20 minutes? Holy shit, where do you work? Antarctica on a 300 baud modem? The time it takes now for infection is on the range of seconds.
When CodeRed came out, some of us actually noted it on the job at UC Berkeley ResComp. The shortest one was on the range of 5 minutes., barely enough time to do an update from windows update.
Years later, for Welchia, etc, it was within 1 minute that we'd see the machine do the reboot by itself. So the infection actually took place before that (since the rest of the minute was the download and install of the virus)
It's a question of bandwidth(cpu power) versus latency. If you have a realtime app, unless very specialized it's not going to take all your CPU power simply because then there's no way to recover if you miss your schedule by a little bit without taking a hit in quality/detail. On the other hand, for computation apps, there's no way it's going to be realtime because they're all CPU bound anyways.
Modern realtime games are running a constant simulation, yes. However that doesn't necessarily mean that it's using all of the CPU time availiable. While the AI is moving around, it doesn't take much time for it to calculate, nor is is necessary. Think about it, does the AI do more thinking my CPU power goes up? If I have some decked out 8-way Xeon for a gaming system, am I going to have any significant experience than a guy playing on a P3 1Ghz? No.
Half-Life 2 was written around the minimum system requirements of a P3 700 or a P3 1.2 depending on where Google takes you but runs best on a P4 2Ghz. Two things are happening, either the P3 will get less detail than the P4 in some unknown aspect and will fly by unnoticed by the user, or it's simply because they expect the P4 will be able to load stuff faster and hence the better experience. Once you get beyond a P4 2Ghz, you're still going to be requiring the simulation run in realtime, but it's not going to take 100% of cputime. Running SETI@home will actually get work done without your system slowing down. And hence my argument stands.
I'm running Warcraft 3, United Devices(like SETI@home), and iTunes at the same time on an old Athlon 2000 which wasn't close to top of the line on release, and yet, it hasn't made a difference.
Now, the majority of average user applications don't have realtime aspects nor are CPU bound, and that's why I classified them to be unimportant in this argument.
But the ones I did highlight, all of them, once the job starts, full utilization all the way no matter what CPU. And given that some of the tasks take days, we're not close to any sort of fix to that.
With the growing amount of ram and processing power on a GPU, along with programmable shaders, we're going to reach a dead end in the realism bound by the output capibility of our monitors. At that point in time, nVidia, ATI, and friends will need to find another use for all that power so they can make money. Since the GPU is a massively parallel processor with fast access to the enviroment being rendered, it makes sense to put physics calculations there because physics arn't CPU intensive because they're complex, they're CPU intensive because there are just a ton of them. Hence perfectly matching the GPU architecture.
In the end, it comes down to this. With the emulated GPU piping stuff to the real GPU, if my dual G5 emulates a P3 700, that's enough to play most games. But if my quad Opteron still only emulates a 1Ghz G3, it's painfully useless.
Maybe it seems there's something I'm failing to explain in detail.
While I can't say that I understand what I'm supposed to be looking for in cs_office, I can still justify how my argument stands. And while I can agree that people are starting to find uses for the CPU in games, it's not going to be significant.
Note that for Half Life 2's system requirements, it will not have the CPU running full blast if I was running it on a Athlon 2400. Nor on a P4 2.4. Nor on a P4 3.0. But it probably will be on a P3 700.
When I meant by CPU bound, I meant really, it just can't get faster because of your CPU. The fact that Half Life 2 depends on human input for data to calculate, that physics models can sooner or later be pushed to GPU shaders, and that there's a limit in which the CPU can contribute to performance still puts games at lower CPU usage than apps like Photoshop, FCP, BLAST, Logic, simply because while HalfLife 2 will run very nicely on a P4 2.4, it will barely be any better on a P4 3.0 because adding CPU power doesn't help much. You'll barely get any more frames because you're already pushing 3D data to the GPU faster than it can handle.
On the other hand, the other apps I mentioned, if you have CPU power leftover, it will be scheduled to push your job faster. That is what I meant by CPU bound. Simply that the speed of the program is bound by the speed of your CPU.
My PC was built using $250 US about a year and a half ago (Athlon XP 2000+ 512 MB, Geforce 3). And if I price a Sempron 2200+ mobo/CPU combo at Fry's for $50, a $20 case, and 512 megs of memory, it's pretty competitive to your CPU.
But anyways. Given the the limitations of emulators, if (one heck of a big if) they do happen to reach 80% efficiency, with the 30% P4 penalty you'll end up with what amounts to roughly a 2 Ghz G4 for your one specific task. Not bad.
However if you're looking at what you'd expect for a real emulator to get, 30% or less, you're pretty much down to what you'll get with a 1 ghz G4 anyways. Or even less than that. And still this is only for one task. If you do anything else with it, the performance will rarely go up, usually down. Not to mention no GUI acceleration.
Sure, all that for $50 (and pirating OSX) is fairly nice. But it's not too far from just building or buying a used mac. With actual Mac hardware, pirating the software will feel like less of a crime.:)
It's possible to price out a used G4 on ebay that will do everything fine, and let you still use your PC for less than the cost of building another PC. While spending $800 for a eMac with a monitor built in is a bit for something that isn't often used, it's not too bad to buy a dual 450 G4 or even a G4 733 tower for less than $500. Even building it yourself wouldn't be too bad since a 1ghz chip is $200, and a board can be bought for another $100. The ram, drive, and PS can be pulled from a PC given the right board, giving you a Mac for $300 plus old PC parts.
In the end, I've got serious doubts to the ability of this emulator to perform. But then again, I am also quite biased towards the idea of having both a PC and a Mac.
Yeah, but because you can't address these extra registers, they're useless for emulation purposes. All this does is let you have more inflight instructions (google for Tomasulo)
And as a side note, the G3/G4/G5 PPCs probably have those as well, since they're not a x86 specific thing. I know that the 604 does, and it's a generation 2 PPC.
You're not exactly right but there's a bit of truth to your analysis. There's 2 main sides, it's either bound by the speed of the computer or the speed of the user's ability to actually input data for the machine to work with.
Here's the problem.
If you're a PC user, you're either using it for specialized apps, say engineering sims. Or you're using it for games. Or you're using because it was cheap.
If you're a mac user, you're either a hardcore video/audio (creative side) user or a my-kid-doesn't-want-to-fix-my-PC-anymore user. In a growing case, you're doing biomed analysis.
For the PC:
Games haven't been CPU bound since the release of the P3. Half-life 1 was bound by the video card already when I was running it on a P2. The AI's, even in Half-Life 2, will not be as CPU intensive as the graphics on the GPU. Games as a whole are simply GPU bound.
Office suites are not hardware bound and haven't been since the 486. Emulating it won't matter.
Engineering suites and specialized apps are typically done in either a portable manner for CPU bound stuff, or user-bound for all the development environments. In most cases, it'll be user bound simply because these are mainly IDEs and not large computation jobs.
For the Mac:
Video and Audio are heavily CPU and Disk bound. In the long run, both of these will be faster on the PPC platform than the Intel platform simply because it's easier to develop and tune these given the instruction sets.
The computer illiterate user doesn't need anything faster than a 500 Mhz G4 because all they'll run is Word, Excel, OpenOffice, and some browser. So this is not hardware bound, it's user bound.
If you're doing genome analysis or other bio med work, you're doing it on a G4/G5 simply because of the vector engine.
------------
Look at this. The typical user's applications are user bound. All the office suites and browsers, it wouldn't matter if we gave them a 3 Ghz P3, a 1 Ghz G4, or heck, even a Pentium 200 from the trash. And because we can give them a Pentium 200 from the trash, emulating them won't be an issue.
If we decide to cross over any of the other apps, we'll have severe problems.
Crossing over a HalfLife 2 will be an issue because most emulators can't pipe OpenGL into real GPU because they emulate some shit video card.
Crossing over Photoshop, Final Cut, or BLAST from the Mac to the PC will all be at least CPU bound. And in these cases, heavily dependent on the cache organization and vector unit because they already stress the Mac. Running it in emulation on the PC will by abysmal even if you optimized the emulator and overclocked the PC.
Crossing over engineering suites from the PC to the Mac will suck for all the simulation software (which you were supposed to buy a Sun station for), but all the home users will tend to be using the development environments and layout tools which are user bound and not a problem.
--------
Overall, having a Mac emulating a PC to play games will be a problem up until emulators get the ability to pipe Direct3D/OpenGL calls into the real GPU which is the same on both sides. And while Half-Life 2 is going to suck until that happens, it stands a chance of being completely playable on current hardware because it's a emulator limitation and because it's not CPU bound.
While on the other hand, running FinalCut, Logic 7, iMovie, Photoshop on the PC using a Mac emulator, we're looking at a limitation that is physically impossible to get over given the current hardware.
--------
In the short run, running HalfLife 2 on PC emu will suck worse than Photoshop on a Mac emu. But in the long run, HalfLife will become fully playable while Photoshop will still suck. (assuming time passes and we stick with the current hardware)
Right now, for most users, getting the Mac to emulate everything but games is a good idea. And you're right, games are a larger market than those other apps because anybody any age can get into it with minimal effort. However not everybody is a hard core gamer, and for anything that isn't a game (aka, actually trying to do work), getting a PC to emulate it just isn't worth the time nor effort.
It doesn't run slowly at all, even on hideously underpowered machines. However, it takes quite a while to compile. Including associated libraries, a full recompile takes a couple of hours on a 3.0 Ghz P4. Last time I checked, it was taking a little over six hours on a 1.0 Ghz G4. This is why I'd prefer OS X running at about 80% of the speed of my Linux system instead of running at 100% the speed of a low-end eMac. It'd also make testing and recompiling quite a bit faster. This all assumes the press release is close when they say 80% (though provided they provide 50% of the speed, that'd be good enough for me).
Nevertheless, you're still assuming that a 3.0Ghz P4 is equal to a 3.0Ghz G4, which it obviously isn't. On average, a G4 is approximately 30% faster over a identical clock speed P4 based on architectural differences.
Even then, you're not expecting to get neither a speed advantage nor a price advantage considering the price of a G4 versus a 3 Ghz P4. Heck, just the 3.0 Ghz P4 chip itself is more expensive than my entire Athlon rig, brand new, and price competitive with a G4 upgrade chip. Attach a decent board will result in roughly the same price. (check ebay for Sawtooth/Gig-E/DigiAudio type G4 boards)
Much easier and more stable to just get the G4 when you want a G4 and Athlon when you want a PC.
There's a basic one up, it does routing, but you'll need to hack around with it. I actually played with the code for a while and got it to route from one place to another, however because it uses Tiger right now, it doesn't consider one way streets.
The main developer is in the process of considering a new backend for it.
The forum is http://thexcar.com/forums/ but nobody's been posting lately.
Nevertheless, most of the gestures from the Newton arn't availiable on InkWell. (At least, when I tried it.)
Some of them, like correcting spacing or punctuation were very nice to have since they felt like natural extensions to the recognizer system. A 'V' between characters would join them in case you wrote them spaced too far apart.
How funny..... At Cal, we have the Forestry major he's looking for. It most likely exists to take care of the Stanford mascot. Yeah.... little do they know the Forestry people are really Cal Black Ops....
(soon it will be the right time of year to kidnap the mascot, mwahahahahaha)
So in the "icing" link, I see a mobo with 4 DIMM slots. One's got a DIMM with heatsinks. The other appears to have an LED segment display and a pair of molex connectors to what looks like a DIMM.
On the other hand, 98 > XP. Definetely. Plus when 98 does fail, it's easier than dealing with NTFS which is what everybody seems to default to these days.
A hole in IE cannot escilate beyond the user's privledges in the process's security context. Your machine be hosed by a hole in IE if you are running it as Admin (just as a hole in Mozilla if it was running as root), but not if it is running as a normal user.
Oh yes it can....... along with anything with a combo box.
So what? I've been using Solaris since 1997, and trying to use Linux since 2001, and I didn't even know that macromedia has packages like that.
I've installed linux on both PPC and x86 architectures, but sure installing is fine. All there is to that is partitioning the drive and getting a kernel on there.
Getting stuff to work beyond the kernel start is the main problem.
Let's see......Here we go: Zoran zr36120 drivers. Where was I supposed to find out (aside from finally grepping the kernal source directory in entirety) that the driver is broken past 2.4?
X11R6 worked fine on my Powerbook 2400c. However since the Linux PPC distro disappeared, no more packages were built and even the isos are hard to find now. So I try Debian. Okay, X doesn't start and no amount of tweaking with XF86Config helps. Yellowdog? They list my Powerbook as supported. Except the installer crashes. But I know X should work since they managed to start it to get me into the graphical installer......which is somehow backed by python and dies. SUSE? Can't find CDs on their site (keep in mind that this is 2 years ago) Vine Linux? Looks almost like LinuxPPC and also supports my specific model laptop, except all of the GUI is in japanese. Oh so close.....
How about PCMCIA? 2 distros I tried installed the kernal modules for a different kernal than the one it set my system to boot with. I started blindly copying files around and trying to force start modules before reinstalling. Same thing again. Eventually, I found that there was another kernal on the install CD and stuff seemed to work to boot that. You know, since nobody explained to me that modules were kernal dependent, I spent 9 weeks, about a day a week working on that. And it was only recently with gentoo, that I found out that modules were built along with the kernal. This is 5 years later. No documentation ever had that written down where a "n00b" would find it until Gentoo.
More PCMCIA. So while trying Debian, every card I insert would kernal panic the machine. Yay. Googling didn't turn much until months later where some obscure memory range change seemed to fix stuff.
Let's try Fedora. To make a mythtv box. On a EPIA board. With a M179 mpeg2 decoder. Let's use the HOWTO...... Where's my ethernet? People said that Fedora supported everything on my specific target platform. All the hardware. the only thing not listed was the power supply but that's not a problem. So where's my ethernet? Apparently an obscure problem with the via-rhine driver requires that I go in there and comment out a check. Great, I'm good at hardware interface stuff and can read code pretty well. So I figure out the line. And I comment it out. And then I type make. Which doesn't make my module. If I have a compiled kernal and want to rebuild via-rhine.c what to do I type? Beats the hell out of me. After roughly a day, I copy over the.config file that came with the distro, rebuild the kernal and modules in entirety and copy out the module.....to find that the module identifies itself to another kernal or rather another kernal name since the Makefile at/usr/src/linux happens to have the wrong name in a variable. (don't ask how I figured that one.)..... anyhow, couldn't get MPEG hardware decoder to run, nor the encoder to be reliable so my friend who I was helping out, returned both and we restarted with a VIA KT266/Sempron2200 and a WinTV PVR-250. 1 month later, we finally started last night, a test run of the mythtv box to see if it crashes. With all the problems we've had, I'm betting it'll die at the half hour recording mark.
How about a PowerMac 6500? listed as supported. 225 mhz. 80 mb ram I think. And interrupt errors when I boot without a SCSI drive mounted. WTF? So I can boot when I have the SCSI cdrom to boot from. I can boot if I stick in a 40 mb harddrive on SCSI and run BootX from that. but if all I have is my 4 gb IDE, I'm fux0red.
Oh.... and ethernet. I don't have a clue as to how I
Of course, being a true mp3 packrat, he'll have used up the space completely so all those Cease & Desist letters will just bounce off like pebbles hitting a shield.
Because of politics and stupidity, you can't discount CDMA's technological merit.
Between GSM/TDMA and CDMA, CDMA is still the superior technology in the same way that BetaMax and PowerPCs are the better technology. And despite the lower sales.
Agilent/HP made the first optical engines for mice which were used by Microsoft, Logitech, Apple, and everybody else.
All that each manufacturer had to do was write their own USB firmware.
If you look at the bottom of a lot of the first and second generation of optical you'll notice that the lens shape is the same. This is because it was speced by Agilent in the spec sheet for the sensor. In fact, I've seen Logitech and Microsoft mice have the sample lens plastic which still has Agilent's logo on it. (Apple's is different because they needed it smaller for the twist tension thingie)
I thought AOL was based off AppleLink or some old Apple service? I also thought that's why the AOL software and Apple's old eWorld software were pretty much the same things just with different content?
Maybe it's my pro-import bias, but doesn't the Corvette smoke the WRX in drag races but totally lose in everything else? Or does the Corvette actually have decent handling? (since I know jack about american cars)
My comparison would have been a Skyline R32 GTR just smoked a Toyota Trueno on a downhill..... but I'd hate to compare anything in my personal car's linage to something from Intel, especially a car as good as a R32:P
Actually..... My dad worked for AMD on the K5 project in FPU design. He got one of those job offers "you can't refuse" from Intel and moved over. They asked him for some of the K5 FPU info and he didn't give it to him. Since there was some sort of recent hire performance eval system, they terminated him on accounts of "bad group cooperation" or something lame like that.
So say what you want, but Intel's certainly got some pretty evil sides to it. There was also an article on theregister.com talking about Intel doing it en mass to Motorola's CPU groups.
On the other hand, my uncle left AMD about 2 years ago being unhappy about some sort of racial discrimination thing, but I don't know the details.
Then again, wtf do I care? My main three computers are Macs.
For starters, we can look at how just being 3Ghz might not do it. Distributed.net's RC5 cracking depends heavily on multiply and accumulate which PowerPCs seem to do very well because there's a instruction that does that. So a 500 Mhz G4 will perform better than a 3Ghz P4 Xeon. Now, this is totally unrelated to doing XORs for parity, except that a 3Ghz processor is getting it's ass kicked by a 400 Mhz one.
Now for the heavy analysis: Let's say we're going to write a 32k block from memory. The P4 will have to pull in 32k from memory. Say we can get 32 bits at a time and put it into a register and do an xor of some sort (I don't know x86 assembly so I'm using mips conventions). We'll need 1 instruction to do the load, and 1 to do the xor, then 1 to do the store to do DMA. Plus, say a 10 instruction overhead to do context switches in every 4 bytes since we're on a multitasking OS and all sorts of stuff's going on. So.....that's 13x8192 = 106496 instructions to do 32k. Which is like nothing still....
But. Say the 3ware designers made their ASICs with at least the knowledge and skill of a new college grad from Berkeley. Use a buffer of flipflops and a xor gate and as the data comes in, do the xor on the fly. So, from the computer side, It's just a straight write. 1x8192. No overhead because you don't need to switch to the RAID code. No extra load/stores cuz you're not having the CPU compute. And there's no delay incurred to calculate it since the XOR happens during transit. The xor finishes as the data gets there. And heck, it doesn't even need to be clocked beyond 2X the PCI bus. (since it's just transit, even just 1X PCI would work.)
And in this simple analysis, you can see why a 33Mhz processor (PCI Clock) can still rape the P4 if done right.
Let me let you in on a secret the PC world doesn't seem to notice.
Apple's Macintosh logic board designs are almost the same across the board. This means that the eMac, iMac, Power Mac G4, and i/Power-Book G4s all have roughly the same design. What's the difference between them?
Simple. The difference is that the other chips (ethernet, sound, cache, etc) are changed. Basically, across all those G4s, there were two main chipset designs. One was the G4s that used PC100/PC133 ram (Uni-North). The other is the G4s that use DDR (I dunno what it's called but the hardware info utils should say it.). These two northbridges were permuted with gigabit ethernet controllers, newer sound cards, and more cache. Oh, and tweaks to increase AGP and memory bus speed.
What does this mean to you? The only thing you need to worry about is new chipset or old, how much cache is included, the CPU speed you want, video card, and how it looks.
An eMac isn't as slow as you think. Inside, it's actually the same as a Power Mac G4 packaged as a lumpy all-in-one. If you find a slow eMac, somebody just installed Norton Antivirus on it. Don't blame the eMac, it'll still be faster than the cheap Dell.
Slashdot Mirror
20 minutes? Holy shit, where do you work? Antarctica on a 300 baud modem? The time it takes now for infection is on the range of seconds.
When CodeRed came out, some of us actually noted it on the job at UC Berkeley ResComp.
The shortest one was on the range of 5 minutes., barely enough time to do an update from windows update.
Years later, for Welchia, etc, it was within 1 minute that we'd see the machine do the reboot by itself. So the infection actually took place before that (since the rest of the minute was the download and install of the virus)
It's a question of bandwidth(cpu power) versus latency. If you have a realtime app, unless very specialized it's not going to take all your CPU power simply because then there's no way to recover if you miss your schedule by a little bit without taking a hit in quality/detail. On the other hand, for computation apps, there's no way it's going to be realtime because they're all CPU bound anyways.
Modern realtime games are running a constant simulation, yes. However that doesn't necessarily mean that it's using all of the CPU time availiable. While the AI is moving around, it doesn't take much time for it to calculate, nor is is necessary. Think about it, does the AI do more thinking my CPU power goes up? If I have some decked out 8-way Xeon for a gaming system, am I going to have any significant experience than a guy playing on a P3 1Ghz? No.
Half-Life 2 was written around the minimum system requirements of a P3 700 or a P3 1.2 depending on where Google takes you but runs best on a P4 2Ghz. Two things are happening, either the P3 will get less detail than the P4 in some unknown aspect and will fly by unnoticed by the user, or it's simply because they expect the P4 will be able to load stuff faster and hence the better experience. Once you get beyond a P4 2Ghz, you're still going to be requiring the simulation run in realtime, but it's not going to take 100% of cputime. Running SETI@home will actually get work done without your system slowing down. And hence my argument stands.
I'm running Warcraft 3, United Devices(like SETI@home), and iTunes at the same time on an old Athlon 2000 which wasn't close to top of the line on release, and yet, it hasn't made a difference.
Now, the majority of average user applications don't have realtime aspects nor are CPU bound, and that's why I classified them to be unimportant in this argument.
But the ones I did highlight, all of them, once the job starts, full utilization all the way no matter what CPU. And given that some of the tasks take days, we're not close to any sort of fix to that.
With the growing amount of ram and processing power on a GPU, along with programmable shaders, we're going to reach a dead end in the realism bound by the output capibility of our monitors. At that point in time, nVidia, ATI, and friends will need to find another use for all that power so they can make money. Since the GPU is a massively parallel processor with fast access to the enviroment being rendered, it makes sense to put physics calculations there because physics arn't CPU intensive because they're complex, they're CPU intensive because there are just a ton of them. Hence perfectly matching the GPU architecture.
In the end, it comes down to this. With the emulated GPU piping stuff to the real GPU, if my dual G5 emulates a P3 700, that's enough to play most games. But if my quad Opteron still only emulates a 1Ghz G3, it's painfully useless.
Maybe it seems there's something I'm failing to explain in detail.
While I can't say that I understand what I'm supposed to be looking for in cs_office, I can still justify how my argument stands. And while I can agree that people are starting to find uses for the CPU in games, it's not going to be significant.
Note that for Half Life 2's system requirements, it will not have the CPU running full blast if I was running it on a Athlon 2400. Nor on a P4 2.4. Nor on a P4 3.0. But it probably will be on a P3 700.
When I meant by CPU bound, I meant really, it just can't get faster because of your CPU. The fact that Half Life 2 depends on human input for data to calculate, that physics models can sooner or later be pushed to GPU shaders, and that there's a limit in which the CPU can contribute to performance still puts games at lower CPU usage than apps like Photoshop, FCP, BLAST, Logic, simply because while HalfLife 2 will run very nicely on a P4 2.4, it will barely be any better on a P4 3.0 because adding CPU power doesn't help much. You'll barely get any more frames because you're already pushing 3D data to the GPU faster than it can handle.
On the other hand, the other apps I mentioned, if you have CPU power leftover, it will be scheduled to push your job faster. That is what I meant by CPU bound. Simply that the speed of the program is bound by the speed of your CPU.
My PC was built using $250 US about a year and a half ago (Athlon XP 2000+ 512 MB, Geforce 3). And if I price a Sempron 2200+ mobo/CPU combo at Fry's for $50, a $20 case, and 512 megs of memory, it's pretty competitive to your CPU.
:)
But anyways. Given the the limitations of emulators, if (one heck of a big if) they do happen to reach 80% efficiency, with the 30% P4 penalty you'll end up with what amounts to roughly a 2 Ghz G4 for your one specific task. Not bad.
However if you're looking at what you'd expect for a real emulator to get, 30% or less, you're pretty much down to what you'll get with a 1 ghz G4 anyways. Or even less than that. And still this is only for one task. If you do anything else with it, the performance will rarely go up, usually down. Not to mention no GUI acceleration.
Sure, all that for $50 (and pirating OSX) is fairly nice. But it's not too far from just building or buying a used mac. With actual Mac hardware, pirating the software will feel like less of a crime.
It's possible to price out a used G4 on ebay that will do everything fine, and let you still use your PC for less than the cost of building another PC. While spending $800 for a eMac with a monitor built in is a bit for something that isn't often used, it's not too bad to buy a dual 450 G4 or even a G4 733 tower for less than $500. Even building it yourself wouldn't be too bad since a 1ghz chip is $200, and a board can be bought for another $100. The ram, drive, and PS can be pulled from a PC given the right board, giving you a Mac for $300 plus old PC parts.
In the end, I've got serious doubts to the ability of this emulator to perform. But then again, I am also quite biased towards the idea of having both a PC and a Mac.
Yeah, but because you can't address these extra registers, they're useless for emulation purposes. All this does is let you have more inflight instructions (google for Tomasulo)
And as a side note, the G3/G4/G5 PPCs probably have those as well, since they're not a x86 specific thing. I know that the 604 does, and it's a generation 2 PPC.
You're not exactly right but there's a bit of truth to your analysis. There's 2 main sides, it's either bound by the speed of the computer or the speed of the user's ability to actually input data for the machine to work with.
Here's the problem.
If you're a PC user, you're either using it for specialized apps, say engineering sims. Or you're using it for games. Or you're using because it was cheap.
If you're a mac user, you're either a hardcore video/audio (creative side) user or a my-kid-doesn't-want-to-fix-my-PC-anymore user. In a growing case, you're doing biomed analysis.
For the PC:
Games haven't been CPU bound since the release of the P3. Half-life 1 was bound by the video card already when I was running it on a P2. The AI's, even in Half-Life 2, will not be as CPU intensive as the graphics on the GPU. Games as a whole are simply GPU bound.
Office suites are not hardware bound and haven't been since the 486. Emulating it won't matter.
Engineering suites and specialized apps are typically done in either a portable manner for CPU bound stuff, or user-bound for all the development environments. In most cases, it'll be user bound simply because these are mainly IDEs and not large computation jobs.
For the Mac:
Video and Audio are heavily CPU and Disk bound. In the long run, both of these will be faster on the PPC platform than the Intel platform simply because it's easier to develop and tune these given the instruction sets.
The computer illiterate user doesn't need anything faster than a 500 Mhz G4 because all they'll run is Word, Excel, OpenOffice, and some browser. So this is not hardware bound, it's user bound.
If you're doing genome analysis or other bio med work, you're doing it on a G4/G5 simply because of the vector engine.
------------
Look at this. The typical user's applications are user bound. All the office suites and browsers, it wouldn't matter if we gave them a 3 Ghz P3, a 1 Ghz G4, or heck, even a Pentium 200 from the trash. And because we can give them a Pentium 200 from the trash, emulating them won't be an issue.
If we decide to cross over any of the other apps, we'll have severe problems.
Crossing over a HalfLife 2 will be an issue because most emulators can't pipe OpenGL into real GPU because they emulate some shit video card.
Crossing over Photoshop, Final Cut, or BLAST from the Mac to the PC will all be at least CPU bound. And in these cases, heavily dependent on the cache organization and vector unit because they already stress the Mac. Running it in emulation on the PC will by abysmal even if you optimized the emulator and overclocked the PC.
Crossing over engineering suites from the PC to the Mac will suck for all the simulation software (which you were supposed to buy a Sun station for), but all the home users will tend to be using the development environments and layout tools which are user bound and not a problem.
--------
Overall, having a Mac emulating a PC to play games will be a problem up until emulators get the ability to pipe Direct3D/OpenGL calls into the real GPU which is the same on both sides. And while Half-Life 2 is going to suck until that happens, it stands a chance of being completely playable on current hardware because it's a emulator limitation and because it's not CPU bound.
While on the other hand, running FinalCut, Logic 7, iMovie, Photoshop on the PC using a Mac emulator, we're looking at a limitation that is physically impossible to get over given the current hardware.
--------
In the short run, running HalfLife 2 on PC emu will suck worse than Photoshop on a Mac emu. But in the long run, HalfLife will become fully playable while Photoshop will still suck. (assuming time passes and we stick with the current hardware)
Right now, for most users, getting the Mac to emulate everything but games is a good idea. And you're right, games are a larger market than those other apps because anybody any age can get into it with minimal effort. However not everybody is a hard core gamer, and for anything that isn't a game (aka, actually trying to do work), getting a PC to emulate it just isn't worth the time nor effort.
It doesn't run slowly at all, even on hideously underpowered machines. However, it takes quite a while to compile. Including associated libraries, a full recompile takes a couple of hours on a 3.0 Ghz P4. Last time I checked, it was taking a little over six hours on a 1.0 Ghz G4. This is why I'd prefer OS X running at about 80% of the speed of my Linux system instead of running at 100% the speed of a low-end eMac. It'd also make testing and recompiling quite a bit faster. This all assumes the press release is close when they say 80% (though provided they provide 50% of the speed, that'd be good enough for me).
Nevertheless, you're still assuming that a 3.0Ghz P4 is equal to a 3.0Ghz G4, which it obviously isn't. On average, a G4 is approximately 30% faster over a identical clock speed P4 based on architectural differences.
Even then, you're not expecting to get neither a speed advantage nor a price advantage considering the price of a G4 versus a 3 Ghz P4. Heck, just the 3.0 Ghz P4 chip itself is more expensive than my entire Athlon rig, brand new, and price competitive with a G4 upgrade chip. Attach a decent board will result in roughly the same price. (check ebay for Sawtooth/Gig-E/DigiAudio type G4 boards)
Much easier and more stable to just get the G4 when you want a G4 and Athlon when you want a PC.
tmrs.sourceforge.net
There's a basic one up, it does routing, but you'll need to hack around with it. I actually played with the code for a while and got it to route from one place to another, however because it uses Tiger right now, it doesn't consider one way streets.
The main developer is in the process of considering a new backend for it.
The forum is http://thexcar.com/forums/ but nobody's been posting lately.
Nevertheless, most of the gestures from the Newton arn't availiable on InkWell. (At least, when I tried it.)
Some of them, like correcting spacing or punctuation were very nice to have since they felt like natural extensions to the recognizer system. A 'V' between characters would join them in case you wrote them spaced too far apart.
How funny..... At Cal, we have the Forestry major he's looking for. It most likely exists to take care of the Stanford mascot. Yeah.... little do they know the Forestry people are really Cal Black Ops....
(soon it will be the right time of year to kidnap the mascot, mwahahahahaha)
http://espm.berkeley.edu/ugmajors/
So in the "icing" link, I see a mobo with 4 DIMM slots. One's got a DIMM with heatsinks. The other appears to have an LED segment display and a pair of molex connectors to what looks like a DIMM.
What is that?
Damnit. It's 3am....and I'm posting to this story.
And the time stamp really says it too.......
On the other hand, 98 > XP. Definetely. Plus when 98 does fail, it's easier than dealing with NTFS which is what everybody seems to default to these days.
In one of the pictures, you can see the plastic frame of an Apple 15 inch LCD panel (white ADC connector type) with the front plate removed.
So..... most likely prototypes or something like that what you said.
A hole in IE cannot escilate beyond the user's privledges in the process's security context. Your machine be hosed by a hole in IE if you are running it as Admin (just as a hole in Mozilla if it was running as root), but not if it is running as a normal user.
t in /MS03-045.mspx
Oh yes it can....... along with anything with a combo box.
http://www.microsoft.com/technet/security/Bulle
So what? I've been using Solaris since 1997, and trying to use Linux since 2001, and I didn't even know that macromedia has packages like that.
..... Where's my ethernet? People said that Fedora supported everything on my specific target platform. All the hardware. the only thing not listed was the power supply but that's not a problem. So where's my ethernet? Apparently an obscure problem with the via-rhine driver requires that I go in there and comment out a check. Great, I'm good at hardware interface stuff and can read code pretty well. So I figure out the line. And I comment it out. And then I type make. Which doesn't make my module. If I have a compiled kernal and want to rebuild via-rhine.c what to do I type? Beats the hell out of me. After roughly a day, I copy over the .config file that came with the distro, rebuild the kernal and modules in entirety and copy out the module.....to find that the module identifies itself to another kernal or rather another kernal name since the Makefile at /usr/src/linux happens to have the wrong name in a variable. (don't ask how I figured that one.)..... anyhow, couldn't get MPEG hardware decoder to run, nor the encoder to be reliable so my friend who I was helping out, returned both and we restarted with a VIA KT266/Sempron2200 and a WinTV PVR-250. 1 month later, we finally started last night, a test run of the mythtv box to see if it crashes. With all the problems we've had, I'm betting it'll die at the half hour recording mark.
I've installed linux on both PPC and x86 architectures, but sure installing is fine. All there is to that is partitioning the drive and getting a kernel on there.
Getting stuff to work beyond the kernel start is the main problem.
Let's see......Here we go:
Zoran zr36120 drivers. Where was I supposed to find out (aside from finally grepping the kernal source directory in entirety) that the driver is broken past 2.4?
X11R6 worked fine on my Powerbook 2400c. However since the Linux PPC distro disappeared, no more packages were built and even the isos are hard to find now. So I try Debian. Okay, X doesn't start and no amount of tweaking with XF86Config helps. Yellowdog? They list my Powerbook as supported. Except the installer crashes. But I know X should work since they managed to start it to get me into the graphical installer......which is somehow backed by python and dies. SUSE? Can't find CDs on their site (keep in mind that this is 2 years ago) Vine Linux? Looks almost like LinuxPPC and also supports my specific model laptop, except all of the GUI is in japanese. Oh so close.....
How about PCMCIA? 2 distros I tried installed the kernal modules for a different kernal than the one it set my system to boot with. I started blindly copying files around and trying to force start modules before reinstalling. Same thing again. Eventually, I found that there was another kernal on the install CD and stuff seemed to work to boot that. You know, since nobody explained to me that modules were kernal dependent, I spent 9 weeks, about a day a week working on that. And it was only recently with gentoo, that I found out that modules were built along with the kernal. This is 5 years later. No documentation ever had that written down where a "n00b" would find it until Gentoo.
More PCMCIA. So while trying Debian, every card I insert would kernal panic the machine. Yay. Googling didn't turn much until months later where some obscure memory range change seemed to fix stuff.
Let's try Fedora. To make a mythtv box. On a EPIA board. With a M179 mpeg2 decoder. Let's use the HOWTO.
How about a PowerMac 6500? listed as supported. 225 mhz. 80 mb ram I think. And interrupt errors when I boot without a SCSI drive mounted. WTF? So I can boot when I have the SCSI cdrom to boot from. I can boot if I stick in a 40 mb harddrive on SCSI and run BootX from that. but if all I have is my 4 gb IDE, I'm fux0red.
Oh.... and ethernet. I don't have a clue as to how I
Of course, being a true mp3 packrat, he'll have used up the space completely so all those Cease & Desist letters will just bounce off like pebbles hitting a shield.
CDMA is a superior technology over GSM and TDMA.
Because of politics and stupidity, you can't discount CDMA's technological merit.
Between GSM/TDMA and CDMA, CDMA is still the superior technology in the same way that BetaMax and PowerPCs are the better technology. And despite the lower sales.
Parent is so wrong.
Agilent/HP made the first optical engines for mice which were used by Microsoft, Logitech, Apple, and everybody else.
All that each manufacturer had to do was write their own USB firmware.
If you look at the bottom of a lot of the first and second generation of optical you'll notice that the lens shape is the same. This is because it was speced by Agilent in the spec sheet for the sensor. In fact, I've seen Logitech and Microsoft mice have the sample lens plastic which still has Agilent's logo on it. (Apple's is different because they needed it smaller for the twist tension thingie)
What's Playnet?
I thought AOL was based off AppleLink or some old Apple service? I also thought that's why the AOL software and Apple's old eWorld software were pretty much the same things just with different content?
Maybe it's my pro-import bias, but doesn't the Corvette smoke the WRX in drag races but totally lose in everything else? Or does the Corvette actually have decent handling? (since I know jack about american cars)
:P
My comparison would have been a Skyline R32 GTR just smoked a Toyota Trueno on a downhill..... but I'd hate to compare anything in my personal car's linage to something from Intel, especially a car as good as a R32
Actually..... My dad worked for AMD on the K5 project in FPU design. He got one of those job offers "you can't refuse" from Intel and moved over. They asked him for some of the K5 FPU info and he didn't give it to him. Since there was some sort of recent hire performance eval system, they terminated him on accounts of "bad group cooperation" or something lame like that.
So say what you want, but Intel's certainly got some pretty evil sides to it. There was also an article on theregister.com talking about Intel doing it en mass to Motorola's CPU groups.
On the other hand, my uncle left AMD about 2 years ago being unhappy about some sort of racial discrimination thing, but I don't know the details.
Then again, wtf do I care? My main three computers are Macs.
Let's see.....
For starters, we can look at how just being 3Ghz might not do it. Distributed.net's RC5 cracking depends heavily on multiply and accumulate which PowerPCs seem to do very well because there's a instruction that does that. So a 500 Mhz G4 will perform better than a 3Ghz P4 Xeon. Now, this is totally unrelated to doing XORs for parity, except that a 3Ghz processor is getting it's ass kicked by a 400 Mhz one.
Now for the heavy analysis:
Let's say we're going to write a 32k block from memory.
The P4 will have to pull in 32k from memory. Say we can get 32 bits at a time and put it into a register and do an xor of some sort (I don't know x86 assembly so I'm using mips conventions). We'll need 1 instruction to do the load, and 1 to do the xor, then 1 to do the store to do DMA. Plus, say a 10 instruction overhead to do context switches in every 4 bytes since we're on a multitasking OS and all sorts of stuff's going on. So.....that's 13x8192 = 106496 instructions to do 32k. Which is like nothing still....
But. Say the 3ware designers made their ASICs with at least the knowledge and skill of a new college grad from Berkeley. Use a buffer of flipflops and a xor gate and as the data comes in, do the xor on the fly. So, from the computer side, It's just a straight write. 1x8192. No overhead because you don't need to switch to the RAID code. No extra load/stores cuz you're not having the CPU compute. And there's no delay incurred to calculate it since the XOR happens during transit. The xor finishes as the data gets there. And heck, it doesn't even need to be clocked beyond 2X the PCI bus. (since it's just transit, even just 1X PCI would work.)
And in this simple analysis, you can see why a 33Mhz processor (PCI Clock) can still rape the P4 if done right.
.....except Carbon apps.
The core of a OSX bundle for a Carbon app is still just like a classic app, resource forks and all.
Example: InstallerVise installers
Let me let you in on a secret the PC world doesn't seem to notice.
Apple's Macintosh logic board designs are almost the same across the board. This means that the eMac, iMac, Power Mac G4, and i/Power-Book G4s all have roughly the same design. What's the difference between them?
Simple. The difference is that the other chips (ethernet, sound, cache, etc) are changed. Basically, across all those G4s, there were two main chipset designs. One was the G4s that used PC100/PC133 ram (Uni-North). The other is the G4s that use DDR (I dunno what it's called but the hardware info utils should say it.). These two northbridges were permuted with gigabit ethernet controllers, newer sound cards, and more cache. Oh, and tweaks to increase AGP and memory bus speed.
What does this mean to you? The only thing you need to worry about is new chipset or old, how much cache is included, the CPU speed you want, video card, and how it looks.
An eMac isn't as slow as you think. Inside, it's actually the same as a Power Mac G4 packaged as a lumpy all-in-one. If you find a slow eMac, somebody just installed Norton Antivirus on it. Don't blame the eMac, it'll still be faster than the cheap Dell.