More 90nm PowerPC G5 Details?

[daveL]

Re: Re: Re: Don't expect a G5 PowerBook

Originally posted by greenstork
Actually the article says that 12 W is at the lowest possible clock speed on the chip. He is extrapolating that this is at idle for a scalable chip speed, and correctly so.

I disagree with the extrapolation. Just my opinion.

 

[daveL]

Re: Re: Re: Don't expect a G5 PowerBook

Originally posted by illumin8
Ok, but do you agree that the chances of seeing a G5 PowerBook in January are pretty slim, based on the numbers you've seen? My guess would be G5 PowerBook (rev. A) first seen next September/October at the earliest. Most likely in 1 year from January.

Sure, I don't expect a G5 PB in January. I agree it may be the latter part of 04 before they appear. The only thing that might accelerate that timeframe would be some new cooling technology, but then you still have the battery life issue.

 

[murak]

"-Oboy! Wet dream again, a 12" PB with a 1.6 ghz G5, aw man that would be so damn sweet... Its never to soon for a PB rumor."
Im such a geek =)

 

[x86isslow]

we're likely to see a revision B of the current powerbooks in January: most likely with faster rpm harddrives, and hopefully none of the screen and 'gap/maw' of the lid issues that this A-round saw.
hopefully apple will keep rev A of a g5 powerbook internal, seeing the nuisance of 10.2.8, panther, and the revA alu pwrbks.
i look forward to a second gen g5 powerbook
i'm betting that said powerbooks will hit the market when i have the $$ to afford them- about a year and a half from now sounds about right. (wwdc 05)

 

[sharky2313]

these new speeds make the Moto g4 look really really slow. Apple has to feel the need to get the powerbooks jump started.

I understand that people like the current batch of books but the disparity between the powermacs and the books looks really bad to me.

 

[Tim Flynn]

Re: Re: Re: Re: Don't expect a G5 PowerBook

Originally posted by daveL
I disagree with the extrapolation. Just my opinion.

I agree with daveL.
The first mention is 96 watts for a 970 130nm at 2.6 Ghz. This not idle, that is flatout (The definition of flatout is ..., but very little agreement on that )
The 68 watts at 2.6Ghz on 90nm is reasonable for a processor running flatout. The 970 (130nm) runs I believe 46 watts at 2.0 Ghz.
So 12 watts at 1.2Ghz for a 970FX (90nm), is again flatout. The 970FX, I think will make it into a laptop.

I have 1.8 G5, which run maybe 30-40 watts flatout, not at idle.

 

[jettredmont]

Originally posted by QCassidy352
I don't understand... I thought that the bus was intrinsically linked to the processor in the G5. How could there be a bus increase to 1.5 Ghz unless the processor hit 3 ghz?

You are correct: the CPU:FSB ratio on the G5 is set to one of 2:1, 3:1, 4:1, or a few higher multiples. You can't have a 2.6GHz G5 running on a 1.5GHz FSB.

HOWEVER, The new system controllers are purported to be stable at *UP TO* 1.5GHz. Which is to say, they won't need to be replaced (and thus the CPU:FSB multiplier won't need to be changed) until the G5 goes abot 3.0GHz.

Of course, on a 2.6GHz G5 they will run at 1.3GHz (or 866MHz , or 650MHz ...)

 

[k2k koos]

Seems to get real interesting guys...

 

[jettredmont]

Originally posted by greenstork
I'm afriad you'll probably lose that bet. Intel would only have to push their processors .8 GHz to reach 4Ghz while Apple would have to jump 2 GHz and at the current rate of speed bumps, could be expected no earlier than late 2005.

Umm, actually, the "current rate" of speed bumps on IBM is 50% raw clock rate in one year. In mid-2005, one would expect IBM to be at 4.5GHz from a straight linear extrapolation, which would put 4GHz around February/March of 2005.

That's hardly "late 2005". Of course, it's relying on the always-faulty linear extrapolation, which can hardly be called an accurate predictor, but that's what you specified in your original comment.

"Late 2004", however, Intel claims to want a 4GHz CPU in production, so based on public claims Intel should hit 4GHz well before IBM.

As you said, however, it has yet to be seen what the real-world performance of those beasties will be. Suffice to say that, right now, IBM has an as-fast-or-faster overall chip than Intel, and that, right now, IBM's chip is advancing at a significantly faster rate than Intel's.

As you and others have said, I think the real speed crown competition for the next period of time is AMD, not Intel. Intel can't be ignored, as its marketing and warchest would let it survive even an obvious performance whipping from AMD/IBM, but I think that for the first time in many years we can honestly say Intel's performance is worse than an Apple machine's.

 

[jettredmont]

Awesome increases ...

These sound like some real heart-stopping (especially within certain executive offices) advances for the G5 line. Let's hope they come to be!

For myself, I'm ready to go home and enjoy the last month of Total World Supremacy with my 1/1100th of a supercomputer ...

 

[greenstork]

Originally posted by jettredmont
Umm, actually, the "current rate" of speed bumps on IBM is 50% raw clock rate in one year. In mid-2005, one would expect IBM to be at 4.5GHz from a straight linear extrapolation, which would put 4GHz around February/March of 2005.

That's hardly "late 2005". Of course, it's relying on the always-faulty linear extrapolation, which can hardly be called an accurate predictor, but that's what you specified in your original comment.

I hate to be picky because you were right to correct me but a 50% year over year increase is not linear, it's parabolic. Plot some hypothetical numbers on a graph and you would see the same thing.

I was thinking of it in terms of a linear progression which is 1 GHz per year making it mid to late 2005 until IBM reaches 4 GHz. As you pointed out, linear progressions are wrought with inaccuracy.

While I'll probably eat my words and be proven wrong by Moore's Law, I would venture to guess that IBM chip speed will continue to increase at 1 GHz a year for a while and not at the 50% a year clip. That said, I stand by my original statement that IBM won't be cranking out a 4 GHz chip until mid to late 2005. Sure, it's just speculation, call it a hunch.

 

[jouster]

Originally posted by jettredmont
Umm, actually, the "current rate" of speed bumps on IBM is 50% raw clock rate in one year.

Unfortunately not. There is no current rate because there have not been any speed bumps yet.

The "hoped for and promised by Jobs rate" is 50% in one year but heh, plenty can happen in either direction that we know nothing about.

Remember, 400--->500--->450?

 

[jettredmont]

Re: Re: Correct Me IF I Am Wrong

Originally posted by illumin8

I did have a question that maybe someone could answer though: Is DDR-2 533 the same as PC4200 memory? PC4200 memory (533 mhz.) is available in a 1GB kit (2x 512MB sticks) for only $168 according to Pricewatch. That's not too much more than PC3200, which is $124 for a 1GB kit.

Yes, PC4200 and DDR-533 are one and the same.

I'm not sure I'd ust the memory at the prices you quoted, though; Macs have a history of being quite fussy with memory quality, G5s as well. I wouldn't use the $62/card 512MB PC3200 memory in my machine (I'd rather pay an extra $30/card and know that I'm not going to end up with system instabilities and such) unless it, for some reason, came from a really reputable manufacturer with a good warrantee (and a good PR policy on accepting bad memory returns instead of blaming it on everybody else and their dog).

I'm just generally suspicious of memory coming at a 33% discount from what the top brands are selling ...

 

[legion]

Re: Re: Correct Me IF I Am Wrong

Originally posted by illumin8
CAS 3 just gives you slightly higher latency. One thing you should be aware of is that unless you buy an OEM Intel 875P motherboard and build your own system, you won't be able to take advantage of CAS 2 or 2.5 memory. Every system manufacturer that has implemented dual channel DDR400 memory so far has locked the memory controller at CAS 3 just to ensure compatibility with all 3rd party memory.

I know this because I have a Dell Dimension 8300 (Intel 875P chipset) and it performs exactly the same in Sandra's memory benchmarks no matter what memory I put in it.

As far as the G5 goes, I know this because on barefeats, he benchmarked both CAS 2 and CAS 3 memory and got the exact same results.

I did have a question that maybe someone could answer though: Is DDR-2 533 the same as PC4200 memory? PC4200 memory (533 mhz.) is available in a 1GB kit (2x 512MB sticks) for only $168 according to Pricewatch. That's not too much more than PC3200, which is $124 for a 1GB kit.

Not true. IBM gives you 2.5CL default and now 2.0CL is available from the bios (and it works!)

 

[Mav451]

interesting...but wuz bout say the p4800? or the IC7? do those also lock it at cl3?

I remember b/c the early 875 chipsets had major dual channel problems--the slots being extremely picky of the ram that was inserted.

 

[Freg3000]

Originally posted by jouster


Remember, 400--->500--->450?

Don't remind me.

 

[JoeRadar]

Re: Re: Re: Don't expect a G5 PowerBook

Originally posted by illumin8
Ok, but do you agree that the chances of seeing a G5 PowerBook in January are pretty slim, based on the numbers you've seen? My guess would be G5 PowerBook (rev. A) first seen next September/October at the earliest.

This seems reasonable. Apple will probably consume most of the available 970FX (or whatever will be the name) for the PowerMac line -- the line for which performance is critical. Next will probably come the xServes. Finally, the PowerBooks will bring up the 3rd spot.

Next fall should be another heady time in the rumor market -- PowerMacs based on the 980 running at 3GHz, PowerBooks on the 970 running around 2GHz, maybe even the next major OS release (although I could use a break on the major OS release rates).

 

[dongmin]

Re: Re: Re: Re: Don't expect a G5 PowerBook

Originally posted by daveL
I disagree with the extrapolation. Just my opinion.

guys guys calm down. the rumors are so vague right now that you can really extrapolate just about anything regarding G5 powerbooks.

until these rumors come flat out and say that the 970fx at such-and-such clockspeed will make it into the next revision of PowerBooks, it's too far away to make 'educated' guesses. But for reference:

early estimates by IBM:
42W @ 1.8 ghz, 1.3v, 130nm
19W @ 1.2 ghz, 1.1v, 130nm
(I read somewhere that these estimates are lower than actual results)

Appleinsider source:
96W @ 2.5 ghz, 1.5v???, 130nm
62W @ 2.5 ghz, ??? voltage, 90nm (35% reduction in power)
12W @ ??? ghz, ??? voltage, 90nm

Extrapolation (based on the logic that going to 90nm saves about 30%-35% in power)

12W @ 1.2 ghz, 90nm
27W @ 1.8 ghz, 90nm

I generally consider these numbers to be best-case scenarios. My guess is that a 1.8ghz G5, even at 90nm, is too hot for a laptop. 1.6ghz might be possible. But then you have to ask yourself, where is Apple gonna go after this? Will the 980 bring even better performance-to-power ratios? Is Apple and IBM working on a separate mobile 9xx chip for the laptops (like it was discussed regarding IBM's involvement in game consoles)? Or will we have to wait for 65nm process to come on line?

 

[ITR 81]

Found out some interesting news today.

Today I was called by AppleCare rep because my PB is about to end it's 1 yr warranty in about another week and he asked if I like to get AppleCare. When I told him not really because I've had no problems with it and I plan on getting a new G5 PB when they come out next yr he said "yeah alot people here have been talking about the new G5 Mac processors." When I asked you mean faster speeds? He was like oh um nah I was talking about the current G5.

I think my AppleCare rep has heard a few things from inside of Apple and kinda of let it slip out new processors were coming for the G5 in Jan.

He did tell me MacWorld is going to be huge and I should go and get passes now.

To me this pretty much says the G5 is getting speed bumps in Jan.

I did however found it nice Apple calls you to see if you like to extend your warranty..never had that happen while owning any PC.

 

[Daschund]

Originally posted by Wonder Boy
haha. actually they'd buy me the computer just so i would stay!

Damn! Than you don't even have to think about it! Just choose the damn model you want!! LOL

Daschund

 

[johnnowak]

Will these chips offer hyperthreading or not?

 

[visor]

Re: Re: Re: Re: Re: Don't expect a G5 PowerBook

Originally posted by Tim Flynn
I agree with daveL.
The first mention is 96 watts for a 970 130nm at 2.6 Ghz. This not idle, that is flatout (The definition of flatout is ..., but very little agreement on that )
The 68 watts at 2.6Ghz on 90nm is reasonable for a processor running flatout. The 970 (130nm) runs I believe 46 watts at 2.0 Ghz.
So 12 watts at 1.2Ghz for a 970FX (90nm), is again flatout. The 970FX, I think will make it into a laptop.

I have 1.8 G5, which run maybe 30-40 watts flatout, not at idle.

I've measured my Ibooks power consumption.
Idle: 8-10 watt, normal use 14-20 watt, full blazing and charging: 40 watt.

 

[daveL]

Originally posted by johnnowak
Will these chips offer hyperthreading or not?

No, not till the 980 comes out later next year.

 

[Sun Baked]

Re: Correct Me IF I Am Wrong

Originally posted by willmg
With a 1.25+GHz bus what ram would equate to full bus utilization? DDR 533 is 1066 efective speed right, so I guess with dual channel that gets around 2132MHz effective? So its fast enough but isnt DDR 533 really expensive and isn't most of it is still CAS 3 with high timings, which actually performs worse than slightly slower memory with faster latency? Anyone know anything about DDR2/3 availibilty by Febuary? I thought it was just recently that standards were even set for both of these formats for ram.

If this is close to http://hardware.earthweb.com/chips/print.php/1562021

Still haven't found a decent article on the current JEDEC DDR-II standard, but I haven't been looking hard either.

 

[nek]

Re: Re: Re: Re: Re: Don't expect a G5 PowerBook

Originally posted by dongmin
12W @ 1.2 ghz, 90nm
27W @ 1.8 ghz, 90nm

I generally consider these numbers to be best-case scenarios. My guess is that a 1.8ghz G5, even at 90nm, is too hot for a laptop. 1.6ghz might be possible. But then you have to ask yourself, where is Apple gonna go after this?

Perhaps if the new 970FX and 980 make use of "PowerTune" to allow the PowerBook to run at a slower frequency and power, then 1.8GHz would be OK. They would be able to use less of the battery and run cooler when not plugged in.

But like you said, where do they go from there? I guess the PowerBooks don't get updated very often anyway, just putting a G5 in there at any GHz would be a good start. Then they can increase the speed when cooler processors become available.