G5 1,8 bench
Just go to www.xbench.com and click on compare results and then G5 you 'll see a new benchmark of the 1,8. As i looked to it, i think xbench isn't really doing it job right. THe 1,6 and 1,8 have the same processing power of 1,14 gflops.
Xbench needs optimisation or if it has it is a bad bench'er
Xbench needs optimisation or if it has it is a bad bench'er
Comments
But you can by a G5 MP2 and ask if they return your money back if Apple can?t deliver there promise that it is the fastest Personal desktop computer in the world.Or the the Del in that matter.
Originally posted by fieldor
Just go to www.xbench.com and click on compare results and then G5 you 'll see a new benchmark of the 1,8. As i looked to it, i think xbench isn't really doing it job right. THe 1,6 and 1,8 have the same processing power of 1,14 gflops.
Xbench needs optimisation or if it has it is a bad bench'er
Did you also read that the 1.6 had more then twice the amount of ram installed as the 1.8?
I'm not a genius (yet), people make mistakes i just pointed it out for some good all fashion disscussion.
Originally posted by fieldor
Yeah I know but the 1,8 has faster ram than the 1,6 so it doesn't make a big difference i think.
I'm not a genius (yet), people make mistakes i just pointed it out for some good all fashion disscussion.
From the other thread about the 1.6 showing up (this in reference to a 1 GHz, 100 MHz fsb G4 result someone posted):
Originally posted by smalM
Let's see:
bandwidth = 800 MB/sec,
memory test = 1400 MB/sec
Is the rest of this "test program" just as good as this result?
-- Mark
Taking 1.1 hours worth of AAC files in iTunes and converting them into AIFF took 49 seconds. Converting those same files back into AAC took 3:02 -- about a 22x rate. Since the files were being read from the hard drive, and not from a CD, this should have been mostly a compute-bound, rather than I/O-bound, test.
My dual 1.25 did roughly the same amount of AAC -> AIFF work (I didn't carefully check how closely the playlist's rounded figure of 1.1 hours really was to 66 minutes on either computer) in 47 seconds -- no significant difference.
Going the other direction, my G4 took 3:40 -- about an 18x rate. Since I believe the AAC encoder is written to take advantage of dual processors, this would appear to mean that the single 1.8 GHz G5 processor ran about 20% faster on this task than two 1.25 GHz G4s. This could mean that a dual 2.0 GHz G5 system could get up into the 40-50x range.
I also ran my own Java speed test applet. The G5 was faster on all of the scores, some just a little, some significantly. This applet doesn't use multi-threading, so it's more of a comparison of single-processor performance -- one 1.25 GHz G4 vs. one 1.8 GHz G5.
(All values in milliseconds, smaller is better.)
double, trig, allocation -- G4: 2581, G5: 2211
double, trig -- G4: 2096, G5: 1550
double, allocation: G4: 790, G5: 660
double -- G4: 477, G5: 166
int -- G4: 860, G5: 819
long -- G4: 7881, G5: 1633
The G5 really shines (at least via Java) in double precision and long integer performance.
The bad news about the G5 didn't come out of my own testing, but in something I heard about Virtual PC from a sales guy. Since that's about software, I've put my comments about VPC in the software forum:
Now I know why Virtual PC won't work on the G5
Originally posted by Addison
I think that without Panther the G5 isn't going to show its real potential.
A good compiler can help a lot. Check out the public beta of IBM's XLC compiler specifically for the G5 on OS X. Apparently it can be much faster than GCC 3.3...
http://www-3.ibm.com/software/awdtools/ccompilers/
-- Mark
Originally posted by fieldor
xbench update: a 2GHz G5 MP is slower than an 1,6 ghz G5. Something really wrong with xbench. PLz Use the Ibm compiler to get real results
That assumes that the problems with Xbench can be fixed by the IBM compiler.
You should always look for algorithmic problems before you go mucking about with the compiler, because that's almost always where the problems are. See the outlandish memory bandwidth results reported for a G4 upthread.
http://www.barefeats.com/g5.html
Their summary:
ANALYSIS
It looks like the dual G5 2GHz has the best bang for the buck. The average speed advantage over the G5 1.8GHz is 45%. Yet it only costs 25% more (factory standard configuration).
We're working on producing results from other applications (UT2003, Quake3, FCP). We'll post them here as we compile them.
"PANTHER PUNCH"
Meanwhile, here's some data on the speed increase that OS X "Panther" (10.3) will provide G5 owners once it's released. We ran Xbench 1.1 on a G5 1.8GHz with 10.3 beta build 7B49. Compared to 10.2.7 "Jaguar"....
....CPU score increased 40%
....Thread score increased 44%
....Memory score increased 38%
Originally posted by Carson O'Genic
"PANTHER PUNCH"
Meanwhile, here's some data on the speed increase that OS X "Panther" (10.3) will provide G5 owners once it's released. We ran Xbench 1.1 on a G5 1.8GHz with 10.3 beta build 7B49. Compared to 10.2.7 "Jaguar"....
....CPU score increased 40%
....Thread score increased 44%
....Memory score increased 38%
Thanks for the link.
However the Punch with Panther is so high, that it demonstrate not only that Panther will be a good upgrade, but that Xbench is not the good benchmarks for a G5.