Slashdot Mirror
New G5 Power Macs "Fastest Desktop In The World"
Beyond the many numbers, the bottom line is that the new machines have a new architecture, and that the memory speed is now the bottleneck, not the processor or bandwidth speeds. So they can have up to 8GB of 128-bit DDR RAM, as it is efficient to keep data in memory. The memory bandwidth is one of the most talked-about features of the new architecture.
USB 2.0 is now included, as are FireWire 400 and 800, Bluetooth, AirPort Extreme, and digital audio in and out. The 4x SuperDrive is now standard, and it can house up to 500GB of internal storage.
For video, the GeForce FX5200 is standard on low-end models, Radeon 9600 Pro on high-end models.
The case of the new machines is redesigned too, from the ground up, focusing on decreasing noise and heat. It is an aluminum enclosure, with ports for FireWire and USB on the front, and a door on the side to get into the box. It has four distinct "thermal zones" with computer-controlled cooling with its nine (yes, nine) independent fans. And it is much quieter than its predecessor.
The G5 is 10 percent slower than the P4 and Xeon in SPEC int scores in single-proc units, but 20 percent faster in FPU scores, and the dual-proc G5 beats the dual-proc Xeon in all SPEC scores.
The models are a single 1.6 GHz ($1999), single 1.8GHz ($2399), and dual 2GHz ($2999). They will ship in August. A 3GHz processor will be available from IBM in 12 months.
Apple notes that recompiling apps for the 64-bit architecture is easy, and in some cases can be done in minutes.
There was no word about the heavily anticipated redesign of the 15" PowerBooks.
Thanks to iPalindrome on irc.arstechnica.com for his running transcript of the keynote address.
Power Mac page
Apple store
http://www.beosjournal.org/wwdc/
for some pictures of the new case.
user@host$ diff
RDRAM last time I checked had higher total bandwidth than DDR, but fails to be faster where it counts - latency. Latency on non-sequential read/write is where the memory bottle neck is.
Why?
If that's not the case, I as a consumer would be confused indeed.
Ñ'
"Sun should be very scared. Their Dual 1.2GHz 64bit offering is $14,995. Ouch!
"
Oh DO try to at least pretend to keep up.
$5,795 for sun's dual 1GHz 64-bit system (with 2GB of RAM, not the 512M G5 Macs start with).
Out of interest, how much is it for the 106-CPU version of the Mac, again?
Clue - Sun and Apple are targeting completely different markets.
People should not be afraid of their governments - Governments should be afraid of their people.
I'd like to see some independently-verified benchmarks before I believe that it's the "Fastest desktop in the world". I seriously doubt ol' bullshitter Stevo would tell the full truth.
Well, the problem is... Steve is telling the truth.
Go to www.spec.org and look at the SpecINT and SpecFP ratings for the Power4 (single core benchmark).
Okay, the PPC970 is based on this core and yes, at 1,6 Ghz it runs around an 3 Ghz P4.
Okay, now take a look at the SpecINT and SpecFP ratings for the alpha 21264 and 21364.
Those processors are a real match for the P4.
With a 1.5 times slower clock they are as fast as most 1.5 higher clocked P4's.
The thing is, that intel doesn't have a decent 64 bits processor.
Their itanium II is a joke with a performance which is equal to most 64 bits processors 2 or 3 years ago.
Contrary to intel ibm knows how to build fast 64 bits processors without all the tradeoff's intel had to make with the P4.
Second, if you look at the price of the PPC970 and compare it with the P4 you will see that the P4 is almost 2 times as expensive as the PPC970.
Let's face it, at the moment there is no 64 bits or 32 bits processor available which is faster than the PPC970 (i mean for desktop systems).
It will take intel at least more than a year to get the itanium near the PPC970 2 Ghz..
But then they are no match to the PPC970 3 Ghz. which will be available then.
I hate to bust your bubble, but there is no such thing as SMP P4. Intel designed the P4 to be only single processor. Xeon is for SMP applications.
Also, with SMP you can't just double the speed of one chip to come up with a benchmark. You double it, and take 10-15% off the top. You see, there is overhead in SMP because the two processors need to communicate to make sure they are on the same page, so to speak.
- FinalCut Pro
- Logic Platinum
- Shake
When you're throwing around cinematic quality film clips, the more power the better.According to Apple's website, they're specing against a dual 3.06Ghz Xeon.
See for yourself.
concrete5: a cms made for marketing, but strong enough for geeks.
but strangely only to a SINGLE-CPU P4 machine?
Gee why could that be?
BECAUSE you can't have anything BUT a single P4 machine. There are no dual P4's - the chip just doesn't support multiprocessing.
Summary: It not only beats up the P4 and Xeon, it takes their lunch money as well.
It did even better at DNA matching: "Testing BLAST with common searches using a word size of more than 11, the Power Mac G5 far outperformed the Pentium 4-based system and the dual Xeon-based system, and nearly five times faster at the long word length of 40."
Lawrence Person (lawrencepersonh@gmailh.com (remove all "h"s to mail)
http://www.lawrenceperson.com/
Sun Blade 2000
The Single 900MHz is 7,595
The Single 1.2GHz is 9,995
The DUAL 1.2GHZ is 13,995 (whoops off by 7.5%)
What about Dual 2.0GHz don't you understand? They may not have been in the same market before now. However, that will soon change (there is your clue). As far as the 106 CPU Version Cluster the XServes the same way Sun does it. I said Sun should be scared. They no longer have a lock on the 64bit market.
I guess things never change in your world. Look out someone may be moving your cheese.
It looks more like an electic razor to me.
Unfortunately, it looks like they've abandoned the easy-access pull-down door that let you add ram and add-on cards with ease.
From the Apple web site:
In other words, they didn't just keep it, they improved it.
Information wants to be anthropomorphized.
I sat next to three guys from Dassault at the keynote, and about two rows back from a whole busload of guys from PTC. CATIA and the Pro-E products are going to be out on OS X by the time the G5's ship, or shortly thereafter.
Safari 1.0 is now available through Apple's software update.
The new version seems noticibly faster and has no bug button, but there is still a "Report Bugs To Apple" option under the Safari menu.
Integrate Keynote and LaTeX
Because the Mac is cheaper... go to dell.com and configure a dual Xeon 3.06 with a DVD burner and the same vidcard/RAM...
Guess what, the Mac is over 1k cheaper! That's why... better performance, and a cheaper price...
_CMK
Bad spellers of the world untie!
Well, Photoshop, for one thing...yes, Macs are still used for graphics, dontchaknow.
Try editing CMYK graphics at 600 or 1200 dpi for high-end print work sometime. With layers. And masks (which are essentially added layers). Running filters. The whole she-bang.
Such a file can easily get into hundreds of megabytes in size, and Photoshop generally needs 2x to 3x as much RAM as the actual file size to efficiently work; even then it starts to bog down at those file sizes.
My dual G4/450 with 1.5 GB RAM and Radeon 9000 already gags on that enough so that it's a hassle when I have to design and edit that kind of stuff. Believe me, I'm going to be first in line as soon as I scrape together the $2500 or so for a new G5 system with added RAM (the more RAM, the merrier -- Photoshop is VERY hungry for RAM).
Not to mention video editing and 3D, both of which are markets that the Mac has generally been strong (if not dominant) in for some time.
I might add that you could ask the same question about P4-based PCs. Who needs that kind of firepower? Not many (mainstream) people, really -- aside from perhaps gamers. The vast majority of users just do e-mail, web surfing and word processing, maybe a little photo editing. A P2 or P3 running Linux or an older version of Windows would be more than enough in those cases. Hell, even an old Pentium with a smallish Linux installation would be enough in many cases.
OTOH if you give users and developers the added power of new processors and mainboards (strange that HyperTransport hasn't gotten much mention here), people will find a way of using it. One example: Apple's predicted that video editing will be the next mainstream computing revolution, like desktop publishing was twenty years ago. If you think about it, they're probably right.
Most newer computers can easily handle basic video editing now; the question is just how to make it easier for Joe Sixpack to edit his family videos (and maybe make Junior a budding David Lynch).
Cheers,
Ethelred
Everyone wants to be Ethelred. Even I want to be Ethelred.
Except that the Octane's bus is theoretically much, much faster. It has an end-to-end point speed of only about 3 and half GB/sec, but it can connect any of the individual systems to each other simultaneously at full speed; the memory can talk to the processor while the processor writes to the disk subsystem while the video card...and none of it ever has a collision and can operate at Crossbar's full point-to-point speed without effect from other subsystems.
Not only that, but as you add processor modules(which if I remember right, have memory on them?), you add Crossbar bus bandwidth; adding modules adds extra Crossbar channels(I think. It's been a long time since that technology briefing).
It's a quad-processor-capable system- so I don't think you are giving it a very fair shake; on a 4-processor system, I think each processor would have about +14GB/sec access to anything in the system(including memory), which is just a few GB shy of double the G5 which can only manage 8GB/sec for access to main memory. Oh, and let me remind you Crossbar is 5-6 years old...
Thanks, but if I want to push around multi-gigabyte datasets, I'll take the Octane. I find Hypertransport, at only 16 bits wide, destinctly unimpressive...
Please help metamoderate.
According to Apple's web site, they tested their machine against two Dell Intel boxes (Dell Dimension 8300 (P4) and Dell Precision 650 (Dual Xeon)) running Red Hat Linux 9.0 Professional (at Apple's request).
.PDF format) including all hardware and software used is available from Veritest's web site.
Intel states that Red Hat Linux 9.0 Professional is one of the Linux OS's currently available that "include optimizations for HT Technology and are currently eligible to carry the Intel® Pentium® 4 Processor with HT Technology logo".
Apple commissioned the benchmark from a company called Veritest. The full report (in
This could make Intel take notice! Of course, this benchmark comes on the same day that Intel announces the 3.2GHz Pentium IV (and Xeon) processors. Go figure!
Of the published data on both (in SPEC processor benchmarks), Apple's Power Mac G5 generated a SPECfp_base2000 score of 840 and SPECint_base2000 score of 800, while Intel claims that their new 3.2 GHz processors get a SPECfp_base2000 score of 1252 and a SPECint_base2000 score of 1221.
And the SPIN goes on!
-Joe
If we're all god's children, what's so special about Jesus? - Jimmy Carr
For students and/or educators (personal purchases), the Powermac G5 line goes like so in prices:
1.6GHz - $1,899
1.8GHz - $2,299
Dual 2GHz - $2,849
The discounts are consistent with previous Apple academic discounts. These are the same configurations as the corresponding non-educational priced retail systems:
1.6GHz - $1,999
1.8GHz - $2,399
Dual 2GHz - $2,999
-Joe
If we're all god's children, what's so special about Jesus? - Jimmy Carr
First off, GCC is probably better optimized for x86 then it is for the PPC 970 by virtue of the fact that it's been running on x86 for so much longer. So, even using the same compiler, the field is still tilted in the direction of x86.
Second, the test is of the speed of the processors, not the quality of the optimizing compilers for them.
Third, the "fastest" comment was made with respect to the dual-processor configurations. The numbers you site are for the single-processor version.
Yes, in single-processor land Apple lost in intspec by about 10%, but won in floating-point land by about 30%. This is using a compiler that is better optimized for the competitor. And they still came out ahead.
In dual-processor land they came out ~10% ahead in integer land and over 40% ahead in floating-point land. A tremendous difference.
The real-world tests they performed seemed to back up these results with Photoshop, Mathematica and a few other programs running an average of 2x faster on the PPC 970.
This may sound incredible, but it is just a matter of bandwidth, and the G5 has plenty of it.
The dual-processors have completely independant busses, a 1Ghz FSB, 400Mhz 128-bit DDR memory, two independant floating-point units and two independant integer units. The PPC970 is capable handling over 120 in-flight instructions, that is, instructions which can be worked on and processed in parallel. In P4-land only a few dozen instructions are can potentially be run in parallel.
Do you really think that Apple would hire a company like VeriTest to verify their results and then lie about them? If they didn't actually have better spec scores they just wouldn't have used those tests...
Justin Dubs
Combo Drive (CD-RW/DVD-ROM) [Subtract $200]
Timeo idiotikOS et dona ferentes
Quote: Except that the Octane's bus is theoretically much, much faster. It has an end-to-end point speed of only about 3 and half GB/sec, but it can connect any of the individual systems to each other simultaneously at full speed Uh, for those of you on the short bus, Apple's new memory chip is also point-to-point. From the G5 (system, not chip) white paper: Advanced System Controller A new system controller is central to the overall performance of the Power Mac G5. This revolutionary application-specific integrated circuit (ASIC)â"one of the industryâ(TM)s fastestâ"is built using the same state-of-the-art IBM 130-nanometer process technology as the PowerPC G5 processor. A superefficient point-to-point architecture rovides each primary subsystem with dedicated throughput to main memory, so massive amounts of data can traverse the system without contention for bandwidth. In contrast, subsystems that share a bus, as on other PCs, must deal with time-consuming arbitration while they negotiate for access and bandwidth across a common data path.
Stand back. I've got a brain and I'm not afraid to use it.
Dude, ain't no such thing as a dual processor P4. They. Don't. Exist.
I am a believer of momentum and curves.