc. The Pentium 4 has done nicely in the engineering workstation market, and the "Extreme Edition" should do even better.

Please check the SPEC web site for a performance evaluation of the Pentium 4's floating-point (FP) performance. In particular, it outperforms the UltraSPARC III even though the latter has a 2-to-1 advantage in the width of its databus -- 64 bits versus 32 bits.

What changed the x86 chips from also-ran losers in FP performance to the kings of the hill? SSE.

The SSE extension to the x86 instruction set architecture (ISA) opened up a whole new world of applications for the Pentium III and successors. Older Pentiums were saddled with a FP stack that hurt their performance. The SSE extension established a directly addressable bank of 8 128-bit registers or 32 32-bit registers for FP operations. As a result, the Pentium 4 outperforms the UltraSPARC III on video applications.

At 3.2 GHz, the "Extreme Edition" of the Pentium 4 should help the Pentium 4 to capture even more of the engineering workstation market. Nowadays, the first-choice workstation among engineers in Silicon Valley and Boston's Route 128 is Linux running on a fast Pentium/Athlon, not Solaris lumbering on a slow UltraSPARC III.

... from the desk of the reporter

We've got the No. 1 64-bit computing architecture out there.

Perhaps, McNealy is referring to #1 in the sense of #1 performance. Again, the #1 in performance is the triad: Power architecture (with implementations being Power4, Power4+, Power5), the Itanium architecture (with implementations being Itanium 2, 3, etc.), and the x86 architecture (with implementations being the Pentium 4, etc.). A quick review of the performance stats at SPEC should clarify any confusion. The SPARC is among the worst processors in terms of performance.

Below is the second key quote.

Shouldn't India be a little upset that we have most of their software programmers here?

Compared to IBM, Sun is #1 -- in the sense that Sun has more H-1B employees. IBM, as a matter of corporate policy, refuses to hire any H-1B workers unless they are applying for a job that requires a Ph.D. The Power4, which handily beats the UltraSPARC III in performance, was built almost exclusively by American citizens or permanent residents. No H-1Bs.

Perhaps, McNealy was referring to the number of H-1Bs when he was talking about the SPARC being the supposed #1 computing architecture.

... from the desk of the reporter

However, Dr. Joy has 1 notable failure. When Sun Microsystems was developing the first SPARC processor and had planned to migrate all its software and tools to the new chip, the engineers designing SPARC came to a critical juncture in their work. That juncture can be summarized in one question: "Should we build register windows into SPARC?".

At the time, Professor Patterson at Berkeley was championing the incorporation of register windows into microprocessors. According to his flawed studies, register windows supposedly give a significant performance boost to C-language code. Patterson was an inspiration for young Dr. Joy.

The SPARC engineers did their own studies but went far beyond what Patterson did. They actually looked at how to build register windows into high-performance circuits, and they concluded that building register windows would be too difficult. They also questioned the supposed performance benefits of register windows. Unfortunately, Dr. Joy overruled them and forced them to incorporate register windows into the first SPARC chip.

The rest is history. It proved that register windows give SPARC no significant performance advantage over other microprocessors. Indeed, register windows appears to reduce performance. Look at the results at the SPEC web site. Power4, Power5, Pentium 4, Athlon, etc. all crush the SPARC chip in performance. When the 80486 became the first 80x86 chip to exceed the performance of the SPARC chip on SPECint, Sun Microsystems should have gotten a clue and dropped register windows. On the contrary, Sun Microsystem stubbornly stuck with register windows. You might say that register windows killed the performance of the SPARC processors.

To this very day, SPARC processors are about 2 generations behind their peers in performance.

... from the desk of the reporter

Others who know more than I do about Myrinet have already pointed out that you don't seem to know what you're talking about, so I'll skip that.

Next point, FP performance. Let's see, Apple managed a score of 840 on Spec CFP2000 base on their 2.0GHz PowerPC using GCC. AMD managed a score of 1209 on their 2.0GHz Opteron using Intel's compiler, while IBM managed a slightly lower score of 1172 using GCC. If the G5 is so superior, it sure isn't doing a very good job of showing it.

The memory bus on the G5 is better? What are you smoking here?! Each Opteron chip has a low-latency integrated memory controller capable of managing 5.4GB/s of bandwidth. Two G5 shares a higher latency, off-chip memory controller capable of 6.4GB/s total. Their bus may allow memory and cache transfers at the same time, the Opteron has a completely separate buses for these tasks! The Opteron also has twice the L1 data cache and twice the L2 cache that the PowerPC 970 has.

I will, however, agree with you that the Opteron is no cheaper than the G5 if you can get Apple to give you a useful configuration of the boxes for a supercomputer (who the hell needs a high-end gaming video card on each super-computer node? Let alone iTunes!). As for building the machines themselves though, why would they do that? Why not just buy a machine from one of the dozens of companies that are selling dual-processor 1U Opteron servers now. Many can and have built clusters of these already. Speaking of '1U' and all, brings me to one fo the dumbest things about this supercomputer, it's ALL being built usind the damn PowerMac tower cases! Now THAT is just a DUMB idea! 1100 giant desktop cases using bog-standard power supplies and the works, now that is just ridiculous!

Others who know more than I do about Myrinet have already pointed out that you don't seem to know what you're talking about, so I'll skip that.

Next point, FP performance. Let's see, Apple managed a score of 840 on Spec CFP2000 base on their 2.0GHz PowerPC using GCC. AMD managed a score of 1209 on their 2.0GHz Opteron using Intel's compiler, while IBM managed a slightly lower score of 1172 using GCC. If the G5 is so superior, it sure isn't doing a very good job of showing it.

The memory bus on the G5 is better? What are you smoking here?! Each Opteron chip has a low-latency integrated memory controller capable of managing 5.4GB/s of bandwidth. Two G5 shares a higher latency, off-chip memory controller capable of 6.4GB/s total. Their bus may allow memory and cache transfers at the same time, the Opteron has a completely separate buses for these tasks! The Opteron also has twice the L1 data cache and twice the L2 cache that the PowerPC 970 has.

I will, however, agree with you that the Opteron is no cheaper than the G5 if you can get Apple to give you a useful configuration of the boxes for a supercomputer (who the hell needs a high-end gaming video card on each super-computer node? Let alone iTunes!). As for building the machines themselves though, why would they do that? Why not just buy a machine from one of the dozens of companies that are selling dual-processor 1U Opteron servers now. Many can and have built clusters of these already. Speaking of '1U' and all, brings me to one fo the dumbest things about this supercomputer, it's ALL being built usind the damn PowerMac tower cases! Now THAT is just a DUMB idea! 1100 giant desktop cases using bog-standard power supplies and the works, now that is just ridiculous!

> Eh, I wouldn't say that GCC is "worthless", it's just that its worth lies in an
> area that has nothing to do with high-performance computing. Or even > mid-performance computing. :)

GCC is really fast.
Look at some SPEC_FP numbers:
(1) http://www.spec.org/cpu2000/results/res2003q3/cpu2 000-20030728-02428.html and
(2) http://www.spec.org/cpu2000/results/res2003q3/cpu2 000-20030728-02417.html

The c++ benchmarks (177.mesa,179.art,183.equake,188.ammp) were compiled with ICC (1) and gcc (2).

At two out of those four benchmarks (50%), GCC 3.3 is FASTER than the best Intel Compiler.

At least for Opteron (and Athlon) GCC is really good. Opteron with GCC and PGI Fortran compiler is excellent suited for High-Performance-Computing.

A big Thanks to the SuSE team for porting GCC to AMD64 and doing some usefull optimizations. http://www.x86-64.org/contributors/gcc

> Eh, I wouldn't say that GCC is "worthless", it's just that its worth lies in an
> area that has nothing to do with high-performance computing. Or even > mid-performance computing. :)

GCC is really fast.
Look at some SPEC_FP numbers:
(1) http://www.spec.org/cpu2000/results/res2003q3/cpu2 000-20030728-02428.html and
(2) http://www.spec.org/cpu2000/results/res2003q3/cpu2 000-20030728-02417.html

The c++ benchmarks (177.mesa,179.art,183.equake,188.ammp) were compiled with ICC (1) and gcc (2).

At two out of those four benchmarks (50%), GCC 3.3 is FASTER than the best Intel Compiler.

At least for Opteron (and Athlon) GCC is really good. Opteron with GCC and PGI Fortran compiler is excellent suited for High-Performance-Computing.

A big Thanks to the SuSE team for porting GCC to AMD64 and doing some usefull optimizations. http://www.x86-64.org/contributors/gcc

The new IBM compiler should rectify the situation. Apple will not need to manipulate the SPEC scores by hiding behind the GCC compiler. In the past, Apple stuck with the GCC compiler because it causes the Pentium to perform much worse than it would perform on code compiled with an Intel-provided compiler.

Of course, both the Power4 (and derivatives like the PowerPC 970) and the Pentium IV crush the UltraSPARC in performance. The new IBM compiler is yet in another nail in the coffin of the UltraSPARC.

OK, here you go:
Sun Blade 2000 (1.2GHz): 722 SPECint2000, 16.4 SPECint_rate2000.
The SPECfp results for the UltraSPARC III are useless due to the art bug.

Somebody should be crying, but it's not the Mac users.

OK, here you go:
Sun Blade 2000 (1.2GHz): 722 SPECint2000, 16.4 SPECint_rate2000.
The SPECfp results for the UltraSPARC III are useless due to the art bug.

Somebody should be crying, but it's not the Mac users.

I wouldn't go so far as to say that the Opteron performs better than any Intel offering. According to the latest SPEC results, the Itanium2 blows past it while even some Pentium 4s beat it out. The Opteron price, meanwhile, is roughly in parity with Xeons (Opterons are not cheap; have you actually compared prices lately?).

You're also comparing apples and... well... ok, 64 bit processors to 32 bit processors. If we roll out the Intel 64 bit Itanic, we discover that a 1.3Ghz Itanic puts up 1770 or so, which is 70% better than the UIII at a slightly slower 1.2 Ghz.

I'm not religious about processors or operating systems, but such a blatantly misguided statement cannot just get ignored. I'm not familiar enough with the IBM offerings to spot the Power 4s, but maybe someone else can look at this list (www.spec.org(scores)) and let us know whether or not they're within shouting distance of the other chips.

The new Mac is awesome. Sun servers are not. Period. No glamour, no speed, and not particularly more dependable than our Compaq boxen.

In fact, in simple userspace terms, I can tell you that a dual P4 Xeon 2.8 compiles the open-source toolkits we use an about half the time required by a V880 with eight processors and 8 GB RAM. Both using make -j.

Linux brings no value to Sun and actually destroys Sun's profits. Why? For years, Sun has hidden its performance-poor servers behind its Solaris operating system. Sun focused its marketing message on "the whole system" and said that performance is only one part of the system value. Most of that system value outside of simple performance came from Solaris.

Now, with Linux, the Sun salesperson can no longer argue that the operating system has some intrinsic value over the operation system of, say, an IBM machine. The IBM machine and the Sun machine are running the same operating system, Linux. Then, the comparison of the two machines comes down to performance. In other words, the customers will be forced to look at the quality of the basic hardware. In this area, Sun falls woefully short. Look at the results for the ""SPEC benchmark" or the "TPC-C benchmark".

What kind of competition does Sun have? Consider IBM's p690 and HP's Superdome. Both are in a neck-to-neck race to be #1 on the internationally recognized TPC-C benchmark by the Transaction Processing Council. Both of their scores is about 750,000. Please read "IBM touts own chips over Itanium". By contrast, Sun's best score is about 250,000 (from the TPC website).

As for SPEC performance, the p690 and the Superdome again crush Sun's best machine.

The only thing left for Sun is to create fear, uncertainty, and doubt (FUD). Sun is hinting that it will soon slit IBM's jugular vein by hinting that Sun may purchase SCO. After all, SCO claims control over IBM's UNIX patents. Sun is trying to create the fear that future IBM customers may be in expensive legal trouble if they run AIX or Linux because Sun-controlled SCO has terminated its UNIX licensing agreement with IBM.

Do you hear "it"? The bell is tolling. It tolls ominously for Sun.

At "SPEC", you can easily find the performance of a Power4 @ 1.45 GHz. Its SPEC2000 rating for floating point is 1097. When you scale that to the 2.0 GHz processor in the G5, you conclude that it has a SPEC score of about 1500.

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.

"Submission to SPEC's review process is not required. Testers may publish rule-compliant results independently. "

In my experience, PC doofuses have always been big with the benchmarks. It's like a bragging right to them. "I tweaked my dual Smockron 4500 and got it up to 313.3 on SPECdickweed_base!"

Meanwhile, us Mac doofuses (and I use the term with the greatest affection) spend that same time actually working. Because we need the extra cash to feed our $4000-a-year Mac habit.


Sure, certain models of x86 boxen costs less than macs, but the benchmarks cost $500.. I mean validation might well be worth something to the insecure, but is it really worth spending 20% of your budget on a benchmark suite? Remember, spec results are only valid for a machine of identical specification.

use this link to spec as the above one doesn't work all to well in all browsers. [missing the "www." some browsers add this if it's missing - others don't]

Why aren't there any SPEC numbers from Apple submitted to SPEC? Usually this means that a company knows they will show poor results. And I suppose in Apple's case they aren't reliant upon fast CPUs since they have those sexy designs and OS X now. But to win converts from the x86 camp they really should have some results submitted soon.

If these are reasonable, how come we don't see this numbers on SPEC's web page?

I rather wait and start to believe them when we see them here at SPEC's own site.

Dell Precision WorkStation 350 (3.06 GHz P4): 1092

Intel D875PBZ motherboard (3.0 GHz, Pentium 4, HT) 1213

HP workstation zx6000 (1000 MHz, Itanium 2) 1356

The numbers, according to Apple, are

3 GHz P4 test system 693

Apple Power Mac 2.0 GHz G5 (single processor) 840.

A few things really stand out here:

Even a 2.0 GHz G5 machine, which isn't even shipping - is only 21% faster than a commodity P4, according to Apple. (Apple only give the numbers using one of the processors. This is on the floating point test, too, which favors the G5. This is absolutely amazing to me. Despite a 1 GHz bus and a 64 bit architecture, and accepting Apple's questionable benchmark, their cpu is only 21% faster? WTF?

But using the specs from the Spec2000 website make the G5 look even more disappointing. Apparently a fast P4 (which is available today) is significantly faster than the fastest G5. According to Dell, it's 30% faster. According to Intel, it's 44% faster. That's a huge difference. And to those of you who say it's not, consider if someone sold you a 2 GHz chip, and when you looked inside the box, you saw it was actually running at 1.12 GHz. You'd be pissed, and rightfully so.

(An Itanium II, btw, is 61% faster, running at half the clock speed. Incredible.)

Why does Intel say its sytems are 75% faster than Apple says they are? I really don't know. But I seriously doubt that it's pure coincidence that Apple is the only company that can't get these machines to run fast.

I'm sure this will be resolved in a few months, once the results can be independently (in the "not under contract by Apple" sense of the word) verified. But I, for one, was expecting quite a bit more out of this chip, and have a very hard time understanding how Apple did not cook the books on this one ...

Unfortunately that really only proves that those results used a better compiler/heap/etc. This is a really simple concept folks. I'm sure if they had used a nice compiler along with all kinds of crazy compiler, OS and system optimizations, an even larger group of people would be saying that Apple was misrepresenting the benchmarks.

I'm going to say this once again: the Apple benchmark chose software consistency over potential performance winds in an attempt to isolate the performance benefits of the hardware.

It's worth noting that IBM's own 1500MHz Power4 CPU "kicks the crap" out of the published results for the G5. Guess what? The Power4 processor is the slower processor. IBM's own SPEC benchmarks show that the G5 at 1.8GHz should "kick the crap" out of that 1500MHz CPU, let alone a 2GHz model.

For the record, IBM's own SPEC tests for the 970 put it at 937/1051 when running at 1.8Ghz. Extrapolating this would suggest that the 970 using an "optimised" test bed, would score slightly below those ICC results for integer, while beating it for floating point. So amazingly, this performance differences demonstrated by Apple's benchmarks actually correspond to performance differences that exist in the case where both sides are using an "optimised" test bed.

Sadly, I haven't seen any 970 SPEC_*_rate benchmarks using IBM's compiler yet, but given that those benchmarks are much more influenced by motherboard design, it's quite possible those numbers would be misleading anyway. That being said, for the SPEC_*_rate benchmarks, IBM's 1500MHz Power4 system definitely "kicks the crap" out of the Dell system your suggesting.

• Okay, if he asserts that redoing the test WITH hyperthreading, and on Windows, will only slow down the Intel scores, then DO IT.

AMD, IBM, Intel, and Sun all measure the best results for their chips, and submit the full disclosure reports to Spec. This allows open comparison of processor performances with a fair and level playing field. All the performances reported are the best possible for that chip, since they are all provided by the chip vendors, and it is clearly in their own self-interest to provide the best possible numbers.

If Apple were to simply get the best numbers for the G5 they can, and submit them to Spec it would be a lot more believable; if they just come up with another set of numbers that are 30% - 40% too low it won't help their cause at all.

Nobody? Sure. I must be imagining all those people who told me how SPEC was such a "real" benchmark because it consists of subtests that are (parts) of actual applications - like gzip, gcc, crafty (chess), perlbmk (something from perl), bzip2, mesa (3D library), and other apps you never heard of, because the don't come from id.

Nobody? Sure. I must be imagining all those people who told me how SPEC was such a "real" benchmark because it consists of subtests that are (parts) of actual applications - like gzip, gcc, crafty (chess), perlbmk (something from perl), bzip2, mesa (3D library), and other apps you never heard of, because the don't come from id.

If you don't trust Dell's reported score for their Xeon line, then you must not trust Spec's online database:

Here

I would believe the results submitted to by Dell themselves before I believed the results subitted by Apple on Dell's behalf.

If Dell is lying on their benchmarks, it will show when other system makers submit their benchmarks. I'm sure spec doesn't want their benchmark database to be tained with bogus results, and they will not allow such activies to persist. That said, I've never seen big discrepencies in their submitted results.

That said, Apple's benchmark of the Dual Xeon system is much different from the ones sitting on SPEC's site.

spec.org. URL missing http:// part.

Yes, it is very strange that they ran the SPEC benchmarks on their competitors machines instead of just using the already published results for those vendors. In particular, using gcc vs. the Intel compiler makes a huge difference. The whole point of SPEC is to let each vendor do the best optimization job possible, so that each system runs as fast as possible on the given tasks. Where it gets misleading (kind of like the video card benchmarks) is when vendors implement compiler 'optimizations' for specific SPEC benchmarks. Even given it's flaws, SPEC is the best game in town.

One thing that the author complained about was the testers turning off hyperthreading (or HYPEthreading as I prefer to call it). In fact, that was almost certainly the correct thing to do on the Xeon system. After all, it has two real CPUs, it has no need to simulate more at the cost of some overhead. It would be interesting to repeat the test with hyperthreading turned on to see the difference.

One thing going against the Mac was that it was running a 32-bit OS that's not yet fully optimized for the G5.

Gone unnoticed in all this is the fact that Opteron is substantially faster than either, with a 2-processor system coming in at 25.0 CINT2000.

At any rate, the new Macs look to be a good value for the money. Even if they are essentially neck-to-neck with the Xeons, they are 64-bit and running a nicer OS than available for that system A dual processor Opteron 244 workstation will set you back substantially more than $3000 at this point. You are also stuck running Windows on it if you want a decently supported commercial OS.

Ah well, at least the world is more interesting than it was two days ago. :-)

Now hot grits, I know that was a joke, but seriously, what can one get in a 2 processor opteron system for USD 3K? One place a checked wanted about $750 for each Opteron 242. That does not leave much of a budget left.

Before people jump on me about comparing a Linux server to a Mac, this is really what I want to buy, a cheap (err, relatively speaking) 64bit box with at least 6G of memory. If it has to come with pretty box and a religon (well, true in both cases I suppose).

The opterons are a bit faster, at least at the top end, but the 970s seem ok. And the Macs are much easier to buy for a University (as in one phone call, compared to chasing down 3 quotes and filling out extra paperwork).

So here is a partial answer to my own question. Looking at one vendor (www.einux.com) then a server with twin opteron 240 (1.4GHz, right?) and 512M memory (for comparison), 120 G disk, prices out at $2500 w/o OS. So the extra $500 is not so bad considering what you get.

My back of the envelope calculations suggest the fastest 970 (2GHz) is about the same speed as the slowest opteron (that I quoted above). Namely, take the Linux gcc specint 2000 1045. Scale down by 0.77 to get a number for 1.4 GHz of 811.

Then of course you have to spend about the same amount on RAM. Sigh.

And just for some more perspective, here's an AMD 3200+ XP system:

SPECint2000 : 1080

Not as fast as the P4, but .... faster than the G5, I'm afraid.

I find it kinda strange how the SPEC results varied quite markedly between what Intel had listed and what was on display on the Apple website. So, I decided to look into it. First thing I did was goto Veritest's site and look at how they did the benchmark.

I noticed something odd... Veritest decided to run TWO different tests in the P4, one with and without Hyper-Threading enabled. Hyper-Threading is enabled by default on the P4 processor. Odd.

Then I decided to goto www.Spec.org and do a benchmark search for Intel P4:

SPEC.org results:
SPECint2000 : 1200
SPECfp2000 : 1229
SPECint_rate2000 : 14.1
SPECfp_rate2000 : 13.7

Apple.com results:
SPECint2000 : 889
SPECfp2000 : 693
SPECint_rate2000 : 10.3
SPECfp_rate2000 : 8.07

And yes, I did choose the latest results for the Intel P4.

It is very clear that the results obtained by Veritest and put forth in their report is of a P4 3.06 GHz with Hyper Threading DISABLED. The last I checked, HT is a feature which is enabled by default. WHY would someone purposely disable HT? Purposely make their CPU run SLOWER? Hmmm...

Come on Apple, do a serious system comparison, the best versus the best. Not the best versus a crippled system.

I find it kinda strange how the SPEC results varied quite markedly between what Intel had listed and what was on display on the Apple website. So, I decided to look into it. First thing I did was goto Veritest's site and look at how they did the benchmark.

I noticed something odd... Veritest decided to run TWO different tests in the P4, one with and without Hyper-Threading enabled. Hyper-Threading is enabled by default on the P4 processor. Odd.

Then I decided to goto www.Spec.org and do a benchmark search for Intel P4:

SPEC.org results:
SPECint2000 : 1200
SPECfp2000 : 1229
SPECint_rate2000 : 14.1
SPECfp_rate2000 : 13.7

Apple.com results:
SPECint2000 : 889
SPECfp2000 : 693
SPECint_rate2000 : 10.3
SPECfp_rate2000 : 8.07

And yes, I did choose the latest results for the Intel P4.

It is very clear that the results obtained by Veritest and put forth in their report is of a P4 3.06 GHz with Hyper Threading DISABLED. The last I checked, HT is a feature which is enabled by default. WHY would someone purposely disable HT? Purposely make their CPU run SLOWER? Hmmm...

Come on Apple, do a serious system comparison, the best versus the best. Not the best versus a crippled system.

I find it kinda strange how the SPEC results varied quite markedly between what Intel had listed and what was on display on the Apple website. So, I decided to look into it. First thing I did was goto Veritest's site and look at how they did the benchmark.

I noticed something odd... Veritest decided to run TWO different tests in the P4, one with and without Hyper-Threading enabled. Hyper-Threading is enabled by default on the P4 processor. Odd.

Then I decided to goto www.Spec.org and do a benchmark search for Intel P4:

SPEC.org results:
SPECint2000 : 1200
SPECfp2000 : 1229
SPECint_rate2000 : 14.1
SPECfp_rate2000 : 13.7

Apple.com results:
SPECint2000 : 889
SPECfp2000 : 693
SPECint_rate2000 : 10.3
SPECfp_rate2000 : 8.07

And yes, I did choose the latest results for the Intel P4.

It is very clear that the results obtained by Veritest and put forth in their report is of a P4 3.06 GHz with Hyper Threading DISABLED. The last I checked, HT is a feature which is enabled by default. WHY would someone purposely disable HT? Purposely make their CPU run SLOWER? Hmmm...

Come on Apple, do a serious system comparison, the best versus the best. Not the best versus a crippled system.

I find it kinda strange how the SPEC results varied quite markedly between what Intel had listed and what was on display on the Apple website. So, I decided to look into it. First thing I did was goto Veritest's site and look at how they did the benchmark.

I noticed something odd... Veritest decided to run TWO different tests in the P4, one with and without Hyper-Threading enabled. Hyper-Threading is enabled by default on the P4 processor. Odd.

Then I decided to goto www.Spec.org and do a benchmark search for Intel P4:

SPEC.org results:
SPECint2000 : 1200
SPECfp2000 : 1229
SPECint_rate2000 : 14.1
SPECfp_rate2000 : 13.7

Apple.com results:
SPECint2000 : 889
SPECfp2000 : 693
SPECint_rate2000 : 10.3
SPECfp_rate2000 : 8.07

And yes, I did choose the latest results for the Intel P4.

It is very clear that the results obtained by Veritest and put forth in their report is of a P4 3.06 GHz with Hyper Threading DISABLED. The last I checked, HT is a feature which is enabled by default. WHY would someone purposely disable HT? Purposely make their CPU run SLOWER? Hmmm...

Come on Apple, do a serious system comparison, the best versus the best. Not the best versus a crippled system.

I find it kinda strange how the SPEC results varied quite markedly between what Intel had listed and what was on display on the Apple website. So, I decided to look into it. First thing I did was goto Veritest's site and look at how they did the benchmark.

I noticed something odd... Veritest decided to run TWO different tests in the P4, one with and without Hyper-Threading enabled. Hyper-Threading is enabled by default on the P4 processor. Odd.

Then I decided to goto www.Spec.org and do a benchmark search for Intel P4:

SPEC.org results:
SPECint2000 : 1200
SPECfp2000 : 1229
SPECint_rate2000 : 14.1
SPECfp_rate2000 : 13.7

Apple.com results:
SPECint2000 : 889
SPECfp2000 : 693
SPECint_rate2000 : 10.3
SPECfp_rate2000 : 8.07

And yes, I did choose the latest results for the Intel P4.

It is very clear that the results obtained by Veritest and put forth in their report is of a P4 3.06 GHz with Hyper Threading DISABLED. The last I checked, HT is a feature which is enabled by default. WHY would someone purposely disable HT? Purposely make their CPU run SLOWER? Hmmm...

Come on Apple, do a serious system comparison, the best versus the best. Not the best versus a crippled system.

It will take intel at least more than a year to get the itanium near the PPC970 2 Ghz..

Pu-leaze... At least do a little research before spouting off on a topic you obviously don't have a clue about...

Go to spec.org and search for "itanium". You'll notice some itanium2 results from HP from last year , that blow apple's (highly biased) G5 results outta the water: Apple got 800/840 SPECint/SPECfp in their testing for the dual 2GHz G5. In contrast, HP acheived 810/1431 SPECint/SPECfp for a dual 1GHz Itanium2 machine, a full year ago ! Since Itanium2 is over a year old now, I'd expect HP/Intel to be releasing a newer/faster Itanium any time now, that will further widen the gap. I don't think Intel or HP are quaking in their boots at apple's meager competition...

According to Apple's web site, they tested their machine against two Dell Intel boxes (Dell Dimension 8300 (P4) and Dell Precision 650 (Dual Xeon)) running Red Hat Linux 9.0 Professional (at Apple's request).

Intel states that Red Hat Linux 9.0 Professional is one of the Linux OS's currently available that "include optimizations for HT Technology and are currently eligible to carry the Intel® Pentium® 4 Processor with HT Technology logo".

Apple commissioned the benchmark from a company called Veritest. The full report (in .PDF format) including all hardware and software used is available from Veritest's web site.

This could make Intel take notice! Of course, this benchmark comes on the same day that Intel announces the 3.2GHz Pentium IV (and Xeon) processors. Go figure!

Of the published data on both (in SPEC processor benchmarks), Apple's Power Mac G5 generated a SPECfp_base2000 score of 840 and SPECint_base2000 score of 800, while Intel claims that their new 3.2 GHz processors get a SPECfp_base2000 score of 1252 and a SPECint_base2000 score of 1221.

And the SPIN goes on!

I would particularly like to see the results at the SPEC org website: All published SPEC CPU2000 results

The PPC970 wit its Power4 core, clocked at 1.6GHz completely trashes a 3GHz P4. Faster bus, faster integer, and completely outclasses the P4 for FPU and SIMD.

Yawn.

The IBM PPC 970 1.8GHz Specint 2000 score quoted at Microprocessor Forum last year was 937 (source)
The current 3.0GHz P4 with Canterwood chipset already sports a Specint 2000 score of 1164 (source)

Wow...20% slower.
As for floating point, the ratio is even worse.

Move along...nothing to see here...

Yup, that is the deal. I think the 179.art benchmark is the one at fault:

SPEC.

Yes, the 179.art is several factor quicker on the US3 than any other processor. Hmm. I think if you knock it off, it drops their SPEC fp score by like 20% and it looks very ordinary all of a sudden.

Especially considering Intel, IBM, IBM and Alpha can all do 1000+ Spec FP easily!

Wrong again. The best benchmarks are 'real world' applications (which SPEC CPU2000 consists of, BTW). I can bench systems for my particular app. In the case of gamers (for instance) every review has benchmarks showing actual results for real games. The clock speed of the processor doesn't matter, what matters is the number of frames per second.

Game performance reviews for Opteron should be interesting versus P4. ;-)

The sad part is that with IBM's lack of progression of the chip, the PPC has not maintained the competitive advances it SHOULD have been implemented years ago that it is capable of doing.

Lack of progression? You mean like #1 in SPEC FP 2000, #4 in SPEC INT 2000 and #2 in TPC-C? IBM has been doing a great job with PowerPCs, but they've been concentrating on server versions.