Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

minimax

macrumors 6502
Original poster
Feb 9, 2005
351
0
Ok, as the Intellification of Macintosh is almost upon us I thought it would be fun to make some predictions on how the G4+ holds up against the Yonah. First benchmarks have been released on Anandtech which show the Yonah performing almost on par with the Athlon 64.
Let's hope there will be proper applications (PS!) and benchmark tools for OSx86 released together with the introduction of the Macintel.

My estimation, based on a few benchmarks (also crossplatform) is that Yonah (single core, 2MB L2) will be about 25% faster on IPC.
I also think that this will be fairly even distributed on FP and integer operations, as Yonah seems to be stronger in integer like the G4.
With vector instructions the G4 will still be a bit stronger me thinks.
 

Dont Hurt Me

macrumors 603
Dec 21, 2002
6,055
6
Yahooville S.C.
minimax said:
Ok, as the Intellification of Macintosh is almost upon us I thought it would be fun to make some predictions on how the G4+ holds up against the Yonah. First benchmarks have been released on Anandtech which show the Yonah performing almost on par with the Athlon 64.
Let's hope there will be proper applications (PS!) and benchmark tools for OSx86 released together with the introduction of the Macintel.

My estimation, based on a few benchmarks (also crossplatform) is that Yonah (single core, 2MB L2) will be about 25% faster on IPC.
I also think that this will be fairly even distributed on FP and integer operations, as Yonah seems to be stronger in integer like the G4.
With vector instructions the G4 will still be a bit stronger me thinks.
Me thinks you have been burning some, how about sharing? G4 stronger then Athlon 64 thats funny, or G4 stronger then Yonah LOL.
 

sethypoo

macrumors 68000
Oct 8, 2003
1,583
5
Sacramento, CA, USA
There is no way the G4 will be stronger than Intel's Yonah. Yonah is going to blow the G4 out of the water, and give the G5 a run for its money.

Apple will not put in a processor that is less powerful than their current lineup.
 

minimax

macrumors 6502
Original poster
Feb 9, 2005
351
0
If both of you read more careful you would have seen I said stronger on *Vector* performance. It is widely known the Altivec instruction engines are stronger as their SSE1/2/3 counterparts. But Vector instructions are only a part of the total palette and like I said, overall the Yonah will be stronger indeed.

edit: As I have the idea both of you totally misunderstood what i was saying: IPC = Instructions Per Cycle used as a measure of overall performance relative to the clockspeed. Technically this is not correct as benchmarks can only show system performance per cycle.
With 'even distributed' I meant both are stronger on integer whereas the G5 and Athlon 64 are relatively stronger on FP operations. So i was not saying the G4 will be as fast as the Yonah, but that the performance difference between both will be fairly even distributed over the spectrum.
 

Flynnstone

macrumors 65816
Feb 25, 2003
1,438
96
Cold beer land
Since we are predicting the future ...
How would a G4 compete if they (Freescale) moved the memory controller on chip? i.e. decrease memory latency and increase bandwidth.
 

GFLPraxis

macrumors 604
Mar 17, 2004
7,152
460
sethypoo said:
There is no way the G4 will be stronger than Intel's Yonah. Yonah is going to blow the G4 out of the water, and give the G5 a run for its money.

Apple will not put in a processor that is less powerful than their current lineup.

I'd be more interested in comparing Yonah to the e600 G4. We're talking about the next gen, scales-to-2-GHz, 64-bit, dual-core, low-power G4's. :)
 

generik

macrumors 601
Aug 5, 2005
4,116
1
Minitrue
Flynnstone said:
Since we are predicting the future ...
How would a G4 compete if they (Freescale) moved the memory controller on chip? i.e. decrease memory latency and increase bandwidth.

Onboard memory controller does not increase bandwidth.
 

Morn

macrumors 6502
Oct 26, 2005
398
0
Onboard memory does increase bandwidth, gives full bandwidth of ram to the CPU and takes load off the FSB.
 

Morn

macrumors 6502
Oct 26, 2005
398
0
Latency is really important, ondie memory controller means the CPU doesn't have to do nothing while waiting for the memory as often.
 

bousozoku

Moderator emeritus
Jun 25, 2002
15,719
1,894
Lard
minimax said:
If both of you read more careful you would have seen I said stronger on *Vector* performance. It is widely known the Altivec instruction engines are stronger as their SSE1/2/3 counterparts. But Vector instructions are only a part of the total palette and like I said, overall the Yonah will be stronger indeed.

edit: As I have the idea both of you totally misunderstood what i was saying: IPC = Instructions Per Cycle used as a measure of overall performance relative to the clockspeed. Technically this is not correct as benchmarks can only show system performance per cycle.
With 'even distributed' I meant both are stronger on integer whereas the G5 and Athlon 64 are relatively stronger on FP operations. So i was not saying the G4 will be as fast as the Yonah, but that the performance difference between both will be fairly even distributed over the spectrum.

Of course, the AltiVec performance is quite good. It should be--it takes up more of the die than it should--at the expense of double precision floating point math.

Considering the number of AltiVec-enabled applications vs. those which require strong floating point performance, AltiVec won't matter, except in extreme cases like the PPC-optimised version of BLAST.
 

tpjunkie

macrumors 65816
Nov 24, 2002
1,251
5
NYC
generik said:
But ram still runs at the same rated speed.

Yeah, and since the FSB bus is limited to 200 MHz in the G4 with the memory controller located off chip, the current G4's are unable to take advantage of that speed.
 

Flynnstone

macrumors 65816
Feb 25, 2003
1,438
96
Cold beer land
generik said:
Onboard memory controller does not increase bandwidth.
It DOES increase bandwidth and decreases latency!
RAM bandwidth is about 3.2 gigabytes per second for single channel 400 MHz DDR. G4 FSB at 200 MHz is I think 800 megabytes per second. No point going with dual channel on a "standard" G4. Kind of like sucking a 10 pin bowling ball through a garden hose :eek:
For more info.
http://www.osnews.com/story.php?news_id=8548
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.