G5 1.6GHz Vs. G4 1.67GHz

[FireArse]

Howdy folks,

I was in bed last night thinking about the G4 and the G5 (how sad?!) , and thought - they overlap in terms of MegaHertz - but i hadnt seen a report anywhere how they match up...

I know how hard it is to test Intel vs PPC chips - but PPC vs. PPC should be ALOT easier - has anyone seen a test between the new 1.67GHz G4 and one of the old 1.6GHz G5 PM's - with 512MB ram? Maybe test for Processor intensive tasks (ignoring the graphhics) - Would be very interested in the results!

Thanks

F

 

[zap2]

well i hope the G5 would be faster

but would be cool to see what would ''win''

 

[risc]

I ran xbench on my GFs iMac G5 1.6 GHz and my (now sold) PowerBook G4 1.67 GHz they both ended up with approx. the same score, the G4 was faster at somethings, the G5 at others. The main bottle necks where the crap video card in the iMac G5 1.67 and the slow notebook harddrives on the G4. Xbench is of course completely unreliable it reports my 6800 Ultra DDL as being slower for 3D than my old 9800 XT, so I'll tell you how both machines fell to me... about exactly the same. The G5 obviously flogs the G4 at processor intensive stuff, but in day to day use I was hard pressed to tell the difference between both machines, unless I was doing 3D stuff then the G4 completely wrecked the iMac.

Not a very professional report but I HTH.

 

[maya]

As mentioned there is little differences, the G5 will excel on some things and the G4 with others. I prefer the G5 better.

 

[FireArse]

As mentioned there is little differences, the G5 will excel on some things and the G4 with others. I prefer the G5 better.

I guess part of me was thinking about the Intel move. I think Apple need to make it much easier to optimize code for Intel. You can write code and hope to hell that if you compile with the XCode gcc you'll get an OK optimization - but part of me thinks all the 'gimmicks' surrounding the G5 were just that. Pure clock speed is what makes the difference when it becomes too difficult or not easy to optimize code for the CPU.

Maybe a more technical test would give us further insight?

F

 

[Morn]

http://www.barefeats.com/mini01.html

 

[FireArse]

Such an interesting read...

http://www.barefeats.com/mini01.html

Thanks so much for the link. The difference between a 1.5GHz G4 and a 1.6GHz G5 isn't all that much - people talk about the FSB of that G4 - but to be honest, it coped quite well considering it was against the 800MHz FSB of the 1.6 G5.

I think il try get the last version of iBook G4 12' when it comes out - run to the UK Regents Street store on the day of release of Intel - and get me a PPC chip.

Makes you wonder how optimized OS X is for the Dual G5.....

 

[Morn]

Yeah, the G4 compares well to G5 at same mhz. But..... when most x86 CPU's will compare more closely to a 2ghz to 2.5ghz G5.... The G4 is lagging behind.

 

[zap2]

but would a 1.6 ghz PowerBook make People run out and buy one?

Yes even thought its about the same


But i find hard to belive that the G4 beat the G5 in anything, perhaps tied but beat thats crazy.

 

[Morn]

Nah, the G4 is not as bad as many people think it is. It has a more efficient altivec unit for one and shorter pipelines meaning a lower cost to branch misprediction.
The G5 advantage is really just mhz, ohh and it's dual issue FPU unit, which isn't used much if the program is using altivec though.
There would be no point in making a G5 1.6ghz powerbook, even though IBM has low power G5's at that mhz. Hence the inablity of IBM to provide a fast CPU for a laptop and the Intel switch. I mean the G5 would do as good a job as a G4 in a powerbook, but won't offer anything more.

 

[Catfish_Man]

I guess part of me was thinking about the Intel move. I think Apple need to make it much easier to optimize code for Intel. You can write code and hope to hell that if you compile with the XCode gcc you'll get an OK optimization - but part of me thinks all the 'gimmicks' surrounding the G5 were just that. Pure clock speed is what makes the difference when it becomes too difficult or not easy to optimize code for the CPU.

Maybe a more technical test would give us further insight?

F

I see your point, but I'm gonna have to disagree. The Pentium-4, which is clearly The clock frequency chip, has all sorts of optimization "gotchas" (non-zero-cycle FXCH, slow rotate instructions, small L1 cache, etc...). I guess I'd say that the more "generically powerful" a chip is, the easier it is to optimize for. So clock speed helps, being similar to older chips helps, not having the occasional weird slow instruction helps, out of order execution helps A LOT. The G4 was tricky due to lack of OOOE, and due to its heavy reliance on Altivec. The G5 made things easier, but added the weird group scheduling thing, as well as the fact that all integer instructions take at least 2 cycles on it. Something like the Athlon64, or probably the Pentium-M, are going to be relatively easy to get good perf out of. It's hard *not* to like big caches and lots of pipes

 

[kwajo.com]

Nah, the G4 is not as bad as many people think it is. It has a more efficient altivec unit for one and shorter pipelines meaning a lower cost to branch misprediction.
The G5 advantage is really just mhz, ohh and it's dual issue FPU unit, which isn't used much if the program is using altivec though.
There would be no point in making a G5 1.6ghz powerbook, even though IBM has low power G5's at that mhz. Hence the inablity of IBM to provide a fast CPU for a laptop and the Intel switch. I mean the G5 would do as good a job as a G4 in a powerbook, but won't offer anything more.

yeah I don't really understand the old clamoring for a G5 Powerbook, I remember people demanding a G5 in there even if they had to reduce the clock speed, and it really didn't make any sense. a G4 at 1.5 would be better than a G5 at 1.2 or something, yet people wanted and demanded it

 

[Lord Blackadder]

In my personal experience the G4 and G5 (at similar clock speeds) are pretty close in performance - close enough that things like video cards and hard drivesgive one the edge over another.

A Dual 1.42GHz MDD G4 tower with some mild upgrades (RAM, video card, maybe a SATA drive) can still match or beat some G5 towers in many tasks, although that is becoming less true as new, faster G5s are released.

 

[MacsRgr8]

Yep, I also believe the G4 and G5 are very similar (Mhz for Mhz) when it comes to 32 bit every-day-life apps.
A G5 is overal quicker due to the "other" updgrades it has received (faster bus, faster RAM, S-ATA.. etc.)

Most obvious:
AGP 8x gave it a GeForce 6800 Ulltra or X800 XT grfx card possibility.

Also the fact that a G5 is now available at speeds up to 3 GHz (oh no!! still at 2.7 GHz.... no! 2.5! ) makes it pretty sweet too.

 

[Morn]

You should in theory be able to put a 6800 ultra in a later powermac g4, I mean AGP 8x cards can work at AGP 4x mode.

 

[edesignuk]

I can't image there being much difference. I don't notice anything at all between my 1.5GHz G4 in the PB, to my 1.8GHz G5 in my iMac

 

[dmw007]

I have to agree, the G4 seems to be pretty close in performance to the G5 in terms of MHz for MHz performance.

 

[dernhelm]

I have to agree, the G4 seems to be pretty close in performance to the G5 in terms of MHz for MHz performance.

Well... it depends on what you are doing. But the biggest problems areas for the G4 is the slower FSB, and smaller on die cache. This shouldn't be too surprising, as the Altivec unit isn't really any different between the two, and so stuff like image processing, etc is hurt more by excessive processor wait states, than by anything else.

 

[FireArse]

Most obvious:
AGP 8x gave it a GeForce 6800 Ulltra or X800 XT grfx card possibility.

Whats the AGP speed on the late G4 towers then - four?! I thought you could stick an ATi 9800 in them? Therefore an Ultra too?

I heard NVida released an AGP version of the 7800 - maybe an AGP version for us AGP G5 ppl?

Please?

6800 Ultra's on eBay are pretty hard to find now. Can you get then direct from Apple?

 

[MacsRgr8]

The AGP 4x on the later G4's will not support the AGP 8x cards like the nVidia 6800 Ultra or ATi X800 XT.
The AGP versions of PC grfx cards are not compatible.

The best card for the G4 is the ATi Radeon 9800 Pro

Whats the AGP speed on the late G4 towers then - four?! I thought you could stick an ATi 9800 in them? Therefore an Ultra too?

I heard NVida released an AGP version of the 7800 - maybe an AGP version for us AGP G5 ppl?

Please?

6800 Ultra's on eBay are pretty hard to find now. Can you get then direct from Apple?

You have a G5 with AGP 8x?
Then get the ATi Radeon X800 XT

 

[eXan]

The G5 starts to shine at 2 Ghz (if single core/processor)

Duals really shine

 

[andrewag]

I ran some tests a while ago calculating Pi. I know it's very basic but seemed to be a good idea at the time.

The results are on http://www.dotmac.net.au/misc/pi.php and show the G5 beating the G4.

Running quicktime broadcaster, encoding mpeg-4 320x240 at 25FPS had the G4 1.67 using over 70% CPU, while the G5 imac 1.6 sits at the mid 40%. Both with 512MB RAM, capturing from a ADC with Sky News Channel as the video.

Really depends on the use in the end.

 

[mad jew]

The results are on http://www.dotmac.net.au/misc/pi.php and show the G5 beating the G4.

Nice work.

Just out of interest, why do you calculate pi to different decimal places, Mac versus Windows?