[galarneau]

Saw this link over at Arstechnica:

http://arstechnica.infopop.net/OpenT...5&r=2880904555
I'm hoping that if this is the case, that Apple goes with dual processors... even on the lowend. That single proc 1GHz still annoys me. The dual 867 was superior in almost every way.

[JB72]

"only" 1.8Ghz? I think that's a nice starting point for this sophisticated CPU.

I also have to say, I don't think we'll be seeing the 970 used in Macs until we see OS X.3 sometime in 2004. And not at MWSF. Later.

[madsenj37]

Comparing a 970 to a G4 is not the same as comparing say an amd to an intel processor at similar speeds. The 970 has different architecture than the G4. when Intel released their itanium it came out at about 800 mhz when its P4 line was much faster in terms of mhz. What I am driving at is with a new architecture and 64 bits the 970 is very different than the G4 at around the same mhz.

[galarneau]

Yes, I'm well versed on the differences between the G4 and 970. The point was that many people were getting excited over the prospect of the 970 being intro'd at 2.5GHz.

The rant at the end of my post was just a little annoyance at how the new lowend PowerMacs are less powerful (IMHO) than the last lowend model. Not really related to the PPC 970 news.

[DrBoar]

Only 1.8 GHz 970 might lag behind the Intel P4 in clock speed but not performance
If the SPEC 200 benchmarks is anything to go buy: (Numbers from Moki in Appleinsider forum)

IBM PPC 970:
SPECint2000 937 @ 1.8 GHz
SPECfp2000 1051 @ 1.8 GHz

Pentium IV @ 3.06 GHz:
SPECint2000 1032 @ 3.06 GHz
SPECfp2000 1092 @ 3.06 GHz

MOT G4 does:
SPECint2000 418 @ 1.4 GHz
SPECfp2000 248 @ 1.4 GHz

For G4 to compete with 1.8 GHz 970 it has to run at 1.4 x(937/418) MHz, that is 3.14 MHz for the integers and for the fp 1.4x (1051/248) or 5.9 GHz.

Even for a pair of G4 using SMP with 90% efficiceny the speeds would be 1.74 and 3.28 GHz respectively.

So let us not be fooled by the minor MHz increase of 400 Mhz or so

As a side note the dual 1.42 G4 then when compared to Pentiums scores as P4 at about 2.23 and 1.26 GHz repectivey for SMP applications for non SMP the results are 1.24GHz and 700 MHz respectively. No surprice that maintoshes get beaten up in video compression tests and other number crunching

For IBM 1.8 or 2.5 matters as they will stack two IBM 970 against two Xeon both running Linux.

For Apple it matters far less at least for the moment, compared with the G4 the 970 is far superior at any resonable speed, furthermore the SMP support in OS X and some applications and the lack of SMP in Win XP home edition just open the door to really slam SP Pentiums wih dual 970

Even if the lowend 970 would be a SP 970 at 1.5 GHz it would be a bit faster than the current DP 1.42 in SMP integer and vastly superior in everything else
I do hope Apple will go dual from the start for the mid and high end to beat the P4 at least in SMP applications but for the low end they really do not need to have a DP

So even if IBM would skim the fastest 970 to use in their own servers, the future looks way brighter than the present.

[Thinine]

And those 970 figures aren't even official yet. They may actually increase, depending on how pessimistic IBM was when they released them.

[OreoCookie]

IBM has clearly said that clock speeds will start at 1.8 GHz and top out at 2.5 GHz. The problem (as mentioned by other rumors sites) is clearly heat dissipation that is quadratic in the frequency. If you assume (unrealistically) that the voltage is kept constant and further assume that the PPC970 dissipates 48 [email protected] GHz, you get that it will dissipate 91.2 [email protected] GHz.

I think a reasonable assumption is that the PPC970 will top out at 2.25 GHz (heat dissipation >= 75 W).

About the SPECmarks: these are estimates (maybe based on prototype results, who knows). I think the actual performance will be a bit higher than that due to optimizations in the cpu and in the compiler.

[Thain Esh Kelch]

Originally posted by OreoCookie:
IBM has clearly said that clock speeds will start at 1.8 GHz and top out at 2.5 GHz. The problem (as mentioned by other rumors sites) is clearly heat dissipation that is quadratic in the frequency. If you assume (unrealistically) that the voltage is kept constant and further assume that the PPC970 dissipates 48 [email protected] GHz, you get that it will dissipate 91.2 [email protected] GHz.

The statement about 1.8 Ghz - 2.5 Ghz processors was found in a PDF file which was later pulled from IBM's servers!

And regarding the effect issue, Macboudille (Or how they spell it..) today stated that a 2.5 Ghz would run around 64W, which is about 30% less than what your saying.. Of course, that was also unrealistic..

[DrBoar]

For some professionals dual 2.5 over 1.8 would be a good thing even if the tower would double as a room heater and sound like a hoover .

Perhaps if IBM do not want the fastest and thus hottest 970 until after the dieshrink, they can be used by an other company ASAP

With IBM showing prototypes of 970 servers and Apple taking bids for manufacturing of a totaly new motherboard, it do look like 970 powermacs in 3 to 6 months time.

Then I might be wrong and the motherboards really are slated for a headless emac named "Performa765432CDRW Extreme"

[Eug]

Why not not just an initial release to scale up to 1.8 GHz, and then a new revision to scale higher?

This happens all the time on the x86 side, even without a die shrink. Wattages increase of course, but it's not a linear relationship.

Personally I hope it doesn't take a 0.090 u process to get to 2.5 GHz, but I guess only time will tell.

[clarkgoble]

I don't think anyone was arguing that the 970 would be initially out at 2.5 GHz. We'd all like it, but the comments about the fab sites came out within a day of the initial reports. Clearly the 2.5 Ghz machine is a prototype.

Having said that though, while the 1.8 GHz 970 beats the Pentium, the really issue is the Athalon-64, Opteron, and other such chips. I've said that before though, and apparently a lot here don't like those discussions, so I'll not say much more. Simply that I think Apple really needs a high end system that beats or is the same speed as high end x86 systems. That is essential (IMO) for Apple's success and will have a trickle down effect for those of us who don't typically buy the high end systems.

[Superchicken]

Apple NEEDS a high end work station? REALLY? Wow... could have fooled me... after all they havn't had one for the past uhh... forever... and theyr'e doing pretty OK...

I mean it'd be nice and it'd certainly help but they don't really NEED much.
Quite frankly I think apple most likely will put at least in the high end PPC 970 systems a Dual setup, unless they can't get the mobo in time or something cause after all... why not? Charge more... blow Dell out of the water... anyone else wanna see a head to head with the top Dell system and a power mac that smacks it around in REAL tests?

[clarkgoble]

Apple NEEDS a high end work station? REALLY? Wow... could have fooled me... after all they havn't had one for the past uhh... forever... and theyr'e doing pretty OK...

Have you seen the sales figures for PowerMacs? If you had you wouldn't be saying that. Yeah Apple still has 4 billion in the bank and financially isn't doing horribly. But if they don't fix their high end sometime this year they will be in trouble.

I'm confident they will. But if you think they can survive on portables and low end systems then I think you incorrect.

Anyway, back to the topic at hand, I'm fairly sure that Apple will announce 1.8 GHz 970's in August (including duals) and that they'll get a speed bump to 2.5 in December/January. (Probably the initial 970's will ship mid to late September)

[MindFad]

Originally posted by clarkgoble:
Anyway, back to the topic at hand, I'm fairly sure that Apple will announce 1.8 GHz 970's in August (including duals) and that they'll get a speed bump to 2.5 in December/January. (Probably the initial 970's will ship mid to late September)

I've been thinking that for a while, too. All evidence seems to point to this, and I really don't see how they can go another year with lame bumps. We can only hope -- I do, for Apple's sake, because their Power Mac sales are in the shitter.

[saltines17]

Originally posted by OreoCookie:

I think a reasonable assumption is that the PPC970 will top out at 2.25 GHz

Wait, am I missing something here... it is supposed to TOP OUT at 2.25 GHz?? As in, that's the highest it will go...??

[Catfish_Man]

Originally posted by saltines17:
Wait, am I missing something here... it is supposed to TOP OUT at 2.25 GHz?? As in, that's the highest it will go...??

I believe he meant that the highest frequency that would be practical is about 2.25GHz because of the power consumption. His estimate is a bit high though (see either the arstechnica or realworldtech.com forums, I can't remember which, they both have good info). It's still heavily debatable whether IBM really meant [email protected] micron. Personally, I doubt it. The POWER4+ (which the 970 shares a core with, roughly) runs at 1.45GHz, tops. Assuming a fairly significant boost (~30%) because of the decreased reliability requirements and tweaking, 1.8GHz sounds about right. A slightly under 50% boost from a die shrink (which is reasonable, from what I've seen) would put it in the 2.5GHz range @ .09 micron (of course, there's a bunch of other fun stuff you can do with a die shrink besides boost the clock frequency).

[JB72]

For some professionals dual 2.5 over 1.8 would be a good thing even if the tower would double as a room heater and sound like a hoover .

Assuming that the professional didn't work with anything involving sound .