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.
PowerPC 970 at 90nm and the Power5
- Thread starter MacRumors
- Start date
- Sort by reaction score
IMO, Apple will not hesitate to jump to G6 to make some marketing noise. Going to a PowerMac G6 by Fall of 2004 to incorporate the 980 would be the right thing to do; it would be great for marketing and accurate. IBM is stating that the Power5 is 4X faster than a Power4 though the frequency increases less than 25% and it stays at 130NM process. If we get 2X on a 980 that would be huge and worth the name change.
Such a move would permanently erase the past reputation and reflect Apple's leading position. It would also encourage G4 PowerMac users to upgrade, as now they are two generations behind! Apple's biggest problem has been getting 1st generation G4 owners to upgrade. They made the damn things too well and the upgrades didn't give enough bang for the buck. Now G5's are helping and the G6 will clinch Apple's crown.
Such a move would permanently erase the past reputation and reflect Apple's leading position. It would also encourage G4 PowerMac users to upgrade, as now they are two generations behind! Apple's biggest problem has been getting 1st generation G4 owners to upgrade. They made the damn things too well and the upgrades didn't give enough bang for the buck. Now G5's are helping and the G6 will clinch Apple's crown.
This sums up the sentiment of many posters in this thread. IBM is not only pushing forward with advances in chip design and fabrication, but also paying due attention to compiler technology. Without a compiler that can properly exploit the features of the hardware, a lot of the promised gain is lost.
Optimizing compilers are an important part of the solution and should not be overlooked. If comments in the following thread are true (URL below), then IBM's XL C/C++ compiler, currently in Beta, is delivering 200-300% performance gains on EXISTING processors. Even a 3 GHz POWER-5 will not be 200-300% faster than today's 2 GHz PPC 970! It will take a LOT of extra gigahertz, cache size, front-side bus speed, memory speed, etc. to return those kinds of performance gains.
https://forums.macrumors.com/showthread.php?s=&threadid=47110&perpage=25&pagenumber=4
Analog Kid
macrumors G3
To give the same answer with more words...
Yes, the relationship to clock frequency is pretty much linear with everything else held constant:
P=(.5*f*C*V^2)+Pq
(sorry, someone asked for a techie answer...)
P=power
f=clock frequency
C=capacitance, both transistor gate capacitance and routing parasitics. More transistors means higher C. Smaller process or better dielectric means lower C for transistors but typically higher C for routing.
V=operating voltage, smaller process typically allows lower voltage operation. Lower clock speeds allow lower voltage operation.
Pq=static power (clock stopped). This is mostly leakage, and goes up appreciably at very small processes. At 90nm, up to half the power can be static power.
So, if you're doing everything you can to cut power you'd reduce both the clock and the voltage for a non-linear power reduction, bearing in mind that voltage can only be reduced slightly.
As an example of the other factors, Apple Insider quotes: "According to preliminary tests, a 2.5GHz Power PC 970 G5 processor based on the 130nm process consumes 96 watts, while 2.5GHz G5 built around the 90nm process pulls a more manageable 62 watts."
Due only to the process shrink (presumably) the power will likely have seen a reduction in C and V, with an increase in Pq giving a net 30% total reduction at the same clock speed.
Trying to answer the original question is complicated by the fact that folks here can't agree on the power of the current G5... 🙂
Analog Kid
macrumors G3
Not sure why anyone would be disappointed with a .13µm process... It's the best anyone has for a processor right now and was smaller than the G4s at that time.
Anyone know when Prescott is due to hit the shelves? Would an early '04 release make the 90nm 970 the first shipping 90nm processor?
Analog Kid
macrumors G3
Where would we send these requests, exactly? I can never find an appropriate email address...
I'd love to see some of the ASIC development EDA tools move to OS X....
11 negatives????
Just wondering how any of this news could be considered Negative? For the first time in ages, there seems to be nothing but good news coming from the Apple/IBM axis, certainly
Who does this? Can they not read properly or something?
Yeesh!
Just wondering how any of this news could be considered Negative? For the first time in ages, there seems to be nothing but good news coming from the Apple/IBM axis, certainly
Who does this? Can they not read properly or something?
Yeesh!
Of course he is not concerned about the process technology itself, but the operating specifications of the chip he is buying. The PPC970 built from a 130nm process consumes a lot of power. This has led to a large case that is itself designed to be an efficient "exhaust" system. Despite the size of the chassis, there is very little room inside for expansion. You are limited to 2 hard drive bays, one optical bay, and 3 PCI/PCI-X slots. Now contemplate a laptop version of the PPC970...
So while the process technology itself is not the issue, my guess (and again this is speculation) is that Steve is not someone who tolerates design "imperfections". Chips that might require liquid cooling or other extreme designs to maintain safe temperatures are not ideal for the mass market.
The challenge of the semiconductor industry is to keep finding ways to increase performance while keeping power consumption in check. Transition to a copper process was a significant step forward (I believe IBM pioneered copper dual damascene) while today the search for low-k and high-k dielectrics is raging. Low-k dielectrics will allow oxide or metal-insulating layers to be made thinner and reduce the height of vias (which in turn reduces the signal path). High-k dielectrics will allow fabrication of thicker gate junctions without compromising switching properties, while reducing leakage and wasted energy by a significant margin (100 to 1 for example).
Last week Intel announced a breakthrough in their search for the holy grail of high-k dielectrics. Of course they did not divulge any details. This type of breakthrough is essential, particularly for the 90- and 45-nm technology nodes.
Briefly, then, my guess is that Steve was enamored over the performance of the PPC 970, but disappointed in the fact that it was not being produced with more favorable operating specifications. A perfectionist like Steve might have wanted more creative freedom over the design of the case, fewer fans, and [maybe] more internal room for expansion.
Analog Kid
macrumors G3
Ok, that makes more sense... Personally, though, I think they lovingly and eagerly made all those compromises when they saw the IBM roadmap.
The decisions Apple made were not about accommodating the G5, if they wanted more space for more components, they would have simply used a fan directly on the processor to minimize the heat sink size and create room for more bays.
The current design allows Apple to quickly assemble and build G5's with minimal labor. At the same time it creates a beautiful wire free interior.
The current design allows Apple to quickly assemble and build G5's with minimal labor. At the same time it creates a beautiful wire free interior.
Which decisions were not about accommodating the G5? I don't mean this sarcastically, but it does appear that the case and airflow designs are heavily influenced by the heat dissipation of the G5 and related chipsets.
Rumors are that the case for the PPC 980-based PowerMac will be redesigned (after all, this is a rumors website). One can only speculate about the changes, but Apple will try to make the best use of space. Filling the chassis with fans and heatsinks is not a good use of space for the customer -- it is a necessary use of space for the base electronics.
Even the G4 case accommodates easy upgrades. One side flips out and upgrades are more or less a snap.
The wire free interior of the G5 is designed to minimize restrictions to airflow. Apple would not spend money unnecessarily to beautify the interior.
g5 2gh HEAT?
Hi,
I just bought a gh 2gh. Does it run so hot that it will shorten the life of the computer/processor compared to a cooler chip at the 90nm process?
I haven't turned on the computer at all yet.
Thanks,
Ian
Hi,
I just bought a gh 2gh. Does it run so hot that it will shorten the life of the computer/processor compared to a cooler chip at the 90nm process?
I haven't turned on the computer at all yet.
Thanks,
Ian
No, as long as the processor remains within its safe temperature window you should not be concerned about this. You raise another issue though: solid state devices will break down eventually with use, but much much before that happens your computer will have already have been melted down and recycled or, less wishfully, found its way into a landfill.
But if temperatures exceed safe margins and assuming the heat-protection circuit fails to shutdown the motherboard, you can expect this (before and after pictures):
http://www6.tomshardware.com/cpu/20010917/heatvideo-01.html
$ 1k to read the article from Microprocessor Watch??? I cant afford that. Does anyone have a copy of that article? I wanna read