Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
I was reading through the IBM 970 initial presentation (the is a link for it somewhere in the posts previous). One thing that popped out at me was on this page... (i think page 13 of the presentation PDF).

It states that normal floating point operations (non-vectorized) can be as fast as 7.2 GFlops/sec while vectorized (SIMD) floating point operations will be able to be as fast as 14.4 GFlops/sec. If this is true, then applications will not be required to get significant improvements in speed by coding with the AltiVec only extensions. Possibly this will mean that a single DOUBLE floating point calculation would take the same time as two Altivec FLOAT floating point calcuations would take.

This would be a significant improvement in speed over the G4 processor.
 

Attachments

  • 970speed.gif
    970speed.gif
    13.1 KB · Views: 679
nighthawk:

If the quick math-in-the-head is right, these figures correspond to 4 flops per cycle scalar, and 8 flops per cycle AltiVec, both on a 1.8ghz PPC-970.

If this is true, then applications will not be required to get significant improvements in speed by coding with the AltiVec only extensions.
Certainly the difference between AltiVec and scalar single-precision floating-point performance on a PPC-970 will be much smaller than that difference on a G4.

Possibly this will mean that a single DOUBLE floating point calculation would take the same time as two Altivec FLOAT floating point calcuations would take.
Well, I have to admit that I'm not sure how IBM gets 8 flops per cycle from AltiVec, but it is quite possible that they are counting the permute-ops as flops, when in my mind they are just memory operations. For whatever reason, it doesn't look like they are counting memory operations in the scalar units as flops, though, so I'm not sure what exactly they are thinking. Anyway, one way to interpret things is that both AltiVec and the scalar units are capable of 4 flops of math per cycle, but in different ways, since the 4 flops per cycle on the scalar units is two fused add-multiplies in either single or double precision, whereas the AltiVec is 4 identical adds or 4 identical mults, all in single-precision only, each cycle. These each will have their uses, and I image that some programmers will find ways to use all the units at once. :)

Also, the picture is complicated by questions about how fast the different units can do more complex float ops like division and square roots, but I don't know much about how they compare there.
 
Re: 3rd Q

Originally posted by Sublime


Doesn't April start the 3rd quarter?

No. A quarter of 12 months would be , anyone? anyone? 3 months so April would be the start of the 2nd quarter. The 3rd quarter would start on ... That's right, July but remember unless they qualified it as the beginning of the 3rd quarter, you're probably looking at September/October time frame at best. Realistically, I'm expecting 1st quarter next year more precisely MWSF '04.
 
Re: Re: 3rd Q

Originally posted by sedarby


No. A quarter of 12 months would be , anyone? anyone? 3 months so April would be the start of the 2nd quarter. The 3rd quarter would start on ... That's right, July but remember unless they qualified it as the beginning of the 3rd quarter, you're probably looking at September/October time frame at best. Realistically, I'm expecting 1st quarter next year more precisely MWSF '04.
But then you have to guess whether they are talking the beginning of the corporations fiscal year (which is not always January) or the actual calendar year.
 
Re: Re: 3rd Q

Originally posted by sedarby


No. A quarter of 12 months would be , anyone? anyone? 3 months so April would be the start of the 2nd quarter. The 3rd quarter would start on ... That's right, July but remember unless they qualified it as the beginning of the 3rd quarter, you're probably looking at September/October time frame at best. Realistically, I'm expecting 1st quarter next year more precisely MWSF '04.

see: http://www.corporate-ir.net/ireye/ir_site.zhtml?ticker=AAPL&script=1800&layout=7#fin3

To quote:
Q: What is Apple's fiscal year and corresponding earnings release dates?
A: Apple's fiscal year 2003 runs from September 29, 2002 to September 27, 2003. See the calendar of events for a tentative schedule of earnings release dates for FY2003.

If you look for 1 minute, you can usually find out actual info rather than assuming.

PS, in IBM's projections, I remember them saying that the 970 would be avialable in quantity in the later half of 2003, meaning calendar.
 
Originally posted by pgwalsh
With all the talk about super fast 970 processors and ultra high throughput, I wonder what video card Apple would use. HardOCP over-clocked a P3 to 4.44 ghz and found that the video card was the bottle neck. The system also had a serial ata drive and other goodies.

http://hardocp.com/article.html?art=NDIy

Have a nice day!

Ha! Look at tom's hardware for the even more insane overclock. (if memory serves 4.6 GHZ with super freezing )

The 970 should kick if they get everything in order soon. Apple you mind spending a couple million for a new chip, hire some over-time workers to work at IBM :D
 
Re: Re: Re: 3rd Q

Originally posted by primalman


see: http://www.corporate-ir.net/ireye/ir_site.zhtml?ticker=AAPL&script=1800&layout=7#fin3

To quote:
Q: What is Apple's fiscal year and corresponding earnings release dates?
A: Apple's fiscal year 2003 runs from September 29, 2002 to September 27, 2003. See the calendar of events for a tentative schedule of earnings release dates for FY2003.

If you look for 1 minute, you can usually find out actual info rather than assuming.

PS, in IBM's projections, I remember them saying that the 970 would be avialable in quantity in the later half of 2003, meaning calendar.

I feel confident that we will get some advance notice here on the Mac Forum! :)
 
Originally posted by Abstract
What's the power consumption of the G4's and G3's, and the Intel P3 mobile cpu's? I would very much like to see a 970 in an ultra-portable 12" Powerbook or iBook form factor: >1.8GHz for any PB or PM, and >1.4GHz for an iBook or iMac. :) If power consumption is even lower than it is now for the G3 and G4's, I hope a 6Hr battery life is possible. That would make my year. :)

*gets down on knees and prays for the very first time*

Excuse me by the way did you know that Macs are more than a competitor in the PC world, they have the edge at the moment, the only reason people buy PCs is because theyre dumb and theyre cheap..... oh and the machines.

Anyway the newest G4 shows a huge performance gap between intels 3.06hz pentium 4 chip.

CHECK OUT THIS WEBSITE : http://forgetcomputers.com/~jdroz/09.html
 
Re: Re: Re: 3rd Q

Originally posted by primalman


see: http://www.corporate-ir.net/ireye/ir_site.zhtml?ticker=AAPL&script=1800&layout=7#fin3

To quote:
Q: What is Apple's fiscal year and corresponding earnings release dates?
A: Apple's fiscal year 2003 runs from September 29, 2002 to September 27, 2003. See the calendar of events for a tentative schedule of earnings release dates for FY2003.

If you look for 1 minute, you can usually find out actual info rather than assuming.

PS, in IBM's projections, I remember them saying that the 970 would be avialable in quantity in the later half of 2003, meaning calendar.

Fiscal:
1st Quarter: Oct, Nov, Dec
2nd Quarter: Jan, Feb, Mar
3rd Quarter: Apr, May, Jun
4th Quarter: Jul, Aug, Sep

Given the current situation, there probably won't be a 970 in any Mac at any time in the forseeable future so it really doesn't matter whether its fiscal or calendar year!

The reason? Too expensive for Apples price point.
 
Re: Re: Re: Re: 3rd Q

Originally posted by sedarby


Fiscal:
1st Quarter: Oct, Nov, Dec
2nd Quarter: Jan, Feb, Mar
3rd Quarter: Apr, May, Jun
4th Quarter: Jul, Aug, Sep

Given the current situation, there probably won't be a 970 in any Mac at any time in the forseeable future so it really doesn't matter whether its fiscal or calendar year!

The reason? Too expensive for Apples price point.

What do you mean by too expensive, amount? Are you predicting 2004, say MWSF?
 
Originally posted by MacCoaster

Check this out: FPMathTest.

PowerPC G4 gets slapped silly by AMD and Intel processors in that open source benchmarker. Don't even get me started on the multi-threaded version.

Yeah because Java doesn't run well on macs. And that it isn't coded to run fast on macs since it is PC based app.
 
-----What do you mean by too expensive, amount? Are you predicting 2004, say MWSF?

Wow, that would suck. If G5/970 doesn't come out this year, many loyal Apple fans will be seriously upset. I don't care how much speed they can juice out of G4, this chip is several years old, probably 95 years old in computer chip age. Not just that, it will give AMD and Intel to come out with 64 bit chip first. Apple will just end up getting behind and lose more market share. Even OS X and cool factor can't hold Apple that long.
 
macphoria:

AMD is almost certainly going to have a 64-bit desktop chip before Apple does. However as was pointed out in some other arguement, it is not entirely clear that AMD will have a 64-bit MS consumer OS to go with it berfore Apple has 64 bits.
 
... low power consumption ... silent desktop ...

can we finally hope to have powermacs without health hazard? or reincarnation of cube? now i have something to look foward in year 2004.
 
Power4+

IBM just released their p series servers with up to 4 way Power4+ configs. Now that they have the chip in production with resonable numbers, (only 1.45ghz tho) I can imagine that a pwer 4+ will find it's way into other machines soon...(mac?)

At least all the development and ramp up is already done. Just time to crank up the clock speed and sample!
 
Originally posted by ktlx
What is the basis for this statement? It contradicts what I'd heard previously.

AMD is delivering their 64-bit processors before IBM does and Microsoft has already said they will support the processor. Microsoft already supports the 64-bit Itanium processor http://www.microsoft.com/windowsserver2003/64bit/default.mspx with is Advanced Server, Limited Edition version.

Apple has not even said they are going to use the PowerPC 970 let alone get a 64-bit version of all of Mac OS X (not just the part inherited from FreeBSD) in the hands of testers.[/B][/QUOTE]

You should know Apple by now....They never say anything. And yes a 64 bit version of windows is comming on April with the release of the Operton!!! Although I would really like to see what they came up with.......remember....sudently everything sucks (Microsoft)
 
64- Bit Mac OS X server

It seems logical that there will be a 64-bit version of Server before a 64-bit version of Jaguar, considering the likeliness of the 970 being used in the xSertve first.
 
Re: 64- Bit Mac OS X server

Originally posted by os10geek
It seems logical that there will be a 64-bit version of Server before a 64-bit version of Jaguar, considering the likeliness of the 970 being used in the xSertve first.

So what do you think the time frame will be? Will we still see it for the Power Mac by August! :confused:
 
os10geek:

It seems logical that there will be a 64-bit version of Server before a 64-bit version of Jaguar, considering the likeliness of the 970 being used in the xSertve first.
I see no reason for Apple to release a PPC-970 server before a PPC-970 desktop.
 
Originally posted by ddtlm
os10geek:


I see no reason for Apple to release a PPC-970 server before a PPC-970 desktop.

The 970, being derived from the power 4 series, is a good server
chip. I see no reason for Apple not to release the 970 in a
server before releasing it in a desktop. But I bet it will find its
way into a desktop first anyways.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.