Weigh in - 3GHz-When, & what Processor will it be?
It's a little early knowing not a thing as to what processors Apple, and IBM, are planning for the next revision, but Do we think Apple is going to Sit back, and jerk off for another year, or will there be a revision to the PowerMac line at MWSF, MWNY, Siggraph, Seybold, you get the idea.
Secondly. I am of the opinion that we will not see a dual core processor in the PowerMac in the next revision, but sometime thereafter, like WWDC 2005 (actually sometime after that).
So, back to topic. 3GHz When, and What processor will it be that takes the IBM processors for PowerMacs beyond the 3GHz Mark?
Secondly. I am of the opinion that we will not see a dual core processor in the PowerMac in the next revision, but sometime thereafter, like WWDC 2005 (actually sometime after that).
So, back to topic. 3GHz When, and What processor will it be that takes the IBM processors for PowerMacs beyond the 3GHz Mark?
Comments
Might see dual core chips before that happens.
Even if they get to 3 GHZ, what's next, 3.15 then 3.2 then...etc, waste of time...
Originally posted by onlooker
So, back to topic. 3GHz When, and What processor will it be that takes the IBM processors for PowerMacs beyond the 3GHz Mark?
I don't think that frequency is the issue anymore. Taking PowerMacs beyond 3GHz would imply deepening the pipelines, in the current implementation of the G5 (970fx), and I'm sure it is out of question for IBM.
Maybe the 975 will scale better. Maybe we'll have to wait for 65nm process to be used. Maybe we'll never have 3GHz in a Mac. And that wouldn't be such a big deal! Frequency used to be a rather good indicator in the time when CPU architecture was extremely simple (think Moto 68000 to, say, 68030, and Intel 8080 to 80386), but now, the difference lies within the number of ALUs, the FPU performances, cache management, branch prediction, out of order execution performances, etc... that's why a 1.5GHz Itanium 2 performs way better than a 3.4GHz P4.
3GHz is only a 20% increase over what the 970FX is at now. If IBM doesn't reach it with the next speed bump in roughly 6-8 months, then it'll be the one after that, at around next year's WWDC.
Originally posted by Gamblor
Jeez, guys, Apple misses one milestone, and you'd think the world was ending to listen to you...
3GHz is only a 20% increase over what the 970FX is at now. If IBM doesn't reach it with the next speed bump in roughly 6-8 months, then it'll be the one after that, at around next year's WWDC.
This is not just Apple... the whole CPU industry seems to be running out of tricks to take the frequency up. Even Intel, a true "MHz machine", can't get its P4 further than 3.4GHz (maybe it will, but then, no need to use radiators during winter anymore).
That's not the end of the world. Only a slight change in how to market the chips.
Originally posted by The One to Rescue
that's why a 1.5GHz Itanium 2 performs way better than a 3.4GHz P4.
Only at FP. The Pentium is significantly faster at int.
Originally posted by Existence
Only at FP. The Pentium is significantly faster at int.
(little confused)
Cheers,
C.
1. Increase the pipes just a wee bit to get more clock.
2. Slap on an ondie memory controller.
3 Add SMT
This means a larger more expensive chip for now. I don't think dual core is coming until late 2005 early 2006.
I figure the improved memory performance going to an ondie controller should ameliorate the lower IPC because of the increased pipes.
Question though. Apple is rumored to be keen on Hypertransport 2.0 (architosh.com). Just where would they use this to realize the biggest performance gains. Chip to Chip communication?
Measuring processor performance in Hz is a bit outdated, yet it is still recognized as a relative speed guide for most users.
Be it as it may, a 3GHz G5 PPC would be an impressive processor if it scaled equally in performance as it did in Hz. Which I think is was one of the intentions that Apple, and IBM were trying to get across last year when they made the announcement of 3GHz as a milestone in a years time. The other being to prove that IBM was a highly formidable player in processor technology, and in the future high performance computers.
Falling short is seemingly much worse for IBM right now in that respect than it is for Apple IMO.
I am still interested in seeing how this processor performs at 2.5GHz, but I read to expect a 5% increase in performance across the board which is not all that impressive. 15 to 20% would have been my guess, but I am no expert.
As for my opinion on seeing a 3GHz processor from these two.
I am of 2 schools of thought not including a 4 way processor setup which I always thought was unlikely anyway.
1st being that the 90 nm process on the 970FX is turning out 3GHz processors, but not in abundance, (or soon will), but IBM is still working out the kinks, and when that happens there will be a quiet update when they have it. (4 to 6 months)
The 2nd is that we'll have to wait for this supposed dual core processor.
Unless IBM is already turning them out nobody seems to think this processor will be ready by the time Apple needs it. I don't think anyone is going to wait another 9, let alone 12 months for a PowerMac update.
Frankly I think Apple is in bigger sh*t right now than they were during the G4 crisis. The reason for this is not that the processor is that bad, but it's the wait, and what we just waited for. I think most decided that one long wait was more than enough. Mac users already went through one processor fiasco with the G4 when the updates were few, and far between, and were always falling way short of expectations. PowerBooks, iBooks, iMacs, and eMacs are not in as big of a world of hurt from loyal Mac customers. It's the PowerMac that looks lackluster to those of us who buy the PowerMac.
Chances are that Apple is working on updating the PowerMac in some ways sooner rather than later. Probably as many ways as possible. We'll see.
I hope they do.
Even though they lost my sale to the AMD PC this round if they can get it together I'll gladly come back for the next.
Not that I wouldn't like to see them earlier..
An (oft-repeated) aside on the "Mhz is meaningless" debate - Mhz is often meaningful in comparing revisions of the same processor. If a 2Ghz 970 w/1Ghz bus ~= 3ghz P4 (800mhz bus)(dunno if it actually *does*) then a 2.5Ghz 970 w/1.25Ghz bus > 3.4Ghz P4 w/ 800Mhz bus.
Please do not interpret this post to contain any facts.
Originally posted by Outsider
... I have no idea if IBM wants to invest in newer versions of the 970 by adding more cache and other refinements.
That's easy. Steve will commit to take 400000 per year and IBM will invest in a new version. Or Steve doesn't want it - IBM doesn't build it.
I'm not as confident as Outsider that the 970FX can do 3GHZ, but considering the POWER5 is planned to run 2.5GHz on the same process, why not?
Originally posted by Onlooker:
I think the 970FX will reach 3GHz, max, at the end of it's life cycle. And we'll see it in November of this year. The only reason why we may NOT see a 3GHz 970FX is if they retool it (970GX with 1MB cache for example) or if the next generation G5 based on the POWER5 is ready to be released by years end.
Once more your expectations are way too ambitious and you're setting yourself up for disappointment yet again. The dual G5s aren't even going to be in customers' hands until the beginning of August. I can see a possible update in January to something like ~2.7 Ghz along with some other system upgrades but then that would push the following refresh to the summer.
I suspect though that 3 Ghz is a milestone they'll want to achieve with the G5 and will likely wait until they can hit that target before refreshing the line. And I don't see that happening until next spring, especially from what the Apple reps have been saying and the measures they're already taking with cooling them down.
Cheers,
C.
Originally posted by Onlooker:
I think the 970FX will reach 3GHz, max, at the end of it's life cycle. And we'll see it in November of this year. The only reason why we may NOT see a 3GHz 970FX is if they retool it (970GX with 1MB cache for example) or if the next generation G5 based on the POWER5 is ready to be released by years end.
BTW. concord I did not write that outsider did.
Originally posted by Concord
I can see a possible update in January to something like ~2.7 Ghz along with some other system upgrades but then that would push the following refresh to the summer.
I suspect though that 3 Ghz is a milestone they'll want to achieve with the G5 and will likely wait until they can hit that target before refreshing the line. And I don't see that happening until next spring, especially from what the Apple reps have been saying and the measures they're already taking with cooling them down.
Cheers,
C.
Above = 3GHz by March - May. ???
--------------------------------
If this turns out to be the correct time table, and performance equation that is correct I think once again Apple will be blowing in perspective PowerMac sales. It's always just short of what is expected from them vs. the competition. If this is the case I think they both should consider skipping right ahead to another processor. Possibly the rumored dual core variant.
Actually that seems drastic, but is always a possibility. (probably the better idea to get a jump on the competition though)
What they ought to do is put their heads together right now, and remember, and think about what works. Doubling Altivec throughput, or output (or what ever is the correct phrasing) is something Mac users can latch on to. Another way to improve on this processors performance is what AMD has done, and is seeing fantastic performance with the Operon 64 that was similar to how the G4 worked with memory on die, or cache controller (I forget exactly what it is, but one of you will probably know what I'm talking about).
A big deal has been made about the actual clock speed, but if IBM were to add these, or other similar processor enhancments, and improvements that Apple, and 3rd parties can compare vs. The 2.0 GHz G5's, and other processors that show considerable performance gains most users wouldn't care about the actual clock speed. Remember Apple started the whole "down with the MHz Myth" Theory, and AMD has certainly helped prove it on the x86 side. 3GHz may have been a milestone, but a few things like using techniques/technology that AMD is finding successful, and stuff that is old hat for Mac users in doubling Altivec performance would not have been frowned upon.
Being that AMD is probably going to jump through the stratosphere in performance with their next processor, and it seems that Intel will just try to gain MHz. (weather it help intel or not in actual performance who knows)
With that in mind again I am of the belief that they should skip ahead to edge in on the competitions performance lead, go with a new processor such as the dual core Power 5 variant, and think about increasing Altivec, and memory controller in die with it. (terminology may be wrong, but you get my meaning)
Of course it's all easily said, and done in a forum. How these things progress in a real processor design, and manufacturing facility is another matter.
My 2¢
Originally posted by Concord
Once more your expectations are way too ambitious and you're setting yourself up for disappointment yet again. The dual G5s aren't even going to be in customers' hands until the beginning of August. I can see a possible update in January to something like ~2.7 Ghz along with some other system upgrades but then that would push the following refresh to the summer.
I suspect though that 3 Ghz is a milestone they'll want to achieve with the G5 and will likely wait until they can hit that target before refreshing the line. And I don't see that happening until next spring, especially from what the Apple reps have been saying and the measures they're already taking with cooling them down.
Cheers,
C.
My expectations are not high. I would be surprised if the G5 or G5 derivative does not get to 3GHz by the end of the year. If not 3GHz then close.
Another option is a 970 chip with some type of spreading out of the circuit layout so the heat concentration isn't so high making for, not a cooler chip, but a wider spread hot chip. I know this involves making inefficient use of the circuit layout but it may allow for all the engineering effort put into the 970 to be used to it's fullest extent.
Originally posted by onlooker:
Above = 3GHz by March - May. ???
--------------------------------
If this turns out to be the correct time table, and performance equation that is correct I think once again Apple will be blowing in perspective PowerMac sales.
Sorry for confusing your post with Outsider's btw... I did want to touch upon this briefly by pointing out that CPU speed won't have as big an impact (positively or negatively) on PM sales as you might think. There are much larger industry forces at work that influence PM sales.
Much more than processor speed and high end graphics cards, what Apple really needs to do is find a way to... generate *excitement* in their computers again. Performance alone isn't going to cut it... most people have all the performance they need for day to day computing.
Being that AMD is probably going to jump through the stratosphere in performance with their next processor, and it seems that Intel will just try to gain MHz. (weather it help intel or not in actual performance who knows)
Intel is moving to lower power, more efficient dual core CPUs in or around the same time AMD will (2H05). I think the writing is on the wall and that IBM will have to move over to dual core CPUs as well. While I would love to see IBM beat AMD and Intel to the punch here... it's a little to tell one way or another.
Originally posted by Outsider:
My expectations are not high. I would be surprised if the G5 or G5 derivative does not get to 3GHz by the end of the year. If not 3GHz then close.
Well I can't say that wouldn't be just peachy but... sorry Outsider - it just ain't gonna happen.
Cheers,
C.