Slashdot Mirror
New Power Mac G4s Announced
benh57 writes "Apple today announced the new Power Mac G4 towers with new faces, running at dual-867MHz (US$1,699), dual-1GHz ($2,499), and dual-1.25GHz ($3,299). All are running DDR, the two higher end models at 166MHz FSB with Radeon 9000, the low end at 133 w/GF4MX." Check it out at The Apple Store, and keep your eyes peeled for an appearance on the Power Mac G4 site.
Thats front page news - apple becomes the first PC maker to go totaly to dual processor in it's pro desktops.
This is an excellent move for Apple. Solid UNIX high-end workstations making use of multiple processors, as a robust, pre-emptive multitasking system should. Let's hope Apple finds some what to make the public aware of this singificant spec of their tower machines.
Sadly...my dual G4 800 may be getting grey hair...
blakespot
-- Heisenberg may have slept here.
iPod Hacks.com
Much like the Xserve, it looks like they've managed to hack DDR onto the G4 processor, and its still running on a bus which is not doubled like the Athlons. Running the bus at 166mhz should make up for that a little bit though. Interesting that after all the commentary on the new cases having huge (7lbs) heat sinks, I couldn't find any images of them on the apple site. They must really be that big.
This page says "The faster-than-light processor speed gets an additional boost with an advanced cache memory architecture that provides ultrafast, dedicated [blah blah blah...]"
Not that I expect much truth-in-advertising from computer vendors, but isn't that a bit much?
Oh, well...
I wonder if they count spring-loaded folders as one of Jaguar's 150 "amazing" new features... not that I didn't think it WAS pretty amazing in OS 8, but...
"How to Do Nothing," kids activities, back in print!
It's very likely that apple are pumping out dual g4 systems to simply get rid of the remaining g4 chips, (like how you get them in everything now except the iBook). This could on the opposite side be apple introducing speed updates and time for an even longer wait for the G5.
The other shoe that dropped today was that they've now gone full-tilt with the eMac, adding a Superdrive and running it at 800 MHz for the same price ($1499) as the 15" Combo drive iMac.
Meaning that unless you really like the cool look of the iMac, you can save a couple of hundred dollars by getting an eMac instead, without giving anything else up (I believe they're based on the same motherboard spec) besides the cool screen. And the eMac has a pretty decent screen.
I've been leaning towards getting an iMac in the fall to replace my wife's old iMac DV 450 (we could use the DVD burner to make movies of the baby), but assuming no other drastic changes I'd be inclined to go with the eMac now instead. And Apple is steadily returning the CRT to it's place as the lower-end anchor even though LCD prices are starting to drop again (they also reduced the prices of all the other iMac configs). That's interesting.
Basically, I'm going to be watching the early fall with great interest - once these new configs are well-established there'll probably be some speedbumping of the whole line around October or so. My guess is that the iMac and eMac could hit 1 GHz, the PowerMac towers will start at 1 GHz and go to either 1.4 or maybe as high as 1.6 (Moto is supposedly sampling the 1.6 part now), and the PowerBook will probably get a speedbump to, say, 933 MHz at that point, too. They may not all be at once, but those are the next logical steps, and I'd expect to see them all before years' end (and before Christmas season, in particular).
-- Josh Turiel
"2. Do not eat iPod Shuffle."
Why would you NEED a Zip drive as a built-in option when you can burn a CD that holds over 6 times as much?
You can still buy a USB or FireWire Zip drive and connect it externally, but now Apple doesn't dedicate a place in the case that is a waste of space for anything other than a Zip drive.
Vos teneo officium eram periculosus ut vos recipero is.
It's all about the software though. 10.1 is superb - 10.2 is alleged to be a big improvement.
Who cares how fast your G4 is clocked? It won't run OS X.
Uh... is there some joke in here that I haven't gotten?
Incidentally, 10.2 is a big improvement. I've been running 6C106 (the release was 6C115) for about 10 days now. It's heaven, really peppy even on 400-500 MHz G3 systems.
If anything, the vents make me feel it is less likely these things will howl like the previous Dual GHz model. The extra holes mean better ventilation, better ventilation means the fans don't have to work as hard. Apple has almost certainly heard the complaints about the previous model and done their best to get rid of the most of the noise. I'm going to buy either the 867 or the GHz model, probably today. The one thing that worries me is the good chance for new model bugs, especially given that these are obviously new mobos and there have been vague reports of Xserve problems.
(Not to mention the fact that when Apple quotes gigaflop figures they are talking about all-in-registers zero-pipeline-stall vectorized operation, not actually doing anything useful -- like reading from memory.)
Um. I'm no expert, but to me that sounds like any cache-resident vector function, like a 5x5 convolve or something. You take a small performance hit when you have to load the next cache line, but if you're lucky your pipeline is deep enough to keep the processor units going while that fetch happens.
I mean, how else are they supposed to quote processor performance if it's not this way? If you want them to talk about performance of the whole system, taking things like memory and busses into account, they're going to have to pick a real-world application to test with. They do that already, using Photoshop as their benchmark the same way the graphics board companies are using Quake as theirs. Apple's test shows the dual 1.2 GHz machine to be about 90% faster (or almost twice as fast) as a single-processor 2.5 GHz P4. And yet Apple still gets hell for using Photoshop as their metric.
Seems like you can never satisfy everybody.
The Achilles heel in all this is backup, especially for Mac OS X Server. Every other version of UNIX out there has a built-in backup solution (except, unaccountably, Linux, which has no dump/restore, last I checked). Mac OS X has dump/restore too, but they only understand the UFS file system. Apple rewrote 'fsck' to understand about HFS+ file systems, but not dump/restore.
That leaves Retrospect as the only sensible solution for backup: a third party product. And the regular Retrospect Mac OS X client won't dump a Mac OS X Server system! Instead you have to spend $800 (!!!) for the Server backup software. That software will also dump Windows 2000 and NT workstations, whoop-de-do.
Whatever happened to UNIX as a self-hosting, self-supporting system? Gaaaah. I'm thinking hard about wiping our Mac OS X Server machine and just installing the regular Mac OS X, where at least we can afford the backup software.
Or maybe just dumping Macs entirely and going to FreeBSD on a dual-processor Xeon box. All hail Amanda! At least I could back up a box like that.
Um. I'm no expert, but to me that sounds like any cache-resident vector function, like a 5x5 convolve or something. You take a small performance hit when you have to load the next cache line, but if you're lucky your pipeline is deep enough to keep the processor units going while that fetch happens.
Sorry, but I happen to be an expert on these things, and those 18 Gflops are just a theoretical peak of perfectly combined multiply-add operations in the vector unit. As soon as you can't match an add and a multiply your performance would drop in half. More important - that's the THEORETICAL peak. You will never see anything close to it in practice. Apple's own version of FFT performs at 1-2 gflops.
And the vector unit can't even do double precision, which is kind of important in science. For a double precision FFT the performance is LESS THAN HALF that of a current Athlon CPU.
And yet Apple still gets hell for using Photoshop as their metric.
Nobody is blaming Apple for using Photoshop. The problem is that they don't provide any details. In practice, it looks like they are only testing 3 or 4 filters that are heavily Altivec-optimized, so it is not typical for Photoshop performace. When third parties perform benchmarks based on a large set of Photoshop actions, the Apple machines are
much slower than current x86 offerings.
But what the $&#*@ is up with that case??
The case is a continuation of an award-winning design, that has proven very popular. It maintains the carry handles (believe me, if you've ever dropped a computer, you appreciate the handles), and the ability to open the case easily.
Speaking as one who had to dismantle and reassemble PC's on a daily basis in my misspent youth, I'd be pretty annoyed if Apple had changed the "put most of the guts on the door" design.
-jcr
The only title of honor that a tyrant can grant is "Enemy of the State."
What were you hoping for? An iMac-style dome?
Information wants to be anthropomorphized.
...all the tests I have read have declared Firewire the winner. Explaination, USB 2 relies on your processor for the spead of throughput, Firewire has its own chip...nuff said..