What is Digg?85 Comments
- JohnBooty, on 10/12/2007, -3/+39Uh, newsflash. Leopard will run on PPC, genius.
- selectodude, on 10/12/2007, -8/+25The POWER5 isn't a PowerPC processor. Buried.
- BlackKnight6, on 10/12/2007, -0/+15Hertz is nothing, it is all about the architecture, look at Core 2 by Intel. Clock speeds went way down, so did heat and energy consumption and performance jumped big time.
I say who cares. - evilic0n, on 10/12/2007, -0/+13"IBM's Power6 processor will be able to exceed 5 gigahertz in a high-performance mode"
Is this like the old-school Turbo button? - unloud, on 10/12/2007, -2/+15"I wonder if Steve Jobs is going to have Switcher's Remorse after seeing this."
I don't. Right now, with dual core processors from Intel, the GHz is comparable with that of a 5 GHz PowerPC. Not to mention, it'll produce nearly half as much heat and IBM doesn't make a wide array of processors like Intel does. Intel already has quad core chips on the market; this won't be out until mid year.
Hell, there are huge specialized chips right now that run at hundreds of GHz. They arn't meant for personal computers any more than this one is. - Judge373, on 10/12/2007, -0/+10The main reason that Apple switched to Intel is laptop sales. The newer PowerPC chips run too hot to put into a laptop, and laptop sales are very significant for Apple. Therefore, Apple had to look for other options for effective, high-powered laptop computing.
- geminitojanus, on 10/12/2007, -0/+10You act as if adding OOO to a processor is as simple as bolting on a "unit" and calling it a night. In-order and Out-of-Order ops are two different processor construction paradigms, and are often mixed (processors today are "in-order" until you get to execution dispatching, when the processor starts selecting the instructions it can execute first and re-ordering them to best fit the available execution hardware).
Theoretically, if both machines have the same instruction set, then there is no difference besides performance whether or not the hardware is in order or out of order, it's just how the hardware operates. At 5+GHz, In-Order "thin" hardware makes sense; the heat production at 5GHz is astronomical, and the best way to save on that is simply to turn off units not being used. Instead of trying to build the most efficient instruction processor (trying to saturate all of the units on the processor), you try to build the most efficient power user (turning off hardware when its idle, and being able to turn off and on with no distinct impact on performance necessitates high clock speeds). It's the "Keep It Simple Stupid" approach that Intel used with the Core architecture; stall as little as possible, don't use what you don't absolutely need to, and turn it off when you're not using it.
The reality is these chips could be much faster than what's currently in our Macs and PCs. The problem is that the engineering required to pull this kind of chip off is much more expensive than the kind of engineering Intel needs to pull off the Core revolution. And on top of that, IBMs play is much more experimental (as in they simply cannot know the results of this shift in design until they've designed it and benchmarked it against the old), whereas Intel's is guarenteed to produce results. All-in-all Apple's move to Intel is a net positive.
And this is _without_ talking about the mobile aspects of the processors. - JoeB4ever, on 10/12/2007, -1/+10I'm sorry but anyone who rates a processor just by the ghz, should read a computers for dummies book.
- trekkie, on 10/12/2007, -0/+9Yes because a chip designed for servers and available sometime in 2007 would have carried the G5 at 2.7GHz since 2005.
- strictnein, on 10/12/2007, -0/+9Why's my computer running so slowly? Oh... I bumped the Turbo button.
Never understood those things. - emmanuelsotelo, on 10/12/2007, -0/+7You can't put a G5 chip on a thin laptop such as an iBook or a PowerBook.
Also, moving to intel has done wonders for Apples market share.
Because of Boot Camp and coherence mode in Parallels I think I will end up getting an Apple next time I purchase a laptop. - geminitojanus, on 10/12/2007, -0/+7The purpose of the "turbo" button was to slow down your PC so you could run 386 software without timing conflicts on 486 hardware. "Turbo" mode is what they called it when your PC was running full speed. If your software was well written and didn't crash on "Turbo" mode, or your games didn't run at 10x the speed they seemingly should have, then you were alright and it was safe to unhook the turbo button and just short the link with a jumper.
- OrangeTide, on 10/12/2007, -0/+6too bad for you that IBM continues to develop cutting edge fab technologies.
funny how Xbox360, Playstation3 and Wii are all running PowerPC cores with major contributions (or outright manufacturing) by IBM. - unloud, on 10/12/2007, -2/+8First of all, it's about the Power6, not the Power five. Second of all, don't bury this unless you just simply arn't interested in the topic; the title and description do not elude to apple and neither does the article so this digg user is inaccurate (not the story). Yes, the submitter did accidentally write PowerPC but other than that it is a good digg.
- fearofcorners, on 10/12/2007, -0/+6Who still believes clock speed is everything? Historically AMD has trounced Intel for performance many times with much lower clock speeds. How about we compare a 2.4gHz athlon 64 to a 3ghz celeron?
- selectodude, on 10/12/2007, -1/+7You're completely retarded. These are not general use chips. Neither the Cell BE nor the POWER5 have an out-of-order execution unit. They'd be dog slow in a Mac.
- inactive, on 10/12/2007, -0/+6What's done is done, and the recent Intel Macs are what switched me.
In other words, it's not just performance that influenced Jobs' decision, in my opinion at least. - BlackKnight6, on 10/12/2007, -1/+5@kirakun
I don't think you understand. If you took a 2.4Ghz P4 single core and benched it against a Core Solo @ 2.4Ghz the Core Solo would beat the P4's pants off. The only thing that changed was the design of the CPU. This is one reason AMD had to name its CPUs with numbers like 3200+ even though its a 2.2ghz. It was because people were looking at the Intel P4s with 3.6ghz and such so they had to make it so consumers could get an idea of performance. AMD was getting more performance per hertz while intel was just bumping the hertz higher and higher until Core 2 came for the desktop. Because if you judged by the hertz, the P4's would have always looked WAY faster even though the performance was comparable to AMD's Athlons.
Now I have an Intel Core 2 duo E6600 (2.4GHz) and it ran circles around my 3.92GHz OCed dual core P4 I had water cooled. - BlackKnight6, on 10/12/2007, -1/+5@kirakun
Again, hertz on the Power6 is not comparable to the hertz on a Core 2. They are very different architectures. It is obvious that a dual core @ 2ghz would be slower than a single core @ 4ghz at a single threaded bench. My point was architecture is a much better determinant of performance, not the clock speeds. - OrangeTide, on 10/12/2007, -0/+4You can run a processor really fast and not have it be very useful for general purpose applications. Take the Xenon tri-core 3.2Ghz powerpc in the Xbox 360. 3.2Ghz is really fast, but the processor does not support out-of-order-execution. This is technology that has been around since the Pentium and oldPowerPC G3. You can certainly argue the advantages and disadvantages, but this "faster" processor is less intelligent and in many work loads will be slower.
In a game system, where developers are going to profile code and always run on a well known predictable platform (all xbox 360s are the same, afaik) it's not really an issue. Except perhaps for compiler authors.
Many embedded processors run extremely fast, but have massive pipeline delays for the wrong combinations of instructions. A desktop processor is supposed to try to do a good job with all combinations of instructions. - flamingmb, on 10/12/2007, -0/+3Oh god I remember those, wtf was the purpose? I bet they did it just to get people to buy it, who wouldnt want a turbo button on everything? even if it was useless.
- MacParrot, on 10/12/2007, -0/+3GFX has this thing about any chip discussion with Apple in it. He continuously posts the same nonsense and then disappears when anyone asks for sources. So no proof, just a constant tirade. Good luck with that bucky!
- superpatty, on 10/12/2007, -1/+4a major problem with the 'higher hertz is better' mentality is two big factors:
1. heat dissipation- the higher the hertz the more electricity being used which means the more heat being let off as a byproduct. This is one of the big reasons why the P4 architecture hit a wall in hertz gains, so much heat was being released that the engineers were having major problems keep the chip cool enough
2. pipelining- with a higher frequency (hertz), to ensure that the processor is always being utilized and not waiting on information to reach the processor, you need a longer pipeline (a pipeline is the 'tube' that carries the information to the processor, it has multiple stages) but with a longer pipeline, sometimes the processor needs a piece of information that is not already in the pipeline and the processor has to wait for the system to retrieve it, which wastes time. So you either need to create better branch prediction (guessing what info the processor needs) or shorten the pipeline, this works better with multiple cores with shorter pipelines.
i.e.- the p4 architecture has 31 stages in the pipeline
the core 2 duo architecture has 14 stages in the pipeline
the amd k8 architecture has 12 stages in the pipeline
hence, currently intel seems to have hit a sweet spot for pipeline length at 14 stages - Solstice, on 10/12/2007, -0/+3@flamingmb
The turbo button was originally needed for 100% IBM PC AT compatibility. IBM PC clones were introduced using an 80286 processor clocked at 8 MHz 80286, which was faster than the 6 MHz clock in the original AT. Some software - many games in particular was senstive to being upclocked (their timing loops ran at processor speed), so they offered a quick way to switch between 8 MHz and 6 MHz. Eventually, they started using even faster processors and the turbo button did other things (some AT boards had jumpers where you could configure what speeds you wanted - say 16 and 25 MHz).
Eventually, IBM offered the 8 MHz processor in the AT, but they didn't offer a turbo button. - Broccoli, on 10/12/2007, -0/+2Yah, but its not like intel has frozen in time and will never make a faster chip. Besides my bet is that you will only find that in servers. The g5 runs really hot at 2ghz, i cant imagin one at 5.
- KMartSheriff, on 10/12/2007, -0/+2Wow so there's even PPC zealots too? You guys are worse than hardcore Mac zealots. Switching to Intel was good, as many have made the jump because of it.
- JoeB4ever, on 10/12/2007, -0/+2wow good comeback
actually I did read the whole 9 paragraph article, which just talked about "increasing chip clock speed" and competition from intel and amd. - inkhead, on 10/12/2007, -0/+2Lol if everything IBM said actually happened we'd all have 10ghz computers already running at 50 degrees F, if I had a $$$ for every lie they told about megahertz.... I'd have about 40-50million dollars
- aegis9975, on 10/12/2007, -0/+2Seems their is a shift to multi-core processors rather then clock speed, but really, advances seem to be made on both sides. Intel showed off their 80-core concept, and IBM expects to have dual-PPU+32 SPE core Cell processors by 2010.
http://img164.imageshack.us/img164/5576/cellroadmapxk3.jpg - pong32, on 10/12/2007, -0/+2prominator, here is a good article on the "megahertz myth": http://en.wikipedia.org/wiki/Megahertz_Myth
There is also a link at the bottom for a video of an Apple keynote presentation on the matter. - flamingmb, on 10/12/2007, -0/+2he never claimed 4ghz, he said "We will have 3ghz by next year" and they only got to 2.5 I think.
- carpespasm, on 10/12/2007, -0/+2i think that they meant numeric demand, not performance demand. though intel seems to have given them both.
- r3tex, on 10/12/2007, -1/+3Why does every Counterstrike nerd have to try and be a processor architect?
Pipelining is a system that lets the processor accept a new instruction at the next clock cycle even if the previous command is not done executing. It allows multiple commands to be in different parts of the CPUs logic even before they have finished and given a valid output. They do this to increase the instruction bandwidth which is useful in high clock speeds.
I'll be the first to say that programming for the ppc, mips, and sparc is about a billion times more fun that programming on any x86 processor (amd64 included). - BlackKnight6, on 10/12/2007, -0/+2Exactly, especially with Intel Core 2's people should see how hertz is not the determining factor, only among the same architecture.
- doushanes, on 10/12/2007, -0/+2I thought I read that one of the reasons Apple went with Intel is that IBM couldn't keep up with demand?
- CaptainMordecai, on 10/12/2007, -0/+2Not just performance, and clock speed differences - there was the inability to deliver a G5 suitable for notebooks, by the end of the Powerbooks like x86 notebooks smashed them. The G5 can still hold its own, but they never got it cool/low power enough to put in a notebook.
- OrangeTide, on 10/12/2007, -3/+5Apple will never admit it, but I think a lot of their marketing plan for switching to Intel has to do with Windows compatibility.
- OrangeTide, on 10/12/2007, -2/+4selectodude - Linux on POWER actually uses the PowerPC subset of POWER. Most of the fancy POWER stuff is a bit too fancy for most Unixes to find practicle. Even IBM AIX is (mostly) PowerPC. These days PowerPC is used interchangable with IBM Power, probably because anytime IBM releases a fancy new POWER series they'd spin off some PowerPCs for Apple and embedded customers based on the same core.
ps - I'm not sure why you think it needs to be buried. your criteria seems arbitrary [and stupid] - ricree, on 10/12/2007, -0/+2@unloud
This is a pretty big limitation of digg, really. Sometimes articles get stuck with bad titles/descriptions, and sometimes we get blogspam that links to good articles. There should really be new bury options for those two occasions. Perhaps after enough buries, people could be able to go back and vote on new titles (or links, in that situation). This would allow us to do something in the case of flawed submissions that are still worth looking at. - fallenone05, on 10/12/2007, -0/+1PHONE!
- yabos, on 10/12/2007, -0/+1I bet Microsoft gave them a huge upfront order for the XBox chips. Apple would never do that, they just want IBM to keep improving their chips. Besides, the PPCs used in XBox and PS3 could never be used for a desktop computer because they're too simple and don't support a lot of features that desktop CPUs require. I liked the PPC before and if IBM had kept up their end and developed lower power versions of the chips then they'd still be a good choice today for Apple. However they had no intentions of doing that so there's not much else to do but use x86.
- yabos, on 10/12/2007, -0/+1Do you think Apple didn't know about IBM's roadmaps before the made the switch? Besides, Apple has never used the POWER line of chips in any of their computers, they always had to wait for IBM to make a lower power PPC version with less cache that ran cooler and used less power. IBM STILL hasn't produced a low power G5 chip and I don't expect them ever to do it unless they have a large monetary incentive.
- BlackKnight6, on 10/12/2007, -0/+1@OrangeTide
Yep, and to try to counter the fact that there is no out-of-order-execution with the 360 CPU, same with the Cell in the PS3, they try to counter with 3 cores each with SMT (6 threads) on the 360 and the Cell with 7 working SPEs on the PS3.
Being in-order CPUs makes both game consoles CPUs much cheaper to mass produce, hence why desktp CPUs, like my Core 2 Duo E6600 is around 300 bucks by itself. - drilldown, on 10/12/2007, -0/+1"Power6 has 700 million transistors and measures 341 square millimeters, according to the program."
So, what else did the program say that should be attributed to it? - gimpbully, on 10/12/2007, -0/+1@unloud
It doesn't matter what version Power chip it is, it is architecturally different than the PPC. You cannot run OSX or Linux on a vanilla Power machine (yes yes, I realize there are builds for ppc64, but those don't use the chip's full instruction set and require IBM approved chips (only AIX and possibly z/OS can run on a unhindered Power chip). That coupled with the cost of the Power line not being anywhere close to consumer tastes really make it a poor choice for Apple. I don't think Steve really cares too much. The compatibility gains Apple gets over everyone else with the switch to x86 (OSX + everything else the closest competitor can run) really outweigh any trickle-down Apple would have stood to gain from the Power6 line.
Also, the Power line was first to market with dual cores (I don't recall exactly what version it was, but iirc, it was either the Power5 or Power5+). The 6 will certainly have at least 2. - BlackKnight6, on 10/12/2007, -0/+1The Cell is an in-order execution CPU while our desktop CPUs are out-of-order execution. Same goes for the 360 CPU, in-order execution. They are alot cheaper and to try and counter the fact that gaming is branch intense which is not good for in-order exec, the Cell and Xenon are multi-core/ have SPE's. Intels newest quad cores are already faster than the 360 CPU and PS3 CPU. The Cell processor will be good for specialized hardware and products, but I believe it will not make a splash in the PC/Mac domain, just the type of CPU it is doesn't go with the wide range of things needed to be done on a PC/Mac.
- dyzlexiK, on 10/12/2007, -0/+1@ kirakun
You're wrong. Hertz is a way of increasing speed, yes, but isnt the only way, and isnt the most effective way either.
A 2.2ghz AMD processor can outperform a 3.8ghz Intel processor (We're talking single core here), and the intel is only 600mhz short of being twice the megahurtz. Why? Because the artictecture in the AMD is more efficient. (I was relating a 3500+ to a P4 3.8 Netburst for those wondering)
Therefore, you could have a 4ghz single core that runs slower than a 2.2 dual core. Not because we are adding 2.2 + 2.2 = 4.4, but because just one of the cores on the dual core can outperform the 4ghz single core processor. - inactive, on 10/12/2007, -0/+1I figured you did, I just thought the comeback was funny =P
- OrangeTide, on 10/12/2007, -0/+1I'm tempted to get a PS3 just to use as a Cell devkit. Just waiting for Linux on PS3 to support the SPUs. (I think the hypervisor locks the "guest os" out of SPU and GPU)
- yabos, on 10/12/2007, -0/+1HAHAHAHA, a Cell based general purpose computer, what are you, retarded?
-
Show 51 - 83 of 83 discussions
Digg is coming to a city (and computer) near you! Check out all the details on our 
© Digg Inc. 2009
— Content created and posted by Digg users is 