[anthology123]

I have the opportunity to buy a new system right now. With all the PowerPC software I have, it's not a problem to not wait for an intel Mac. However, I can get either the older Dual processor 2.7 for about $2350 or the Dual Core 2.3. Any opinions, if you had to choose between these two, which would it be? Both have standard 512 RAM, 250GB HD. The Quad 2.5 is out of my budget right now, even with an education discount.

[mduell]

For the life of this system:
Are you going to need a better video card than 6800U/X850XT?
Are you going to need more than 8GB RAM?
Are you going to need any PCIe only cards?
Are you going to need any PCI-X only cards?

[Lateralus]

Here we go again Mark...

You claim repeatedly that the 6800U/X850 XT will be the best graphics cards to ever grace an AGP Mac since ATI and nVidia are 'done' with AGP. Except that they're not. nVidia announced an AGP version of their 7800 GPU last week and most industry watchers are expecting ATI to follow suit.

And, again; The AGP G5s are capable of taking 16GBs of RAM. 2GB non-ECC DIMMs are on the market and are available, among others, from CompUSA and ZipZoomFly.

That said (again), I say go for the Dual 2.7GHz. It'll be the faster machine all-around and they're really quite solid.

[mduell]

Originally Posted by Lateralus

Here we go again Mark...

You claim repeatedly that the 6800U/X850 XT will be the best graphics cards to ever grace an AGP Mac since ATI and nVidia are 'done' with AGP. Except that they're not. nVidia announced an AGP version of their 7800 GPU last week and most industry watchers are expecting ATI to follow suit.

And, again; The AGP G5s are capable of taking 16GBs of RAM. 2GB non-ECC DIMMs are on the market and are available, among others, from CompUSA and ZipZoomFly.

That said (again), I say go for the Dual 2.7GHz. It'll be the faster machine all-around and they're really quite solid.

I saw the 7800GS AGP release, and it has fewer processing units and lower clockrates than the PCIe cards. I also saw the benchmarks; it's not much faster than the 6800U or X850XT (2 - 10% faster in Doom 3).
Given that the performance gain is so small, I still don't think we'll see an AGP version for Mac.

I forgot the pre-PCIe PowerMacs could support 2GB DIMMs. I'd still be wary of doing it, since Apple doesn't mention or support it.

The 2.7 is the way to go unless your apps are memory bound (2.3 has faster memory) or you want a better video card today (and maybe in the future).

[gudrummer]

Well,i have the 2.7+X800XT and it's a hell of a machine,never let me down once on any projects i did with FCP.

[Lateralus]

Originally Posted by mduell

I'd still be wary of doing it, since Apple doesn't mention or support it.

What's it gonna do? Blow the machine up?

There are very few Macs for which Apple lists the correct RAM capacity on their site.

[mduell]

Originally Posted by Lateralus

What's it gonna do? Blow the machine up?

I was thinking more along the line of hardware bugs that only show up with more than 8GB RAM, since they may not have tested it.

[Chinasaur]

Anthology,

Apple Refurb has the 2.7 at $2,149. Save you some dough if you go that way.

[Big Mac]

Originally Posted by mduell

I was thinking more along the line of hardware bugs that only show up with more than 8GB RAM, since they may not have tested it.

That would be a very unusual, if not unheard of, occurrence.

[mduell]

Originally Posted by Big Mac

That would be a very unusual, if not unheard of, occurrence.

I was thinking of the problems Apple laptops (both lines, IIRC) had with >1GB RAM a few months back, so I should have said software bugs in addition to hardware bugs with an untested configuration.

[anthology123]

Anthology,

Apple Refurb has the 2.7 at $2,149. Save you some dough if you go that way.

Well, that's a tough choice because i would prefer a brand new system over a refurb, even though I am aware that most refurbs are just opened boxes. The one I found for $2300 is brand new.
I would use this system to make home DVDs, convert videos for iPod, manage a large iTunes collection. I would not buy a screen larger than 20" and would not need more than 4GB RAM, very unlikely I will need 8.

I don't think I will be adding PCI e or X cards except to add more USB2 or firewire ports.

[thejam]

I have the dual 2.5ghz water-cooled, and if I did it again I'd avoid the water-cooled since mine leaked. Also, I'd go for the 2.3ghz since it takes error-correcting memory, so you can guarantee rock solidity. The small performance difference is absorbed by the increased memory speed of the 2.3. (Remember, Moore's law holds for processors, not memory speed, so memory speed is the primary bottleneck for most compuations, unless they have been optimized. So while certain photoshop operations may be faster on the 2.7, generally there won't be much difference.) I think the 2gb chips will be much cheaper and easier to get in DDR2 for the 2.3 as opposed to non-ecc DDR for the 2.7. Also, the PCIe bus on the 2.3 is better for core image, since PCIe is fast in both directions from the GPU, while AGP is only fast from CPU to GPU. The 2.3 includes the dual link dvi so you can get the 30inch display down the road.

[Lateralus]

Originally Posted by thejam

I have the dual 2.5ghz water-cooled, and if I did it again I'd avoid the water-cooled since mine leaked. Also, I'd go for the 2.3ghz since it takes error-correcting memory, so you can guarantee rock solidity. The small performance difference is absorbed by the increased memory speed of the 2.3. (Remember, Moore's law holds for processors, not memory speed, so memory speed is the primary bottleneck for most compuations, unless they have been optimized. So while certain photoshop operations may be faster on the 2.7, generally there won't be much difference.) I think the 2gb chips will be much cheaper and easier to get in DDR2 for the 2.3 as opposed to non-ecc DDR for the 2.7. Also, the PCIe bus on the 2.3 is better for core image, since PCIe is fast in both directions from the GPU, while AGP is only fast from CPU to GPU. The 2.3 includes the dual link dvi so you can get the 30inch display down the road.

1) You admitted in the other thread that when you took your machine into be repaired, the techies had never seen a system that had its water cooling system fail. So it's obviously rare. Then is it fair to use that as a reason to not go for a 2.7?

2) ECC memory costs more, performs more slowly and the stability difference really isn't something that shows in standard use. The money would be better spent ensuring the purchase of name brand standard memory. Though, it is arguable.

3) The memory controller of the PCIe G5s is indeed faster. But the memory itself really is not. DDR2 has a higher latency than DDR. So DDR2 533 performs nearly on-par with DDR 400. In order to see an appreciable speed boost over DDR 400 with DDR2, you would need to go with DDR2 667, which Apple did not.

4) The Dual 2.7 includes a Dual-DVI as well.

[anthology123]

Can someone here chime in on fan noise? Would the fan come on more often on a DP 2.7 than a dual 2.3?
And about the video card, can I do dual monitor on the 2.3 or the 2.7? I don't intend to do 2 large monitors, maybe at most 2 20" DVI, but one 20" is fine.

Do the Super drives on both units have DL burning, or is the 2.7 only single layer?

[Lateralus]

Fan noise should be comparable. Though if I had to guess, I'd bet that the Dual-Core 2.3 would be louder due to the lack of liquid cooling.

Either Power Mac can drive two DVI displays.

Both include Dual-Layer SuperDrives.

[rhashem]

Originally Posted by Lateralus

Fan noise should be comparable. Though if I had to guess, I'd bet that the Dual-Core 2.3 would be louder due to the lack of liquid cooling.

Either Power Mac can drive two DVI displays.

Both include Dual-Layer SuperDrives.

The 2.7 GHz 970FX is a very hot chip compared to the 2.3 GHz 970MP, both because it runs at a substantially higher clockspeed, and because its two seperate cores without the power optimizations IBM included in the MP series. I wouldn't be surprised if the power requirements for the 2.7 GHz system are double that of the 2.3 GHz system. Since airflow requirements are mainly a function of the amount of heat that needs to be dissipated, the 2.3 GHz system should significantly reduce airflow requirements, thus require the fan to ramp up less often.

[Lateralus]

That's why the Dual 2.7 has liquid cooling; so the fans don't have to ramp up. I've never heard the rans ramp up on my Dual 2.5 970FX and I run the CPUs at 100% 24/7.

Whereas the DC 2.3 relies entirely on the cooling fans, which do have to ramp up at times.

[thejam]

Originally Posted by Lateralus

1) You admitted in the other thread that when you took your machine into be repaired, the techies had never seen a system that had its water cooling system fail. So it's obviously rare. Then is it fair to use that as a reason to not go for a 2.7?

Yes, since even rare is worse than "never". Why take chances, unless you're sure you've got the extra warrantee? Basically, I regret getting water-cooling.

2) ECC memory costs more, performs more slowly and the stability difference really isn't something that shows in standard use. The money would be better spent ensuring the purchase of name brand standard memory. Though, it is arguable.

http://cr.yp.to/hardware/ecc.html

3) The memory controller of the PCIe G5s is indeed faster. But the memory itself really is not. DDR2 has a higher latency than DDR. So DDR2 533 performs nearly on-par with DDR 400. In order to see an appreciable speed boost over DDR 400 with DDR2, you would need to go with DDR2 667, which Apple did not.

According to xbench site: 2.7GHz has triad bandwidth of 2.0GB/s, while 2.3GHz has 2.9GB/s; that's a pretty big increase! But I agree that specific specs can be misleading. Unfortunately, we can only choose based on data we have!

4) The Dual 2.7 includes a Dual-DVI as well.

My bad, I was projecting it's lack from my 2.5 onto the 2.7.

Originally Posted by Lateralus

That's why the Dual 2.7 has liquid cooling; so the fans don't have to ramp up. I've never heard the rans ramp up on my Dual 2.5 970FX and I run the CPUs at 100% 24/7..

You run them on energy saver=highest, not =automatic, I take it? My 2.5ghz G5 winds up periodically when on automatic.

The 2.3GHz has 1MB L2 cache per core, while the 2.7GHz has only 512kB per processor. Cache rules.

[MRTrauffer]

i use dual 2.7's at work...FCP editing. Generally, they run really quiet, but obviously, the fans do rev up when under a heavy processing load.

If price is a factor in the OP's purchasing decision, then I would recommend a refurb 2.7. I bought one about a month ago...a very good deal, in my opinion.

[Lateralus]

Originally Posted by thejam

According to xbench site: 2.7GHz has triad bandwidth of 2.0GB/s, while 2.3GHz has 2.9GB/s; that's a pretty big increase! But I agree that specific specs can be misleading. Unfortunately, we can only choose based on data we have!

Please don't quote Xbench.

Originally Posted by thejam

You run them on energy saver=highest, not =automatic, I take it? My 2.5ghz G5 winds up periodically when on automatic.

I run them on Highest, correct.

Originally Posted by thejam

The 2.3GHz has 1MB L2 cache per core, while the 2.7GHz has only 512kB per processor. Cache rules.

And the DC 2.3GHz has one 1.15GHz system bus. The Dual 2.7GHz has two 1.35GHz busses. It's a trade off.

[rhashem]

Originally Posted by Lateralus

That's why the Dual 2.7 has liquid cooling; so the fans don't have to ramp up. I've never heard the rans ramp up on my Dual 2.5 970FX and I run the CPUs at 100% 24/7.

Whereas the DC 2.3 relies entirely on the cooling fans, which do have to ramp up at times.

The liquid cooled G5s rely entirely on the cooling fans as well. It's simple thermodynamics --- if your processors are putting out 100W, and you want to maintain the case internals at a constant temperature, that heat is going to have to leave the case. Unless your liquid loop is leaking very badly, the liquid won't do that. What gets the heat out of the case is air from the fans. The liquid loop isn't a heat dissipation mechanism, its a heat transfer mechanism. It is there to get heat more efficiently from the processor to the radiator, where the fan airflow can take it away.

Since the airflow required to transport a given amount of heat is roughly constant, your airflow requirement, in terms of volume flow rate, is proportional to the heat output. If your processors put out a lot more heat, you're going to need to run more air past the radiator to get rid of it. The reason liquid cooling loops are often associated with low noise is not because liquid reduces airflow requirements, but because liquid loops enable heat to be spread efficiently through large radiators, on which larger but slower-spinning fans can be used to achieve the same airflow rate.

[joe]

Originally Posted by Lateralus

I've never heard the rans ramp up on my Dual 2.5 970FX and I run the CPUs at 100% 24/7.

I'm jealous! My Dual 2.5GHz (950FX) Tower fans always ramped up under heavy load until I installed min-o-max. Of course that means my Tower's processor performance is set to "reduced" most of the time. Min-o-max just makes it a lot easier to switch to automatic or highest for those times when I need max performance and am willing to put up with the fan noise (ie Doom3).

[mduell]

Originally Posted by Lateralus

3) The memory controller of the PCIe G5s is indeed faster. But the memory itself really is not. DDR2 has a higher latency than DDR. So DDR2 533 performs nearly on-par with DDR 400. In order to see an appreciable speed boost over DDR 400 with DDR2, you would need to go with DDR2 667, which Apple did not.

3 cycles at 400Mhz is the same as 4 cycles at 533Mhz, so the latency is the same, but DDR2-533 should offer more bandwidth than DDR-400. If you're going to argue that latency is what matters, and not bandwidth, then I'd point out that DDR2-667 has the same latency (typically 5 cycles).
However the memory controller in the dual 2.7 has poor random access times, that may have been corrected in the DDR2 PowerMacs.

Originally Posted by Lateralus

That's why the Dual 2.7 has liquid cooling; so the fans don't have to ramp up. I've never heard the rans ramp up on my Dual 2.5 970FX and I run the CPUs at 100% 24/7.

Whereas the DC 2.3 relies entirely on the cooling fans, which do have to ramp up at times.

While the water cooling system certainly provides a heat reservoir to keep the fans down during brief periods of high CPU usage, during sustained high CPU usage you should still need to spin up the fans to get the heat out of the water cooling system (as other users report).

Originally Posted by Lateralus

And the DC 2.3GHz has one 1.15GHz system bus. The Dual 2.7GHz has two 1.35GHz busses. It's a trade off.

Again you claim that FSB has a nontrivial impact on performance, and again I ask for the proof.

[Lateralus]

Originally Posted by mduell

While the water cooling system certainly provides a heat reservoir to keep the fans down during brief periods of high CPU usage, during sustained high CPU usage you should still need to spin up the fans to get the heat out of the water cooling system (as other users report).

Again you claim that FSB has a nontrivial impact on performance, and again I ask for the proof.

My Dual 2.5 is nearly silent. Assuming the DC 2.3 is at worst on-par, I don't see why noise is even becoming a part of this debate between G5s.

I said it's a trade off. Take away half the cache on the 970MP and tell me running both cores on the single bus doesn't have any performance drawback.

[mduell]

Originally Posted by Lateralus

My Dual 2.5 is nearly silent. Assuming the DC 2.3 is at worst on-par, I don't see why noise is even becoming a part of this debate between G5s.

I'm with you there. I was just drawn into the noise topic by other posters.

Originally Posted by Lateralus

I said it's a trade off. Take away half the cache on the 970MP and tell me running both cores on the single bus doesn't have any performance drawback.

More assertions, still no proof.

[Lateralus]

Originally Posted by mduell

More assertions, still no proof.

No assertions, just common sense.

[mduell]

Originally Posted by Lateralus

No assertions, just common sense.

assertion (n) a declaration that is made emphatically (as if no supporting evidence were necessary)

It's like we're not even speaking the same language...

[Lateralus]

Originally Posted by mduell

assertion (n) a declaration that is made emphatically (as if no supporting evidence were necessary)

It's like we're not even speaking the same language...

Did you need to refer to a dictionary? Because I sure don't.

And you're right, we're not. I don't speak SPECese.

[Todd Madson]

Re: 16 GB on DP Macs:
There's a PDF included with DP machines that indicates Apple engineers
tested this and it works fine to run the larger PC3200 DIMMs on the machine
- the tests came after the print runs for the manuals that came with the
machine so was included in the .PDF file.

[Todd Madson]

I've been running my DP 2.5 since 6/14/05 with nearly 100% processor
use 24/7 and no leaks, no problems. Fans ramp up when they need to
but it's so much quieter than my old sawtooth machine it doesn't get
noticed much.

[thejam]

Originally Posted by Lateralus

Please don't quote Xbench.

Why not? At least for the stream sub-benchmarks, I think the Xbench code is just an adaptation of David McCalpin's stream suite: http://www.cs.virginia.edu/stream/
From what I can tell, stream is standard (add, copy, triad) for bandwidth.
I think Xbench added some vector processing optimizations to squeeze out more performance on some processors.

And the DC 2.3GHz has one 1.15GHz system bus. The Dual 2.7GHz has two 1.35GHz busses. It's a trade off.

Again, I'd say an almost 50% _observed_ bandwidth improvement for the 2.3GHz over the 2.7GHz blows away any speculations of potential benefits from the specs you've reported, though I have often been seduced by such manufacturer specs myself. Let's join Francis Bacon's empiricist revolution instead, no?

[Todd Madson]

Let's see what really happens:
-set up a dual 2.7
-and a dual core 2.3
-run the seti at home/boinc client (see the long Altivec optimized client thread in
Team Macnn on this forum to get the appropriate software)
-which machine gets the best performance with the new 5.2a optimized client?
-fastest is best.

While you're at it, it would be very interesting to see the cinebench results
on both a DC 2.3, DP 2.5 and DP 2.7. And a Quad.

[thejam]

Originally Posted by Todd Madson

-run the seti at home/boinc client (see the long Altivec optimized client thread in

Benchmarking using optimized code isn't representative of many programs; only a relative minority of software is optimized since it's so much manual work. If the OP actually cares about seti/boinc, then fine, but if s/he's a developer, then compile time would matter more, etc.

[Lateralus]

Originally Posted by thejam

Why not? At least for the stream sub-benchmarks, I think the Xbench code is just an adaptation of David McCalpin's stream suite: http://www.cs.virginia.edu/stream/
From what I can tell, stream is standard (add, copy, triad) for bandwidth.
I think Xbench added some vector processing optimizations to squeeze out more performance on some processors.

Again, I'd say an almost 50% _observed_ bandwidth improvement for the 2.3GHz over the 2.7GHz blows away any speculations of potential benefits from the specs you've reported, though I have often been seduced by such manufacturer specs myself. Let's join Francis Bacon's empiricist revolution instead, no?

Xbench has its uses, but for the most part it is one of the most tempermental, fluctuating benchmark suites on any platforms. I've seen things like a reboot have a 40% effect on test score. It also puts way too much emphasis on cache.

And I'd happily give up on this argument if you could show me a set of tests that show the DC 2.3GHz to be anything more than inferior to the Dual 2.7GHz. But you can't.

[Lateralus]

Originally Posted by thejam

Benchmarking using optimized code isn't representative of many programs; only a relative minority of software is optimized since it's so much manual work. If the OP actually cares about seti/boinc, then fine, but if s/he's a developer, then compile time would matter more, etc.

So you're complaing about SETI BOINC, which is becoming one of the most quickly adopted forms of benchmarking across most platforms due to its nearly totally CPU-centric nature, yet you want to rely on Xbench scores when most techies consider Xbench to be a joke?

[Todd Madson]

Cinebench is a better indicator of performance than Xbench IMHO.

And for me, Seti performance would matter.

[thejam]

Excellent description of the issue!

[rhashem]

Originally Posted by Lateralus

And I'd happily give up on this argument if you could show me a set of tests that show the DC 2.3GHz to be anything more than inferior to the Dual 2.7GHz. But you can't.

The argument about the busses is sort of irrelevent. The 970 is, in most code, latency bound rather than bandwidth bound. Since both the dual core and the single core have the same atrocious memory latency, the deciding factor is going to be the clockspeed, in which regard the dual 2.7 has a significant advantage.

[mduell]

Originally Posted by Lateralus

Did you need to refer to a dictionary? Because I sure don't.

And you're right, we're not. I don't speak SPECese.

I didn't need to, since I know what the word assertion means, but you're disputing the definition, so I provided it for you and cited a credible source.

I never suggested using SPEC to show that the FSB matters as much as you claim it does. Use TPC, use Cinebench, use Linpack, use BOINC, use dnetc, use video encoding, use compiling, use anything you'd like. Just back up your assertion or drop it.

[anthology123]

thanks for all your help on this thread. I have decided to buy the 2.3 dual instead of the 2.7, I can get the 2.3 for under $1900

[Todd Madson]

I should mention, sorry to say, that hardmac.com has been reporting that the power
supplies in some of the DC G5 macs have been exploding (I wish I was kidding) after
powerup.

Check it out, it appears some screws installed in a particular way causes a very
spectacular failure.

[rhashem]

Originally Posted by Todd Madson

I should mention, sorry to say, that hardmac.com has been reporting that the power
supplies in some of the DC G5 macs have been exploding (I wish I was kidding) after
powerup.

Check it out, it appears some screws installed in a particular way causes a very
spectacular failure.

Jesus. 710W in the single dual-core machine? And there are still people who believe IBM could've gotten a dual-core G5 into the iMac, or a G5 into the PowerBook? How does that even work? The thing only has two drive bays, which means less than 40W from hard drives, and the PCI-E bus is limited to 150W in the SPEC. What is the rest of that power for?

I've got a 385W PSU in the Opteron machine that sits next to my DC PM, and I'm convinced even that is overspecced. The current draw never gets high enough to cause the PSU fan to ramp up, even with an X800 card and two HDDs (one of them a Raptor).

[Todd Madson]

You have to admit, the G5 Powerbook would have melted in your lap
as well as your hands. It would be slag.

[booboo]

2.7 DP is less than 5% faster than the 2.3 DC, according to MacWorld tests.

Dual Core's don't 'chirp' at all, are quieter/run cooler (except for the Quad, which is roughly the same) and use almost half as much power (except for the Quad, which is roughly the same . . . )

And the audio is much improved too . . .