After almost five years and with over 330,000 participants Distributed.net has finally solved the RC5-64 64-bit encrypted key Challenge. The best day was equivalent to the power of “32,504 800MHz Apple PowerBook G4 laptops or 45,998 2GHz AMD Athlon XP machines”! To solve this challenge huge teams were formed, the most powerful team was assembled by the Dutch Tweakers.net website, while my personal favorite, Team RC5 Amiga, earned a honorable 7th place.
rc5-64 Project Completed
2002-09-27 Benchmarks 26 Comments
You know it’s weird… I was browsing the stats just yesterday, and mentioned in BeShare how I thought it was interesting that the key would be in the last 12%, and if people would have joined it at the beginning if they knew it would be this tough.
I wonder what people will do now that they can’t have a moo cow on their taskbar?
Hopefully, we’ll see some of the enourmous horsepower put towards REAL science experiments, such as Distributed Folding, or the Genome project.
>Hopefully, we’ll see some of the enourmous horsepower put >towards REAL science experiments, such as Distributed >Folding, or the Genome project.
Yeah, join folding@home and genome@home at
(both projects share the same distributed network)
Above all this challenge was more like a fun experiment. The daily and overall statistics offered an important competitive element. This competitive element was very important for its success.
Websites like AnandTech and Slashdot.org tried to demonstrate the power of their faithful following. The same goes with regard to platforms, teams were formed for almost any operating system or online communities.
Tweaker.net was able to attract over 6,422 members!
Amiga RC5 team 2,197
Team BeOS 624
IMO these are very interesting statistics.
So what are you getting at? There are 2,197 Amiga nerds out there with nothing better to do than engage in some meaningless race of computation?
Yeah, we Dutch rule with rc5-64!
Why do people sheer for their football teams?
Why is there so much interest for the Eurosongfestival?
Why do people count the medals earned by their country’s athletes, and so forth?
Competitions are fun for many people, it’s as simple as that. However when they loose they mostly react similar as you do. It’s just a game… Who care about Songfestivals… And when they win => 🙂 😮 🙂
No, football competitions do not save lives, nor are they useful for anything other than Entertainment.
It simply is our human nature.
> The best day was equivalent to the power of “32,504 800MHz
> Apple PowerBook G4 laptops or 45,998 2GHz AMD Athlon XP
So 800 MHz G4 is faster than 2GHz AMD?
and a crippled g4 at that, dont understand this at all.
Yes for searching keys, PPC processors are much faster. With proper code optimisations, PPCs are much faster per clock cycle than compared to equivalently clocked x86 CPUs.
Even classic Amiga hybrid machnines, equiped with old ~200 Mhz PPC CPUs, can do quite alot of work. Sadly however MacOS X is extremely inefficient and bloated (multiple CD!?) and therefor offers a rather sluggish overall computing experience. MorphOS for example already flies on low-end PPC hardware, probably AmigaOS4 will do the same.
Clock Cycles as an indicator, are only good as a very rough indication of true computing speed *within* a similar range of CPUs. However even then, there can be enormous differences, for example AMD solutions are mostly faster than Intel solutions, compared clock by clock.
Similarly it is very hard to explain to PC users, interested knowing more about classic Amigas, that a 25 Mhz 040 (68k CPU) Amiga is really alot faster than a 50 Mhz 030 (68k CPU) powered classic Amiga.
At some point in time, something within the computing industry went horribly wrong and the race for the highest amount of clock cycles was started and not for the fastest CPU. This is sadly a result of much hyped misinformation within the x86 market, resulting into big overheated CPUs and thus the need for noisy cooling fans.
resulting into big overheated CPUs and thus the need for noisy cooling fans.
Look on the bright side : this way you can cook dinner while never having to leave your pc ( http://www.handyscripts.co.uk/egg.asp )
I wonder if it is 2000+ or 2000MHz. There’s a big difference you know…. 🙂
Anyway, I noticed Distributed.net code is optimized for AltiVec, and because there is currently no altenative on current Athlon XPs (Clawhammer would introduce stuff like SSE2), it is slower….
Of course this doesn’t matter for Athlon XP’s main target audience – gamers.
Sadly, even though thousands of participants worked hard up until the day they finally closed the project (Sept. 25 2002), last night they deleted all stats data that came after the winning key which was actually submitted July 14, 2002.
Just because the last few months of crunching wasn’t “necessary” doesn’t mean it should be ignored and forgotten. Some people do it for the “competitive element” and whatever rank they’ve earning in stats over the last few months is being rolled back.
Decibel’s .plan mentions the deletion of stats data:
I still managed to graph the entire history data for myself and some friends before they destroyed it. http://dnetc.org/?s=stats
Mike Bourma: Yes for searching keys, PPC processors are much faster. With proper code optimisations, PPCs are much faster per clock cycle than compared to equivalently clocked x86 CPUs.
But of course, AMD and Intel no long make 800MHz processors….
But I wonder how fast Pentium 4-M, using something optimized for SSE2, would be compared to the G4…
> But I wonder how fast Pentium 4-M, using something
> optimized for SSE2, would be compared to the G4…
They will be alot faster. I stated this many times before: The currently available top range of x86 CPUs outperform the top range PPC CPUs. I believe this is mainly due to the need for legacy x86 compatibiliy for desktops, thus alot more money is being spend on x86 CPU development.
But I also believe:
1) For 99% of computer users the current top PPC CPUs would be more than enough, *provided* an efficient PPC OS would exist.
2) PPCs generally allow smaller devices or for instance less noisy computing systems.
I believe there is also a matter of taste, for instance from a hardware perspective I think the GameCube is alot more elegant than a big, heavy and rather bulky xBox.
> They will be alot faster.
I meant the top models not per clock cycle, else the G4 will easily win.
I did a quick search on the net and here’s an interesting comparison article:
Here’s another article I found at TechTV, a quote:
“In 1999, for example, Intel’s Pentium 3 chip reached 500 MHz, compared with the Apple G3’s 400 MHz. Today the performance gap is even wider, with Intel’s Pentium 4 chip running at 2 GHz, compared with the Apple G4’s 876 MHz”
So now we know at least one source where misinformant information comes from with regad to clock frequencies! (Maybe not suprizing as Paul “Mircorsoft founder” Allen is the owner of TechTV)
I think the cracking speed has a lot more to do with client optimization and OS overhead than actual processor speed. I remember running RC5 clients a few years ago on an Amiga and an old Mac, both of which had the same CPU at the time (68030/50MHz). The Amiga client was about 3 times faster…
I am sure client optimization plays an important part as well. But if that would be the dominant reason, then that would in efect imply that PPC coders are better than x86, and Amiga coders better than Mac coders.
I believe that can’t be true. And I don’t think there is a smaller OS overhead with MacOS X or PPC Linux in comparison to WindowsXP.
Overall, the newer x86’s perform better than even the 1.25ghz G4’s but NOT for d.net’s rc5 algorithms. You’ll notice in the Client speed database (link at the bottom of http://stats.distributed.net ) that g4’s blow away the x86’s in rc5, but go to another contest like the currently running OGR, and G4’s are far slower comparatively…
Distributed net is and has been doing OGR which searches for Optimum Golomnar rulers
on distributed nets site it says this
“OGR’s have many applications including sensor placements for X-ray crystallography and radio astronomy. Golomb rulers can also play a significant role in combinatorics, coding theory and communications, and Dr. Golomb was one of the first to analyze them for use in these areas.”
But OGR has no prize attached, they are prepping for RC5 72.
The amount of research done by all of these projects is astounding. Not just in the calculations but in the software and infrastructure to do them.
Mike Bouma: 1) For 99% of computer users the current top PPC CPUs would be more than enough, *provided* an efficient PPC OS would exist.
But with similar performance as lower end *cheaper* processors, a good OS on low end x86 is better than on PPC.
Mike Bouma: 2) PPCs generally allow smaller devices or for instance less noisy computing systems.
Try Transmeta Crusoe and VIA C3. Or Intel codename Banias when it comes out.
Mike Bouma: I believe there is also a matter of taste, for instance from a hardware perspective I think the GameCube is alot more elegant than a big, heavy and rather bulky xBox.
Funny you should compare this, XBox takes less space than GameCube. And because it has more features (for example, its graphics and sound chipset), you can’t blame the amount of space XBOX took based on comparing with something that is much much more underfeatured.
As for elegant – that has nothing to do with x86, and it is up to the eye of the beholder (personally, I find GameCube the ugliest among its competitors, notably PS2 and XBox).
Mike Bouma: So now we know at least one source where misinformant information comes from with regad to clock frequencies! (Maybe not suprizing as Paul “Mircorsoft founder” Allen is the owner of TechTV)
LOL… Paul Allen was once of PPC’s biggest cheerleaders.
Anyway, if you watch shows like Call For Help, Freshgear and Screensavers, you would see they are more lopsided to the Mac (especially Brett Larson).
I am surprised that the usually well-informed people here actually fall into the trap of mistaking the RC5 client for a benchmark.
The RC5 client is a small piece of highly optimized assembler with some auxiliary code (sending and recieving keys etc). As it happens, the RC5 algorithm is based upon a specific assembler instruction (or sequence of instructions, dunno, this is third-party information) that the PPC CPUs execute much faster than the IA32.
This has *nothing* to do with real life performance. That’s why real benchmarks go to great lenghts to use a similar pattern of assembler instructions as used by real-life applications.
Back then, my Amiga (MC68060@50 MHz, MPC603e@240 MHz) topped at > 950 kkeys/sec, outperforming the Pentiums ~233 MHz at the office by a factor of 3. I do tell you, this Amiga was *not* three times faster in everyday use, not by a long shot.
RC5 is not a benchmark period.
> But with similar performance as lower end *cheaper*
> processors, a good OS on low end x86 is better than on
Equivalent x86 CPUs all need coolers. Also many Amiga developers prefer to write their code in PPC assembler. Which suits your needs better depends on your personal preferences.
> Try Transmeta Crusoe
The Transmeta isn’t really a x86 CPU, it can translate other CPU architectures as well. The transmeta currently offers no significant advantage within the PPC market and therefor isn’t used for related PPC (i.e. iBook, GameCube, Embedded) purposes.
> VIA C3
This is one of the worst performing CPUs within the x86 market. Per clock cycle these are even worse than Celerons! Not recommended IMO. Generally I would say, go for AMD or for Transmeta (for small mobile devices) solution instead.
> XBox takes less space than GameCube.
Huh? Are we talking about the same devices here? The GameCube is about half the size and weigh and is around 5 time lighter.
> And because it has more features (for example, its
> graphics and sound chipset)
On paper the xBox seems like a brute-force machine and actually it is, however in reality the real bottleneck lies with ordinary TV sets.
Dolby Digital is cool but only if you own a similarly as cool Dolby Digital stereo HiFi set.
> As for elegant
I stated that I used the term “elegant” from a hardware perspective, it’s a matter of personal taste however. IMO, under the hood the GameCube looks alot a better designed, than a xBox.
> LOL… Paul Allen was once of PPC’s biggest
I think you mean Transmeta!?
> I do tell you, this Amiga was *not* three times faster
> in everyday use, not by a long shot.
The main AmigaOS parts were still running on your 50 mhz 060 and not on the PPC. But PPC software running on Amigas is in general pretty fast compared to similarly clocked PCs, but for reallife games/applications performance the used graphics chips are extremely important as well. AGA (1992) for example is very old, adding a more modern graphic card to your Amiga often does wonders.
> RC5 is not a benchmark period.
No, but it is one of many reallife indicators. Although not something one should solely focuss on.
How can you say the Xbox takes less space than the GameCube? I could see the GameCube being a little taller than the Xbox, but the Xbox has gotta be 3x the width and at least equal depth, if not bigger.
And the power difference is almost nothing.
Wow, this is so earth shaking. They decrypted some meaningless message. Hope they won a candybar.
People are so frivolous. Why not use computers for a useful task?