IBM to unleash new PowerPC chip

[Hammer]

Sounds good. Yet another “wait” (sigh)…

"losing the gigahertz race, but with all the performance (IBM) packed into this architecture, it's going to be competitive with the performance of the Pentium 4 in 2003."

I just hope the corresponding PPC motherboards are just as completive to the X86 motherboards.

I wonder, IF next release PPC is really completive, why not Big Blue abandon X86-32 PCs for PPC PCs?

In the grand scheme of things, the current Pentium 4 core (even next hyper-threading enabled cores) is predicted to lose with the arrival of AMD's Hammer (X86-64) class CPU, unless Intel surprises the world with Yamhill/AMD X86-64 clone like X86-64 project.  

Interesting to note that IBM's  SOI .13 micron fabricated chip is clocked ~1.8 Ghz, while AMD’s non-SOI .13 micron fabricated chip is clocked at ~2.2 Ghz now.

Although IBM will disclose technical details on the chip's architecture at the Microprocessor Forum, pricing and other commercial details won't emerge until the chip ships in the second half of next year.

IF next-PPC release is at the second half of 2003, it would be competing with AMD’s Hammer class (X86-64) chips instead of Pentium 4. Pentium 4 wouldn't be considered the X86 flagchip when the Hammer is release. Intel’s IA-64 move has basically abandon X86 leadership to AMD.

Shooting the “Pentium 4” in mid-2003 is like shooting a sitting duck in the light of AMD Hammer class CPUs.

[Hammer]

Well, if IBM release a chipset as well then there's hope. Otherwise this processor will most likely power iMACs in 2003/4.

I don't think one could built a cheap PPC white box PC from Apple's Hardware...

Let's hope companies like MAI are interested in developing chipsets and motherboards for this cpu.

PPC market needs a cheap X86 style motherboard manufactures (I’m thinking more inline with **MSI/Gigabyte/Asus/PCChips/PC Partner/Acer/Tyan style companies).  

I don’t think “Big Blue” sell their X86 desktop motherboards directly end users (try looking for an IBM branded X86 motherboard as an example).

Having VIA/SIS level companies is nice, but the chain wouldn’t be complete with  cheap board manufactures**.

It would excellent proposition if Big Blue deliver near Nintendo Game Cube priced PPC motherboard (e.g. AUD $250~$320 range).

Highly unlikely for IBM to abandon the Wintel platform for PPC PC's unless M$ release PPC versions of their Operarting Systems like the old NT 3.51 days.

I recall, MS  Windows NT 4.0(up to Service Pack 3) was available for PPC (the company I work has a PPC edition (via old MSDN subscription).  NT 3.51 days is long dead by time of Windows NT 4.0.

As this news indicate
http://www.theregister.co.uk/content/61/27413.html
IBM, MS and Intel are working again for 16 XEON based MS Windows Servers.

I guess they have almost buried their OS/2 vs MS Windows disagreements. Remember, IBM/Intel/MS are the original unholy trinity during 80s.

[Hammer]

@IBM Press Release

52 million transistors would be in the range of Pentium 4’s transistors count. It’s more than the current Athlon XP(~37 million transistors).

I don’t think that chip is for cool computing when compared with the current PPCs.  

Prototype Hammer based motherboards is said to have 800Mhz FSB (via AMD’s hyper transport tech).

[Hammer]

The big problem is that when the 970 lands, the performance of the x86 may already have crawled far ahead of it again.

This press pelease may re-assure PPC fans, but it may give the competitors a target to beat.

IBM should have learned from ATI’s product release style.

[Hammer]

I'm happy to see new PPC chip and I hope they do surpass Intel some day..

A high clock speed (e.g. 1.8Ghz) ~52 million transistor fabricated chip is well within the abilities of Intel i.e. current product range is @ 2.8Ghz(soon to be 3Ghz) clock speed for ~55 million transistor fabricated chip.

Can IBM complete with Intel IF Intel have a licensed (or cloned) PPC-64bit chip (e.g.. Intel has license of ARM CPU designs and it’s currently the world’s fastest ARM chip when compared to Motorola’s ARM chips)?

[Hammer]

Helk even a G4 at 1 GHz is comparible to a P4 at 2 GHz in most real world benchmarks... these were quite interesting with the new 1.25 GHz G4s:

http://www.barefeats.com/pentium4.html

I think the new Power4 based PowerPC 64bit CPU will do much better!

That's hardly the "most real world benchmarks".  Any reasonable statistics must have at least n>30.

Refer to
1. http://www.geocities.com/sw_perf/
2. http://www.macosxhints.com/article.php?story=20020113045343563
3. http://www.barefeats.com/graf31.html

Interesting that barefeats has tested a current PPC flagchip VS non-X86 flagchip .

Better range  of benchmarks
1. Photoshop  
2. Maya  
3. Lightwave  7.5
4. Cinema4D  
5. Mathematica  
6. OGR  
7. SETI
8. RC5  
9. ArsTestBench
10 OpenSSL
11. Quake3 Arena

One has to look at the overall perspective i.e. soon to be released Barton Core Athlon XP with L2 512kb (last Q4 2002 release), which may give Athlon XP some extra IPC jump.

I can't wait to see some benchmarks of AOS4 and the AmigaOne in action!

Are you implying that AmigaOne HW can mount IBM’s new chip?

[Hammer]

Well PPC surpassed x86 when it was conceived

So you comparing a 8086 to Power1? Intel back at time didn't apply RISC techniques to their CPUs.  

and it will do it again since the emphasis is more focused on performance than energy efficiency!

“Energy efficiency” was a recent re-positioning for the said product.

"Energy efficiency" was not even an issue during the launch of PPC 601(max IPC) vs Pentium Classic (max compatibility).

[Hammer]

Well PPC surpassed x86 when it was conceived and it will do it again since the emphasis is more focused on performance than energy efficiency! The only reason Intel has a leg up on IBM/Motorola is the total I/O implementation built around the CPU, which is more up-to-date and has 3 times the bandwidth (RAM/FSB)...

You realise Pentium III(considered to have a better IPC than Pentium VI) communicates with outside world @133Mhz. Anything greater is useless for this chip.

imagine if the G3s and G4s had that bandwidth to play with?!

It would be pointless since the current G3 and G4 communicates with the out world side @133Mhz.  

Secondly, G4 has a massive 512kb L2 cache to compensate some of the problem e.g. OpenSSL's core routines should fitting nicely.

Most games store their textures within GPU’s RAM, both modern Apple and old Intel Pentium III has AGP 4X.  It reduces texture fetch action from the main ram.  

Note that Athlon XP has yet to have the 512 L2 treatment unlike the G4.

More news about the 64bit CPUs
http://www.eet.com/semi/news/OEG20021014S0059

[Hammer]

Actually I'm talking about the PowerPC 601!

You did not state it.

Intel hasn’t implement their post-RISC architectures at that time(mid 90s).  

But it has been the last few years due to targeting the embedded market and has been the primary focus of IBM (and even Motorola) until now!

What was the "PC" stands for in "PowerPC" label again?

The Pentium 4 has 512kb L2 cache as well!

Refer to Pentium III with L2 512kb  vs Pentium 3 with L2 256kb examples. Pentium III with L2 512kb has a better IPC compared the older Pentium III.

 

Yeah but the Athlon MP is the only one that is designed with its memory cache cross-strapped... I think that's a great advantage!

I don't recall 512Kb worth of Cache memory  was onboard a single CPU.  In some bench test MP solution didn’t perform as good when compared to a single Athlon XP solution. Dual CPU has to deal with SMP overheads.

Well unless EyeTech has changed its CPU strategy, I am assuming the G3s/G4s will come as expected!

What’s the point of this statement?
==========================================
I can't wait to see some benchmarks of AOS4 and the AmigaOne in action!
==========================================
The tread was about IBM's new chip not the existing solution.

Doubt that the A1+AOS4 solution will significantly increase their IPC?