No they will just toss on a L4 cache and call it the Pentium 4 Hyper Threading Extreme Edition Super Ultra XP
edit: added quote
edit: added quote
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
90nm PowerMac G5s in March 2004?
- Thread starter MacRumors
- Start date
- Sort by reaction score
Intel and IBM
Intel has BILLIIONS to plow into chips. I wouldn't be confident in IBM delivering something that radically outperforms Intel by end of next summer. I'd sure like to see it. I'd sure like to see all IBM's efforts in SoI, high K, strained silicon (Intel doing this too), SMT, dynamic power and frequency management, and whatever else.
Speaking of whatever else, I hope we see an on-chip memory controller, more L2, radically better Integer performance (another unit or just better use of existing), and improved AltiVec (include some of those functions in the G4+ that the G5 lacks). Oh yeah, one more thing, wider pipes for main memory (provided there's a practical main memory and motherboard solution). I'm not so gung-ho about a dual core 980 unless IBM and Apple can feed the CPU enough data from main memory to make it worthwhile.
As for the next three months, I agree that we'll see a speed bump. I don't accept the notion that heat dissipation alone precluded Apple from throwing 970's into iMacs.
Anyway, I wonder if IBM offered Apple a choice regarding production volume: higher volume in 130 nm 970s now, pushing 90 nm 970s out. Or, just enough volume for the PowerMac line now, with a quicker availability of 90 nm 970s so Apple can offer faster units for PowerMacs, lower power units for PowerBooks, and plenty left over for iMacs.
Well hell, I don't know jack about IBM's production capacity. Who here really does? Remember, Fishkill is not dedicated to 970 production. Other products are manufactured there as well.
As for xServe, IBM announced 1.6 GHz 970s for its blade offering. Apple could almost certainly do the same. Question is, why not already? At 130 nm, the 1.6 GHz 970 heat dissipation is comparable to Intel/AMD beasts. So why is IBM delivering its first 970 finished product in January? Could it have something to do with optimizing Linux (and AIX) and other software for the 970? I wish I knew. Another point, with servers, heat dissipation noise (fans) aren't a problem. Though, the electricity costs do matter.
I've ranted too long again.
Eirik
Intel has BILLIIONS to plow into chips. I wouldn't be confident in IBM delivering something that radically outperforms Intel by end of next summer. I'd sure like to see it. I'd sure like to see all IBM's efforts in SoI, high K, strained silicon (Intel doing this too), SMT, dynamic power and frequency management, and whatever else.
Speaking of whatever else, I hope we see an on-chip memory controller, more L2, radically better Integer performance (another unit or just better use of existing), and improved AltiVec (include some of those functions in the G4+ that the G5 lacks). Oh yeah, one more thing, wider pipes for main memory (provided there's a practical main memory and motherboard solution). I'm not so gung-ho about a dual core 980 unless IBM and Apple can feed the CPU enough data from main memory to make it worthwhile.
As for the next three months, I agree that we'll see a speed bump. I don't accept the notion that heat dissipation alone precluded Apple from throwing 970's into iMacs.
Anyway, I wonder if IBM offered Apple a choice regarding production volume: higher volume in 130 nm 970s now, pushing 90 nm 970s out. Or, just enough volume for the PowerMac line now, with a quicker availability of 90 nm 970s so Apple can offer faster units for PowerMacs, lower power units for PowerBooks, and plenty left over for iMacs.
Well hell, I don't know jack about IBM's production capacity. Who here really does? Remember, Fishkill is not dedicated to 970 production. Other products are manufactured there as well.
As for xServe, IBM announced 1.6 GHz 970s for its blade offering. Apple could almost certainly do the same. Question is, why not already? At 130 nm, the 1.6 GHz 970 heat dissipation is comparable to Intel/AMD beasts. So why is IBM delivering its first 970 finished product in January? Could it have something to do with optimizing Linux (and AIX) and other software for the 970? I wish I knew. Another point, with servers, heat dissipation noise (fans) aren't a problem. Though, the electricity costs do matter.
I've ranted too long again.
Eirik
Hey, with a smaller design you may be able to hold more internal hard drives like the g4's always could. Just dream...
dual 3ghz G5
8gb 533 DDR II
4 x 250gb 7200rpm
readeon 9800 pro
superdrive
and of course dual 30" apple monitors. (when are these coming out.)
dual 3ghz G5
8gb 533 DDR II
4 x 250gb 7200rpm
readeon 9800 pro
superdrive
and of course dual 30" apple monitors. (when are these coming out.)
Altough I agree that it would be very hard for IBM to significantly outperform Intel on the pure horsepower basis, it would be interesting to look at bigger picture here. Ask yourself one question? What is the PC of the near future? In my opinion near- future will belong to neatly designed, silent, small factor PC's with preference put to energy conservation and ease of use. Here I think Apple-IBM(or even Moto) could have a clear winner. PowerPC design allows more bang per Watt used, combined with Apple expertise in ergonomics and ease of use could make really interesting alternative to Intel based PCs. Most Intel CPU offerings are consuming >90 W and therefore require massive cooling effort that results in something like big G5 box 😛(just much uglier). If Intel continues to go down this road, I am afraid that point will be reached where significant market segments won't be able to justify gains in speed at the cost of above mentioned benefits. So as long as IBM (maybe even AMD) are chuggin along I will be perfectly happy chap.
Who knows they could come out in Jan. for all we know right now. Apple is very hush hush about it's Xserve.
The answer is no, I don't think you'll see L3 cache again for a while unless the memory speed stays the same and the processors get to 8 ghz. Think about this for a moment:
The fastest G4 processor ran at 1.42 ghz.
The L3 cache ran at 1/2 processor speed, so about 700 mhz.
On the dual G5 your memory is running at 800 mhz. (2 channels of DDR 400) and your bus is 1 ghz. so your memory is effectively faster than L3 cache was on the G4. There's really no need for the L3 at all at that point. The L3 cache was just a "short-term" measure to get around the fact that bus speed was so slow on the Motorola processors.
Flynnstone
macrumors 65816
My understanding of the bus multiplers is : 1:1, 2:1, 4:1, 8:1.
Of course we are only interested in 1:1 and 2:1 😀
The bus interface between the processor(s) is Hypertransportish. They don't say it's HyperTransport, but looking at the datait sure looks like HyperTransport.
So my theory is its effectively HyperTransport, but by not calling it HyperTransport they don't have to conform to the spec. Easier to crank up the speed 🙂 and speed is good 🙂 🙂
ImAlwaysRight
macrumors 6502a
Reality check
Reality check, people. Apple announced the dual 2.0 this summer (June/July, me forgets already), and doesn't ship in volume until late Sept, yet some think based on this rumor we will see 3.0 GHz G5 in Jan? Unless you mean 2005, I really don't think so.Shipping in July 2004, maybe. But, a boy can dream!
there will be one more 970 G5 at .13 in January/February around 2.4GHZ.
The .09 970 G5 will be not be ready until next summer when with a 1.5GHZ bus implies a 3GHZ G5 like Steve promised us.
Go Apple
Tyler
The .09 970 G5 will be not be ready until next summer when with a 1.5GHZ bus implies a 3GHZ G5 like Steve promised us.
Go Apple
Tyler
dongmin
macrumors 68000
Re: Reality check
true, but this was the this past summer was the first installment of the G5. The first iteration always has a lot of production issues and software issues to work out. The second round should come more immediately. As long as IBM can supply the chips, Apple shouldn't have too much trouble pumping out the boxes. And from all we've heard, IBM is getting good yields on the 970s.
manitoubalck
macrumors 6502a
Deos anyone make DDRII DIMMS yet, I know that they exist on some graphics cards but It's unlikley to go into system memory when DDR chips are up to 850MHz (2x425 MHz) which is in the GeForce FX5900.
Also "The DDR-II memory is expensive, not yet available in sufficient quantities and only supports a 128-bit bus," from www.tomshareware.com
Trend of PowerMac
Has anyone seen a picture of the trend of PowerMac capabilities, i.e., the combination of CPU speed, bus speed, and maximum RAM? Would that give us an even better indication of short-term and long-term Apple hardware futures? The reason I ask is because I'm advising some of my friends to wait on buying iMacs until Apple bump up their CPUs to G5s and bus speeds to something far beyond the miserly 167 MHz.
If memory serves, previous MacForums threads have said that that Apple tend to make significant PowerMac annoucements twice a year or so.
Has anyone seen a picture of the trend of PowerMac capabilities, i.e., the combination of CPU speed, bus speed, and maximum RAM? Would that give us an even better indication of short-term and long-term Apple hardware futures? The reason I ask is because I'm advising some of my friends to wait on buying iMacs until Apple bump up their CPUs to G5s and bus speeds to something far beyond the miserly 167 MHz.
If memory serves, previous MacForums threads have said that that Apple tend to make significant PowerMac annoucements twice a year or so.
Yeah I was thinking about mentioning an *Sparc driven laptop - but they don't really seem to be in the running. I guess technically WinCE is also 64bit when it comes to the Mips cores... but then that's not what I'd call stable. 😀
-Wyrm
Analog Kid
macrumors G3
Re: PB G5?
Might, but I'd still think late summer... If this is accurate they're making 3 major heat improvements:
Shrinking the CPU
Shrinking the system controller
Moving to DDR2
I did the calculations a while back and don't care to do it again, but there was a huge reduction in power going from DDR to DDR2. Micron has some decent data sheets.
And DDR2 will get cheaper as volumes go up.
How sweet is it that we're talking about using technologies as they emerge rather than having to hunt and scrounge for the obsolete stuff?!
Don't expect a 1.5GHz bus in a laptop, obviously...
Might, but I'd still think late summer... If this is accurate they're making 3 major heat improvements:
Shrinking the CPU
Shrinking the system controller
Moving to DDR2
I did the calculations a while back and don't care to do it again, but there was a huge reduction in power going from DDR to DDR2. Micron has some decent data sheets.
And DDR2 will get cheaper as volumes go up.
How sweet is it that we're talking about using technologies as they emerge rather than having to hunt and scrounge for the obsolete stuff?!
Don't expect a 1.5GHz bus in a laptop, obviously...
Analog Kid
macrumors G3
I hope we don't need L3 anytime soon... It's hot and expensive.
AMD isn't the opponent-- they're an ally. The greater the diversity of systems, the more likely people are to be shaken from their stupor and consider new options.
Before there were two choices Wintel and Apple, and it was easy to believe that going with the machine that had greatest market share was the right choice-- because everyone else was making it. Now Intel looks weaker than it has in a very long time... The hounds are closing in-- testing its strength.
If it was only Intel and Apple, people might take the "there's no need for 64bits on the desktop" as the wisdom of a market leader. Now, people view it as what Intel has to say since they don't have a reasonable 64bit offering. At least not yet.
I'm for anything that mars the monolith.
Analog Kid
macrumors G3
Analog Kid
macrumors G3
As long as we're basing our reality on the rumor:
The system controller is capable of 1.5GHz, but that doesn't mean the first units will use that capability. The first ones could just as easily run at 2.4GHz...
Was PowerTune one of the 980 features? Nothing in the article says this is a 970...
The system controller is capable of 1.5GHz, but that doesn't mean the first units will use that capability. The first ones could just as easily run at 2.4GHz...
Was PowerTune one of the 980 features? Nothing in the article says this is a 970...
I buy fpga chips from Xilinx that are produced in Fishkill as well. They are 90nm and very fast. The Fishkill plant is very flexible and has a lot of capacity. This fab will not hold back faster Macs, the designs will but not the plant.
Dont Hurt Me
macrumors 603
look at the specs of the g5 and bus then look at g4 and bus??The Gap or rather crater is getting bigger and bigger, 1.5 ghz bus vs .167 ghz? 2 or perhaps 2.5 ghz vs 1.25 or is it 1.3. a single 1.6 g5 is matching a dual 1.42 g4. lets hope all this G5 stuff comes to pass soon and Apple gets those G4's out of most the line up. If i had to say in the last 2 years what has hurt apple sales more then any single thing it would have to be the lame G4,say what you want but in the computer world speed sells. the sooner the 90 nm g5's get into production the better.😎
The 130nm 970 can still be clocked to 2.5 Ghz you know, they'll run damned hot but it's still perfectly doable.
Plus it'll make a nice and smooth transition from 2Ghz to 2.5Gz to 3,0Ghz(1,5Ghz fsb).