Cherry OS is People
An article on MacWorld regarding the announcement of Cherry OS seems to have thrown a bee in the bonnet of a lot of mac users. This is only going by my mail, but they're predominantly skewed towards Cherry OS at the moment.
The claims are pretty impressive: OSX running at 80% of the speed of your Windows PC, hardware translation (plug in a USB mouse, etc and it gets mapped to OSX environment), etc., and all for $50.
Most of it emails seem to be wigging out with the impression that the Mac platform is the equivalent of a three-legged table, and this could kick one of them out, causing the whole platform to topple. I.E., it'd adversely affect hardware sales if people can just run OSX on their PC with a 20% speed hit instead of having to buy a mac, and that's where most of Apple's money comes from.
There really isn't a whole lot to worry about here. Basically, 99% of the people using Virtual PC on the Mac are doing so for two reasons:
- To occasionally use a must-have application that doesn't exist for the Mac
There are a lot of these... sometimes there are apps for which there are no real alternatives on the mac, sometimes you are just bound to using a specific application by your company, etc. Most often its a very niche app that the Mac just has no correlation for. Visio might be a good example for many of programmers until OmniGraffle started getting up to speed. - Various forms of testing
If you're a web designer on the Mac, you need to test with a PC. While you can buy a cheap PC to sit on your desk for testing pages, this can really start to suck if you are trying to work out of a hotel room. If you're writing software with a PC port, you're pretty much in the same deal, etc.
It's worth noting I said on the Mac, as there are a whole bunch of other reasons people use virtualization technology within the same hardware (think of partitioning a server to use Windows Server 2003, 2000, and Linux all at once for various applications), but on the mac, the above covers 99% of it.
Remember, VPC doesn't exactly have stellar performance on the mac and, if anything, performance has degraded under OSX. It barely has functional G5 support (And no, I do not hold Connectix nor Microsoft at fault for this and those who do are asshats... the deal completely changed with the G5. The fact that they released a G5 version at all surprised me) and if you're lucky you are running at about one third of the speed of your hardware.
If you're lucky. And it isn't as though you have great 3D support, so games are out... no one is running Windows under VPC as their default; it's used to augment, and those who use it are in a specialized market... people who need to use specialized apps, and only occasionally.
If you assumed this thing was the real deal and could give you OSX on x86 hardware at 80% of your hardware's native speed, the Mac, by and large, still wouldn't have much to worry about simply because outside of some pro apps there is remarkably little that the Mac has there isn't an equivalent for on Windows, even if it's not quite as shiny.
No PC user is pining after some niche Mac app they have to use occasionally (occassionally being the operative word) and hence will forgo buying a mac to use it and will instead use this. There are some high-end pro apps, like Final Cut Pro, but for a variety of reasons they aren't that big of a danger as things currently stand.
As far as this situation is concerned, the Mac is about a whole bunch of little things hopefully acting in tandem to improve the experience as a whole, not any one app. If this was Linux you were talking about we might be onto something, but not Windows. Even there, there are already existing solutions for using Photoshop within Linux on x86... and let me tell you, depressingly few web designers/developers go out of their way to buy a mac to test on.
And, at the end of the day, none of the above really matters. Unless these guys are geniuses beyond geniuses who are bending entire computer science conventions, what they are saying they've done (OSX running at 80% of the native speed of your x86 hardware) isn't technically possible as things currently stand. I went over some of why this is back when PearPC came out, and little has changed.
This probably is PearPC repackaged in a very un-nice way, if it even exists at all, but due to the register starvation and the other issues this is either a complete hoax just for the sake of a hoax, or a scam to see how many people's hard-earned-clams they can rake in before people start mentioning they are not seeing the promised performance.
All-told, if MacWorld had spent some more time chewing on this cherry instead of swallowing it straight they'd probably have found a really big pit.
Posted by drunkenbatmanOctober 13, 2004, at 07:42 PM
Comments (8)
Posted by: Sam Pullara at October 13, 2004 10:32 PM
My gut reaction was "PearPC" repackaged or a complete hoax.
Posted by: P and T at October 14, 2004 12:04 AM
They do say they use a virtual machine to emulate a G4 so I assume it is a just in time compiler which is something you say holds back Pear PC performance?
Posted by: at October 14, 2004 12:18 AM
From their website:
" You can expect to get about 80% of your processors power when working in the Apple Environment. For example a 3.2ghz P4 would run as fast as a 800 MHz G4 machine."
They're crooks. Just by MHz 80% of a 3.2 ghz Pentium4 is a 2.6 ghz Pentium 4. So an 800 mhz G4 equals a 2.6 ghz Pentium 4? Hogwash. It is a scam.
Posted by: Ross at October 14, 2004 02:55 AM
The CherryOS thing is a fake. It looks to me like they were trying to get free stress testing on their streaming video server (vx30.com) and slashdot gave it to them :)
Posted by: kristine at October 14, 2004 09:07 AM
Hoax. Total hoax. If it were real, and really being released, why not take preorders with credit cards? Hoaxy hoax hoax.
Posted by: lightningrod at October 15, 2004 03:32 PM
Ross is right. Why else would they have a totally out-of-place link to vx30 on the CherryOS site? It doesn't make sense. Geniuses who could get 80% speed wouldn't do that, because they know that it's out of place.
Strangely, I don't remember that link being there before the slashdot attack. (or during, either. they put a simplified page up during the slashdotting, because their ASP system couldn't serve up the STATIC pages fast enough! why would they use ASP for static pages??)
Posted by: drunkenbatman at October 15, 2004 05:43 PM
P >> They do say they use a virtual machine to emulate a G4 so I assume it is a just in time compiler which is something you say holds back Pear PC performance?
A JIT (just in time compiler) would be of immense help in a situation like this, as basically you're able to take the hit once but then the VM doesn't have to go through the painful translation for every subsequent call that's the same. But even still, I'd be impressed if they were getting 30% of native hardware with it.
The problem is really architectural; think of an Xbox ramped up to ungodly MHz trying to emulate a PS2. The architectures have such vast differences that the code would constantly be 'breaking' and slowing everything down. You can see examples of this with something like the G5: if you try to run code that has been heavily optimized for the G4+ (and altivec) the G5 starts tripping all over itself and performance takes massive hits, but everything is solved with a few tweaks and a recompile. The G5 example is just that, an example, but it should give an idea.
The PowerPC was just a great machine to emulate x86, especially the G4 on down (which allowed endian flipping, as macs and pc's use different endians) and the x86 is just a horrific machine to try to emulate PPC. And even with that, PPC emulating x86 is slow, slow, slow even with huge amounts of optimizations.
If Altivec wasn't involved at all, my position might have to be recalculated but not by much. The same if you're talking about AMD-64 specifically over the Pentium4.
Now, if you had a JIT, they might well be able to get some best-case-scenario where on one specific opcode they were able to see 80% of the speed... but even that is doubtful, and the fact that they'd have a JIT at all would be doubtful.
Virtual machines are hard, hard, hard, especially VMs with any amount of performance. Just ask a project like BOCHS which, while a great project, is incredibly slow. There aren't many of them out there, most come from very large well-funded companies and even fewer allow what Cherry OS says they have.
They may well have something, but when you're dealing with such extreme claims the liability of proof is in their court... and they haven't stepped up.
Oh, oh, FIRST POST. ;)
I wonder if these guys are the same guy(s) that started macMice.... It WOULD be possible to get that level of performance, but they would have to be "geniuses beyond geniuses who are bending entire computer science conventions" and make Hawkings looks like a drunk monkey.
Frankly, the best option is to buy a Mac and a cheap PC, and use RDC.