Dual Core PowerMac G5 Updates?

[AidenShaw]

AFAIK, Linux 2.6 has very good support for physical and logical CPU's. Even better than XP IIRC. 2.4 treats them as regular multiprocessor-machines.

It could depend on the disto....

The original 2.6 kernel was OK. I haven't looked at what all the distos are shipping today, and verified what updates have been made since then.

This article ( http://lwn.net/Articles/80911/ ) implies that significant changes have been made (and that the original 2.6 scheduler had some problems with HT).

Do you know if 2.6 exposes HT through an API to the developer? Win2k3 does, and lets the programmer see the actual topology (physical CPUs, logical CPUs, NUMA structure) and set affinity for CPU scheduling and influence memory allocation.

 

[AidenShaw]

Apple is telling developers that it will be Dothan

I think the Mac Mini and iBook will be the first systems to feature Intel processors. Probably Yonah in both cases. I just hope that Apple doesn't mess up and use the single core Yonah product, I want the dual-core Yonah in at least the Mac Mini.

As I said earlier, I expect MacIntels in January.

The first ones will be Dothan-based, Apple is telling its developers that.

 

[notheremuch]

A lot of people are repeating themselves in this thread (powerbooks first!).
What I'm interested in is the future line-up of the powermac. A dual dual 2 Ghz would have no purpose in the line-up as it is about as fast as the dual 2.7 GHz.
So that leaves the discussion: will Apple replace some dual models for a single dualcore or will there be only dualcore at the high-end? And in that case, will there be one or two new models added?

I am thinking that the 2.7 will become dual-dual only.
Maybe the 2ghz and 2.3ghz will get bumped.... but I don't think they will be dual-dualed.

BTW When the doors open tomorrow, are the new releases in the open? Or, could we possibly have to wait until the end of the expo (worst case) for an intro?

 

[blitzkrieg79]

I wish you would stop repeating this. You haven't described CELL correctly in any of your posts. Please read http://arstechnica.com/articles/paedia/cpu/cell-1.ars and http://arstechnica.com/articles/paedia/cpu/cell-2.ars. CELL has a single general purpose core with multiple attached DSP cores.

What Intel is moving toward is a multiple general purpose core design. Please read http://www.intel.com/technology/computing/multi-core/. This is similar to Sun's Niagara plan. Please read http://www.sun.com/processors/throughput/.

There are lots and lots and lots of potential workstation applications that would benefit from CPUs with multiple general purpose cores. There are lots and lots and lots of existing server applications that would benefit from CPUs with multiple general purpose cores. There is more to life that Microsoft Word and H.264 encoding.

Described it wrong??? You have your arstechnica and I have http://www.blachford.info/computer/Cell/Cell0_v2.html which goes in to a greater detail about the ENTIRE ARCHITECTURE AND CONCEPT, you see, CELL is not only the CELL processor found in PS3, the beauty of CELL is that it will be rather easy to design it to specific customer needs, its a versatile technology...

There will be PS3 and there are also gonna be IBM Cell based workstations running Linux so I am guessing that Cell and its variations are not only for consoles but thats just me...

And Intel is moving (or at least has it in its plans) to multiple cores with specialized DSP-like cores too if you would actually looked at the pic I attached a while ago...

 

[Hattig]

As I said earlier, I expect MacIntels in January.

The first ones will be Dothan-based, Apple is telling its developers that.

We will agree to disagree then.

If anything comes out, it will be an Intel Mac Mini because that is meant for 'consumers'. Shame really because a lot of us 'consumers' actually do more than browse the web and chat, and unless the switch means a $399 Mac Mini I could care less about Dothan.

 

[willyjsimmons]

Dual Xeons.....ha ha never

(comments on thread starting https://forums.macrumors.com/posts/1740048/)


Maybe you've noticed that the PowerMac is already sucking up so much power that it has a water cooling system? What's a few more watts among friends?

This is the kind of emotional reaction that I implied with the "eating crow" comment.

Dual dual Xeon systems would make a lot of sense as a transition system...

• Apple could continue to sell PPC970 and PPC970MP PowerMac G5's alongside the quad Xeons
• The quad Xeons would be very useful for developer systems - how's anyone going to port and code 3D apps and games against the embedded graphics on the DTK? Give them real PCI Express x16 graphics and PCI Express I/O slots! (And Apple could *sell* these to any developer...)
• Quad Xeons would give the companies who have ported their high end apps a high end market to sell to, instead of having the x86 ports sitting on a shelf (or not being ported).
• The development costs would be tiny - like the DTK systems, Apple could put a standard Intel workstation motherboard into the PMG5 case - there's plenty of room.
• The quad Xeons would be good for developers who are working on the 64-bit port to x64 - Apple can't wait for 64-bit Pentium M systems to *start* the x64 porting process. They can *sell* the systems to developers now.

Just like Apple had a transition period where some Macs could boot both OS9 and OSX, Apple will need a transition period where it sells PowerMacs that can run both CPU architectures.

Dual Xeons would be the worst choice in the world.

They're crazy expensive.

Intel already sells dual-core 64bit CPUs.

And I'm sure they have a 2 cpu board in the pipeline.

 

[willyjsimmons]

In theory.

Like I've already said before, Intel already has the chips for Apple to release a single or dual, dual-core 64-bit PowerMac using the same enclosure.

The boards are the exact same size as those in the current developer machines.

If someone here is correct in that Apple is 'dumbing down' to 32bit, then there is no reason why an x86 based PB couldn't be released by Jan/Feb.

If Apple wants everything to be 64bit capable out of the gate, then don't expect a PB until fall of 2006 when the new chips are available.

For the mini, eMac and iMac, it really depends again on if Apple wants everything to be 64bit capable from the jump.

If not, then it really comes down to how fast motherboards could be produced.

I've already voice my opinion on third-party controllers in another thread, and IMO I'd rather see apple 'get it right', then to try and rush something to market.

 

[wdlove]

For the mini, eMac and iMac, it really depends again on if Apple wants everything to be 64bit capable from the jump.

If not, then it really comes down to how fast motherboards could be produced.

I've already voice my opinion on third-party controllers in another thread, and IMO I'd rather see apple 'get it right', then to try and rush something to market.

Apple definitely needs to make sure that they 'get it right.' The world will be watching the releases of the Intel Mac. They will have a golden opportunity to shine.

 

[~Shard~]

Apple definitely needs to make sure that they 'get it right.' The world will be watching the releases of the Intel Mac. They will have a golden opportunity to shine.

Exactly. Whether they like it or not, (or it is realistic or not to expect), Apple has to hit a home run with their first Intel Mac to meet everyone's high expectations - no real second chances. As a result, I can't see them going with an "older" chip such as the Xeon, rather they'll be going with "newer" chips, or the future chips such as the Merom, Conroe and Woodcrest chips.

Guess we'll see!

 

[DavidCar]

I am thinking that the 2.7 will become dual-dual only. Maybe the 2ghz and 2.3ghz will get bumped.... but I don't think they will be dual-dualed.

Taking a simplistic approach, the ThinkSecret article said a 970MP was said to be 50% to 80% faster "in many tasks." That would make a dual dual 2.0 about the same as a dual 3.0 to 3.6 "in many tasks" and therefore better than a dual 2.7. So there is reason to think a dual dual 2.0 could sell better than a dual 2.7, depending on the price points.

(By the same logic, a dual dual 2.5 would be like a dual 3.75 to 4.5.)

 

[Xephian]

Taking a simplistic approach, the ThinkSecret article said a 970MP was said to be 50% to 80% faster "in many tasks." That would make a dual dual 2.0 about the same as a dual 3.0 to 3.6 "in many tasks" and therefore better than a dual 2.7. So there is reason to think a dual dual 2.0 could sell better than a dual 2.7, depending on the price points.

(By the same logic, a dual dual 2.5 would be like a dual 3.75 to 4.5.)

That's nice. If true, then I think a Dual Dual 2.5 would be faster than any AMD X2.

 

[AidenShaw]

That's nice. If true, then I think a Dual Dual 2.5 would be faster than any AMD X2.

... and slower than a dual dual AMD ...

 

[blitzkrieg79]

... and slower than a dual dual AMD ...

For everyday computing tasks, maybe... For specialized tasks(altivec apps-and thats the reason people buy Powermacs) definately not...

Considering the fact that G5s were doing just fine with the competition when they cam out, the addidtion of a dual core and extra of 512 kb of cache would mean to me that clock for clock it would be about the same as AMDs offerings and considering the fact that the rumors are saying that Apple could release a dual core processor running at 2.5-2.8 GHZ I would say that would definately put them on top, but then again, AMD is already been out for a while...

 

[wdlove]

Exactly. Whether they like it or not, (or it is realistic or not to expect), Apple has to hit a home run with their first Intel Mac to meet everyone's high expectations - no real second chances. As a result, I can't see them going with an "older" chip such as the Xeon, rather they'll be going with "newer" chips, or the future chips such as the Merom, Conroe and Woodcrest chips.

Guess we'll see!

Now we have to wait until June.

Come on Steve we need a PowerMac update, ASAP.

 

[~Shard~]

Now we have to wait until June.

Come on Steve we need a PowerMac update, ASAP.

I dobn't think we need to wait until June to see something - Jobs could still unveil an Intel Mac at MWSF and indicate a shipping of date of 6 motnhs out.

As for the PowerMacs, perhaps dual-core 970s are just around the corner...

 

[Lacero]

Now we have to wait until June.

Come on Steve we need a PowerMac update, ASAP.

It was June 2004 when we saw the last real update. Last April's update bump of 200Mhz was terrible. I am expecting to see at least a dual dual-core machine around this timeframe. PM Intels are coming up in 2007, so that's still a long ways away. This is getting ridiculous.

 

[wdlove]

I don't think we need to wait until June to see something - Jobs could still unveil an Intel Mac at MWSF and indicate a shipping of date of 6 months out.

As for the PowerMacs, perhaps dual-core 970s are just around the corner...

Just a view won't be that satisfying. To actually have one in hand there will still be the nine months from now. That is quite a wait from introduction to delivery. Will be interesting if Apple allows orders that far in advance.

Hopefully this week.

 

[~Shard~]

Just a view won't be that satisfying. To actually have one in hand there will still be the nine months from now. That is quite a wait from introduction to delivery. Will be interesting if Apple allows orders that far in advance.

Hopefully this week.

Yes, I agree. In that respect, it will be a long wait, regardless of when the official announcement is made. Jobs will no doubt make mention of it in some aspect at MWSF, so I guess we'll see what comes out of that...

 

[BGil]

For everyday computing tasks, maybe... For specialized tasks(altivec apps-and thats the reason people buy Powermacs) definately not...

Considering the fact that G5s were doing just fine with the competition when they cam out, the addidtion of a dual core and extra of 512 kb of cache would mean to me that clock for clock it would be about the same as AMDs offerings and considering the fact that the rumors are saying that Apple could release a dual core processor running at 2.5-2.8 GHZ I would say that would definately put them on top, but then again, AMD is already been out for a while...

Not really. AMD procs have integrated memory controllers and NUMA support in Windows so they tend to scale better than other processors do. With the introduction of Windows 64-bit the gap gets even wider. A dual core 2ghz A64 (X2 3800+) with the smaller cache size will match a dual proc 2.7ghz G5 in Cinebench, for example. A dual dual Opteron system @ 2.2ghz (x75) more than doubles the 2.7ghz G5 in Cienbench.

 

[notheremuch]

So, I'm ready to buy a new PowerMac. How long should I wait - in the boards opinion!

(I don't need it now, but my store has an offer of discounted ram and 8% off Apple displays until the end of September.)

 

[powerbook911]

So, I'm ready to buy a new PowerMac. How long should I wait - in the boards opinion!

(I don't need it now, but my store has an offer of discounted ram and 8% off Apple displays until the end of September.)

I think we may very well see the Power Macs updated in September. "Updates," often do not occur at the actual big event, but a week or two afterwards. If it was a huge makeover, they get announced at the events, but otherwise they are announced quietly at Apple.com.

 

[Mr. Anderson]

I think we may very well see the Power Macs updated in September. "Updates," often do not occur at the actual big event, but a week or two afterwards. If it was a huge makeover, they get announced at the events, but otherwise they are announced quietly at Apple.com.

Well, the Expo is going on right now and next Tuesday is the last Tuesday of September - so why delay it one week when it could have been announced for the Paris event? If they do update the PowerMacs, I'm thinking now we won't see what we're all hoping for......

D

 

[~Shard~]

Well, the Expo is going on right now and next Tuesday is the last Tuesday of September - so why delay it one week when it could have been announced for the Paris event? If they do update the PowerMacs, I'm thinking now we won't see what we're all hoping for......

D

The only thing I can tthink of is that there was a slight delay in the PM update such that it wouldn't be available to ship for another week or so. Jobs was going to unveil it at the Paris Expo keynote, but when he found out he would have to annoucne a shipping time of 1 or 2 weeks instead of immediately, he said screw it, I'll announce it when it is actually shipping, and as a result axed the keynote.

Far-fetched, yes, but it's fun to speculate.

 

[Lacero]

That hasn't stopped him before. Remember the iPod mini fiasco of 6 months for non-US residents. I remember that very bitterly!!

 

[blitzkrieg79]

Not really. AMD procs have integrated memory controllers and NUMA support in Windows so they tend to scale better than other processors do. With the introduction of Windows 64-bit the gap gets even wider. A dual core 2ghz A64 (X2 3800+) with the smaller cache size will match a dual proc 2.7ghz G5 in Cinebench, for example. A dual dual Opteron system @ 2.2ghz (x75) more than doubles the 2.7ghz G5 in Cienbench.

I am guessing that you are getting your benchmarks from http://www.digitalproducer.com/articles/viewarticle.jsp?id=32951-1 which, up to a release of dual dual core AMDs, they didn't use Adobe After Effects because Apple was wiping the floor with x86 world...

Anyway, for one, dual cores are more efficient than dual processor systems as the cores communicate directly without having go through busses... Secondly, Cinebench is an x86 optimized benchmark plus OpenGL implementation in Mac OS X ain't nothing to brag about either... Third of all, Mac OS X isn't yet fully 64 bit optimized... Fourth of all, considering that AMD released a dual dual core system that isn't able to soundly defeat an Apple based system running on two processors in every test means that PowerPC is more than holding its own...

And integrated memory controllers maybe makes sense for servers, in workstations you don't see a whole lot of everyday benefit as the speed is limited to the speed of RAM anyway...

So in conclusion, I still think that in raw processig power when fully optimized, PowerPC is still superior to x86 in workstation market... Portables are a different story though...