Skype 5-way Calling Limit Cracked

BobPaul writes "It turns out when Skype limited 10 way calling to Intel Processors only it really was arbitrary! Maxxus has a patched version of Skype that allows 10-way calling regardless of the processor installed. There's also info about the patch: "The patch is the result of two phases: code analysis and design of the patch. The code analysis, or reverse engineering, reveals the relevant code block, which overrides Skype's limitation for Intel's dual-core CPUs. The patch design isolates the minimal set of instructions that need to be modified to cancel this limitation." Windows only so far."

Lawsuit

I think this shows this was done on purpose to lock out amd users. A lawsuit by amd should be succesfull.

Re:Lawsuit

Yeah, except they're not locking out just AMD users. They're also locking out anybody who has an Intel chip that doesn't meet their arbitrary requirements. So to me it sounds more like a forced obsolescence plan to get people to upgrade to higher end PCs.

Re:It -is- AMD theyre locking out.

this could easily allow AMD to have their chip be configurable to return GenuineIntel

there was already a case decided involving sega that using a trademark to lock out interoperability is not permitted.

Re:Law suit bullshit...

Of course, I'm ignorant. But how come a law suit? Companies make marketing arrangements all the time.

The rules change slightly when you've got a near-monopoly. This is part of what tripped up Microsoft in their anti-trust trials.

The problem is that it's far easier to convince someone to exclude "the competition" from the market when the competition has a disproportionately small portion of the market.

For the ease of math, let's say that the Skype market is 90% Intel, 10%AMD. If Intel had to pay Skype 10million to compensate Skype for the lost market in excluding AMD then AMD would have to pay 90million to get Skype to do the same thing. Add to that the fact that Intel has 10x as much income from their larger market share (presuming the same gross profit margin -- which is rarely accurate in a near-monopoly situation) and you have a 90-1 difference in impact on their profit margins.

Or - - to put it another way, between gross profits and market share, Intel could afford to buy off 100 market slots for every one that AMD could afford to.
If it came to open warfare like this, AMD would be reduced to a tiny portion of the market and customers would be effectively unable to even find business that deal with AMD. (Dell anyone?). Once you further reduced AMD's market share like this, Intel's ability to further marginalize them would increase until AMD was reduced to an insignificant market access independent of the relative quality of their products.

It's basically a market-ratio squared relationship which can easily spiral into a near-absolute market ownership, denying customers any real choice in the market no matter how good the competition is. (MS/Linux, anybody?)

It's actually a worse than ratio squared relationship because we haven't taken into account the probability that, if Intel has a 100-1 ratio of market-exclusionary agreements, they can now charge a higher profit margin without significantly affecting customers' willingness to buy AMD. That, however is harder to quantitize, so I'll only mention it, rather than including it in my math.

About the only real way to avoid this problem is to create artificial rules designed to stop such market-killing agreements when the market gets too lopsided, to prevent market choice from getting totally destroyed.

Re:Law suit bullshit...

A marketing arrangement, as you call it, is one thing. But in this case, it goes beyond a mere marketing arrangement...it is a very specific type of marketing arrangement in which one company (Skype), in exchange for cash, artificially cripples their product to only work with a certain other (unrelated) product (in this case, dual-core Intel CPUs). The key element here is that the product (Skype) would, if left to it's own devices, work with either Intel or AMD CPUs, but has been crippled in it's use with a certain subset of those artificially.

This is really a form of product tying [wikipedia.org], which was made illegal by the Sherman Antitrust Act (1890), but only if a non-trivial amount of business is affected by the tying. This last requirement will likely be the reason the suit isn't successful, but that certainly dosn't mean that the behavior isn't borderline, at best.

Again, this isn't a compatibility issue, as you said, "Why should Skype have to write software to work on AMD?" The real question is "Why should Skype be allowed to artificially exclude AMD users in exchange for money from AMD's competitor?"

Re:Law suit bullshit...

The problem is that Skype tried to blow smoke up our arses, and obviously we don't like that, and get suspicious, and think that something illegal is going on.

In this case, I would say "get a life." Honestly, getting hot and bothered because "Skype tried to blow smoke up our arses" sounds like childish foot stamping. Show your displeasure with Skype by not using their product, spend your valuble time (it is, right?) doing something productive rather than reactive. By the way, all this frothing and arm waiving will accomplish nothing at all, it's wasted energy. Move on to some other VoIP app in the secure knowledge that due to their greed, Skype will soon be dead. Or part of some spyware package...

Aaaah Maxxuss

Ah, Maxxuss. Is there anything you can't do? First you crack OS X, now Skype. You and DVD Jon should team up and become some sort of cracking superheros.

Re:Aaaah Maxxuss

Next stop: HDCP!

(I hope.)

Re:Aaaah Maxxuss

Seems it has allready been done [securityfocus.com].

should teach intel a lesson

people really don't want to pay money to have less features in a product you dimwits [intel]. I hope AMD sues ya ass to kingdom come for anti-competative practices. where are maxxas based, somewhere where they can't be served with a DMCA order?

Re:should teach intel a lesson

I think an xp2000 should be enable to push 10 streams via vnware.

I mean, wtf is supposed to take the cpu power?
An xp 2000 can compress mp3 in 10 times realtime, for example. Plus in a conference call, you actually send THE SAME to everybody...
Mixing audio is quasi cpu free (with less than 50 channels or so)
Sending data over the net is nearly cpu-free.

So what needs a dual core cpu for 10 connections?
I would understand that for VIDEO streaming, but its simply inconcievable for audio...

Re:should teach intel a lesson

I wonder what your reaction would be if microsoft announced they believe Intel's CPUs aren't powerfull enough to run Vista in all its glory. So if it detects an Intel CPU, you can only open 2 windows at a time. It's microsoft's product, so their free to market it as they like. So either you don't buy Vista which has previously informed you about this limitation (there are plenty of alternatives), or you throw away your Dell... How about if Oracle started thinking the same way and limited the conections to their DB buy CPUID (just so people don't exclude MS for their monopoly)? With all do respect, it isn't so funny now, General 8883 sir, is it? :)


One thing I also haven't seen pointed out yet is this: Let's say that AMD's CPUs really can't handle 10 conferences at the same time. How can skype guarantee that this will be the case six months down the road. Determining the CPU's capabilities by its manufacturer is lame...

"Arbitrary"?

Since this limit was "arbitrary", that means the only deciding factor was not technology, but money. I wonder how much the block cost Intel?

And now that it's in the open, (like that was going to take very long?) I wonder if they'll remove the block?

Re:"Arbitrary"?

It'd be an interesting test case for the DMCA, wouldn't it? In this case it's not specifically copy or content protection software that's being circumvented, but a feature designed to maintain (potentially) a marketing agreement, if in fact that's what this turns out to be.

Re:"Arbitrary", but they already admitted it

In the first article [com.com]on this deal that slashdot linked, Intel admitted the limit was arbitrary, and the result of a marketing deal:

But there are no specific instructions in Intel's current Pentium D or Core Duo chips that enhance the performance of VoIP applications, an Intel representative said. Skype is using an operation called "Get CPU ID" to identify the type of processor running on the PC. The Skype software has been preset to only accept Intel's chips as having the performance necessary to host conference calls of more than five people, the representative said.

Re:"Arbitrary", but they already admitted it

The Skype software has been preset to only accept Intel's chips as having the performance necessary to host conference calls of more than five people, the representative said.

If that was said by a representative from Intel then that statement quite qualifies as misrepresenting a competing product. Comparison is perfectly fine, misrepresentation is definitely not and Intel should be forced to compensate for it.

Now all that's missing

..is someone demonstrating that the X2 can in fact handle 9+1 persons at once, which I have no doubt it can. Then it's time for Intel to open up the wallet and give AMD some nice $$$ and some even nicer PR. Stand by to bend over!

Victory for AMD? Not so fast!

I know what you all are thinking. Great! He hacked it with three bytes, and showed his work. Now all AMD needs to do is get a 10-way conference call going on an X2 and they'll have another strike in their lawsuit against intel.

But wait -- there is a way out. See the code is written to identify CPUs, and to run on dual core CPUs, but it doesn't make that distinction for AMD. So all the defense needs to do is set up an XP box running an AMD 1.4 GhZ "Firebird", next to some oily rags, get a 10-way conference call going, and simulate a CPU heatsink failure. Clearly they were blocking AMD 10-way calls out of product liability concerns.

Optimization is where?

Skype made a lot of noise in their press release saying that the 10-way feature was "optimized" for Intel chips. This was picked up by the media of course as well as evidence of AMD's poor performance.

I'm having trouble understanding what this optimization that used the special features of Intel chips (presumably their high power) was. It looks from the patch that they just check who the manufacturer is, and if it is not AMD, they pretend your computer doesn't have the power to host 10 participants.

What's also interesting is that folks likely signed up for SkypeOut and other paid products not realizing that they would be treated differently depending on what chipsets they happen to use, especially as that choice matters almost no where else. They should give more warning about this to paid users.

This focus on locking software into specific vendor chips seems a dangerous one. No longer will it be the best chip that will win, but the focus goes to competing on locking up software applications. The proprietary unix'es went down that path, and it would be sad if Intel managed to get that to happen here.

Re:Optimization is where?

I'm having trouble understanding what this optimization that used the special features of Intel chips (presumably their high power) was. It looks from the patch that they just check who the manufacturer is, and if it is not AMD, they pretend your computer doesn't have the power to host 10 participants.

My guess, like yours, is that this is blatant marketing crap. But it would be nice to see some tests of how many people can be usefully conferenced on different hardware. Skype is a CPU pig, and it's possible that they really have optimized it for some Intel-specific feature.

Re:Optimization is where?

To anyone doing such testing, make sure the code running is the same that would run on a dual-core Pentium. I didn't follow the patch in detail, but you'd have to make sure that any flags changed after CPU detection (like for instance the one at 0xB8E6DC) is the same for both cases. Else you might find yourself in the situation that, yes, the limit is removed, but you're still running a different ("unoptimized") path. In the (very interesting case) that the code crashes on AMD (due to use of Intel-only prefetch instructions or whatever, I don't even know if such still exists?), such crash can be used to land smack boom right in the relevant code to analyze.

A good reverse-engineer could then fix the code if needed (substituting or even noping the faulting ops) -- the theory being that the major optimizations are in fact portable.

In fact, demonstrating that there truly really is only one code-path would be pretty damning too; that's evidence this is just pure PR with no grounding in tech at all. So either case makes for an interesting evening in front of IDA.

Re:Optimization is where?

due to use of Intel-only prefetch instructions or whatever

The AMD instruction set is a strict superset of the Intel instruction set. There are no Intel-only instructions anymore. There are however many AMD-only instructions (3dnow, 3dnow+, etc.), so if the situation were reversed, there might have been a legitimate claim, but since the AMD CPUs were locked out, it is clearly a bribe^Wmarketing descision.

Skype Patch... from Russia with Trojan??

Yeah right.. I'm gonna' d/l this patch and app from www.silo.ru (in Russia!!!), they even say on their aite that that "There is no virus or backdoor added!". You've got to be kidding me!

Poor programmer at Skype

Will we see transcripts from depositions done by AMD?
I'd bet that they will be as funny as some of the SCO transcripts.

I'd bet that they will depose the programmer who wrote the code encryption and the GenuineIntel check, and then continue with his supervisors.

Who authorized to add code encryption?
Who approved it?
How were the limits to 5 or 10 concurrent connections determined?

An encrypted binary?

Anybody know anything about their encrypted binary? I can't figure out what they were trying to achieve with that. Some sort of misguided anti-hax0r protection? Or perhaps they're trying to conceal something...

Limit

So what is now limiting the conference calls to 10 people now? Is that a phone company limit, or another arbitrary limit?

Probably a limit due to bandwidth or latency...

Unless Skype's playing reflector for the whole conference, each peer's connectivity limits what you can/can't do.

At 128kbps (the average upstream speed on broadband these days in the US...), you can typically host a four to six way voice conference or a 2-3 way video conference. This is because you have to provide the outbound traffic for each of the peers and control traffic. With a reflector system, you can host larger conferences, limited only by the inbound bandwidth because the reflector is flipping the traffic from your mic (and possibly camera...) to all the participants. However, that's REALLY bandwidth intensive, so to keep it economical, you'd probably limit it to 10 participants or so to limit hogging of that limited resource.

Now, this is all due to everything being unicast UDP. If we had IPv6 and Multicast support for the same available, one could handle at least up to the 10 without needing a reflector as the router infrastructure would handle it right along with the video on demand, etc. streams. However, since this is not likely to happen in our or several generations' lifetimes at the rates things are going, waiting or wishing for that is a waste of time. :-)

Maxxuss

Maxxuss is definitely making a name for himself (if it really is only one person). He is already heavily involved with removing Apple's restrictions on Mac OS X for Intel too.

BitTorrent Mirror

BitTorrent Mirror here [mininova.org]

Don't Believe the Skype

The code seems to be calling the cpuid instruction, so as far as the "Windows-only" patch, could anyone comment on patching the Linux kernel to essentially lie to the Skype client?

Or, so as not to break other programs that use cpuid (to determine which instructions they can run, for example) perhaps this could be done in a user-space way.

I'm thinking of artsdp as a model, so you would just launch your Skype client with something like "cpufake --cpuid='Genuine Intel Dual Core We Like Skype' skype.bin" (or whatever it's called.)

I've got no idea how such a program would work, but the article did say the code was encrypted so I wonder if that would be an issue.

Indeed...

Considering that SIPPhone already HAS voice conference calls of at least 10 or more, works with ANY SIP enabled device that's not crippled to a single provider (Vonage devices come immediately to mind...), and costs nothing for VoIP calls- I'd say, skip Skype all together, especially after this little stunt.

Re:Indeed...

Svartalf, you are indeed insightful today. Personally I use Asterisk meetme conferences for all my conferencing needs. Only limited to my server horsepower and available bandwidth. Can have almost an unlimited number of people call into it through SIP, IAX2 or an IAX2 trunked PTSN #.

I tried skype a couple times (mostly because some girl talked me into it), but she wasn't worth it. The lack on interoperability totally killed it. The last thing I need is yet another app running on my main console all the time. Asterisk runs happilly on my server in the corner and rings my normal home phones all over the house if someone is trying to reach me. I might even pay for a skype IAX2 or SIP access account. But being a closed system they are too much trouble to deal with.

Need Open Standards

Just goes to show why we need open protocols and open code for the future of VOIP. It's too important to leave to a single company, which is why I prefer SIP and clients like Google Talk and Gizmo [gizmoproject.com] where possible.

Re:Need Open Standards

In fact SIP has supported dial in/out for years -- you can get termination for SIP compliant phones from Vonage, using Free World Dialup, or from smaller termination only providers (similar to Skype Out/In) like EXGN -- there are literally hundreds of them. ALL the commercial gateways sold by Cisco and the other major players are SIP to PSTN (regular telephone system) gateway (or Cisco proprietary Call manager -- but not Skype). Even Skype themselves in the backend is almost certainly using SIP to get to the public phone network for their Skype In/Out system, since none of the major gateway companies build anything else, and Skype isn't building one off hardware, it simply wouldn't be economically practical.

There is also signifcant work to make SIP P2P to eliminate the central servers http://www.p2psip.org/ [p2psip.org] from SIP going on right now. As an aside, Skype isn't really even that P2P -- it uses central auth servers, so it is more of a hybrid system -- ala Napster -- in reality.

And with a SIP phone you can use *any* of those SIP providers. With Skype, you have one choice.

Skype is very good at making things work out of the box, hence the popularity, but there really isn't much (if anything) it can do that SIP can't. It isn't even that the P2P mattered. Skype's success is a matter of a very nice UI and user experience. They gained market on ease of use and marketing -- not bad things mind you -- not better technology. Kudos to Skype for making it easy for users to use VoIP, which was (and still is) notoriously hard to use with other providers. But the technology is different to allow Skype to lock up users, not to make things better from a technical standpoint.

Hackers/Crackers... The last revolutionaries?

You gotta love their efforts. Keep it up.

so what

well what can I say ... Intel is about to take one more for the team ... a law suit for unfare competition that is ... Other than that I really could care less about skype and their limit of 5 people on a conference call. The only cituation that you would use a conference call with more than 5 users is a corporate environment and I have yet to hear about one that uses Skype.

No shit

It was a made-up limit? No kidding.

Remember folks; Asterisk. Skype isn't open source, and the company behind it has it's own motives. Asterisk is open source, has a good community behind it, and can do *anything* you want it to. Regardless of the hardware behind it.

Optimizing for AMD

Being in a company who worked exclusively on Intel and nVidia chips until recently, it is possible to have horrible performance when switching to AMD and ATI. In our case, we didn't use any nVidia specific GL calls. As for SS2, it is supported on both platform so in theory it shouldn't be an issue. The reality is, unless you are making a game and using what I'd call "game-oriented opengl calls", the performance is going to vary a LOT between ATI and nVidia. Don't believe the hype of these companies when they say that they support full OpenGL. Some either have very bad hardware for 2d ops with OpenGL or literally do software "decelleration". Benchmarks have shown speed dropping as much as 200% in some areas. As for AMD and Intel, after patching the executable, the performance was different, sometimes in favor of Intel, sometimes on AMD.

With that being said, no platform specific instructions or features were used. I suspect the Skype guys may have simply used Intel machines for so long and never bothered using AMD machines for development and then were too lazy to simply rewrite some of the code so that it runs normally on AMD. This happens especially when you write tight assembly loops by taking into account instruction latencies for one processor and then realize the performance sucks on another platform. You then have the choice, rewrite it so that the performance is similar, or slap a OPTIMIZED FOR INTEL on the box.

Thankfully we rewrote.

Not "optimized"...

Sound software typically is optimized quite well for both company's offerings, esp. if you don't use special features of the other brand. To be sure, there's edge cases where you need that- but Skype's NOT one of them nor is there evidence that there's the degredation you claim happened with your app experience. (Not to mention that there's tons of other P2P VoIP applications that use SIP and Jabber technology that work as good or better than Skype, and there's other commercial proprietary systems (such as Eyeball Networks' stuff...) that DOES handle up to 10 people (or more as bandwidth will allow...) without needing a dual core anything, let alone an Intel one.)

This is plain and simple being bought to support one over the other. Please don't try to defend this- it's not something that has much of any good explanation for this, especially considering that they actually DO appear to be just CPUIDing and crippling the app if it's not a dual core Intel CPU...