May 25, 2012 - 05:30 PM EDT — AAPL: 562.29 (-3.03, -0.54%) | NASDAQ: 2837.53 (-1.85, -0.07%)

64-bit Power Mac G5 the world’s fastest personal computer

Thursday, July 17, 2003 · 9:53 pm · 9 Comments

“Based on IBM

Thank You for supporting MacDailyNews!

♥ Adobe Photoshop CS6 Extended for Mac - Student and Teacher Edition DVD (65171323) only $249!
♥ Blowout: Apple 15.4" MacBook Pro i7/4Core/2.0GHz/4GB/500GB (MC721LL/A) only $1,549.99 + Free Shipping from MacMall
♥ Mac Madness Save 25% on Parallels Desktop 7!
♥ 15% Off Compatible Ink. 10% Off All Others (excludes OEMs). Free Shipping over $55. Coupon: 123BLOSSOM, Expires 6.18.12
♥ Save up to 60% at the Hammacher Schlemmer Special Values Sale-While supplies last.
♥ Dragon Dictate for Mac 2.5 - New! Simply Smarter Speech Recognition
♥ Up to 30% OFF Accessories from Verizon Wireless
♥ Limited Time Offer: Get free ZAGGsmartbuds when you buy a ZAGGsparq!

Comment Feed

9 Comments

Fred Mertz

Thursday, July 17, 2003 - 11:13 pm · Reply

Good, clear, detailed and concise article from Helm. Funny, though, that all of a sudden, after Apple blows the doors off Wintel’s best, “SPEC benchmarks are arguably a slightly unfair�not to mention unrealistic�measure of performance” whereas before the G5 announcement, “Wintel Sufferers” ate up the SPEC benchmarks as if they were manna from Heaven.
Stingerman

Thursday, July 17, 2003 - 11:57 pm · Reply

What is even funnier is how Intel proponents fail to admit that their SPEC benchmark results were fudged to use their SIMD instead of the FP Unit the SPec was supposed to be testing. So the Intel ICC results are rigged, since they really show vector unit. Apple is only showing the Scalar results in their tests. However, if this was a vector unit verses vector unit comparison, the G5 would also win by a landslide.
NoPCZone

Friday, July 18, 2003 - 1:05 am · Reply

Wait until we see the benchmarks of the Dual G5 WITH the release version of Panther. My understanding is that the Benchmarks are with a special release of Jaguar (10.2.7), and that 10.3 boots and runs faster even on older iron (400mhz G3, etc). Imagine what the new faster OS is going to do with this monster. And consider aside from benchmarking that Adobe and others are prepping updates for software to take advantage of the expanded capcities of the G5. With the arrival of 10.3, OS X will finally have arrived as a mature OS. The Wintel partisans can gripe for a little while longer, but when Panther comes out, it’s gonna be over. The fat lady isn’t singing yet, but she IS running scales warming up.
Cliff Goolsby

Friday, July 18, 2003 - 1:14 am · Reply

Am I wrong here? I thought that 32-bit architectrues have a 4GB limit because they have a 32-bit wide bus (2 to the 32nd power is 4.29GB). Given a 64-bit wide bus we are not limited to 8GB. 2 to the 64th power is 18.4 x 10^18. That’s not 18 gigabytes. Not 18 terabytes. Not 18 petabytes. It is 18 exabytes. It’s huge. This is the addressing range of a 64-bit bus. It is only Apple’s current board architecture that limits us to 8GB…yes?
Aryugaetu

Friday, July 18, 2003 - 1:52 am · Reply

Cliff, it’s physical space and design not just math. But to answer your question, Apple did so at http://www.apple.com/g5/architecture.html …

Mindboggling Memory
Even elephants would be hard-pressed to keep track of as much data as the PowerPC G5. With 8GB of main memory, the PowerMac G5 can hold a gigantic 3D model, a complex scientific simulation or a sequence of HD video entirely in RAM � drastically reducing the time to access, modify and render such data. That�s because it�s 60,000 times faster to access RAM than your hard drive, and you can stream 40 times more data between RAM and the processor.

Theoretically, the G5 could address up to 18 billion billion bytes of virtual memory. Or 18 exabytes for short. But in practice, the PowerPC G5 has 42 bits of address space for memory, which means it supports just 2^42 or 4 terabytes of system memory. You can�t put that much RAM in a desktop computer � yet � but it sets the stage for untold computational feats.
Frank Williams

Friday, July 18, 2003 - 1:59 am · Reply

While the tone of the article is great for Apple/Mac users (me being one of them since the late 70s), there are inaccuracies which those adamantly opposed to the Mac platform will use to label the article to be worthless, i.e., “The author does not know what’s really going on. He can’t get the facts straight. Therefore nothing he says is worth while.”

We’ve all seen it happen before. Wintel die hards will pick at the smallest flaw so discount the entire message. For example: “…in contrast, 64-bit devices can access up to 8 GByte.” Wrong. The G5 actually has a 42 bit address path allowing a theoretical limit of 4 terabytes. A machine with a 64 bit address space would be able to have, theoretically, 16 exabytes of memory. (If I recall correctly, the original POWER chipset had a 48 bit address space allowing it to address up to 256 terabytes of memory — a truly astounding amount back in 1989!) It is only Apple’s design which limits the PowerMac G5 to a mere 8 Gigabytes. (This is much like the PowerMac G4s could theoretically address up to 4 Gigabytes. However, all but a few G4 machines were limited by Apple’s design to no more than 1.5 Gigabytes of memory.)

Another example is “a 1-GHz frontside bus, allowing data to be delivered to the processor at up to 8 Gbit/s”. Actually its 8 GB/s (bytes not bits). Even with this the 8 GB/s is a little bit superfluous and misleading. It is only 4 GB/s into the processor and 4 GB/s out of the processor. (They are really only 32 bit wide in and out data busses running at 1 GHz.) Thus for very compute intensive vector processing operations or even very compute intensive double precision floating point operations (two 64 bit results every clock cycle) the G5 chip (aka PowerPC 970) is still starved for data.

There are other errors besides these.

Don’t get me wrong. I love the overall message. I just wish he’d have checked his facts a bit more closely and proofed the article a bit better. It would then be much more difficult to attack it.
john

Friday, July 18, 2003 - 7:06 am · Reply

This is a great chip. Why not port it to windows so PC users can see what they are missing?
no1allowed

Friday, July 18, 2003 - 7:55 am · Reply

“This is a great chip. Why not port it to windows so PC users can see what they are missing?”

Tell it to Microsoft. Nobody but Microsoft can port it to Windows.

“it�s 60,000 times faster to access RAM than your hard drive,”

It’s actually a little faster than that. 400MHz memory is accessed in 2.5 ns whereas a typical hard drive is accessed in 12 ms. Closer to 4.8 million times faster
Jocknerd

Friday, July 18, 2003 - 8:44 am · Reply

I’ve never heard of a chip being ported to software. Its the other way around. Maybe Microsoft should port Windows to the G5.