Power 5 to hit 3GHZ, but thats just for starters

[The_Editor]

My Truck ..



Believe it or not ... Car drivers actually DO try to muscle us out of the way !!   :-o    They fail to grasp that it wont be just a little bump ...... More like .. Globules of human flesh scattered over the highway !!  ( Ask a Fireman)

[Kronos]

@Editor

That doesn't count !!

Get rid of 2 wheels and your allowed to play with ... :-D

Must admit that the biggest truck I ever drove was a 7,5 t + 7t trailer, but it was just one meter
shorter than the big ones, and driving that one through inner-city Berlin (including small living-quarters-streets) on saturday
morning/ noon was a special way of fun  :-D

[The_Editor]

@ Kronos     :-D

All in a days Yakka  (work)

[Hammer]

The P4 actually decreased IPC by around a third. The P4 at 1.5GHz was beaten by a P3 at 1GHz.

That true to a limited extent but your not encapsulating the whole picture.

Pentium 4 (Northwood A core) increase their IPC compared to Pentium 4 (Willamette core) due to minor 512Kb L2 cache increase and switching to .13 micron  processes.

IPC is further increase in Canterwood (Intel’s answer to nForce2) due to FSB increase.  The next minor IPC increase will be due to 1Mb L2 cache increase. Intel is not that stupid when they reached a clock speed barrier with the current designs.  Not factoring in hypertreading (with Northwood B core) due to problems with some applications.

Further minor IPC increases will be applied in the
Prescott  (planned Q4 2003 release)          
http://www.chip-architect.com/news/2003_03_26_Prescott_clues_for_Yamhill.html

Thats how the 970 at 2GHz gives the P4 a good run for it's money at 3GHz.

Depends on application being run…
e.g.
1. Apple’s Quake 3 number is in conflict with most major X86 PC sites (most of them running at 400+FPS.
2. Pentium 4’s problems with Portland group’s complier V4 i.e. structural problem with Intel's SSE2.
3. Pentium 4’s narrow application in regards to SSE2 (generally benefits games not scientific applications).
4. Problems with Apple's claims.

References
http://www.overclockers.com/tips00408/

http://www.amdzone.com/articleview.cfm?articleid=1296&page=2

http://www.theinquirer.net/?article=10292
(PGI compiler V5.0(cite 32bit version) improvements over V4**).

http://www17.tomshardware.com/graphic/20021218/vgacharts-05.html
(nForce2/AthlonXP vs Pentium 4(533FSB)(mostly  entertainment bias benchmarks)

http://members.cox.net/craig.hunter/g5/
(PS; Portland Complier V4** for Pentium 4)

http://www.theinquirer.net/?article=10155

http://www.amd.com/us-en/Processors/ProductInformation/0,,30_118_1274_3734^3750,00.html
(AMD Athlon™ XP processor 3200+ in entertainment bias benchmarks).
 

When the POWER5+ "consumer version" appears (second revision of the 980?) it should be I'd guess around 3.5GHz, don't know where Intel will be but I suspect by then the IBM will be outgunning them performance wise.

How can you conclude "IBM will be outgunning them performance wise" then you haven't satisfied "don't know where Intel will be"?

[Hammer]

A 1.5GHz POWER4+ can outgun a P4 running at 3.2GHz...

"Power 4" out guns "Pentium 4" in price… Some people didn’t learn from Alpha > Pentium Classic days….

[Hammer]

And 'reputable' is irrelevant here since SPEC.org doesn't take any responsibility/accountability for results submitted READ THE DISCLAIMER!

"Accountability" did not appear in http://www.spec.org/spec/disclaimer.html

SPEC only state "... submissions by member companies and the contents of any SPEC reporting page are the submittor's responsibility. SPEC makes no warranties about the accuracy or veracity of this data."

[AmigaMac]

Accountability" did not appear in http://www.spec.org/spec/disclaimer.html

Never said the word 'accountability' was in the disclaimer :-P

[Waccoon]

Hammer:  "Power 4" out guns "Pentium 4" in price…

Too bad that's only the CPU.

[minator]

How can you conclude "IBM will be outgunning them performance wise" then you haven't satisfied "don't know where Intel will be"?

I didn't conclude, I said "I suspect".

New CPUs are always weak due to weak compilers, P4 was a case in point, it was outperformed by the P3 at the begining but then Intel got into gear, shrunk the die, added enhancements and got a better compiler it started to fly.

The 970 has shown it is has legs despite the fact it's brand new.  When it gets a die shrink, better compiler etc. I think things will be very interesting.
The next generation (980 I assume) based on the POWER5 will hopefully get the same enhancements that chip shall have - POWER5 is expected to perform 4X faster than POWER4.

As for where intel will be - well the next gen P4 is now said to be at 100Watts and Intel are talking about water cooling.  If they can't cool the P4 they'll not be able to scale it.  Indeed, notice how the clock rate increases have slowed recently.

"Power 4" out guns "Pentium 4" in price… Some people didn’t learn from Alpha > Pentium Classic days….

Problem with the Alpha was although they had the best technology DEC didn't know how to sell it.  IBM on the other hand do know how to sell which is why things they are spending millions developing things like POWER5 and POWER6 - BTW they are not optimied for speed, they use larger transistors than usual for extra reliability.

Alpha was a market failure, it's technology was -and still is- the best going, which is why Intel were so keen to get their hands on the Alpha team.

[minator]

BTW Alphas are cool:

http://amiga.org/gallery/index.php?n=360

:-D

[downix]

That's an itty bitty one compared to the mobo in my Alpha XP10,000

[Hammer]

As for where intel will be - well the next gen P4 is now said to be at 100Watts and Intel are talking about water cooling.

Notice it’s non-mass market release…

If they can't cool the P4 they'll not be able to scale it.

They have to find an engineering solution e.g. .09 micron die strink.

Addendum :

Refer to http://pc.watch.impress.co.jp/docs/2003/0812/kaigai011.htm (in Japanese) for the Prescott core @3.4Ghz’s planned release for Q4 2003.  

 Indeed, notice how the clock rate increases have slowed recently.

Notice Intel's alternative "Pentium M" release...
Refer to http://www.theinquirer.net/?article=10964

The 970 has shown it is has legs despite the fact it's brand new.

Not quite brand new, refer to "Power 4" ISA.

When it gets a die shrink, better compiler etc. I think things will be very interesting.

Note that both is at .13 microns.

[Hammer]

Problem with the Alpha was although they had the best technology DEC didn't know how to sell it. IBM on the other hand do know how to sell which is why things they are spending millions developing things like POWER5 and POWER6

How could one equal 'R&D' to 'selling'?  Selling/Marketing is quite different to R&D.

[minator]

Quote:
As for where intel will be - well the next gen P4 is now said to be at 100Watts and Intel are talking about water cooling.

Notice it’s non-mass market release…

Since when has Intel done anything other than Mass market CPUs?

The 100W refers to Tejas, and the water cooling is for everything from what I read.
Intel have already talked about how they expect mainstream CPUs to hit 140 Watts on EE Times.

Nvidia said the same for graphics chips.

They have to find an engineering solution e.g. .09 micron die strink.

Refer to http://pc.watch.impress.co.jp/docs/2003/0812/kaigai011.htm (in Japanese) for the Prescott core @3.4Ghz’s planned release for Q4 2003.  

Which will mean they have managed to up the clock 350MHz in a year, not exactly the stellar scaling we are used to.

What happend to moores law???

Not quite brand new, refer to "Power 4" ISA.

The 970 isn't a POWER4, it's based on it yes but there are differences, not least Altivec. Also there are additional instructions for runing 32 bit OSs.

How could one equal 'R&D' to 'selling'?  Selling/Marketing is quite different to R&D.

Where exactly did I equate R&D to selling?

[Hammer]

Since when has Intel done anything other than Mass market CPUs?

That was not point. The point was; economically viability for the mass market. This includes ISV/distribution channels support infrastructure..

Refer to http://www.theinquirer.net/?article=2462
(Intel demos fanless, cool 5GHz chip).
 

What happend to moores law???

This is not first time that they hit a speed wall. Refer to Intel’s Pentium III 1.13Ghz (pre-Tualatin release) era for their attempts to break the barrier.

Also there are additional instructions for runing 32 bit OSs.

What about the earlier 64bit/32bit PowerPC 601?

 

The 970 isn't a POWER4, it's based on it yes
but there are differences, not least Altivec.

Note; Altivec tweaked programs there are available for the PowerMac G4...

Note that Apple used tweaked library for their bench test……

[Hammer]

The 100W refers to Tejas, and the water cooling is for everything from what I read.

What was "Power 4's" watts number again?... (rhetorical question)