Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
I'm sure that Apple's made quad processor Macs for testing, and other stuff. I am pretty sure that there will be a G5 within a year, though.
 
There is an unfortunate lack of understanding here of the problems associated with a quad G4, or quad anything really.

Sure, a quad G4 may even be possible, and it could happen, but lets look at some problems:

1) The FSB. Right now the dual G4 systems share their FSB, and as far as I know a quad would have all four sharing. Now when I say sharing, it's like kids and fancy toys share. Sometimes they play nice, sometimes there is some crying and fighting.

By this I mean that if you think G4's are held back by RAM now, imagine doubling the number of G4's expected to share! Ow!

2) Other than endless rumors that people repeat as fact, there is no reason to believe that there is a new revision of G4 with DDR support coming next month. I maintain that it is too soon, seeing as how the Apollo was just released like 6 months ago, and seeing as how Apple's fancy new Xserve chipset is brand new (and uses a SDR FSB, not a DDR one).

3) While Apple could implement multiple busses for their chips (like AMD with their dual Athlons), this is no small task. You can't just take out the marker and draw it on the motherboard, it would take a lot of time and money to design and test and custom mobo and custom chipset, and worse, it would not be a part shared with other products anywhere in the world. This is just far too expensive for Apple. They can't make wacky custom chips for a small slice of their tower line... not at that rumor-as-fact $3500 that got tossed around.

4) Performance: Quads are not gona blow you away. I'm happy to see this is generally realized in the thread, but consider this: Going from one chip to two increased performance on a *specially* *coded* task by 50% or less. Lets be nice and assume we can get another 50% by doubling again, to quads. 4ghz of chips doing only twice (or so) the work of a single 1ghz chip. Sheesh. This is worse efficientcy than an uber-clockspeed chip like the P4, and I know how much people love that chip around here! Hows that for perspective?

Oh, and that only works if the app is specially optimized, unlike the situtation for the much-reviled P4. Owch.

Don't bother giving me the multitasking line either... lets see how far you can get with four chips sharing a bus. I am guessing it won't be far.
 
I don't know think you realize something: the only way that a quad-G4 system would ever be released, was if dual or even quadruple system buses were in place (which would mean a new motherboard and some tricking the processors). Maybe dual-133MHz system buses or dual-166MHz system buses (as in the case of the rumored 7470s) would provide ample bandwidth.

Edit: Remember the rumors of the NForce2? The NForce2 has dual-channel DDR support. Though I think it's unlikely we'll see NVidia chipsets, it's some food for thought.

As for jumping from the Apollos... what makes you think that we couldn't use the Apollos in the iMacs, and the 7470s in the towers? That would extend the life of the Apollo significantly.
 
Originally posted by ddtlm
There is an unfortunate lack of understanding here of the problems associated with a quad G4, or quad anything really.

Sure, a quad G4 may even be possible, and it could happen, but lets look at some problems:

1) The FSB. Right now the dual G4 systems share their FSB, and as far as I know a quad would have all four sharing. Now when I say sharing, it's like kids and fancy toys share. Sometimes they play nice, sometimes there is some crying and fighting.

By this I mean that if you think G4's are held back by RAM now, imagine doubling the number of G4's expected to share! Ow!

2) Other than endless rumors that people repeat as fact, there is no reason to believe that there is a new revision of G4 with DDR support coming next month. I maintain that it is too soon, seeing as how the Apollo was just released like 6 months ago, and seeing as how Apple's fancy new Xserve chipset is brand new (and uses a SDR FSB, not a DDR one).

3) While Apple could implement multiple busses for their chips (like AMD with their dual Athlons), this is no small task. You can't just take out the marker and draw it on the motherboard, it would take a lot of time and money to design and test and custom mobo and custom chipset, and worse, it would not be a part shared with other products anywhere in the world. This is just far too expensive for Apple. They can't make wacky custom chips for a small slice of their tower line... not at that rumor-as-fact $3500 that got tossed around.

4) Performance: Quads are not gona blow you away. I'm happy to see this is generally realized in the thread, but consider this: Going from one chip to two increased performance on a *specially* *coded* task by 50% or less. Lets be nice and assume we can get another 50% by doubling again, to quads. 4ghz of chips doing only twice (or so) the work of a single 1ghz chip. Sheesh. This is worse efficientcy than an uber-clockspeed chip like the P4, and I know how much people love that chip around here! Hows that for perspective?

Oh, and that only works if the app is specially optimized, unlike the situtation for the much-reviled P4. Owch.

Don't bother giving me the multitasking line either... lets see how far you can get with four chips sharing a bus. I am guessing it won't be far.

Ah, you had to come out and say it didnt you. The nagging little voice of reason in the back of all of our minds. Most of us (myself included) hase pretty much come to expect a huge upgrade in august, but the things you mention in your porst have been nothering me as well. It certainly seems odd that the xserve's lack the ddr fsb. when i first saw it i immediatly assumed apple would take the same route with the new powermacs. the response to this thought has been that raw processing power in a server is less important than memory access, and so the xserves can get away with a slower fsb as long as they have ddr ram, while the powermacs rely on their raw processing power more, which has been hindered for some time by the slow fsb. im not sure if i buy the retort or not, and although all logic seems to indicate that you are right about a more conservative release come august, my only hope is that no matter how much we think we know apple does tend to be good at surprises. the g4 lcd imac, is a great example. sure we all basically knew we would get a 15" lcd, but only the crazies thought there would be a g4 in it as well. but sure enough apple pulled it off and shocked us all. im hoping apple will be able to again meet the expectations of the crazies, because it is a sweet song they sing.
 
rice_web:

Dual-channel DDR RAM does not equal dual front side busses. The nForce cannot run dual Athlons because it only has one bus (and Athlons can't share) even though it has two channels of RAM. On the flip side, the 760MP chipset can run dual Athlons cause it has two busses, but it only has a single DDR RAM channel.

All other chipsets for the Athlon have one channel for RAM and one FSB, but they could have made some with dual FSB's and dual-channel RAM at a high cost (and for a small market).

I rate it a vitrual impossibility that Apple will roll out an expensive custom chipset with multiple FSB's in a month, or ever, given their current target audience. G4's are chips that naturally share a FSB, and doing otherwise would a mess, although stranger things have been done for P3-Xeons and Xeons. Of course, notice that those custom setups are very expensive and used only in servers.

AmbitiousLemon:

I wouldn't buy those excuses about the Xserve... plenty of people want to use them for very intensive tasks such as render farms. I think the problem is that Moto does not have a G4 that supports a DDR FSB, so Apple did the best they could.

In any case, I think that DDR FSB or no, in many many tasks the G4's are not yet using up their full memory bandwidth, because in many tasks my PC133 1.53ghz Athlon is much faster. If I was in the market, I would definately consider a dual G4 with SDR FSB and DDR RAM, hopefully at > 1.2ghz per G4.
 
Originally posted by ddtlm
2) Other than endless rumors that people repeat as fact, there is no reason to believe that there is a new revision of G4 with DDR support coming next month. I maintain that it is too soon, seeing as how the Apollo was just released like 6 months ago, and seeing as how Apple's fancy new Xserve chipset is brand new (and uses a SDR FSB, not a DDR one).

Even some Apple employee's who are sworn to secrecy (I'm not talking about the sales people at the Apple stores, but true Apple Reps who are pretty high up on the information ladder at Apple) are not arguing with the rumors about 1.4-5 Ghz G4's with DDR RAM. I asked one of these guys recently, and he just said to expect something in August with the 1.4-5 Ghz G4 and DDR, and a G5 sometime in the middle of next year. It seems that he just looked at the rumor sites, and just repeated what he saw there, but I talked to another Apple guy who was there (not as high up on the info ladder. He actually tries to pry info from this other guy, but he's cool...) and he said that the guy I talked to does know about future Apple products, and that he's probably used them and seen them before. Probably the only reason that he was able to say what he did, was because those specs were already plastered all over the rumor sites, and it wouldn't hurt him to tell, but that's all he said. Of course, If you were in his position, you wouldn't say much either, at the sake of losing your job...

But, I do imagine that quad's are a pipe dream, for now...
 

Attachments

  • bus.jpg
    bus.jpg
    68.1 KB · Views: 495
quad G4s are an utter waste of apples time, there is no marktet for them. How much of apples market do the Dual 1ghz take up??

plus, hmmmmmmmmm MOSR... what happened to the Dual G4 TiBooks they rumored before???????/hhhmmmmmmmmmmmmm...
 
Originally posted by ddtlm
Going from one chip to two increased performance on a *specially* *coded* task by 50% or less. Lets be nice and assume we can get another 50% by doubling again, to quads. 4ghz of chips doing only twice (or so) the work of a single 1ghz chip. Sheesh. This is worse efficientcy than an uber-clockspeed
[/B]

All of your other points are good, but this is just flat out wrong. The only reason you are seeing such horrid numbers as 51% increase per cpu on the macs is because the bus is abosulutely atrocious. A single 1Ghz G4 is aleady starved for memory bandwith on the current motherboards apple is using. If apple had a real bus behind it you would see near linear scaling as CPUs were added. No you would never see 100% but with a decent bus (see Alpha/AMD) you can get upper 80s out of another CPU and a degneration of 1-2% on each additional CPU on the bus.

The G4 is fine. The current bus being used by Apple is what is killing everything. Go over to Arstechnica and read some of the posts by BadAndy or if thats too much trouble check out the peformance specs on dual 450s vs dual 500s vs dual 800s vs dual 1Ghz... the curve goes down horrible fast on gain from performance... much more so than if you compare single g4s.

A 4-CPU 1Ghz g4 on something like HyperTransport would see something along the lines of relative 3.36Ghz performance when lined up against a single 1Ghz cpu.
 
Hey guys, my name is Steve, but I go by Syco. A friend of mine directed me to this website. I've been interested in getting a new computer (I have my eyes set on the Dual 1GHz G4 with a GeForce4 Ti :D ) once I get the money scrounged up. However, these rumors of Quad-processors has really set my eyes ablaze. With OSX's symmetric multiprocessing, having 4 G4s would be an amazing thing to play WarCraft 3 on. However, I'm not the biggest genius when it comes to computers, so I have a few questions regarding multiprocessing altogether.
1) Would it be feasible to, maybe, include 16 100MHz processors into a computer? Heat-wise and cost-wise? If this is possible, it would make the system more efficient if OSX did what they do with memory - denote which apps get which processors to fulfill their speed need.
2) Why doesn't Apple test their Dual 800 MHz G4 with their 800 MHZ SP G4 to test the advantage of DP? Even showing benchmarks of 933MHz SP to Dual 1GHz has the discrepancy.
3) How does MP support for X have an effect on apps that already support MP in their own way? For instance, the program SNES9X, which is a SNES emulator, uses one processor to handle the ingame processing, and the other is used to apply the ingame video modes. How would MPing affect the program in that sense?

Thanks for your time, and sorry if I sound like a total newbie.
 
Here's my take

Here's my take on what the next PowerMacs will be with a bit of ftaok-logic thrown in.

Apple didn't release a new PowerMac at MWNY because (a) the double (or quad) pumped bus designed to work with the 7470 G4 was not ready, or (b) the bus is ready but Motorola wasn't able to get satisfactory yields for the 1.6ghz chip, or (c) Motorola doesn't have a chip that's faster than 1ghz and Apple hasn't been designing a new bus.

Being the optimist (and Motorola supporter) that I am, I'm inclined to say that it's either (a) or (b). So when the rumored August date rolls around, the PowerMac that will be intro'd will have the following specs:

Dual 1.6ghz 7470 (with 2 or 4MB of L3)
333mhz effective bus speed
Support for lots of DDR RAM
Big Hard Drives
FireWire2 - maybe
USB 2.0 - only if Firewire2 is there

plus 17in & 19in Cinema Displays to go with the 22 and 23in jobbers.

Oh, and built-in Bluetooth.
 
Re: Here's my take

Originally posted by ftaok


Dual 1.6ghz 7470 (with 2 or 4MB of L3)
333mhz effective bus speed
Support for lots of DDR RAM
Big Hard Drives
FireWire2 - maybe
USB 2.0 - only if Firewire2 is there

Oh, and built-in Bluetooth.

--

warning - I am typ i ng on a u nfamiliar key board in an i nternet cafe in Bali (i ndonesia) ri ght now, and this stupid PC keyboard is deciding to put spaces everywhere for me :confused: ) - warni ng

I think that you are t oo optomistic. But I would be VERY angry if apple di dnot i nclude USB 2 on th ep owermacs. P ro users need th e newest technology, and USB 2 is hardly new technology. I will not be mad at apple if they di dnt i nclude FW2/gigawire, but I would like it :D . I am mad at apple fo r fall i ng b ehi nd in speed, but i guess it is motorola i should be mad at. I really hope apple goes with IBM's scaled down Power4. Then we would go fr om the g4/5 to the P4. That might get co nfusing with pentium 4 and stuff tbet that is irrelevant.

I quad p roc essor is useless, 2 is the only ammount of processo rs that is worth the extra co st. And 16 100 MHz pr ocessors would not wor k. Quad procesors would be a bad mood. Apple needs megahertz. That is the only thing p eople look at, and they will never gai n any mar ket share u nless th ey are competing with the Wintel processor speeds.

:rolleyes: i find my s elf co ncludi ng i n half my posts that apple needs to start concentrating on speed, because the iApps wo nt make up for their lack of speed in the lo ng run...:(
 
Re: Re: Here's my take

Originally posted by scem0

But I would be VERY angry if apple did not include USB 2 on the powermacs. P ro users need th e newest technology, and USB 2 is hardly new technology. I will not be mad at apple if they didnt include FW2/gigawire, but I would like it.

I think this statement couldn't be farther from the truth. Pro users use the fastest technology, with the greatest reliability. Pro users aren't complaining about the IDE bus speeds since they use U160 SCSI. Pro users aren't complaing about USB 2.0, because there are no devices capable of saturating FireWire 1, let alone USB 2 of FireWire 2. Anyone silly enough to build a FireWire RAID, and complain about saturating the bus is a gadget freak, not a Pro. Pro users are further silent on the matter, because very few Pro devices are released which take advantage of a SERIAL bus. Furthermore, the FireWire bus was designed for speed + ease, and USB was designed for ease only. USB 2.0 is an attempt to turn a SERIAL bus into a storage bus. USB was designed for keyboards, cameras, MP3 players, semi-pro shuttles, speakers, mics, etc. FireWire was designed for digital video, CDRW, DVDRW, etc...hmmm


I am mad at apple for falling behind in speed, but i guess it is motorola i should be mad at. I really hope apple goes with IBM's scaled down Power4. Then we would go from the g4/5 to the P4. That might get confusing with pentium 4 and stuff but that is irrelevant.

Everyone would like to see more speed from Apple hardware, but I am seeing Pro users considering the platform more than ever before. When data/security/reliabilty/licensing flaws can cost a Windows box 10 days/year of operation, a Mac looks much better. Furthermore, linux and unix users are seeing OSX as filling a void that Linux could not fill. The Xserve, and future hardware, regarless of processor, will help prove Apple as an enterprise solution (for the first time in a long time).

Apple needs megahertz. That is the only thing people look at, and they will never gain any market share unless they are competing with the Wintel processor speeds.

Speed in MHz is what consumers look at. Professionals look at things like productivity, TCO, reliability, licensing & support costs, etc. Ask yourself why Sun powers 60% of the web, even though the UltraSparcIII is on par with the G4 for MHz. Even the Itanic II is not MHz driven. CTOs, and IT pros don't look at MHz. Professional end users value MHz less than you might think. Consumers think the sun shines out of the MHz hole.

i find myself concluding in half my posts that apple needs to start concentrating on speed, because the iApps won't make up for their lack of speed in the long run

Again, while everyone would like to see a faster Pro line, a professional looks at the iApps as cute toys for the kids. PhotoShop matters. Oracle matters. FCP matters. Pro audio software matters. Apache matters. Exchange support (sadly) sometimes matters. Networking matters. iApps waste drive space.
 
true

Well, that is all true, but pro users arent what apple needs more of, because there are more regular consumers, then pro consumers. They need megahertz if they want more market share. They need a better company if they need more megahertz, because motorola isnt pulling their weight. I think IBM or AMD are good choices, but that is just an opinion. While it is true that a true pro consumer doesnt need USB2, a lot of pro users would like it because of its ease. I would want it on my computer because most CD burners out right now require USB2, and firewire devices arent being made as much now, with FW2 coming out (hopefully).
Yes, you are correct - to a professional, iApps waste space. But that is the reason that they wont make up for lack of speed.

I agree with everything you said though. :) pretty much:D
 
Originally posted by dukestreet
Cool, but this is also interesting

the PPC 7500, is purported to be less than a year away with its 500MHz RapidIO bus

That means that the G5 won't be around for quite some time - if they're already looking at a replacement to the 7470 in less than a year? Damn, not good. If this rumor is true, the life expectancy of the G4 is much longer than I ever expected.

D

Any appearance of quad g4s is indeed its own bad news- confirmation that a g5 is yet a considerable ways off.
 
Originally posted by szark
The 7500 was on the older version of the roadmap in the G5 category, along with the 8500.

Finally found a copy of the old roadmap where Motorola showed the 75xx as a G5:
 

Attachments

  • powerpc roadmap -- v3.jpg
    powerpc roadmap -- v3.jpg
    74.7 KB · Views: 180
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.