Xenon= Modified G5 Triple Core Theory

[Fox5]

It's not possible to 'disable the OOE'. It's a fundamental and integral part of the instruction pipeline. You can as little 'disable the OOE' as you can 'disable' (remove) the chassis of a car and still have a driveable vehicle.

So what you propose is pure nonsense really.

I would have thought they'd share the instruction set, schedulers, and execution units from the G5/Power4. Not necessarily a G5 based cpu, but that they share common transistor blocks, like a Pentium 3 and a Pentium 4.

Is good for performance that an In-Order CPU has this long pipeline?

My logic says no, but a part of me says that I am wrong.

I think they're depending on finding highly parrelizable and non-branchy tasks and efficient multithreading to get good performance?

Apple's beef was with the power use, not clockspeed.

It was both. They cared about power use on the laptops, but considering how they've made no attempt yet to replace their G5 processors with Core Duos (despite that core duos actually would be faster in just about anything not vectorized), I don't think apple cares about the power use for their desktops, rather more that they were promised a 3.2ghz processor to launch back in 2003.

 

[Asher]

Trivia: IBM never made the 3.2GHz promise to Apple; Jobs made that promise publically ("we will have it by next summer") in an effort to pressure IBM publically. Unfortunately for Apple, they began to become more trouble than they were worth at a time where IBM's semiconductor business was very busy with other customers...

And I reckon the reason why there's no PowerMac replacement yet is CoreDuos don't currently work properly in SMP (2 CoreDuos), or they're waiting for the Core 2 Duos in July.

Putting a 2.16GHz Core Duo in PowerMacs isn't any different than their iMac.

 

[Fox5]

Trivia: IBM never made the 3.2GHz promise to Apple; Jobs made that promise publically ("we will have it by next summer") in an effort to pressure IBM publically. Unfortunately for Apple, they began to become more trouble than they were worth at a time where IBM's semiconductor business was very busy with other customers...

And I reckon the reason why there's no PowerMac replacement yet is CoreDuos don't currently work properly in SMP (2 CoreDuos), or they're waiting for the Core 2 Duos in July.

Putting a 2.16GHz Core Duo in PowerMacs isn't any different than their iMac.

I thought they didn't plan to replace the G5s until sometime in 2007?
Anyhow, CoreDuos don't support SMP at all, Intel has always left that to their Xeons.
They could replace the low end single G5 models though, but do they even still sell single G5 systems? For that matter, they could replace all their single core and dual core systems with the Core Duo right now, the only reasons I think they might not are...
1. G5s mainly specialized in vectorized quantities, which the Core Duos (and all Intel processors) are weak at. Opterons might make a decent replacement though.
2. Most likely the apps that run best on the PowerMacs are customized for G5s and Altivec, and would take a while to port, or at least port and get running at the same speed. The software currently plays to the strengths of the G5s, and the Core Duos are still weak at those strengths.
3. They already have the contract for the G5s, so why waste money?

 

[Asher]

I bet the new PowerMacs will be based on this: http://www.dailytech.com/article.aspx?newsid=2625

 

[Fox5]

I bet the new PowerMacs will be based on this: http://www.dailytech.com/article.aspx?newsid=2625

Any reason for that bet? I'd bet it's more likely the new PowerMacs would be based on Xeons, considering so far the PowerMacs have been heavily pushing multicore.

 

[Asher]

Because Xeons are still Netburst, which Apple isn't optimizing around. I'd very much doubt that.

Otherwise they'd launch Xeon PowerMacs a long time ago, no?

 

[Fox5]

Because Xeons are still Netburst, which Apple isn't optimizing around. I'd very much doubt that.

Otherwise they'd launch Xeon PowerMacs a long time ago, no?

What I meant is they'll wait for Woodcrest Xeons, the ones based on the Core 2 Duo. I think multicore is too big of a marketting thing in the PowerMacs to go back to only 2 cores, plus they wanna push the machines as workstations and servers, the same market that Xeons are aimed at.

 

[Asher]

Isn't that what Xserves are for (servers)?

http://www.thinksecret.com/news/0605intelxserves.html

Xeons and Extreme Edition processors aren't so different.

Xeons may end up in PowerMacs too, I didn't realize they were out at the same time...which makes sense.

 

[Fox5]

Isn't that what Xserves are for (servers)?

http://www.thinksecret.com/news/0605intelxserves.html

Xeons and Extreme Edition processors aren't so different.

Xeons may end up in PowerMacs too, I didn't realize they were out at the same time...which makes sense.

That's really interesting that the Xserves would have Woodcrest Xeons before PowerMacs got Conroe; Intel's consumer line usually launches first.

Oh, and Xeons are usually clocked much lower than either the Extreme Edition models or the top of the regular line up. They usually feature a slower memory bus too, and they also have dual processor capability. They're meant to be 100% stable and error free after all.
I find it strange apple would go from a quad core PowerMac to having a dual core PowerMac. Of course, actual performance could still be higher, and they could always just go longer without retiring the top of the line powermac until a quad core single chip processor comes out. (or just use xeons just for that model

 

[archie4oz]

It was both. They cared about power use on the laptops, but considering how they've made no attempt yet to replace their G5 processors with Core Duo (despite that core duos actually would be faster in just about anything not vectorized), I don't think apple cares about the power use for their desktops, rather more that they were promised a 3.2ghz processor to launch back in 2003.

It wasn't really so much about power consumption, but more about Apple not wanting to fund processor development, and haggle for processor pricing.

As far as the pro towers go, they're less of a priorty to update than were the consumer lines. The Banias/Dothan/Yonah cores were simply too much to overlook compared to the 7447 offerings that was getting Apple pummeled in the mobile and consumer range (970 powered iMacs withstanding). The 970MP powered desktops are still excellent machines (I'd love to trade in my dual 970FX for a quad 970MP) that are still relatively competitive with high-end x86 offerings, thus the update priority is lower.

What I meant is they'll wait for Woodcrest Xeons, the ones based on the Core 2 Duo. I think multicore is too big of a marketting thing in the PowerMacs to go back to only 2 cores, plus they wanna push the machines as workstations and servers, the same market that Xeons are aimed at.

IMO, the likely scenario is that Xserves and the top 1-2 spec PowerMacs get Woodcrest, while the lower end tower(s) get a comparable clocked Conroe (the larger cache and higher speed FSB of Woodcrest would the higher spec marketing point). Of course it also depends on how long the iMac and Mini sit with Yonah too. Those both go Merom, or the Mini (and MacBooks) go Merom, while the iMac goes Conroe, while the PowerMacs and Xserves all go completely Woodcrest.

Anyways, why the hell did this thread get to 2 page? Any notion that the PPE is related to the 970 is totally absurd. ADEX summed it all up in a nice tidy manner...

 

[Fox5]

IMO, the likely scenario is that Xserves and the top 1-2 spec PowerMacs get Woodcrest, while the lower end tower(s) get a comparable clocked Conroe (the larger cache and higher speed FSB of Woodcrest would the higher spec marketing point). Of course it also depends on how long the iMac and Mini sit with Yonah too. Those both go Merom, or the Mini (and MacBooks) go Merom, while the iMac goes Conroe, while the PowerMacs and Xserves all go completely Woodcrest.

Is it confirmed Woodcrest will have an advantage over Conroe, performance wise? Previous Xeons were lower clock speed, lower FSB, and had an L3 cache, and were generally inferior in performance to the top end P4s.

 

[Gubbi]

Is it confirmed Woodcrest will have an advantage over Conroe, performance wise? Previous Xeons were lower clock speed, lower FSB, and had an L3 cache, and were generally inferior in performance to the top end P4s.

Not inferior in the server space. The big L3 makes a big difference on a whole host of workloads.

Cheers