970MP - Still Coming?

[JonaGold]

more series?

maybe it's time to introduce a second pro-line: just as there are two lines for consumers now (mini & iMac) there could be two pro-lines: powermacs & '...'macs??
that would make it possible to have 2 different architectures at 1 time.

 

[deputy_doofy]

maybe it's time to introduce a second pro-line: just as there are two lines for consumers now (mini & iMac) there could be two pro-lines: powermacs & '...'macs??
that would make it possible to have 2 different architectures at 1 time.

Supermacs?

 

[Platform]

maybe it's time to introduce a second pro-line: just as there are two lines for consumers now (mini & iMac) there could be two pro-lines: powermacs & '...'macs??
that would make it possible to have 2 different architectures at 1 time.

So where do you say the eMac falls......not consumer of something else

 

[~loserman~]

The rumor sites seem to disagree with you. The consensus is the 970MP starting at 3.0 GHz.

Conincidentally, this is likely why the top of the line PM is 2.7 GHz now. This revision gives Apple the time it needs (6 months?) to get IBM producing 3.0mp's outside of prototype land. In that revision, we'll probably see all the niceties that got cut for this revision: PCI-E, DDR2 667, HT 2.0, Dual Core 970's, and maybe even Blu-Ray.

Sorry but the rumor sites are wrong.
I'll place a bet with anyone here that you won't see a dual core 970MP at 3GHZ within 18 months.

 

[FlyNolJ]

?

Sorry but the rumor sites are wrong.
I'll place a bet with anyone here that you won't see a dual core 970MP at 3GHZ within 18 months.

Do you have inside information, or is this just a little theory? That would really really blow if it were the case. =( I need a new computer...

 

[myapplseedshurt]

Sorry but the rumor sites are wrong.
I'll place a bet with anyone here that you won't see a dual core 970MP at 3GHZ within 18 months.

well, look at it this way... if 970GX isn't at 3G, and it took 6 months to get a 200 MHz increase just to achieve 2.7G, then what rationale does anybody have that a 970MP will magically attain an additional 300 MHz??

Maybe the timeframe lends itself to speculation, but MHz increases are NOT linear (i.e. based on 6 mos. for 200MHz increase, then can say 12 mos. for 400). So it's a toss up. but I can tell you that when an MP system qualifies, and performance is given even a little boost, you'll see rumblings all over the place.

 

[daveL]

~loserman~

Do you have inside information, or is this just a little theory? That would really really blow if it were the case. =( I need a new computer...

Well, he does know more than most; he works in a very significant Xserve installation. Beyond that, he has a fair amount of common sense. One of these days, someone on this thread will realize that IBM/Apple/G5 is not going to advance at a significantly faster pace than AMD64. So, where are the AMD64 DC cpus now? Answer: 2.2GHz, although they aren't actually available until mid-June (Athlon64, which isn't suited for SMP) or late May (Opteron 200 series, ~$2k per cpu). Given the pace we've seen clock rates improve on AMD64, how long will it take to hit 3 GHz? Answer: not any time in the foreseeable future, so maybe like 18 months.

 

[Frobozz]

Sorry but the rumor sites are wrong.
I'll place a bet with anyone here that you won't see a dual core 970MP at 3GHZ within 18 months.

Well, I guess it depends on what the foundation technologies are. I believe the IBM documentation that was temporarily online said 3.0 GHz, too. So, the rumor sites could just be basing their predictions on that document and some of their sources.

I do know one thing, though. The 970mp is not just two 970fx cores slapped together. Also, I don't think it's fair to compare the Dual core Athlon's clock speed to the results IBM will have. x86 cores have very different considerations than PowerPC cores do. There are logistical differences in their architectures which lend the PowerPC to work better in multi-core environments. At least, this is what the consensus of rumor, opinion, and briefly leaked documentation suggests.

I hold firm that the above information, combined with the seemingly odd 2.7 GHz high end PowerMac released last week, gives us a dual core 3.0 Ghz machine by this fall. At worst, by MWSF '06.

Today:
single 1.8
dual 2.0
dual 2.3 <--- this would (should?) have been dual 2.5
dual 2.7 <--- this would (should?) have been dual 3.0

Next Release *Speculation*:
single 2.0
dual 2.3
dual 2.7
dual 3.0

The other thing to consider is the use of water cooling in 2.3 Ghz + 970fx chips. It seems like a dual core will spread out the heat better and maybe they can remove the water cooling from the mid range?

 

[auxplage]

To speak for those of us editors out there...any one of us with half a brain (with the exception of those who absolutely MUST use Avid Symphony or DS suites) edit video on a Mac. I've tried it on a PC, and it isn't pretty. I would rather cut on a Mac that is slower, than try to edit on a PC.

Exactly, one must balance the positives and negatives. To me, losing in a benchmark by 10% is not that big of deal if you are completing the project faster because you are more efficient with the OS.

 

[daveL]

Also, I don't think it's fair to compare the Dual core Athlon's clock speed to the results IBM will have. x86 cores have very different considerations than PowerPC cores do. There are logistical differences in their architectures which lend the PowerPC to work better in multi-core environments. At least, this is what the consensus of rumor, opinion, and briefly leaked documentation suggests.

Then why has the current G5 just been updated to 2.7 GHz, will AMD64 has been at 2.6 GHz for a few months. It sure doesn't appear that the x86 baggage has kept AMD64 from modestly staying ahead (I did say *modestly*) of the G5. Believe me, I'm not arguing that we will see a 970MP within 6 months or so, I just don't think it will also be accompanied by an increase in clock rate.

 

[Kaborka]

x86 cores have very different considerations than PowerPC cores do. There are logistical differences in their architectures which lend the PowerPC to work better in multi-core environments. At least, this is what the consensus of rumor, opinion, and briefly leaked documentation suggests.

Consensus? This theory is new. At least to me. Let's talk about Opteron, the P4 has reached it's end of life. These are the differences between G5/PPC970 and Opteron Multicpu systems:

Opteron:
- 8 CPU configuartions available
- Existing system CPUs can easily replaced with dual core CPUs
- NUMA architecture
- Each CPU equipped with three 1000Mhz Hypertransport links and a 800Mhz local memory interface, so one Opteron can communicate with it's environment with 32.4 GB/s
- Direct Hypertransport links between CPUs
- On-die crossbar
- Direct Hypertransport link to I/O
- Ultra low latancy On-die memory controller

G5:
- 2 CPU only configuartions available
- classical northbridge design
- Each CPU equipped 1/2 * #Ghz bus, so one PPC970@2.3 Ghz can communicate with it's environment with 9.2 GB/s when both directions are fully utilized, in one direction only with 4.6 GB/s
- Both CPUs connected to the northbridge (G5's "system controller"). No direct connection between CPUs possible
- Routing of all traffic takes place in the northbridge (CPU<>CPU, CPU<>MEM, AGP<>MEM, CPU<>IO). A northbridge was and still is a bottleneck.
- I/O connected via crippled 800MHz Hypertransport to the northbridge
- AGP connected to the northbridge
- Memory connected to the northbridge
- Shared high latency northbridge memory controller for both CPUs

Think I don't need to explain which one is the superior topology. Opteron systems live in the same spheres as Power4/5 and Itanium systems whilst PPC970 setups do not.

Kaborka

 

[FlyNolJ]

Almost there...

Then why has the current G5 just been updated to 2.7 GHz, will AMD64 has been at 2.6 GHz for a few months. It sure doesn't appear that the x86 baggage has kept AMD64 from modestly staying ahead (I did say *modestly*) of the G5. Believe me, I'm not arguing that we will see a 970MP within 6 months or so, I just don't think it will also be accompanied by an increase in clock rate.

People are saying that dual core will debut at 1.8, 2.0, and 2.2. (They are saying also that the MP was supposed to debut at 3.0ghz, which I don't believe.) This is reverting back to the first release and revision b of the Powermac G5 right?? I don't understand the difference between dual processor 2.7 ghz, and dual core 2.2. Is there a direct link for communication or something? I don't understand why people would want to revert back in mhz, or are they hoping for multi cores? I think that would cost upwards of 4,000 just from looking at today's models. daveL, clarify for me please.

 

[daveL]

Consensus? This theory is new. At least to me. Let's talk about Opteron, the P4 has reached it's end of life. These are the differences between G5/PPC970 and Opteron Multicpu systems:

Opteron:
- 8 CPU configuartions available
- Existing system CPUs can easily replaced with dual core CPUs
- NUMA architecture
- Each CPU equipped with three 1000Mhz Hypertransport links and a 800Mhz local memory interface, so one Opteron can communicate with it's environment with 32.4 GB/s
- Direct Hypertransport links between CPUs
- On-die crossbar
- Direct Hypertransport link to I/O
- Ultra low latancy On-die memory controller

G5:
- 2 CPU only configuartions available
- classical northbridge design
- Each CPU equipped 1/2 * #Ghz bus, so one PPC970@2.3 Ghz can communicate with it's environment with 9.2 GB/s when both directions are fully utilized, in one direction only with 4.6 GB/s
- Both CPUs connected to the northbridge (G5's "system controller"). No direct connection between CPUs possible
- Routing of all traffic takes place in the northbridge (CPU<>CPU, CPU<>MEM, AGP<>MEM, CPU<>IO). A northbridge was and still is a bottleneck.
- I/O connected via crippled 800MHz Hypertransport to the northbridge
- AGP connected to the northbridge
- Memory connected to the northbridge
- Shared high latency northbridge memory controller for both CPUs

Think I don't need to explain which one is the superior topology. Opteron systems live in the same spheres as Power4/5 and Itanium systems whilst PPC970 setups do not.

Kaborka

Could you price these other CPUs for me? Thanks.

Can you point me to an SMP workstation that doesn't have the architectural limitations you ascribe to the DP G5 PM, other than the lack of an on-chip memory controller? (Obviously, there are things you can do better with a proper dual core chip than with a DP setup.) Thanks.

Also, all your posts seem to be x86 fanboy stuff. Since this is a Mac forum, don't you think members here would go to a x86 oriented site for this kind of info? In addition, and I'm sure you'll be surprised by this, the MR forum archives are filled with the same information you're beating everyone over the head with. In fact, I'll go so far as the say it appears over and over and over again. If you don't like Macs, why do you post here? Just curious. There always seems to be someone out there that feels a need to come to MR to educate us ignorant Mac folks.

 

[daveL]

People are saying that dual core will debut at 1.8, 2.0, and 2.2. (They are saying also that the MP was supposed to debut at 3.0ghz, which I don't believe.) This is reverting back to the first release and revision b of the Powermac G5 right?? I don't understand the difference between dual processor 2.7 ghz, and dual core 2.2. Is there a direct link for communication or something? I don't understand why people would want to revert back in mhz, or are they hoping for multi cores? I think that would cost upwards of 4,000 just from looking at today's models. daveL, clarify for me please.

Dual core, properly done, ala AMD64, does have the potential to perform a bit better than a similarly clocked DP system, mainly due to direct core-to-core interfaces on-chip for cache coherency and such. A DP system has to use an external bus and controller to interface the two processors. The dual core CPUs run slower, because you have twice as many transistors as a single core, so you have a lot more heat and power to deal with. However, there is a modest win: A single dual core machine should be cheaper than an equivalent DP system to manufacture. I suspect Apple will satisfy the high-end with dual dual core systems, thus mitigating the fact that each core will clock somewhat slower than single core processors. The low and mid-range should see similar performance for a lower price. Of course, I'm just giving you my best guess; who knows?

 

[fpnc]

Think I don't need to explain which one is the superior topology. Opteron systems live in the same spheres as Power4/5 and Itanium systems whilst PPC970 setups do not.

The Opteron and Athlon systems are nice. However, show me an Opteron system that can compare with the form factor and price of a G5 iMac and then maybe you'd have something.******To a large degree, the same goes for the dual-processor Power Macs.

As for the possibility of a 3GHz 970MP system this year, well DaveL, ~Loserman~ and I often do not agree, but here I must say that I too am very doubtful that we'll see a 3GHz dual-core system from Apple this year. In fact, that may be one of the reasons why we haven't had a 970MP announcement from Apple thus far -- the clock speeds and yields may be too low to compete with the existing dual-processor Power Macs. Is that necessarily a bad thing? I don't think so, since the current 970FX processors are probably pretty cheap and Apple already has a long and reasonably successful history of shipping dual-processor systems. I can't see Apple going dual-core until the 970MP is both cheap enough and running at high enough speeds to clearly beat the existing dual-processor Power Macs. With a redesigned motherboard and chipset I suspect that something in the 2.4GHz range would be good enough, but a lot of people are going to see that as a somewhat "suspect" upgrade from the current 2.7GHz duals. Of course, if Apple builds a twin dual-core system that would more than make up the difference in clock speed, but it could also lead to price increases (as has been discussed, $3500 or $4000 hard to know).

Finally, there is some speculation that the 970MP will be pin compatible with the 970FX, that is, it could be a drop-in upgrade similar to the dual-core Opterons and Athlons. I find that somewhat unlikely, but it may be a possibility. In any case, even if it is true I suspect that Apple will put some type of block in the Power Mac firmware to prevent someone from just removing the 970FX processor and dropping in a 970MP.

 

[fpnc]

As for Opteron prices and how the current Power Macs compare in performance, below is a post from Ars that gives a pretty good overview (contains performance graphs and system prices). Note, however, that the dual-core Pentium and Opteron systems that are shown in this comparison haven't even shipped yet -- so you may want to take the quoted prices and performance on those systems with a "grain of salt."

http://episteme.arstechnica.com/eve...24/m/867008203731/r/895009013731#895009013731

And note, that this data was prepared by someone who is not a particularly big fan of Apple -- so it might represent the "best" side of the Pentiums, Xeons, and Opterons.

 

[blitzkrieg79]

Although I agree with looserman and daveL upon the clock frequency of the first generation of 970MPs I still dont know why people assume that it will be basically a 970fx chip x2, I mean the current 970fxs cant go past 2.7 but whose to say on what basis the unannounced 970mp is actually based upon, what if it will borrow some of the power saving techniques of the Power5 and that would make it easier to clock it higher I guess... Anyway, the price of dual core processors is probaby what kept Apple from releasing it in the latest PowerMac update BUT what happened to 970GXs? From my understanding they do in fact have better power saving features (which makes them cooler to run so therefore easier to higher clock plus they supposed to have 1MB of cache, so 970GX is still an improvement over 970fx but why arent they yet released?) Anyway, I do believe that a drastic change in processor plans happened to IBM and Apple probably in early 2004 (due to rather disappointing 970 research and development and a fact that we will never see a mobile G5) and I think something radically new is comming up but if its a Power5 derivative or Cell-like architecture (or maybe some kind of hybrid) all I can do is speculate as I have no real info on that matter... Anyway I just ordered my dual 2.3GHZ PowerMac as I need it for my profession...

 

[whooleytoo]

well, look at it this way... if 970GX isn't at 3G, and it took 6 months to get a 200 MHz increase just to achieve 2.7G, then what rationale does anybody have that a 970MP will magically attain an additional 300 MHz??

It may be that the 970MP was delayed, so the 2.7GHz 970GX speedbump was introduced as a stop-gap solution. Or it may be that the 2.7GHz 970GX was delayed, whereas the 970MP is proceeding nicely and will be launched soon.

The only thing we can deduce without any incisive insider info is that we can't deduce much at all!

 

[poundsmack]

It may be that the 970MP was delayed, so the 2.7GHz 970GX speedbump was introduced as a stop-gap solution. Or it may be that the 2.7GHz 970GX was delayed, whereas the 970MP is proceeding nicely and will be launched soon.

that sounds good and all as long as you switch 970GX with 970FX

 

[whooleytoo]

that sounds good and all as long as you switch 970GX with 970FX

Err.. yeah <cough> What he said!

 

[eric67]

give it a try..;

It may be that the 970MP was delayed, so the 2.7GHz 970GX speedbump was introduced as a stop-gap solution. Or it may be that the 2.7GHz 970GX was delayed, whereas the 970MP is proceeding nicely and will be launched soon.

The only thing we can deduce without any incisive insider info is that we can't deduce much at all!

possible hypothesis
this guy has similar hypothesis; and predicted the iMacG5 + Radeon 9600 with a logical explaination. a bit long, but interesting points are mentioned. he thinks that MP will be there in September, with clockspeed probably similar to AMD and possibily new PB with a new processor too.

 

[fpnc]

Okay, I went and read the Hardmac blog (link in previous post), IMO it contains nothing new or original and you'd get as much or more from just reading this thread on MacRumors. As for the recent graphics card updates in the eMac and iMac, I think that has far more to do with Tiger's new Core Image and Quartz 2D Extreme than it does with gameplay. In this case, the effect on games is just a happy side effect, the real "target" is enhancing the Tiger experience.

 

[eric67]

Okay, I went and read the Hardmac blog (link in previous post), IMO it contains nothing new or original and you'd get as much or more from just reading this thread on MacRumors. As for the recent graphics card updates in the eMac and iMac, I think that has far more to do with Tiger's new Core Image and Quartz 2D Extreme than it does with gameplay. In this case, the effect on games is just a happy side effect, the real "target" is enhancing the Tiger experience.

it has nothing to do witj rumors I guess...
The guy was mostly pointing out some interesting points ,such as the revision of the entire Apple hardware in less than 3 month...except for iBook but it is coming!!
I thought that this was interesting regarding strategy and explaining maybe why Apple did not announce MP because could not produce them before September; or made a simpl speed bump of 200MHz, because the future MP will top at 2.7GHz...
I think the guy wrote a comment not related to rumors , but looking beyond them...

 

[deputy_doofy]

The other thing to consider is the use of water cooling in 2.3 Ghz + 970fx chips. It seems like a dual core will spread out the heat better and maybe they can remove the water cooling from the mid range?

Unless I'm misunderstanding what you're saying, the mid-range does not have water cooling. Per the Apple website, only the 2.7 has it. I just got the 2.3GHz and there is no mention of water-cooling that I can find.

 

[mcdawson]

What kind of sense does it make to update the Power Macs and then one month later have Steve say at WWDC, "One more thing, we have a Dual Core 3GHz Power Mac that we will bring out in September."??????

NO ONE would buy any of the newly upgraded G5's....plain and simple. To think otherwise is just plain retarded and I recommend that any of those that think this go take an Economics 101 course at their local Community College.

Well, I would say that Apple HAD to revise their line because of the iMac revision (it would be just too close). I would agree that economically, it wouldn't make total sense for Apple to turn around and annouce a big speed bump only a month later. HOWEVER, I don't think that the speed-bumped models are "important" enough to worry about--Apple, for marketing reasons, had to spend $$ to release an updated PM line. I do believe that its only a "placeholder" (and expensive one) until the dual-core PMs are available. If the dual cores are a big performance gain, Apple will bring them out when they are avaliable, even if that means throwing out a recent investment. That's more marketing than economics, but I believe that's where Apple's roadmaps lies. That said, they're likely to wait until there's an event, such as WWDC (June) or MacWorld France (Sept/Oct) to annouce them. I would also expect that, even if they ship in Sept-Nov, they'll be in short supply until MacWorld SF--there'll be the same pent up demand as the original G5s had.