Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
2.6GHz PowerMac G5s at MacWorld SF 2004?
- Thread starter MacRumors
- Start date
- Sort by reaction score
- Status
Leet1 needs to look a little deeper into the benchmarks before he/she goes off on a Mac board about how superior the Opteron/Athlon64 is.
try this on for size...
FFT Benchmarks
It's a link to some Fast Fourier Transform benchmarks (set up by some guys at MIT). The G5 eats up the Opteron, sometimes with nearly double the performance. FFTs are pervasive in research computing.
This just goes to show that you can pick out a benchmark to make just about any processor faster than another. Benchmarking Opterons against G5s in Word and Premier is a total joke.
The facts are, Opteron/Athlon64 is a great processor. The 970 is also a great processor. The Opteron has the on-die memory controller, The 970 can retire MORE instructions per clock cycle than an Opteron... Both architectures have their advantages and dis-advantages.
The reality is, the 970 was picked over the Opteron, Itanium, and Xeon because of price, availability, AND performance benefits.
G5s really are wicked research machines, even if they are somewhat lacking in MS Word performance :snicker:
I hate to beat a dead horse here... but as far as i've been able to tell, the G5 requires more optimization (over previous architectures) than the Opteron requires (over older x86 32bit code). This isn't because the 970 runs 32bit code worse, but because it is based off a different (Power4) architecture, so it 'does some stuff differently' than the G4 and G3. it's also much wider... FP code compiled for G4s are only expecting half the hardware to be there...
Pretty much all Mac software is effectively crippled on the G5 right now. OS X and Mac OS X programs run really fast on the G5 right now, but they will only get better as GCC comes up to speed, and as developers get their hands on truely optimized compilers like IBMs xlc and xlf. xlc sometimes generates code that is 70% faster than gcc, and xlc for the 970 is still in beta!
2003 was a good year for Mac users. 2004 will be unbelievable.
jmho.. ffakr
Re: you can't have dual AGP ports
Uhm, actually, AGP is a bus with a single port. You can run multiple AGP buses in a system, but there are design considerations to be had (since an AGP card can access RAM directly, you make your controllers more complex and expensive as you add more devices that can access RAM, not to mention degrade performance).
So yeah, a quad monitor video card is more likely, or just get a dual head AGP card and a dual head PCI card.
Uhm, actually, AGP is a bus with a single port. You can run multiple AGP buses in a system, but there are design considerations to be had (since an AGP card can access RAM directly, you make your controllers more complex and expensive as you add more devices that can access RAM, not to mention degrade performance).
So yeah, a quad monitor video card is more likely, or just get a dual head AGP card and a dual head PCI card.
Hmm My Predictions
MWSF 2004
What i expect
SP 2.0Ghz G5 $1799
DP 2.25Ghz G5 $2499
DP 2.5Ghz G5 $2999
Optomistic/Hopefull
SP/DP 2.2Ghz G5 $1799
DP 2.4Ghz G5 $2499
DP 2.6Ghz G5 $2999
Either way i will probably be buying whatever is the low end tower(so as you might guess im REALLY Hoping for DP 2.0-2.2Ghz on the low end, as that would really be nice).
MWSF 2004
What i expect
SP 2.0Ghz G5 $1799
DP 2.25Ghz G5 $2499
DP 2.5Ghz G5 $2999
Optomistic/Hopefull
SP/DP 2.2Ghz G5 $1799
DP 2.4Ghz G5 $2499
DP 2.6Ghz G5 $2999
Either way i will probably be buying whatever is the low end tower(so as you might guess im REALLY Hoping for DP 2.0-2.2Ghz on the low end, as that would really be nice).
For anyone interested in dropping off
G4 400 or so Mhz, i'll be accepting these machines for the everyday great price of $5 per usable CPU to start the ultimate G4 400Mhz CLusteR. WIll be called ClusTer's Last Stand and that puppy is gonna be 5000 CPUs strong by the end, and oh yea, it'll almost be as fast as the new dual G5 3.2Ghz due next summer.
For the last time, the dude was joking! Oh what the hell,
"Are you crazy? YOu expect Apple to slow down for you, you slimebag!"
I'm sure the author of the original post knows i'm kidding.
Cheers!~
Edited for stupidity.
G4 400 or so Mhz, i'll be accepting these machines for the everyday great price of $5 per usable CPU to start the ultimate G4 400Mhz CLusteR. WIll be called ClusTer's Last Stand and that puppy is gonna be 5000 CPUs strong by the end, and oh yea, it'll almost be as fast as the new dual G5 3.2Ghz due next summer.
For the last time, the dude was joking! Oh what the hell,
"Are you crazy? YOu expect Apple to slow down for you, you slimebag!"
I'm sure the author of the original post knows i'm kidding.
Cheers!~
Edited for stupidity.
Re: Re: you can't have dual AGP ports
I was mistaken, but you're not right either.
The new PCI 3.0 Spec does, in fact, support multiple AGP Ports. It's a change introduced in the new spec. AGP 2 and AGP 1 specs didn't support multiple ports (there was a technical reason for this that was too detailed for me to bother remembering ;-))
Page 14 of the spec.
So, if you read the AGP 3.0 final Spec from Intel's web site, you'll see Accelerated Graphics Port is a Port not a bus, and it finally supports multiple AGP 3.0 ports. AGP 3 has been designed to be backward compatible with AGP 2 (to some extent). I haven't seen yet if you can have multiple 4x/8x ports (that's AGP 2 and AGP 3 compatible ports) or if you have to have multiple ports that ONLY have compatibility for AGP 3 spec. I'm too tired to read the rest and find out. ;-P
So, you are saying that the Accelerated Graphics Port is actually a bus?
I was mistaken, but you're not right either.
The new PCI 3.0 Spec does, in fact, support multiple AGP Ports. It's a change introduced in the new spec. AGP 2 and AGP 1 specs didn't support multiple ports (there was a technical reason for this that was too detailed for me to bother remembering ;-))
Page 14 of the spec.
So, if you read the AGP 3.0 final Spec from Intel's web site, you'll see Accelerated Graphics Port is a Port not a bus, and it finally supports multiple AGP 3.0 ports. AGP 3 has been designed to be backward compatible with AGP 2 (to some extent). I haven't seen yet if you can have multiple 4x/8x ports (that's AGP 2 and AGP 3 compatible ports) or if you have to have multiple ports that ONLY have compatibility for AGP 3 spec. I'm too tired to read the rest and find out. ;-P
I actually think that we'll see these, either very close to MWSF or shortly after.
I would guess that if we see single/dual/dual, then we'll see a price cut across the board. It could come at the same time, but Apple may try to milk early adopters of the Rev2 G5s, then cut shortly after.
If Apple goes all dual (which is possible), then I expect to see a price cut on the mid/high end... again at the same time or shortly after.
I'm ever the optimist, but I'm guessing that near or at MWSF, we'll see the above machines, maybe a price cut, and a G5 iMac.
Slightly different note. I'm pretty sure Apple won't pre-announce a faster G5. If it isn't ready at MWSF, they'll announce it at an event shortly after, when it's ready to ship. Apple needed to pre-announce the G5 in June because there was serious discontent over the desktop line. The G4 wasn't shipping anyway so it didn't hurt much.. and the developers needed an early look at the G5 architecture. Now, the G5s are shipping well, it wouldn't be a good idea to cause buyers to put off purchases and wait for the next great thing.
i'm glad you posted this, because code optimization is the first thing i thought of after reading fukui's post, and i think you are dead on.
a friend of mine worked on an existing cluster at los alamos (he's at UC-berkley now), and it's all about optimization. they're doing nuclear sims, where saving .00000000000001 (or so
) seconds on one cycle can equal weeks or more in the whole scheme of things. (you should talk to this guy with a couple of bourbons in him--it will boggle your mind!)they have sooooo many lines of code already there, and it just wouldn't make sense to have to go back and re-optimize the existing stuff for a different chip. the g5 *may* be faster, but i don't know if it would matter for los alamos compared to the time for a re-write. it's kind of like me switching to a pc *if* it were a bit faster--i've already got a lot of money invested in mac software so it wouldn't be financially sound.
Re: 2.6GHz PowerMac G5s at MacWorld SF 2004?
that seems to make sense and 90 nanometer chips should run cooler
...which could also make a powerbook g5 a possibility
that seems to make sense and 90 nanometer chips should run cooler
...which could also make a powerbook g5 a possibility
Re: Re: 2.6GHz PowerMac G5s at MacWorld SF 2004?
I'll be surprised if there's 2.6Ghz chips.. That's a major jump. 2.4 wouldn't be a major surprise.
Re: Re: Re: 2.6GHz PowerMac G5s at MacWorld SF 2004?
It is a major jump, I agree, but keep in mind we're dealing with IBM now, and not Motorola!
It is a major jump, I agree, but keep in mind we're dealing with IBM now, and not Motorola!
Re: Re: Re: Re: 2.6GHz PowerMac G5s at MacWorld SF 2004?
moto-who?
hey wait, didn't ibm invent the PC?
moto-who?
hey wait, didn't ibm invent the PC?
Re: Re: Re: Re: 2.6GHz PowerMac G5s at MacWorld SF 2004?
True true.. I just don't think we should set our expectations too high. In magine how annyoing it would be if people are disappointed with 2.4 ghz at the top of the line. The rumor gets the best of them.
Re: Re: Re: Re: Re: 2.6GHz PowerMac G5s at MacWorld SF 2004?
Very true, and I am by no means expecting 2.6 GHz @ MWSF, but I'm just thining not that Apple is dealing with IBM and not Motorola anymore, the likelihood of it is definitely more plausible. Honestly though, I don't think anything @ MWSF will disappoint me, regardless of the announcements - it's all good! But you're right, there are those who take rumors as gospel and are extremely upset when they're 20" dual G5 PowerBook doesn't get announced.
Very true, and I am by no means expecting 2.6 GHz @ MWSF, but I'm just thining not that Apple is dealing with IBM and not Motorola anymore, the likelihood of it is definitely more plausible. Honestly though, I don't think anything @ MWSF will disappoint me, regardless of the announcements - it's all good! But you're right, there are those who take rumors as gospel and are extremely upset when they're 20" dual G5 PowerBook doesn't get announced.
Originally posted by skymac
They wont make a 17" g5 PB without changing thre other two to G5 at the same time, it does not make any sense.
Anyways i think the 2ghz will be thrown in the xserve and the other three will be used for the PM. We will probablty see a PB G5 or iMac G5 at WWDC. Of course there is always the chance that a new not existing yet product could be released
I guess this means i'll have to tell my dad to hold of his purchase for a few months. A 2*2.6ghz PM would be sopo kick A@$
Still even though the pb might not be updated this cycle it should happen before the imac. Unless they change the form factor of the imac it remains quite difficult to cool.
G5 Semi-Early Adopter
I just got my Dual 2Ghz G5 in October and this machine screams. I certainly won't be complaining when Apple does a speed bump, that's the nature of the biz. This is my first Mac and I bought it for amatuer video editing, and it does what I need it to do very well. (Using Final Cut Express, great software) Will my 2Ghz machine be any slower because a 2.6Ghz machine exists, no. I agree though, I think I am in the minority, most people will be bummed.
I just got my Dual 2Ghz G5 in October and this machine screams. I certainly won't be complaining when Apple does a speed bump, that's the nature of the biz. This is my first Mac and I bought it for amatuer video editing, and it does what I need it to do very well. (Using Final Cut Express, great software) Will my 2Ghz machine be any slower because a 2.6Ghz machine exists, no. I agree though, I think I am in the minority, most people will be bummed.
- Status