Slashdot Mirror

• Bridge & board: "The chipset is an Apple design"
• Servers
• Peripherals and Standards

User-serviceable parts (RAM, HD, AGP, etc) are commodity, but the hard stuff is designed in Cupertino.

I was interested to find out find out they used the older Altivec instruction unit rather than the one from the G4e. Is there anyone that can comment of differences between the two Altivec units?

gere are some articles that have some quotes about the G4:

ths is a refrence about plans for the G4
there ya go

this is a nice articel with a nice table for ya.
here it is

here is one with a discusion of just the x86 line.
and that should do it for ya

let me know if you need anything else on the subject.

oh...in case you would liek to know, I went here to find it all.

While I agree that it's a bit premature to call the new plant a total failure, orders have been smaller than IBM expected.

Check this out for suggestions on building system that'd probably blow yours out of the water for half the price (if you remember to subtract the cost of the monitro). You're either going through the wrong vendors or just buying the most expensive available components (which isn't a good idea when building a $1200 machine).

As far as I can tell from various reports out there, there are enough technical difficulties with cellphone-to-ground service on planes that they are not simply going to allow regular cell phone use:

Ars Technica reports that there are two bans in place: the FAA for flight safety reasons, and the FCC for cellphone network interference reasons. (A cell phone can reach too many towers at once, thus interfering with towers other than the one it's actually communicating with) The USA Today article quoted by the Ars article discusses all this pretty well.

So, companies are trying to come up with a solution. One company, AirCell, has been granted "a patent" (US Pat 6,408,180) for facilitating cell phone use in planes. AirCell has a press release touting their patent and technology. USA Today said Aircell would charge a roaming fee to use their network.

The patent discusses the various methods used to reduce interference with ground stations, like antenna polarization.

It also seems to discuss an additional cell site on the plane itself, that is designed to convince all the passengers' phones to talk to it, so that it can efficiently relay the signals in a non-interfering mannor down to the ground.

It ALSO seems like they talk about redesigned ground sites to facilitate this, so you have to wonder what the involvement of each of the cell phone providers will have to be.

(skimming patents is not easy... I might be wrong, but the images help. In TIFF format: diagram of relay system on the plane, special cell sites vs normal cell sites)

It just hurts to think about the infrastructure investment in all these different, competeing cell technologies in the US. Wouldn't it make lots of sense to just GIVE UP and legislate/regulate a single standard. Say, GSM for example. Works for Europe... :)

- Peter

"...(they melted off from extreme heat -- look up the professor who got 3rd degree burns from his Vaio) ..." Tried it but only found this ArsTechnica newsdesk item (about a professor who received 2nd degree burns) And do you really think it's possible to receive 3rd degree burns from a computer if it isn't outright burning?

Reminds me of a comment I once saw on Slashdot claiming that there are places on our planet where one could receive 3rd degree burns from the Sun...

Each g5 dissipates a whopping 97 watts

I couldn't figure it out until I read this Ars technica article on .NET. Highly recommended.

I would recommend the Ars Technica OpenForum

Yes, when you consider that he's making this trip in late August and the G5 he's purchasing isn't available until September 1 it seems pretty silly to worry about shipping an item he won't have yet to ship!

Assuming it was an issue, my advice would be to ship ground in the original packaging. With insurance.

Other, more reputable hardware sites:
ExtremeTech (broke the story on nVIDIA 3dmark2k3 cheating, uses good benchmarking methods with the GameGauge)
FiringSquad (good mainstream site, quicker more casual reading before getting into the really in-depth stuff)
ArsTechnica (excellent for info on more fundamental aspects of hardware)

Check out Arstechnica.com.
They have a nice page devoted to cases and cooling.

Check out Arstechnica.com.
They have a nice page devoted to cases and cooling.

I'm a MacHead, and i'm reading ArsTechnica on a regular basis. Go see Jon Stokes conclusions about the PPC 970 chip.

BTW, the Mac is a PC.

Well, the PowerMac G5s ahve only been out for a few days, but a lot has been known about the processor for awhile. Ars Technica has done 2 articles on it already (here and here), and I don't think it's a stretch to assume that Apple and IBM let Adobe in on much of the structure for the 970 so they could produce an optimized version of PhotoShop in time for the WWDC bake-off.

And yes, I realise that Adobe optimizes PhotoShop for the Pentium series as well, I was just pointing out that their optimization for the G3/G4/G5 processors makes it a better choice for speed demos than arbitrary benchmarks.

As for programming, vim may run the same, but compilation varies. IANAHCP (where HCP=hard-core programmer), so I'm not sure what the mix of integer vs FP operations is in compilation, so I don't know how much the speed will vary between a P4 (better INT scores) and a G5 (better FP scores).

Well, the PowerMac G5s ahve only been out for a few days, but a lot has been known about the processor for awhile. Ars Technica has done 2 articles on it already (here and here), and I don't think it's a stretch to assume that Apple and IBM let Adobe in on much of the structure for the 970 so they could produce an optimized version of PhotoShop in time for the WWDC bake-off.

And yes, I realise that Adobe optimizes PhotoShop for the Pentium series as well, I was just pointing out that their optimization for the G3/G4/G5 processors makes it a better choice for speed demos than arbitrary benchmarks.

As for programming, vim may run the same, but compilation varies. IANAHCP (where HCP=hard-core programmer), so I'm not sure what the mix of integer vs FP operations is in compilation, so I don't know how much the speed will vary between a P4 (better INT scores) and a G5 (better FP scores).

Well, the PowerMac G5s ahve only been out for a few days, but a lot has been known about the processor for awhile. Ars Technica has done 2 articles on it already (here and here), and I don't think it's a stretch to assume that Apple and IBM let Adobe in on much of the structure for the 970 so they could produce an optimized version of PhotoShop in time for the WWDC bake-off.

And yes, I realise that Adobe optimizes PhotoShop for the Pentium series as well, I was just pointing out that their optimization for the G3/G4/G5 processors makes it a better choice for speed demos than arbitrary benchmarks.

As for programming, vim may run the same, but compilation varies. IANAHCP (where HCP=hard-core programmer), so I'm not sure what the mix of integer vs FP operations is in compilation, so I don't know how much the speed will vary between a P4 (better INT scores) and a G5 (better FP scores).

Then you need to actually READ ArsTechnica, rather than just try to be cool be referencing them.

Search for 'IBM 970' or 'Jaguar'.

Or just click HERE

But it's not the world's first 64-bit personal computer. It ships in September. AMD releases the 64-bit Opteron in August. By the time Apple's machine exists (i.e., you can have one on your desk) 64-bit PC's will have been available for a month.

But if Jobs' "announce now, deliver later" bullshit doesn't bother you, I'm accepting pre-orders via paypal for the world's first 128-bit 30Ghz personal computer*.

*it ships in 2009.

Seriously. Apple thinks that just because it announces something that wont be shipping for 3 months, it can claim the title of the "only" 64-bit desktop machine.

Apparently, either Jobs hasn't been paying attention to AMD, or AMD doesn't count because it's not Intel.

Right Here: http://arstechnica.com/wankerdesk/3q02/wwdc-622.ht ml.

Note the (idiot)slashcode addition of the space in the ht ml.

The guys at ArsTechnica wrote an awesome article last year, comparing the Honda Insight to the 2003 Honda Civic. I didn't come across it until a few days ago, and I have to admit that it really got me thinking about a hybrid, whenever I decide to purchase a new car. HTH.........

The guys at ArsTechnica wrote an awesome article last year, comparing the Honda Insight to the 2003 Honda Civic. I didn't come across it until a few days ago, and I have to admit that it really got me thinking about a hybrid, whenever I decide to purchase a new car. HTH.........

I have. Read the line about integer performance and you'll see why Apple will still be playing catchup with Intel and AMD. Most people are going to be doing integer and not floating point calculations when they are running their systems. Those that do benefit from floating point are likely not "Switch" candidates anyway. Either way, it's difficult at best to just drop one system and replace it for another when it comes to FP calculations as you not only need to purchase new hardware, you have to purchase new software and even with Adobe allowing crossgrade licensing, it's going to be a big hit to the wallet.

OK. I cede the point on renderfarms; I was really talking about web/db servers. Still, there are probably better modifications to be made for rendering, ie huge video cards and shitloads of (fast!!) RAM.

I can indeed point you to a reference on Altivec, namely this page in the 970 ArsTechnia article. Quote from the article: It appears that in the 970 the Altivec unit is sort of "tacked on" to the core. While the vector register file sits alongside the general purpose and floating-point register files for the purpose of keeping LOAD and STORE latencies down, the actual vector execution hardware is off on a different portion of the die, away from the vector register file and away from the rest of the execution core. This necessitates the addition of at least two extra stages to the vector pipeline: one stage at the beginning of the execution phase to actually move the instructions out to the vector execution unit and another stage at the end of the execution phase to move the instructions from the unit back to the group completion queue.

Also note that it says on the site that while the G4e can issue any two different vector ops in one cycle, the 970 can do it only if one of them is a permute, and in addition, the pipelines are longer.

Here, here, and here are just some of the reasons! ;-)

[...]
In the cell, DNA is modified biochemically by a variety of enzymes, which are tiny protein machines that read and process DNA according to nature's design. There is a wide variety and number of these "operational" proteins, which manipulate DNA on the molecular level. For example, there are enzymes that cut DNA and enzymes that paste it back together. Other enzymes function as copiers, and others as repair units. Molecular biology, Biochemistry, and Biotechnology have developed techniques that allow us to perform many of these cellular functions in the test tube. It's this cellular machinery, along with some synthetic chemistry, that makes up the palette of operations available for computation. Just like a CPU has a basic suite of operations like addition, bit-shifting, logical operators (AND, OR, NOT NOR), etc. that allow it to perform even the most complex calculations, DNA has cutting, copying, pasting, repairing, and many others. And note that in the test tube, enzymes do not function sequentially, working on one DNA at a time. Rather, many copies of the enzyme can work on many DNA molecules simultaneously. This is the power of DNA computing, that it can work in a massively parallel fashion.
[...]

From here, someone else mentioned it in another post. Biology isn't the same, but it does have properties that are very close -- and they can be intergrated as the Author implies with his questions. Your evolution argument has nothing to do with the fact that biology can be applied to computers and that -- as the Author brought up -- computing theory might change thanks to biology. Read this page to get a better idea of where the Author is coming from.

Jake Lead
Salk Institute

That isn't the whole story, and the running backwards analogy is just plain wrong.

Have you heard of something called parallel computation? RSA is doing it right here with DNA computing.

I suggest you read some background on what this means in terms of the nature of modern day computing, there's a good article here. Here's something from the second page of the article:


Now let's consider how you would solve a nontrivial example of the traveling salesman problem (# of cities > 10) with silicon vs. DNA. With a von Neumann computer, one naive method would be to set up a search tree, measure each complete branch sequentially, and keep the shortest one. Improvements could be made with better search algorithms, such as pruning the search tree when one of the branches you are measuring is already longer than the best candidate. A method you certainly would not use would be to first generate all possible paths and then search the entire list. Why? Well, consider that the entire list of routes for a 20 city problem could theoretically take 45 million GBytes of memory (18! routes with 7 byte words)! Also for a 100 MIPS computer, it would take two years just to generate all paths (assuming one instruction cycle to generate each city in every path). However, using DNA computing, this method becomes feasible! 10^15 is just a nanomole of material, a relatively small number for biochemistry. Also, routes no longer have to be searched through sequentially. Operations can be done all in parallel.


This is a huge deal for computing. Huge.

I went to Berkeley too. Have you heard of The Berkeley Initiative in Soft Computing (BISC)? Read their website, it will also increase your understanding as to how fuzzy logic translates into efficiencies and more to the point, performance. Not to mention the potential for efficent and high levels of data storage in DNA. The possibilites are amazing! A detailed understanding of evolutionary biology in the context of fuzzy logic and modern day computer computation (especially parallel) will blow your mind in terms of how things came to be, and how they fit so perfectly with certain operations. This is really the next big thing.

G.R. Bouchard, PhD
Associate Professor of Biophysics

2) - Dual rpocessors give a 70% speed increase at best. Few programs are optimised for them so the biggest benefit you get is when running multiple programs, so going with a 30% increase would be a tad more realistc.

I was under the impression that current PowerMacs supportsymmetric multiprocessing of multithreaded applications (even Java threads), which means that in iTunes, one processor can be ripping a CD while the other can be importing a library of MP3s. I wouldn't be surprised by a 70% figure for multithreaded apps.

Many of my predictions are indeed based on a number of different rumor sites, but not copied directly. Some of the rumors I've read I disagree with, and some I agree with.

For example, I think it was Mac OS Rumors that said the G5 will not be called the G5, and I disagree with them. This issue was also mentioned by As the Apple Turns, who said that according to AppleInsider, it would be called the PowerMac G5. I agree with them. That doesn't mean my prediction is based on theirs, merely that we both made the same prediction.

The nature of the PPC970 chip, and that Apple will use it at all, is based largely on a couple of articles at ArsTechnica, but they didn't say anything about when it would ship.

The 1.4, dual 1.6 and dual 1.8GHz clock speeds are consistent with Mac OS Rumors, although I'm sure I've seen other speeds suggested elsewhere. I believe I've heard 2GHz suggested, and I don't agree with that (not for WWDC). I forgot to mention pricing, but I predict the low-end and mid-range models will be $1499 and $1999 respectively; this is based on Apple's current pricing, not on any rumor site.

USB2 support I heard somewhere, but don't remember where (it had to do with motherboard specs). Bluetooth, FireWire 800 and Airport Extreme are currently shipping features.

I've heard about the 15" Aluminum PowerBook from a few sources I think. The PowerBook G5 has also been mentioned in multiple places including this Slashdot article, but I don't expect to see it until next year, possibly announced at MacWorld San Francisco but probably not.

The G5 shipping with 10.2 was a possibility I had been considering, but was confirmed by ThinkSecret and eWEEK. Same source for gcc 3.3. Pricing is based on Apple's history.

The multiple simultaneous users feature I heard from a few places quite some time ago; I don't remember where. Apple's WWDC material says Panther and WebCore will be demonstrated at WWDC; that's no secret. As for PAC and WPAD, I haven't seen that suggested anywhere.

In any case, a rumor is "A piece of unverified information of uncertain origin usually spread by word of mouth." Many of my predictions are based on rumors. The sites I got the rumors from are mostly just passing on rumors they've heard. I don't feel that not citing sources was inappropriate, since these are MY predictions, BASED ON what many others have said, not simply a copy of someone else's predictions. I would expect others to be able to make similar predictions, based on overlapping sources.

It's the architecture. I remember reading an article on Ars-Technica that discussed just how different the architecture is for the PS2 vs. Xbox/PC. And how it was causing some issues with programmers not being able to fully comprehend and exploit the system. I just got done playing Midnight Club II, and I can tell you that they comprehend it now -- that machine is over 3 years old and still puts out amazing graphics.

And IBM calls it VMX (Vector Multimedia Extensions, IIRC) because Motorola holds the trademark on the name "Altivec."

Actually, VMX is a different architecture than Altivec, although the implementation in the 970 is Altivec instruction compatible (at Apple's request). The guys at Ars Technica seem to believe that VMX might actually underperform Motorola's current Altivec implementation.

Given the effort IBM has gone to in order to make the 970 perfect for Apple, I'm left with little doubt that all high-end Macs will soon be powered by the 970. Motorola's last-ditch efforts to boost the G4 are simply too late to dissuade Apple from that course of action.

But the 970 will almost certainly be more expensive than the G4 at its introduction and possibly for some time afterward. Therefore, its probable that Apple is already planning to relegate its low-end offerings (iBook, eMac) to the G4 initially. These rumblings from Motorola are probably meant to persuade Apple to keep things that way for longer than it was perhaps already planning to. And if Motorola really does ramp the G4 to 3GHz in the near future (somehow I have my doubts about that) then they might just succeed.

In any case, I still believe that it is only a matter of time before the the Mac line is converted entirely to the 970. But what I want to know is this: will Motorola gain access to the 970 design specs because of the Apple-IBM-Motorola (AIM) PPC contracts. If so, will we see Motorola 970s in the future? I hope so; competition of this sort always benefits the consumer.

Ars Technica reported on this yesterday, and the center of the story was the blog posting you see above. All the AP post on the NYTimes has done is read something into that blog post that isn't there. Nowhere in Frankel's blog does he state that he is going to resign right now, or that he has tendered his resignation. He simply said that he would probably leave eventually.

This is a very misleading story from the AP.

http://www.arstechnica.com/archive/news/1050185127 .html

Power 4 has thicker gate oxides than the 970. The 970 is a desktop chip, the Power 4 is a server chip.

Apparently, in order to increase the reliability of the Power4 for the high-end server market, IBM used much thicker gate oxides on the chip's transistors. The trade-off for this decreased failure rate and improved reliability was that the Power4's transistors have slower switching speeds, so even with process shrinks it's harder to push the design to higher clock speeds. Since the 970 is made for the desktop market, there's no need for such measures and therefore the new chip's clock speed will scale much higher than the Power4's. In sum, the 970 is made to be faster, cheaper, and significantly less reliable than the Power4."

Inside the IBM PowerPC 970

Um yeah, that's a valid concern.

Would the SPEC INT and FP numbers projected by IBM for the 970 take into account programs optimized for the Altivec/Velocity Engine?I know ars tecnica has compared the SPEC benchmarks to the current P4 3.06 processor and found it wanting, but if the bus can feed with the VPU with 4x more bandwidth than the current crop of G4's, then we might have the ultimate photoshop bake-off machine on our hands :)

The trade-off for this decreased failure rate and improved reliability was that the Power4's transistors have slower switching speeds, so even with process shrinks it's harder to push the design to higher clock speeds. Since the 970 is made for the desktop market, there's no need for such measures and therefore the new chip's clock speed will scale much higher than the Power4's. In sum, the 970 is made to be faster, cheaper, and significantly less reliable than the Power4. (Of course, when I say "significantly less reliable than the Power4," you have to understand that this puts the 970's product life and failure rate on par with other mainstream CPUs, since the Power4's increased gate oxide thickness makes it significantly more reliable than most mainstream CPUs.)

This kind of reminds me of the whole ArtX Ars Technica thing awhile back... Is there anyway to verify that this AC didn't post from SCO? (Sorry if this is a troll)

Cache is only useful when you're working on a small data set. Multimedia applications tend to constantly move through a large data set, which makes the cache all but useless. Full details here.

You're assuming that the videocard will be maxed out, this doesn't appear to be the case.
You're also assuming that the majority of graphics/video tasks hit the videocard, most seem to be software at this point (and dripping with altivec code)

give this a look

having the video subsystem handle things that were previously handled by the processor (like window composition) is faster than the cpu doing it, and also frees up the cpu do throw horsepower at an FCP render :)

What the /hell/ are you on?

1. NO 970 MACHINE HAS BEEN ANNOUNCED BY APPLE YET. Say it with me, dammit. While it may be likely, don't take as canon rumor sites and IBM press releases that don't even mention Apple Power Macs. Jeez. You're already a Mac user, eh? (And I say that being one.)

2. 980? 990? WTF? At what data are you looking? Search Google for "ibm 970 chip" and the only info you find are two random comments in some forum somewhere; search IBM for roadmap info on PowerPC, and you will find their "9xx" selection, and the only thing under that is this:

http://www-3.ibm.com/chips/techlib/techlib.nsf/tec hdocs/A1387A29AC1C2AE087256C5200611780

Lastly, with the release of the 970 being sometime in the second half of this year , don't you think saying we'll probably have a "990" by 2005 is a little premature?

Meh.

...I take prompts from the fellows at Arstechnica.

The post-sweet chairs they picked out a few months ago are still very cool...

As Hannibal (Jon Stokes) notes in the article in question:

"The fact that the Altivec unit was slapped onto the design, leaving some room for improvement in future iterations, leaves no doubt that the 970 achieved its present form under pressure from Apple and that Apple will be rolling out systems based on the new processor. This is the most plausible and reasonable explanation for the way the vector unit looks. If the 970 were solely intended as a Linux desktop platform for IBM, they would've preferred to reduce the 970's die size, power consumption, time-to-market, etc. by just leaving out the Altivec unit altogether, instead of shoehorning it into the design the way they did."

Most Linux variants and apps aren't Altivec-optimized, so there wouldn't be very much incentive for IBM to include the functionality in a Linux-only box given the engineering work involved in doing so. It makes much more sense to do it when you know that you could easily sell hundreds of thousands of these CPUs to another company whose customers are desperately eager for that level of performance, i.e. Apple.

Interesting, if you look at the pipeline design of the PowerPC it is much closer to Intel than AMD. The PowerPC pipeline has sixteen stages, the Pentium 4 twenty, and the Athlon ten.

Presumably the P4 can reach higher clock speeds than the Athlon because there is less work to do at each pipeline stage. On the other hand a longer pipeline increases the probability of a stall, so the work done per clock cycle goes down.

I'd speculate that the PowerPC ought, therefore, to be able to achieve clock rates approaching but not equalling the P4, since they are both comparatively "over-pipelined". At the same time, the PowerPC ought to deliver slightly more throughput per clock cycle because the pipeline is slightly shorter.

Meanwhile, the Athlon will be running at a significantly lower clock rate, but delivering comparable throughput.

Here is a handy CPU and Chipset Guide
page to look at

How about the recently made Ars Technica Guide to Capturing, Cleaning, & Compressing Video? It was made with exactly what you want to do in mind.

He is still fighting. He is putting the search engine back up and defying the RIAA as soon as he is sure he is legally protected. We are all threatened by this kind of behavior by large corporations, so we should try to support the individuals that have been singled out. This could easily have been you that is being sued. If you download music using any technology, you are trackable, and the RIAA may one day randomly select you for prosecution. Don't say it can't happen, because they have already privately messaged thousands of music traders. This means that they can find you, and they can prosecute you. We need to help the people who are being prosecuted, and we all need to join together to make sure that even though the RIAA temporarily shuts down music finding services such as Phynd, we will all join together to pay their settlements and continue on the way we were, legally.