IBM has today presented various new versions of their G5 processor at the Power Everywhere Forum in Japan. Firstly, it introduced the much-anticipated PowerPC 970MP, the dual-core version of the G5. In addition, they also announced 3 low-power G5s, ranging from 1.2Ghz at 13W to 1.6Ghz at 16W. These processors will most likely find their way into Apple’s Macs.
Where it the 3ghz computer promised last year to Apple?
Having dual G5/3000Mhz for mobiles would be a bit too much, no? Also, what with the Mhz Myth? From my point of view, Apple move is all about cash and nothing about technology.
Apple can switch even to Z80, IBM and Freescale couldn’t care less.
Sorry but the current powerbook G4 technology doesn’t come close to Sonoma, much less the upcoming Napa (2nd and 3rd generation Centrino).
“Apple move is all about cash and nothing about technology. ”
Oh, and YES! They want to make more money, what did you expect? How can they achieve that? By having more competitive hardware! That’s debatable from the PowerMac’s point of view but I think it’s clear they will be much better off with Intel when it comes to the Laptops.
,,Oh, and YES! They want to make more money, what did you expect?”
I’m have nothing against that, we’re all working for cash. 😉 But I’ve mentioned it because Apple would love to claim that there’s more technical merits in that.
“I’m have nothing against that, we’re all working for cash. 😉 But I’ve mentioned it because Apple would love to claim that there’s more technical merits in that.”
But there is! There is! Yonah and its family
I really can’t wait for Yonah-based Mac laptops.
And can you imagine what Apple can do with Ultra-Low Voltage Pentium M processors at their disposal? How about an ultra thin laptop? A tablet mac, you name it! All of this is technology IBM couldn’t provide to them.
In the same time that the 970 made it from 2.0GHz to 2.5GHz, the P4 made it from what – 3.2GHz to 3.7GHz? AMD is mentioning the clock speed in the fine print for quite a while.
The days of MHz counting are over. CPU battles are fought on bandwidth, multicore or power consumption. Workstation and Server CPUs (POWER, MIPS, Itanium or SPARC) have done this for quite a while, now it’s the consumer grade hardware that’s following.
“The days of MHz counting are over. CPU battles are fought on bandwidth, multicore or power consumption”
Well, I hope we are not at the dead end for what concerns clock speeds and that we will continue to count MHz and GHz…
Don’t forget that MHz is not a mith, computational efficience is the amount of calculations possible at each cycle and more MHz means more cycle so are the straightest way to more computing power.
A way with may tradeoffs, i.e. power consumption and noise in dissipating the thermal output, but MHz demistification doesn’t mean at all that MHz aren’t useful, increase the MHz and you will increase computing power for any cpu intensive task, while increasing power by increasing parallelism is nearly useless for the plethora of application that are mathematically not parallelizable, increasing bandwith will be useful only for memory hungry applications and useless for cpu intensive tasks, increasing cache is useful only if the task must keep a fast access to limited amount of data that will fit into the cache but tush useless for cpu intensive tasks.
A powerful system is a system with as few bottlenecks as possible, recent tendences are about “fixing” those bottlenecks (dual core to not have the system stalling if some task goes crazy, bigger cache to limit access to ram, faster and twin rams and faster fsb to limit performance loss for cache misses etc… disk… don’t even talk about it, disks ARE the bottleneck of today’s computing) but don’t forget that everything you will do will be faster if you do it at a faster clock (of the system, not only of the cpu clock)!
IMHO I think they are trying to save face, after apple’s switch and their reasons.
Really think they can ship new processors in one month or less?
They made a CPUs, not bread…
#429
I thought Steve said IBM didn’t have a roadmap for the low power flavors. Seems like _somebody_ isn’t telling the whole truth…
I thought he said Intel had a better roadman than IBM, not that IBM didn’t have one at all. Either way, I hope some of these do make it into Macs prior to the Intel switchover.
What is the difficulty in switching a system from G4 to G5? Would this be a relatively simple processor swap for the powerbooks?
I’m guessing if it is anything more than that, we won’t see a G5 powerbook.
,,What is the difficulty in switching a system from G4 to G5?”
G5 is hotter, bigger and more power hungry. Models avalible nowdays will not fly in laptop devices. But, as you can see, IBM have something that would fit mobile market better.
Freescale’s dual-core G4 is on the way, too.
That’s pretty sweet. but it’s too little too late. though a dual core G5 powerbook would be sweet.
Now that Apple has switched to Intel isn’t IBM a bit too late. Maybe we would have a G5 powerbook afterall
I highly doubt it. Since Jobs apologised profusely for not being able to deliver such machine, I doubt that Apple is just now going to invest a lot of money in producing a machine that for many will be a dead end. At most, Apple will just marginally upgrade the current offers with slightly faster G4 variants in the fall before phasing out production of these machines entirely in early 2006.
I agree too little too late. I doubt very many people are going to buy into a EOL product such as the PPC Macs.
“I doubt very many people are going to buy into a EOL product such as the PPC Macs.”
You watch. Millions like me will buy a PPC Mac the last day they are on sale, and pay next to nothing for them too. 🙂
Count me in!
IBM didn’t develop those CPUs in a few weeks just to show they’re able to make better CPUs. And, as some people have pointed out before, IBM doesn’t care if they loose Apple: how many IBM cpu-based was apple selling (so buying the cpus to IBM)?
Apple’s decision to move to Intel is not 100% based on faster cpus, lower compsumition and IBM’s (un)ability to deliver more cpus.
It’s too little, too late… Yonah will blow away anything IBM has to offer, it seems to be a really well designed chip. Intel will be switching all their Desktop processors to this architecture, in the future, and those are the chips Apple will use, so don’t talk about Pentium 4’s, those suck, actually.
It’s too little, too late… Yonah will blow away anything IBM has to offer, it seems to be a really well designed chip.
I see the Mac folks have overcome their cognitive dissonance in record time! Tell me, what are your thoughts on the G3 versus the Pentium II? Were you one of the Mac folks who though that the G3 was better? As I recall, most Mac folks didn’t think much of Yonah’s architecture back when it was called the PII.
I’m being a bit sarcastic, of course, Yonah isn’t exactly like the PII, but it’s safe to say that the evolutionary jump any bigger than the one from the Athlon (K7) to the Athlon64 (K8). Mac folks cracked jokes about the aging x86 architecture with regards to the PII, and are now hailing Yonah as the second coming, and it’s irking me off a little.
In any case, the multicore G5 looks like a nice chip. It seems about time we saw one, given the G5’s small die size. The 970MP should weigh in at about 170M transistors, making it closer in size to a single-core prescott than a dual-core prescott. Additionally, it’s very likely that the 2.5GHz multicore chips coming out now will be faster than the first iteration of dual-core Yonahs that’ll be coming out next year. Those will debut at 2.13GHz, and Yonah’s IPC isn’t good enough to overcome that difference in most cases.
“I’m being a bit sarcastic, of course, Yonah isn’t exactly like the PII, but it’s safe to say that the evolutionary jump any bigger than the one from the Athlon (K7) to the Athlon64 (K8). Mac folks cracked jokes about the aging x86 architecture with regards to the PII, and are now hailing Yonah as the second coming, and it’s irking me off a little.”
1st: I don’t even own a Mac, I never had one for more than a decade.
2nd: Pentium M’s are based on Pentium III’s architecture, yes. They have numerous enhancements also like a beefier Brach Predictor, really good power management, among other things I really don’t feel like explaining here. If you read some recent articles comparing a recent Pentium M running with an adapter on a Desktop motherboard (not a very recent one) you’ll see that these mobile chips hold very well against even the best Athlon 64.
3rd: the Yonah processor will not only be a dual-core version of the Pentium M. It will have many important enhancements to it, including SSE3, better power management and power consumption. It will also feature an enhanced floating point engine and a beefed up instruction decoder that can handle “several” SSE instructions in parallel.
4th: It’s dual core and both cores share 2MB of L2 cache which doesn’t happen with current dual-core processors out in the market. This feature can be a very high performance bonus since the two processors can communicate with each other faster and easier.
Read about Yonah, I suggest. http://www.arstechnica.com is a good start.
Oh, and BTW: Yonah is being designed since the Pentium M was released! I really encourage everybody to read stuff about the Pentium M and Yonah to understand that this is going to be a really great mobile (desktop too) chip!
If you read some recent articles comparing a recent Pentium M running with an adapter on a Desktop motherboard (not a very recent one) you’ll see that these mobile chips hold very well against even the best Athlon 64.
You obviously just skimmed the articles. The Pentium M, when overclocked by at least 50%, can “hold very well” against a MID-LEVEL Athlon 64 in the demo benchmarks of three games. In all other tests, the Pentium M has its butt handed to it by all other processors, including low-end Athlon XPs.
“Additionally, it’s very likely that the 2.5GHz multicore chips coming out now will be faster than the first iteration of dual-core Yonahs that’ll be coming out next year. Those will debut at 2.13GHz, and Yonah’s IPC isn’t good enough to overcome that difference in most cases.”
The 2.5Ghz multicore will be ‘probably’ faster than the Yonah’s incorporated in the first Mactel Laptops, mac Mini and iMacs. Howhever, you are comparing high-end liquid cooled CPU with a mobile chip, and that is where the Macs are currently lacking most in terms of CPU performance.
I believe Yonah is a Pentium III derivative, which is a Pentium II derivative, which is a Pentium Pro (notice I didn’t say derivative).
There is almost no difference between a PIII and a PII aside from the on-die cache (which some P2s had IIRC) and SSE. SSE is so bolt-on in the P6 architecture that its hardly worth considering as a new generation of chip.
Never understimate a Macintosh apologists ability to repeat Steve Job’s verbatem.
I’m not trying to be rude or insulting; but it does start to seem like that sometimes. This is one of those times where Mac fans cried for so long about how G5 was better than Xeon (we all know anything beats Prescott) and Opteron and now that Apple is leaving it suddenly Intel is going to produce an amazing x86 chip for them? Do you guys just trust Steve Jobs implicitly?
If I had been a big Mac fan (and I was a big fan of the G5 and the G4 actually) when this happened I’d have shut-up about processor superiority and started talking about hardware quality and OS superiority.
Don’t own a Mac, never have! Never said G5 was better than Opteron! It is better than Xeon though..
The G5 in a Power Mac has got nothing to apologize for.
A G5 in a laptop was the problem. And IBM’s time to market wasn’t as good as one would hope. Had IBM started with the G4 and built it’s 64bit replacement, then there might not have been this change to Intel.
I guess there are just so many good ideas when it comes to chip design.
But, it would have been interesting to see what IBM could have done had they started with the G4, added 64bit regs, and only increased the 7 stage pipeline to 10+, dual core, better Altivec…
All this brow beating about “how dare mac zealots sing the praises of intel when they bagged them for so many years” is starting to get me annoyed.
Each porduct should be judged on its own merits. I am an unashamed G4 fanboy, but in the last 12 months I have been getting increasingly jealous of my centrino toting friends. The P4 is a pilo ‘o crap, but the Pentium M is really sweet.
Some intel chips are good, some are bad, some PPC chips are good, some are bad. It seems to me apple is just betting that future intel chips will be the better. There is a lot of water to go under the bridge before we can make any judgement as to wether or not this was a good idea.
Each porduct should be judged on its own merits.
The problem is that years ago, when Apple was spouting the “MHz myth” and “the G3 is faster than any Pentium II” the Mac fanboys bought it wholesale. They continued to believe it long after the G3’s limitations became painfully apparent.
I am an unashamed G4 fanboy… but the Pentium M is really sweet.
The ironic thing is that back when the P-M was called the Pentium II, the G4 fanboys thought it was a piece of shit. If the P-M is really sweet (better than a G4, certainly), wouldn’t that have made the PII “pretty sweet”? That’s not what the G4 fanboys said. Nope, all I heard was “MHz myth, nyah nyah nyah, I can’t hear you!”
It’s true, it doesn’t matter anymore. The reason apple switched to Intel was:
1. They wanted to introduce tmac (TabletMac) and other new small devices (hmac=handmac) with powerful efficient CPUs.
2. They wanted to pumb the speed of their historic slow laptops
3. They wanted to advance more in the powermac arena, which is slowing down again
They will probably be using these processors now. Jobs said that their are “plenty of excellent PPC products coming.” And even after they start manufacturing Intel products they will still be producing PPC based Macs. The switch is about future products well down the road, not the ones that are out NOW.
13W…..
Is that best or average?
According to the slides it’s 13W (typical)..whatever that means
.
LOL, ‘typical’ sounds like the old joke, “how long is a piece of string?”
I would say that Apple already knew about this product well in advance; the issue isn’t so much the power consumption NOW but the power consumption later on – when the clock speed starts going up, and the wattage goes up as well, the two will get out of wack.
From what I understand, as Apple shows it on their PR, Apple need something that has equal performance to IBMs future road maps, but without the watt consumption hanging over it; Intel has shown its cards, and Apple likes it – Apple was probably also persuaded by the completely integrated solution – video, audio, wireless etc etc The video may have looked weak now, but I’d say that the future video chipsets by Intel will be a massive improvement – equal to that of the ‘value end’ Nvidia and ATI priced chipsets.
I think people forget, Intel offered the whole deal – they not only offered the foot long sub, the included a drink and 4 biscuits at no extra charge – thats in comparison to IBM who were just selling the Sub, and required the customer to find the other parts of different vendors.
IBM, like Intel, reports average thermal design power. I couldn’t tell you how much higher max thermal design power is, but for reference, a Prescott with 130W TDP can use over 180W under load. It should be noted that AMD deserves kudos for citing max thermal design power in their specs.
how do i get a system that runs one of theese that isnt a mac and is below 10,000$?? anyone? anyone?
What about the Pegasos ? ( http://www.pegasosppc.com )
There is no G5 one yet … but i think there will be …
But i am getting a mac soon too … stil want a PPC mac before it is too late ..
But i am getting a mac soon too … stil want a PPC mac before it is too late ..
I hear that. I also am currently saving up the $$$ to get a PPC mac before it’s too late.
Gah! I reckon IBM did this to spite Apple!
/me chuckles at corporate pram-toy throw-fest
Gah! I reckon IBM did this to spite Apple!
Silly. Developing a new processor takes rather longer than a few weeks.
As the biggest customer, Apple would have been well aware of those new processors for some time and is probably just now putting the finishing touches on PowerBooks and PowerMacs based on them.
As the biggest customer, Apple would have been well aware of those new processors for some time and is probably just now putting the finishing touches on PowerBooks and PowerMacs based on them.
Hehe, these people act like Jobs and co. are finding out things on OSNews.
Apple would have been well aware of those new processors for some time
…and just like me, Apple was quite unimpressed.
Bring on the Intel chips!
Apple would have been well aware of those new processors for some time
I think it would be a medium to major pain to keep producing both x86 and PPC Macs. Developers would hate that. On the other hand, these new IBM chips look really nice for servers. Even low power consumption could come in handy – not everybody has heavy duty cooling for their server racks! It would be a great complement to the Intel Mac line.
A day late and a dollar short. They should’ve had these out a year ago.
IBM should sell (and/or license to chinese manufacturers) ATX-form mainboards for geek people as us to buy and mount a linux/*BSD/* G5 box, like Pegasos did with G4/G3.
IBM should destroy Apple creating an alternative personal computer with open architecture that could run open operating systems and, maybe, a PowerPC Windows version. If Mactel succed, M$ could want to do this.
They did (to an extent). It was called PReP, later CHRP. Destined to run AIX, Windows NT PPC, Workplace OS (OS/2 PPC), MacOS and other operating systems. In reality, the only company to use the full specification was IBM itself with its RS/6000 (pSeries) line, running AIX only.
simon vannarath.
i want a 970mp desktop…. not because it’s better of faster then amd or intle’s offerings but….. because it would be sooo cool lol
IBM’s profit from Apple was NeXT to nothing.
They will sell these chips to other users.
As for Apple using an x86 is well, stupid. Why be “different” for so long just to get like all the rest?
1st: I don’t even own a Mac, I never had one for more than a decade.
LOL, my apologies. Though, you have to admit that your post had a rather “fanboy” tone…
Pentium M’s are based on Pentium III’s architecture, yes. They have numerous enhancements also like a beefier Brach Predictor, really good power management
Pentium Pro architecture, actually, and yes it has macro-ops fusion and a better bus and a number of other things. I’m familiar with the architecture. It’s still the same exact execution engine behind all the front-end stuff. It’s quite a stretch to say that it’s better than anything IBM has up its sleeve, at least in the desktop space, especially since its not 64-bit, and especially since it is missing a lot of the niceties that the amd64 instruction set brings to the table (no more x87 FPU, no expanded register file, etc).
3rd: the Yonah processor will not only be a dual-core version of the Pentium M. It will have many important enhancements to it, including SSE3, better power management and power consumption.
For all intents and purposes, it will be. Power management and power consumption is really quite irrelevent (to a point) in desktop space, and SSE3 by itself does not by itself buy you more than 1-5% on most code.
It will also feature an enhanced floating point engine and a beefed up instruction decoder that can handle “several” SSE instructions in parallel.
That statement, while true, is mostly irrelevent. The decoder will be updated to remove some bottlenecks that currently exist. However, the execution core will be unchanged. That means that the performance of the chip will only improve to the extent that the decoder was the bottleneck in current versions of the chip. Yonah will still only be able to execute half a 128-bit SIMD instruction per cycle.
Yonah is a well-designed chip, no doubt. But its not a leap ahead of the PPC970 of K8 architectures (or even ahead, depending on the code), especially considering its 2006+ release date.
“Power management and power consumption is really quite irrelevent (to a point) in desktop space”
To me, the main reason for Apple to switch to Intel is the Laptop market. I think you agree that it is a very good CPU for laptops.
I agree with you that PPC970 is in some ways a better arch than what Intel has to offer currently. I also would prefer to see AMD chips in the Desktop Macs.
“Yonah is a well-designed chip, no doubt. But its not a leap ahead of the PPC970 of K8 architectures (or even ahead, depending on the code), especially considering its 2006+ release date.”
It is, howhever, better than anything IBM or AMD can offer by Q1 2006 and in the near future, in terms of mobile CPU’s.
As for the PPC970 architectural advantages, I don’t think, in practical terms, it is better than what AMD has to offer (I know, Apple is going with Intel :). Of course, the G5 wins in some departments, but you have Intel/AMD winning in many other..
“As for the PPC970 architectural advantages, I don’t think, in practical terms, it is better than what AMD has to offer (I know, Apple is going with Intel
”
What I mean is: in real-life usage the 970 can’t really ‘show’ in most software usage that it is all that better than the current x86 processors.
IBM should build a open source operating system which is game friendly since almost all consoles use IBM processors in one form or another of the powerpc architecture. With such a OS if done well will attract the game developers if they can have a it easy to use graphics cards for fast visualisation and network cards fully for multiplayer games. Ofcourse security is off the essence but using a pico kernel with monitoring of hardware access it could work even with more control to developers. And make a universal standard like a system.h instead of all operating system have own like windows.h etc.
As for the PPC970 architectural advantages, I don’t think, in practical terms, it is better than what AMD has to offer (I know, Apple is going with Intel
. Of course, the G5 wins in some departments, but you have Intel/AMD winning in many other.
I agree with that statement completely. From a programmer’s perspective (not an ASM programmer’s perspective, mind you), I think the K8 is the nicest architecture out there. Opterons are among the fastest chips available for both integer and FPU code, they’ve got a good SMP arrangement, they’ve got nice big caches, a decent number of registers, and are relatively insensitive to instruction scheduling. Probably not the best in any one category, but a very good point in the design space.
Exactly, and that really is what matters. Not that it is theoretically much better and has a greater potential as an architecture but that it effectively offers better performance overall and shows most promise of improving in the future. The G5 have hit 2.7Ghz and are liquid-cooled!
Apple could always change its mind and **also stay** on PPC, rather than phasing out PPC entirely for its OS. It could reside on both–make one of them the ‘high-end’ Mac.
–EyeAm
http://newwestrecords.com/alice/
Any chance they could put a 1.8 970MP into a Powerbook?
Not a chance. period.
Yeah, if they wanted to piss off all their developers…
4th: It’s dual core and both cores share 2MB of L2 cache which doesn’t happen with current dual-core processors out in the market.
If i’m not wrong, the actual dual-core microprocessor from AMD shares the cache among the cores. If we see the actual implementation of dual-core from Intel is “lets shrink 2 P4 and put them on a dice”.
If it seems that Intel and AMD are toe to toe, it’s because Intel is somehow cheating on the implementation of the tecnologies so the broad public actually belives that they are a the same place than AMD.
Actually, the current dual-core processors from AMD give each processor a dedicated L2, and allow them to communicate at high-speed over the on-chip SRQ (system request queue). Having dedicated L2 caches has some advantages, in that it allows greater bandwidth per processor, and easier scalability to quad core configurations. It also has some disadvantages, namely more cacheline ping-ponging when both processors are working on the same data.
having dedicated L2 caches has some advantages, in that it allows greater bandwidth per processor,
Depends on whether the shared caches haven’t got a wider bus. Does anyone know?
It also depends on whether those buses are anywhere near fully utilised.
and easier scalability to quad core configurations.
Why? It just moves the work of integrating more cores to the other side of the cache.
It also has some disadvantages, namely more cacheline ping-ponging when both processors are working on the same data.
Plus the fact that 1 MB of cache goes unused when there’s only one thread or when one of the threads doesn’t need much L2 cache.
Intel was REALLY caught with their pants down. The only advantage they have over AMD (technology-wise) is their chipsets and mobile platform. We mustn’t forget that they have a lot of resources so we should see them recovering in the following year or two. Of course, AMD is not going to sit still.
Oh well, we all win with such competition!
Indeed.
Hopefully, Genesi will soon have such a board running one of these new processors soon. I really need a dev board to try porting SymphonyOS to PPC, although basic work has been started by Ryan Quinn.
Really, although it’s fine for IBM buisness-wise, I’m sad to see them leave the PPC Desktop idea. All my experiences with the architecture have been fantastic, ever since I got my first Bondi Blue powermac.
Although I now will concentrate on a Nano-itx system from VIA, I hope IBM still advance their RISC chips ‘ere. The performance is still supurb, and I know plenty of devs who would buy cheap G5 boards, if avaible.
It’s funny to see how now Apple can do magic with that horrible thing that is the x86 architecture. Ah, zealots…
“It’s funny to see how now Apple can do magic with that horrible thing that is the x86 architecture. Ah, zealots…”
Can you tell me of a better mobile chip than the Pentium M with another ISA? No one is saying x86 is a good architecture but IT IS the most pervasive in the Desktop and mobile market. Moreover, there are some nice x86 chips such as the upcoming Yonah and Athon 64 X2 (OK, that’s AMD64)
This is GREAT news for Apple.
Even though they were building OSX for Intel for years and not telling anyone, they were always letting that fact slip out as rumor to keep IBM and Motorola in check.
It wasn’t working.
Now that Apple has announced their shift, IBM realizes they can’t play hardball with Apple, and they have to win back over some lost trust.
Why does IBM need Apple? PPC advocacy came from them. Sure Amiga and PegasOS were PPC, but the only major PC deployment came from Apple. But that’s not all; IBM is a Linux company, and what system was getting the most open source developers on board? Macintosh.
Without the large pool of PPC developers creating ready-port products for IBM Linux boxes and Apple’s contributions, there is a huge support for ports that would really hurt IBM.
Without desktop chips, they loose their gateway and loose a certain amount of respect. They become a footnote after Intel and AMD. They loose an incredible amount of free publicity and are effectively shut out of the mainstream computer market.
Of course the end of PPC in Macintosh will also herald the end of it for many alternative OS’s also. I hope that the final version of Apple’s Intel design (if it ships… but I’m betting on 95% it will) completely bypasses the BIOS for their autoconfig motherboard design. I still get IRQ conflicts and I don’t want that in my Mac.
IBM dosn’t actually make a large amount of money from desktop PPC use.
PPC makes its money off the server segment, which is not motivated by advertising or use by desktop as much as performance-per-dollar.
Power is now relegated to two segments, sadly – Gaming, and upper-mid/high level servers.
Dark,
I think you’re underestimating the power of the Gateway.
IBM hasn’t got a cheap entry point into Power with Apple gone.
Most of the managers I know hold onto the Windows platform with a death grip BECAUSE they can run the OS on their laptops.
IBM should make them an offer they can’t refuse.
( The good kind, not the Horse Head in the Bed kind. )
[…] which is a Pentium derivative, which is (…) a x86
The Pentium Pro is not a Pentium derivative. The two are internally completely different chips, as much as an PowerPC 601 and a PowerPC 970 are completely different chips. The Pentium is a 2-way in-order chip with a non-pipelined FPU and direct execution of x86 instructions. The Pentium Pro is a 5-way out-of-order chip with a fully-pipelined FPU and conversion of x86 instructions into internal RISC operations.
Laptop sales are outpacing desktop sales…
Apple truely shines in its laptop sales and IBM wasn’t delivering.
Personally, they should have AMD for the desktops, iNtel for laptops.
The best of both worlds…
[PowerMac]
You know, I don’t think it was EVER a question of whether or not IBM could make a 3.0 Ghz PPC G5 chip, or a low-power G5… It was more a question of whether or not they WANTED to do it fast enough, and I think until this latest announcement they didn’t want to. It’s a shame that IBM ignored Apple’s G5s, because Apple abandoned it lest they end up with slower processors than the competition
Who knows what the future will bring? Maybe Intel will make a Pentium M based chip with EM64T. Could be dual core 3 GHz each
. I do agree that IBM could have produced these chips faster. Let’s not forget that IBM has Power 5 chips, that they keep to themselves.
The one angle, that I haven’t seen anyone write about, is a conspiracy theory between IBM, Apple, and Intel. Maybe Apple made the push to Intel, with telling IBM very late into the game, in order to force IBM to relese new G5s for cheaper? or force better G5s? How about a theory with all of them? IBM, and Apple needed to calm down the fire of no G5s or the 3GHz G5. Apple does a PR stunet and “goes” to Intel. IBM then says they can do the chips Apple is looking for, saving some face. Apple, “goes” to Intel to save some face. Thus, creating an conflict of the truth. Intel, comes up as a “new hope” and gains some face among the Apple crowd. Just some ideas from a person, that is very, very tired. It could happen
.
16W “typical”? Try 17-21W under full load for Dothan. 3W while idle. And those are actual numbers on shipping systems.
BTW, to the poster who mentioned “it takes more than a month to come up with a chip”: who said IBM actually has these chips? The picture of the chip in the corner? This is still just a slide at a conference, not something shipping in machines today.
Welcome IBM to the land of being Sony. Always overpromising, continually underdelivering.
The G4s you find in notebooks have an absurdly slow bus. 166mhz and so forth.
The G5s were introduced with custom ASICs IIRC that support up to halfspeed busses. (1Ghz on a 2Ghz chip). The iMac G5 and the now dearly departed single-chip PowerMac G5 however used a variant that I believe was a 1/3rd clocked bus.
So…
(1) heat and power concerns
(2) different controller / bus design
IBM paper launches chips that it should have shipped 1-2 years ago.
This is not impressive.
It is clear that outside of IBM’s mainframes which use POWER5/5+ and custom designs for big monied players (Micro$oft , $ony), there is not much going on in POWER/PowerPC land.
The game consoles could have been made as custom x86 machines and still would have been fantastic.
That leaves POWER for IBM mainframe. Otherwise known as a vast sinkhole of time, money, and corporate energy. The margins are good in mainframe land, otherwise POWER would simply be abandoned. There really is no customer value being delivered.
In sum, IBM’s 970PL chips (970 Paper Launch) don’t make a difference. They are as dumb as the poorly made G5 and should have been shipping a long long time ago. IBM’s time in the sun as a microprocessor vendor is gone.
These new low-power G5s are probably just underclocked G5s with reduced core voltage and perhaps some extra power saving features thrown in.
Nothing wrong with that, that’s how laptop processors are usually developed. Intel only did the Pentium M because the Netburst pipeline was simply too long to get decent performance at laptop-compatible clock rate and power consumption.
Trouble is, the clock rates of those not-yet-delivered G5s are pretty disappointing even compared to Pentium Ms shipping today. (And no, the G5 has no big clock-for-clock advantage over either Pentium M or Athlon64, except for SIMD code).
IBM just aren’t making volume in the Apple/Desktop market.
They’re already making more volume in the embedded sector than they ever were from the Apple desktop sector. To be honest, our firm alone has shipped many more PPC processors in our DTT decoder product line this year in the UK than Apple has shipped Mac’s – and we’re not the only ones using embedded PPC chipset’s from IBM.
And that’s nothing compared to the volume that’ll be shipping in the next-gen consoles.
The Mac folks have this vision that they are the centre of the PPC universe, but really that’s nothing like the truth…..the Apple desktop is largely irrelevant to IBM’s PPC platform.
i think the low power g5s will have typical 16w and around 20watt at full load. i wonder, what ibm will do agains the “Risc problem”…. i mean the fixed lenght 32-bit instructions in 64 bit mode. the ppc970/FX are designed to group 5 instructions (5 instructions=1*64 bit instruction and a shift instruction) and load them at once. i wonder, if the lack of registers,that makes the ppc64 processors slower in 64 bit mode, then in 32 bit mode will be fixed. i mean athlon64 has doubled registers (including double sse2 reg) at 64 bit mode. xeon/and any other em64t intel processor is slower in 64 bit mode because, they do NOT have double registers like the amd processors. amd is faster at 64 bit mode. intel at 32 bit. ppc64 is also faster at 32 bit mode. i hope they will double the altivec etc. registers on the new processors. i heard ibm will do that on the 3core ppc64 processor for xbox360. instead of 32 128bit altivec registers, the processors will have 128 128bit alivec registers. and i think it will have at one of this vmx 128 units on at least 2 of the 3 cores.
i also read somethink about a hack, that makes the ppc64 load 64 bit instructions directly. (not with this instruction groups). but thats a guess.
i can’t say much about cell’s ppc core, but i think the spus will make the processor inredible powerful. i mean look what the emotion engine from the ps2 can do. its a 300mhz 64-bit mips IIIcore with two vector units. and i think it rocks. it can never beat a p4 but hey, haveyou ever seen a 2ghz emotion engine?
Please use the shift key when writing, proper capitalisation makes things much easier to read.
i wonder, what ibm will do agains the “Risc problem”…. i mean the fixed lenght 32-bit instructions in 64 bit mode.
How is that a problem? Do you want them to waste even more bandwidth and extend instructions to 64 bits? If anything, they should introduce a 16-bit instruction format like ARM and MIPS have.
i wonder, if the lack of registers,that makes the ppc64 processors slower in 64 bit mode, then in 32 bit mode will be fixed.
Lack of registers in the PowerPC? You must be joking. The 32 registers it already has are hardly ever fully utilised anyway.
The 64-bit performance hit isn’t due to lack of registers, but due to the fact that pointers and register spills take up more cache space and memory bandwidth.
As for extra Altivec registers: they can be useful for holding more transformation matrices and such like, but somewhere there’s a point of diminishing return where the extra performance isn’t worth the extra hardware.
oimg wtf?!!? sif u int31, l1nUx 4 l4f3
yeah like #77 said. intels p m is a really good processor. i am currently working on a celeron m with 1.5ghz and its all i need. it can play h264 without much problems. at low resolutions
its much fater than my old amd athlon 1ghz. (desktop)
but wo don’t speak of pentium m, we speak of pentium 4 3.6ghz intel macs. and we speak of intel pentium m intel powerbooks…..but wait, there is one problem. the pentium m family doesn’t have em64t. that means it will be a step back. powerbooks will stay 32 bit. with ppc or with intel..it doesn’t matter. (maybe with ppc970 low power 64 bit chips..i don’t think apple will use these.)
and we have to think what MAKES a pentium m so fast. is it the architekture? is it the fpu/sse2? no because the celeron m has the same sse2 units but celeron m is still much slower. so what is it? right it is the 2mbyte l2 cache.
i wonder how a ppc processor will perform with 2mb l2 cache. (hmm power 5 has 1.9mb l2)
we will see how the ppc970mp will perform with 1mbyte l2 per core.
i am a ppc fan, but i x86 architekture is still a bit faster. why that? because is not boundet to the 32 bit fixed lenght adresses. i mean look at ibm power5. it is a powerfull processor. more powerfull than opteron xeon etc. BUT not more powerful than the intel itanium 2. the itanium 2 is not boundet to 32 bit fixed lenght instructions. a 1.5ghz itanium 2 is faster than a 1.9ghz single core power5 (not much but it is faster).
i don’t say ppc is a slow architekture, i like ppc, but its time to extend the architekture with direct 64 bit instruction (like i said at #76 i don’t know if the instruction groups of the power4/ppc970 slow the processors down, but i think they will be faster with direct 64 bit instruction loading.
also im should improve vmx/altivec with 64 bit like sse2 has. the x86 world extends their architekture with new simd units like sse4 (its only a matter of time) and what does ibm/freescale? they don’t extend altivec/vmx. like i said, i am really interested how the 3 core ppc64 processor of the xbox360 will perform with its improved altivec/vmx unit.
so what is it? right it is the 2mbyte l2 cache.
Yep, the 2MB of very low latency L2 cache.
the itanium 2 is not boundet to 32 bit fixed lenght instructions.
Itanium’s instruction format is even more cumbersome and cache-hostile than a RISC’s!
i don’t say ppc is a slow architekture, i like ppc, but its time to extend the architekture with direct 64 bit instruction (like i said at #76 i don’t know if the instruction groups of the power4/ppc970 slow the processors down, but i think they will be faster with direct 64 bit instruction loading.
I don’t think you really understand the concept of a 64-bit instruction. 64-bit instructions are 64-bit because they operate on 64-bit operands, not because they are 64-bits in length. Generally, the only constraints 32-bit instructions pose for 64-bit code is the inability to load 64-bit immediates with a single instruction. Generally, code doesn’t do this nearly enough for it to be a significant bottleneck.
1. Pentium M IS a very good mobile CPU. It will be even better when Yonah arrives in Q1 2006. Does anyone dispute that?
2. The current K8 processors can GENERALLY outperform the current G5’s, despite the G5’s “better architecture”, as some have pointed out. There are exceptions, of course, but I’d say that on the whole, for most people the best AMD processors are “better” than anything IBM has to offer in this segment. They don’t need to be liquid-cooled, also.
3. The Intel PowerMacs won’t arrive until 2007! Since Intel has completely rewritten their roadmap some time ago after AMD started spanking their ass, deciding to drop the Netburst architecture for something closer to the Pentium M, you won’t see Pentium 4 or anything based on Prescott in your beautiful PowerMac’s.
4. It’s really too soon to judge how good will the Intel chips be by 2007. I expect them to catch-up to AMD.
5. As for the G5, do you think that it can keep scaling up in terms of performance as Apple would expect? They are having trouble getting out faster G5’s and I’m not even talking about the mobile ones. Yes, Intel is having the same problems also, but as I said, they are scrapping Netburst from the map and by 2007 their chip offer will be considerably different.
– Again, some of this is just speculation. Let’s stop speculating about why Apple did this and that because we just DON’T KNOW! Apple knew what IBM had to offer, they knew about Intel, about AMD. If they chose to make this big transition then MAYBE they had good reasons.
To all those that haven’t YET really got it, from a power consumption/cpu speed ratio point of view, please take a look here http://www.theregister.co.uk/2005/07/08/ibm_powerpc/
Now, find all the excuses/justifications you want for Apple’s-Steve’s switch decision, but just stop fooling yourselves with “superior Centrino/Yonah blah-blah GHz output AND less heat/power consumption” guys… It’s just not true.
What’s your point? That Register article has the same information as the macrumours article.
You can get a 90nm Pentium M now at 2.1GHz and 21W thermal design power. In normal use it draws significantly less than that.
The article talks about a 1.6GHz G5 at 16W that hasn’t been delivered yet and where we don’t even know what that 16W number actually describes.
From your link: “A new, low-power 970FX consumes between 13W and 16W at frequencies of 1.2GHz, 1.4GHz and 1.6GHz. That’s more than the 10W that the Freescale MPC7448 found in today’s 1.5Ghz PowerBooks consumes, but around half the maximum power consumption of Intel’s Pentium M, which powers today’s Centrino laptops.”
The chips don’t consume from 13W to 16W! You have the 1.2Ghz which consumes 13W typically (not under load) and the 1.6Ghz which consumes 16W typically (not under load). You obviously can’t compare those numbers with the MAXIMUM Pentium M power consumption because typical load and maximum load generate very different power consumption. Besides, these chips aren’t released yet!
Please tell me why the 1.6Ghz G5 (which won’t be here at least until the end of the year) is faster than the faster Dothan Pentium M? And what about Yonah? I’m sorry, you can argue that G5 is a better chip for workstations/desktops BUT I find it hard someone finds me a better mobile chip than Dothan (currently) and Yonah (early next year). Please, please, show me I’m wrong!!
Turion is a better mobile chip (for me at least) because it can run linux x64 (and Windows for x64) while Dothan simply can’t.
You see there is no absulute truth, only relative…
Yeah, there’s the Turion but still in the PowerPC camp there’s nothing competitive in the mobile market.
First, “typical” power dissipation is not the same as idle – it’s TYPICAL – say an average given typical usage of the system. I don’t know about you, but my systems are rarely idle. So it’s not idle, but it’s not max – it’s in-between.
Second, a 1.6 GHz G5 will beat a 2 GHz Pentium M at most tasks. Pentium M’s have received a lot of attention lately because enthusiast sites have shown that overclocking a Pentium M to 2.5 GHz allows it to play a few game demo benchmarks at close to Athlon 64 speeds. However, if you check all the other benchmarks, Pentium Ms score very poorly, even overclocked by 50%.
A G5 is a monster chip on par with the Opteron and Xeon EM64T chips. It’s not faster than the AMD64, but it’s in the same ballpark. The fact of the matter is, the Pentium M is not in that ballpark at all.
It is too late.