Slashdot Mirror

As a side bonus, you can find SPEC benchmarks for Itanium and Itanium IIs on that chart (search for the word Itanium - Dell and HP have both submitted results).

They were just keeping with the grand tradition of not releasing the pertinent info, like most Apple benchmarks (photoshop, for instance). Truly, another Apple innovation (c)(TM). Notice how barefeats actually tried several different things, rather than just one narrow task like Apple likes to. Also, see spec for instance. Notice who doesn't want to play ball? Hmmmm....

Running a search on the processor 'itanium' or 'power4' will provide some pretty interesting marks. The Itanium 2 platforms are very significantly better than the Itanium. The Power4 does have very high specint2k scores, but as you guessed, it's the xeon that takes the trophy.

Also, I don't think AMD has gotten Hammer to a point where they are ready to publish benchmarks, so there won't be any comparisons. (Also, I don't think I've ever seen Intel compare itself to AMD!)

CINT2000
Comp Base Processor
Dell 824 Intel Xeon (400 MHz system bus)
HP 807 Intel Itanium 2
IBM 804 Power44
Sun 537 UltraSPARC III Cu
HP 379 Intel Itanium

CFP2000
HP 1356 Intel Itanium 2
IBM 1202 POWER4
Dell 803 Intel Xeon (400 MHz system bus)
Sun 701 UltraSPARC III Cu
HP 701 Intel Itanium

If you actually bother to go to the full list [spec.org] you'll find that linux servers are generally faster than IIS running on the same hardware. Sometimes being over twice as fast.

If you actually bother to go to the full list [spec.org] you'll find that linux servers are generally faster than IIS running on the same hardware. Sometimes being over twice as fast.

If you actually bother to go to the full list [spec.org] you'll find that linux servers are generally faster than IIS running on the same hardware. Sometimes being over twice as fast.

If you actually bother to go to the full list [spec.org] you'll find that linux servers are generally faster than IIS running on the same hardware. Sometimes being over twice as fast.

How about here and here ? Ok, Linux is on the Thunder MB where Win2k is on a Tiger. Then there is this and this . Those are identical hardware, except the number of clients is different.

How about here and here ? Ok, Linux is on the Thunder MB where Win2k is on a Tiger. Then there is this and this . Those are identical hardware, except the number of clients is different.

How about here and here ? Ok, Linux is on the Thunder MB where Win2k is on a Tiger. Then there is this and this . Those are identical hardware, except the number of clients is different.

How about here and here ? Ok, Linux is on the Thunder MB where Win2k is on a Tiger. Then there is this and this . Those are identical hardware, except the number of clients is different.

I am not going to go through every point MSFT has on that page, but what I hope to show is that they are not changing their strategy, just their tactics to carry out the strategy.

It used to be that MSFT FUD was a set of bald faced lies, apparently the hope was that no one would check them out at all. Well people did, found out they were lies and went to Linux.

All of the new MSFT FUD is now more subtle, and appears to have supporting material in some cases. Now you can even check this stuff out, and if you are not very knowledgeable about software you can be fooled.

Point 1

Sum up as "Linux/Samba is not really compatible with Windows networking."

In fact Linux, and Samba do support almost all features of CIFS. When Samba has been incompatible it is because MSFT changed their implementation. For this matter, Win95 and Win98 are incompatible with W2K CIFS networking.

Point 2

Sum up as "Linux is not fully compatible with Active Directory"

True enough, but Linux is compatible to the extent that Active Directory is compatible with LDAP. In truth, MSFT is the one failing to comply with existing standards...neat how they twist this one around.

Point 5

Sum up as "IIS 5 is faster than Linux for SpecWeb99"

This is just FUD. The link they point to seems to agree with their assertion but how about this link instead. It sure seems to tell a different story on identically configured hardware.

Point 11

Sum up as "Windows has reliable drivers that are signed by MSFT, Linux doesn't"

Windows has NEVER had reliable drivers. Not all the best drivers are signed by MSFT if at all.

The situation is only somewhat better than Windows for Linux to be fair. First off most drivers are delivered with an MD5 checksum, which is good enough for most uses. Secondly you get the source most of the time. Finally, since when has NASA written drivers for MSFT ? (Thanks to Don Becker, NASA GSFC.)

Point 16

My favorite..."The GPL is nasty and dangerous and can force you to give away all your secrets."

First off the GPL is easy to understand, and very consistent. You get quite a lot for a simple price, "our changes to the code are to be made public with your codes binary release."

MSFT has a problem with this because they are in the business of keeping code secret, not open sharing of ideas. Frankly that is OK, and can be a fair way to do business, despite what many OSS evangelists will tell you. What it fails to be however is an advantage to the consumer of the final product.

From the m$ site:
A SPECweb99 study found that a Windows 2000 Web server could process more requests and serve more users than a similarly configured computer running Linux.
The study compared this with this. I dont know about you, but I find those machines to be vary dissimilar (a 667MHz w/ 2G Ram vs. a 533 w/ 1.5G). I find these (ms/rh) comparisons much more compareable.

From the m$ site:
A SPECweb99 study found that a Windows 2000 Web server could process more requests and serve more users than a similarly configured computer running Linux.
The study compared this with this. I dont know about you, but I find those machines to be vary dissimilar (a 667MHz w/ 2G Ram vs. a 533 w/ 1.5G). I find these (ms/rh) comparisons much more compareable.

From the m$ site:
A SPECweb99 study found that a Windows 2000 Web server could process more requests and serve more users than a similarly configured computer running Linux.
The study compared this with this. I dont know about you, but I find those machines to be vary dissimilar (a 667MHz w/ 2G Ram vs. a 533 w/ 1.5G). I find these (ms/rh) comparisons much more compareable.

From the m$ site:
A SPECweb99 study found that a Windows 2000 Web server could process more requests and serve more users than a similarly configured computer running Linux.
The study compared this with this. I dont know about you, but I find those machines to be vary dissimilar (a 667MHz w/ 2G Ram vs. a 533 w/ 1.5G). I find these (ms/rh) comparisons much more compareable.

The whole IP thing is just FUD. If yuo use linux to run your servers you are much less likely to fall foul of IP laws than if you use Windows in the same situation. Compare the usage restrictions in MS's EULA and in the GPL (for the uninformed, there aren't any in the GPL).

And then there's the SpecWeb99 link. The machines compared is Windows 2k with RH 6.1 in Q4 1999. If you actually bother to go to the full list you'll find that linux servers are generally faster than IIS running on the same hardware. Sometimes being over twice as fast.

So no, this isn't any "fairer" than the last page. It's just less full of complete untruths. Instead it has things that are technically true but not the whole story. A quite nice example of content free marketing.

So, according to some study on webservers (probably funded exclusively by M$), IIS 5 performs better than Zeus 3.3.2. Yes, Zeus. Seriously, who in their right mind would compare Zeus to IIS rather than Apache and IIS? And I love how they use different hardware for the comparisons... kinda trying to imply that Linux doesn't work on "normal" Dell hardware, but only "expensive" IBM hardware.

Here's the quote from M$:

Server appliances built on Windows 2000 perform better versus Linux on similar equipment in SPECweb tests. A SPECweb99 study found that a Windows 2000 Web server could process more requests and serve more users than a similarly configured computer running Linux. The Windows 2000-based server with Internet Information Server (IIS) 5.0 handled 707 concurrent connections, compared to 545 connections for the Linux-based.

So, according to some study on webservers (probably funded exclusively by M$), IIS 5 performs better than Zeus 3.3.2. Yes, Zeus. Seriously, who in their right mind would compare Zeus to IIS rather than Apache and IIS? And I love how they use different hardware for the comparisons... kinda trying to imply that Linux doesn't work on "normal" Dell hardware, but only "expensive" IBM hardware.

Here's the quote from M$:

Server appliances built on Windows 2000 perform better versus Linux on similar equipment in SPECweb tests. A SPECweb99 study found that a Windows 2000 Web server could process more requests and serve more users than a similarly configured computer running Linux. The Windows 2000-based server with Internet Information Server (IIS) 5.0 handled 707 concurrent connections, compared to 545 connections for the Linux-based.

Apple is notably absent from SPEC's list--they never submitted results.

However, third parties have run the SPEC benchmarks. A 1GHz G4 seems to perform about as well as a 1GHz Pentium III: decent but not overwhelming. See also this Register article.

Apple should move to the G5 quickly. Or, perhaps, Apple should even switch to some 64bit Intel or AMD processor--Motorola is likely going to keep remaining behind the curve a bit.

That 450 Mhz risc machine would run rings around any PC I could get (with exception of a Itanium, properly configured...). RISC machines are proof that doubling MHz will not necessarily double performance.

You dont visit SPEC much, I take it.

It depends on what your simulation code is doing. (Surprise !)

If you look at the SPEC CPU2000[1] scores for AMD vs. Intel you'll see that for, say, an XP2000 vs a 2G P4, the AMD is faster at integer work and the Intel is faster for floating point. Note though, that the Intel scores are with the Intel Reference Compiler, which will probably be generating SSE code. If you're running non-SSE code, then Intel stuff is considerably weaker for floating point work.

The second thing to consider is how the pipeline length affects the execution of your code. The longer pipeline in the P4 means that, roughly speaking, the P4 is faster in a straight line, but Athlons corner better.

For the simulation stuff I'm doing, which involves huge amounts of conditional integer and bit-twiddling operations and next-to-no floating point, I use a dual MP1800 box[2]. Getting the equivalent performance - for my code - would have cost a huge amount more with a P4-based solution and may not even be possible with the current P4 range.

[1] If you're doing real computing rather than fragging folks, SPEC is probably a better place to get your information from than Tom's Hardware.

[2] One SETI work unit every 80 minutes. Yummy.

You called me a liar in your response.

No, I called Steve Jobs a liar. You, I think, are just a bit too credulous for your own good.

How is excluding floating point performance relevant?

Stop right there.

Altivec is not a floating point unit. Really. Go read Motorola's own documentation if you don't believe me. The kinds of operations that Altivec can speed up are not general-usage cases, and it's fiendishly difficult to do even when the math is the right type. (The same can be said, btw, of the x86 world's SSE and SIMD units, but they get used a lot more often because Intel actually wrote their own C/C++ compiler to produce code that's automatically tuned for them. Unfortunatly for Apple, Motorola and IBM have not been quite so generous.)

The P4 and Athlon XP (especially the latter) clean the G4's clock at normal floating point ops. That's what the SPECfp benchmark measures, and the figures are in black and white.

Yes, the G4 is faster at equal clock speeds. That's nice, but not very relevant when you can buy faster Athlons at the same price point.

Stop listening to marketers and start listening to engineers. Facts are facts.

Now, the nice thing about the xServe is that it's got those huge-ass 2MB DDR L3 caches, which means that you're getting the equivilant of a P4 Xeon or a Sun UltraSparc IIIi for the price of a normal P3 server, which is a good deal. Raw CPU speed is often a secondary concern in a server environment: I/O bandwidth, context-switching speed and physical robustness can trump it in a lot of cases, and the xServe appears to be very well-positioned there.

despite the fact that the powerPC has been beating the pentium in every reasonable performace comparison for years

That's not a "fact". That is what we in the business like to call a "marketing claim," or what normal human beings call a "bald-faced lie."

Outside of a small number of benchmarks that make extensive use of the G4's vector units, the Athlon XP and Pentium 4 are faster than the G4 at every equivilant system price point -- usually a lot faster. This is a cold, hard fact.

That's not to say that the xServe isn't a nice box. Hell, it's a great little server, and I can't wait to get my hands on one. But the reasons for that have everything to do with the OS, the case design and the management infrastructure, and nothing to do with CPU speed.

And USB2.0 is going to be widely accepted since it's forward and backward compatible.

IBM is introducing low-end servers based on the Power4. They compete head-on with the V880 from Sun Microsystems (NASDAQ: SUNW). The V880 is the only profit generator for SUNW right now; the high-end SunFire 15K is almost unsellable. The new IBM servers will dent the sales of the V880 and bring its days as a profit stream for SUNW to an end.

With the new p670, IBM now has the entire range of servers from the low-end to the high-end covered by Power4-based systems. Each system outperforms the equivalent (in terms of the number of processors) system from SUNW. Worse, at the end of May, the Oracle DBMS will run on the Power4. Finally, the TPC-C numbers will be out. SUNW's systems will be crushed.

Just read the performance numbers at SPEC and TPC .

If you look at the related articles at CNET, you will notice something that is terribly wrong with SUNW. The spokesman for SUNW is Shahin Khan. The spokesman for IBM is Ravi Arimilli. Shahin Khan is a marketing drone. Ravi Arimilli is an IBM Fellow, firmly grounded in engineering. Khan is telling Arimilli why SUNW's systems are technically superior! (Read "IBM, Sun to release dueling servers".)

Finally, on the day that IBM unveils its arsenal of Power4-based servers, SUNW rushes to cut prices. (Read "Prices Lowered on Popular Sun Fire Server Family".) SUNW appears to be running scared. You should dump SUNW stock because SUNW will not achieve profitability by 2002 June. How do you achieve profitability by slashing prices and, hence, margins?

"vast majority of the sites out there that serve out mostly static and simple dynamic traffic"

About 30% of all traffic is 'uncachable' dynamic content. (Cachability can be argued)

"I think apache is more than sufficient (these sites tend to be bottlenecked by the n/w, not by the server)"

no! no! no! CPU utilization is typically the bottleneck. Also, Apache is considered one of the most CPU intensive and (sorry) slowest HTTP servers. That is why you will not find it amoung the SPECweb99 Results [spec].

Indeed, both the Power4 and the Pentium 4 (P4) significantly outperform an UltraSPARC III. Just visit the SPEC web site and the web site of the Transaction Processing Council (TPC) . The performance results at those web sites show that the UltraSPARC III significantly underperforms against the competition. For this very reason, Sun is refusing to run Linux on the UltraSPARC III. Running Linux on UltraSPARC III would allow an even more direct, head-on comparison between the UltraSPARC III and the Power4 or P4. Same OS (Linux). Yet, Power4 and P4 outperform UltraSPARC III. There would be no way for Sun to say, "Well, Solaris causes the UltraSPARC III to run slower than the competition because Solaris is using the extra CPU cycles to give you that much more reliability."

As the official line, Sun Microsystems derides with the SPEC benchmarks and the TPC benchmarks as being unfair and unrepresentative of the "real world". How can any company utter such asinine comments? Both SPEC and TPC are fair, reputable organizations that have set forth to provide a fair and unbiased means to compare a range of computer systems from various companies. You might say that both SPEC and TPC are the "Consumer Reports" of the server market.

To look at something that is really asinine, I highly recommend Big Blue Smoke , which is a childish web site that Sun established to ridicule IBM. Sun must be getting really desperate.

If you're willing to spend the money to get the speed, the nVidia Quadro4 900XGL is the current SPECviewperf record holder, supports two displays (2048x1536 each, better than the XVR's dual 1280x1024), and costs well under half the XVR-1000. It also supports stereo viewing and a programmable vertex & pixel pipeline.

True, its DACs are 24 bits instead of 30 bits (SGI workstations are still the go there, with 36 bit RGB DACs), but the NV30 may change that. It also does multisampled anti-aliasing (currently 9-tap 4-sample, though older drivers did offer a 16-sample mode too).

As for raw compute performance, if you believe Sun's SPEC ratings from their product site, a 1.05GHz SPARC CPU is only just lagging behind an Intel 2.2GHz PIV on integer performance and beating it on FP.

Where do they claim that? According to the SPECcpu website, a 1.05 GHz SPARC III Cu gets 537 base SPECint and 701 SPECfp, while a 2.2 GHz P4 easily beats it with 790 SPECint and 779 SPECfp.

Intel is way ahead in integer, and although the Sun catches up somewhat in FP, if you look at the individual results, it's entirely due to one massive spike on the art test. They recently figured out a (controversial) compiler trick that gave them nearly an order of magnitude increase on that one SPECfp test, and doubled their overall SPECfp score. Sun are known for their stability & scalability, but not their CPU speed.

Of course, if you have 106 of the things, that's different. But you'll be paying over US$4M for it, which isn't exactly workstation class anymore.

As for raw compute performance, if you believe Sun's SPEC ratings from their product site, a 1.05GHz SPARC CPU is only just lagging behind an Intel 2.2GHz PIV on integer performance and beating it on FP.

Where do they claim that? According to the SPECcpu website, a 1.05 GHz SPARC III Cu gets 537 base SPECint and 701 SPECfp, while a 2.2 GHz P4 easily beats it with 790 SPECint and 779 SPECfp.

Intel is way ahead in integer, and although the Sun catches up somewhat in FP, if you look at the individual results, it's entirely due to one massive spike on the art test. They recently figured out a (controversial) compiler trick that gave them nearly an order of magnitude increase on that one SPECfp test, and doubled their overall SPECfp score. Sun are known for their stability & scalability, but not their CPU speed.

Of course, if you have 106 of the things, that's different. But you'll be paying over US$4M for it, which isn't exactly workstation class anymore.

As for raw compute performance, if you believe Sun's SPEC ratings from their product site, a 1.05GHz SPARC CPU is only just lagging behind an Intel 2.2GHz PIV on integer performance and beating it on FP.

Where do they claim that? According to the SPECcpu website, a 1.05 GHz SPARC III Cu gets 537 base SPECint and 701 SPECfp, while a 2.2 GHz P4 easily beats it with 790 SPECint and 779 SPECfp.

Intel is way ahead in integer, and although the Sun catches up somewhat in FP, if you look at the individual results, it's entirely due to one massive spike on the art test. They recently figured out a (controversial) compiler trick that gave them nearly an order of magnitude increase on that one SPECfp test, and doubled their overall SPECfp score. Sun are known for their stability & scalability, but not their CPU speed.

Of course, if you have 106 of the things, that's different. But you'll be paying over US$4M for it, which isn't exactly workstation class anymore.

As for raw compute performance, if you believe Sun's SPEC ratings from their product site, a 1.05GHz SPARC CPU is only just lagging behind an Intel 2.2GHz PIV on integer performance and beating it on FP.

Where do they claim that? According to the SPECcpu website, a 1.05 GHz SPARC III Cu gets 537 base SPECint and 701 SPECfp, while a 2.2 GHz P4 easily beats it with 790 SPECint and 779 SPECfp.

Intel is way ahead in integer, and although the Sun catches up somewhat in FP, if you look at the individual results, it's entirely due to one massive spike on the art test. They recently figured out a (controversial) compiler trick that gave them nearly an order of magnitude increase on that one SPECfp test, and doubled their overall SPECfp score. Sun are known for their stability & scalability, but not their CPU speed.

Of course, if you have 106 of the things, that's different. But you'll be paying over US$4M for it, which isn't exactly workstation class anymore.

AMD will enjoy a short period of incredible success with Hammer, its 64-bit 0x86 architecture. Intel will see the success and immediately release its own 64-bit extension to the basic IA32 architecture. Both chips will be the foundation of commodity, ultra-low cost, servers.

These servers will annihilate Sun in the low-end to mid-range portion of the server market. These servers will gradually creep into the high-end of the server market, where machines having 32 or more processors dominate.

Sun has seen the writing on the wall. As a last desperate measure, Sun has announced that it too will sell Intel/AMD-powered servers running Linux more than 1 year after IBM has been successfully doing the same.

Just look at the performance data at SPEC and TPC . The x86 processors crush UltraSPARC III across a broad range of benchmarks.

The P4 is a Dell Precision Workstation 340.
The Itanium is what mentioned on the intel site.

So it isn't really worthwhile to use an Itanium to run compilers or similar things.

I didn't. These are the numbers from official submissions to the spec organization. (If they can't bother to submit results then they don't count.) I took the machine from with the HIGHEST Int performance - as I said in my post. The FP is the result for the same machine - as I said in my post.
There is a >700 SpecFP machine claimed by Dell but there is no corresponding SpecInt submission. I think Intel claimed nearly 800 Spec2000FP for the Itanium but no one else has been able to re-create those results. That's why non-submitted results don't count.

Don't write off an entire architecture because you didn't like how the experimental implementation came out.

But it wasn't supposed to be a proof of concept chip. It was supposed to be the future of computing.

An exec actually admitted this in 2000.

Which is at least 3 years after they knew it. Intel instead spread FUD around while refusing to talk performance numbers.

From Intel Press Release:
SANTA CLARA, Calif., Oct. 4, 1999 - Intel Corporation today announced it has selected Itanium(TM) as the new brand name for the first product in its IA-64 family of processors, formerly code-named Merced. The Itanium brand extends Intel's reach into the highest level of computing enabling powerful servers and high-performance workstations which will address the increasing demands that the Internet economy places on e-Businesses. "The Intel Itanium processor represents a new level of processor capability that will be the driving force for the Internet economy,"

Ummmm... okay. I see, by "highest level of computing" and "new level of processor capability" they meant "proof-of-concept place-holder chip". It's all clear to me now...
The full text is here on Intel's site since you seem to think I make this stuff up.

Now McKinley is supposed to be the next, big thing.

From an article about McKinley previously on /.
"Applications will be about one and a half to two times faster than what you get on a (current) Itanium," said John Crawford, an Intel fellow in the enterprise platforms group.

The additional bus and processor speed and 3 megs of on-chip on speed cache should deliver nearly a 50% boost all by itself. If the "new" features of the McKinley don't add much more beyond that then where are they going?

Forgive me if I appear skeptical...

=tkk

Tux is a kernel mode web server, but X15 is a user mode web server and has the same performance.

According to SPEC CPU2000 Results, The Itanium at 800 Mhz performances equivalent to a Pentium III 800 Mhz. In fact the Pentium III is a little faster.

Do you have any other figures to substantiate your claim ?

The 1 GHz G4+ that powers the current generation of Macs would probably score about the same as the R14k on SPEC, or a bit lower

The 1 GHz G4+ that powers the current generation of Macs would probably score about the same as the R14k on SPEC, or a bit lower

I think you're right. There are published results at the SPEC website for the R14000 at 500 MHz. Here's the bottom line (CPU / SPECint / SPECfp - all rates are base):

R14000 500---------- 410 / 436
Athlon 2000+-------- 697 / 596
Pentium 4 2200------- 771 / 766
IBM Power4 1300----- 814 / 1169

Looks like SGI should consider joining Apple in the PowerPC world...that Power4 looks pretty awesome!

299,792,458 m/s...not just a good idea, its the law!

Here are some numbers from Spec.org

This is for Spec CFP 2000 (i.e. floating point). I picked the SGI Origin 3200, which has a similar processor (although I'm not sure if its identical or not).

Advanced Micro Devic Epox 8KHA+ Motherboard, AMD Athlon (TM) XP 2000+ 1 596 642
Intel Corporation Intel D850MD motherboard (2.2 GHz, Pentium 4 pro 1 766 777
SGI SGI Origin 3200 1X 500MHz R14k 1 436 463

The Spec CINT 2000 numbers look similar, I just didn't feel like cutting and pasting.

So, sure your average P4 or Athlon is faster, but its not as simple as a matter of Megahertz.

My concern about SGI is that these machines have the same graphics V10 and V12 that they've been using for years now. I heard that these were designed as the last hurrah of designers who have since gone on to Nvidia or ATI.

I wonder if SGI has the manpower left to design new, innovative graphics architectures, or will they be just slapping more texture and cranking the clock on old designs.

dave

No, Intel, like some other chip/OS manufacturer (former DEC w/ Alpha), has its own compiler. Which is heavily worked on, to get decent SPECint/SPECfp figures ; with the added bonus that they can both test the chips and the compilers to see what chip design ideas are good and what are wrong "good ideas" in practice.

The best performance measure is running your code on a variety of systems. Because most people can't do that, it may make sense to look for standard benchmarks that look like your code, and then make analogies based on the similarities of those loads to what you want to do. It's critical to pick the right benchmarks to have a good analogy; if you're interested in 3d performance, it doesn't make sense to make performance comparisons based on the number of rc5 keys per second.

Unfortunately, the Open Magazine article doesn't give any information on what exactly their tests are doing. So it's not possible for you to figure out which, if any, of their tests will be analogous to your code. :-(

As I've mentioned before, I'm mostly interested in integer performance. From what I've read about the Intel C compiler, its strength is floating point. If I did a lot of FP work, I'd be sending Intel a credit card number about now, and I imagine many FP people will.

But for integer work, I think it's not so clear. Andreas Jaeger has a nice page benchmarking versions of GCC. On Athlon processors, SPEC CPU2000 CINT2000 base looks like it's around 10% faster when built with the Intel C compiler than with GCC 3.0.1. I think I can live with that.

It's a lot easier to modify gcc than icc too, and yes, I really do hack on gcc from time to time.

The best performance measure is running your code on a variety of systems. Because most people can't do that, it may make sense to look for standard benchmarks that look like your code, and then make analogies based on the similarities of those loads to what you want to do. It's critical to pick the right benchmarks to have a good analogy; if you're interested in 3d performance, it doesn't make sense to make performance comparisons based on the number of rc5 keys per second.

Unfortunately, the Open Magazine article doesn't give any information on what exactly their tests are doing. So it's not possible for you to figure out which, if any, of their tests will be analogous to your code. :-(

As I've mentioned before, I'm mostly interested in integer performance. From what I've read about the Intel C compiler, its strength is floating point. If I did a lot of FP work, I'd be sending Intel a credit card number about now, and I imagine many FP people will.

But for integer work, I think it's not so clear. Andreas Jaeger has a nice page benchmarking versions of GCC. On Athlon processors, SPEC CPU2000 CINT2000 base looks like it's around 10% faster when built with the Intel C compiler than with GCC 3.0.1. I think I can live with that.

It's a lot easier to modify gcc than icc too, and yes, I really do hack on gcc from time to time.

Distributed.net's RC5 client is _not_ and integer benchmark. It was not written to give any kind of useful comparison of different CPUs. The Athlon RC5 core is hand-tuned specially for that CPU. There is no hand-tuned core for the P4, which is one reason it does so poorly. Search the rc5 mailing list: This has already been discussed. If you want an integer benchmark, stick with SPEC marks.

1. The rumored G5 is a joke. No, bear with me. The numbers bandied about for the (illusory) G5 at 1.6 ghz are preposterous (1342 SpecInt2000 and 1364 SpecFP2000 -- ha!)

Given that SPEC CPU2000 v1.2 int_base numbers for a G4/867 under OS X 10.1 are roughly 320, a 1342 result on a G5/1.6 would be a truly...astonishing improvement.

Apple is a member of SPEC; it would be nice if they were to submit results on their hardware instead of sitting on them. But their membership does give me hope that they are concerned about performance on workstation loads, and will continue to improve.

1. The rumored G5 is a joke. No, bear with me. The numbers bandied about for the (illusory) G5 at 1.6 ghz are preposterous (1342 SpecInt2000 and 1364 SpecFP2000 -- ha!)

Given that SPEC CPU2000 v1.2 int_base numbers for a G4/867 under OS X 10.1 are roughly 320, a 1342 result on a G5/1.6 would be a truly...astonishing improvement.

Apple is a member of SPEC; it would be nice if they were to submit results on their hardware instead of sitting on them. But their membership does give me hope that they are concerned about performance on workstation loads, and will continue to improve.

1. The rumored G5 is a joke. No, bear with me. The numbers bandied about for the (illusory) G5 at 1.6 ghz are preposterous (1342 SpecInt2000 and 1364 SpecFP2000 -- ha!)

Given that SPEC CPU2000 v1.2 int_base numbers for a G4/867 under OS X 10.1 are roughly 320, a 1342 result on a G5/1.6 would be a truly...astonishing improvement.

Apple is a member of SPEC; it would be nice if they were to submit results on their hardware instead of sitting on them. But their membership does give me hope that they are concerned about performance on workstation loads, and will continue to improve.

1. The rumored G5 is a joke. No, bear with me. The numbers bandied about for the (illusory) G5 at 1.6 ghz are preposterous (1342 SpecInt2000 and 1364 SpecFP2000 -- ha!)

Given that SPEC CPU2000 v1.2 int_base numbers for a G4/867 under OS X 10.1 are roughly 320, a 1342 result on a G5/1.6 would be a truly...astonishing improvement.

Apple is a member of SPEC; it would be nice if they were to submit results on their hardware instead of sitting on them. But their membership does give me hope that they are concerned about performance on workstation loads, and will continue to improve.