(If you own a new Quicksilver system and would like to run tests - see the bottom of the page for links to benchmarks. I'd also welcome real apps tests such as those shown in my review of the G4/733 w/GeForce3. For instance the iMovie 2 tutorial export time (stacked clips, no transitions, export to "CDROM medium" settings - using stopwatch to time how long it takes to complete the export.)
G4/867 (2MB L3 cache) Tests:
I bet it has been really busy. All tests were ran on the standard
config of an 867 with VM off, under 9.2.
As for the PPCchecker it reports the following info:
PVR = 0x80000200
As for the GaugePro info it reports the following:
Memory Performance 155.5 MB/sec, using the stock 128 MB from Apple and
but I would take that number with a grain of salt as it also says the CPU
is running at 398.6 MHz.
[I noted the same issues in my previous 2001 Digital Audio G4/733 w/Geforce3 performance tests. However other memory bandwidth benchmarks like Stream also seem to show lower than 7400/7410 rates with the 7450 CPUs, at least the current 2.0 steppings. The first G4/733s (Digital Audio systems) were also rev 2.0 CPUs.-Mike]
As for the Altivec Fractal Demo the results are:
Elapsed Time 4.4 seconds - Achieving 3072.1 MegaFlops
(using the defualts when the program is first dl and opened.
[the G4/733 w/1MB L3 system delivered 2,650.5 MegaFlops. A
Dual G4/500 Dual delivered 3,283.6 MegaFlops in my prev tests. -Mike]
The G4 TimeDemo set to amazing and millions was 97.6 fps, this was using
hardware rendering, software did not work.
[The G4/733 (1MB L3) w/GeForce3 delivered 84.8 FPS in my prev. tests-Mike]
That is all the tests I could run this morning before going off to work.
G4/733 (no L3 cache) Fractal Tests:
The only test I had handy to compare the new 733 to your older tests was the
Fractal Carbon test. In OS9 the 733 did repeatable 2.7 Gflops and with
certain settings got up to 2.9. In OS X (128MB Ram) the highest I got was
2.5. I'll try to do the other tests tonight when I get home. On this test
anyway, It does not seem like too bad a performance hit re lack of L3.
Datalux Tech Support
Just did the Throughput test ( 1.5.) on this GeForce 2MX Quicksilver 733.
CPU 32 bits 191.6 MB/sec
[a G4/867 owner reported 213.0 MB/sec-Mike]
FPU 64 bits 211.6 MB/Sec
[a G4/867 owner reported 213.2 MB/sec-Mike]
Altivec 128 bits 216.8 MB/sec
[a G4/867 owner reported 220.3 MB/sec-Mike]
Copybits 256 Busmaster 224.3 MB/sec
[a G4/867 owner reported 252.7 MB/sec-Mike]
Compared to the "old" 733 with the Geforce *3* results that you got:
[The tests I ran used OS 9.1/2.1 nvidia drivers, new systems ship with OS 9.2 and 2.1.1 drivers, which may or may not affect Throughput results-Mike]
CPU = 213.9
FPU = 214.5
Altivec = 221.6
Copybits = 252.1
[Note these results are about the same as the G4/867's scores-Mike]
Note that the only difference worth noting is the slightly higher GF3
busmaster score. I would say they further increased the bus performance of
I had another thought on the lack of L3 Cache: I bet they decided they didnt
really *need* it. Here's why.....the 867 and dual 800 have a L3 cache
running at 1/4 of the processor speed....or around 216 Mhz. The dual 800
even less. When you do this to the 733, you wind up with a cache speed of
183 or so. Thats only 50 Mhz above the System Bus!!!!
[However note as mentioned here before - the L3 cache is DDR, so the effective rate is 1/2 the CPU speed, not 1/4. I suspect the removal of the L3 cache on the new 733 is just a cost cutting measure.-Mike]
I remember reading
that they did away with a lot of chips in the new board related to the bus
in an effort to reduce latencys, etc. in a interview done with one of the
Apple higher ups at MWNY.
Since this machine is just as fast as the former 733 with the 1 MB cache AND
the GF 3 in this test, I dont feel so bad about the missing cache. It may
have been redundant in the 733. I bet that why they went to 2 MB's in the
other boards too.
(From yesterday's news)
Mac G4/867 vs. PC P3/866 GeForce2MX Performance:
I thought you might be interested in this: I was
looking at Tom's Hardware today, and just for fun, I
decided to compare Oliver's results from the
QuickSilver G4 867 and one of Tom's tests.
mac os 9.2, default (fresh) config + 32bit color +
32bit textures + SOUND OFF:
640x480 122.4 fps
800x600 100.2 fps
1024x768 65.5 fps
1280x1024 51.7 fps
Tom tests with a Pentium 3/866, and in February, he
tested 5 GeForce2 MX cards. He was using Win 98, and
set no sound via the command line in Q3:
These are his 32-bit results:
640x480 ~118 fps
800x600 ~83 fps
1024x768 ~52-53 fps
Not exactly a scientific comparison, but I'd say that
Apple is making great strides in gaming performance!
some of the forum readers have been speculating on the new Quicksilver G4 as
to if they have this or that whatever. Went to the Apple Store today and got
a 733 model. I'm not interested in DVD's or video, and the 733 will surely
be quick enough for anyting I want to do for at least 2 years. I only do web
stuff, mail, and the usual games + I wanted to get into cd burning. So the
733 was a decent value for me and what I do with a computer.
The machine has 9.2 on it, with OS X 10.0.4 Build 5R14. [He later noted this was a typo - it's 4R14-Mike] The PPC checker
program says the Processor is a Power PC 7450, Vger, Rev 2.0. (G4Plus)
Not the 7440 everyone seems to think becuase of the lack of L3 cache. It
will hurt speed some but how fast does everyone think the machine has to be
when 7 out of 10 are still using 44K dial up connections? Makes no sense.
Have only used it for a few hours but it is so far ahead of the G3 upgraded
machine I had before there is no comparing the two machines.
The HD is an IBM 40 GB. The CD RW is a Sony. The "out of the box" experience
was good with no problems. There is no question but that Apple is going for
the Titanium look on the future desktops in my mind. Thye tried to make this
one look as much like metal as possible. The case is more attractive in
person than on the web. Dont judge it until you see it. The front speaker
actually sounds very good BTW, they upgraded that little item a lot.
Anyways, I am satisfied. If you want me to check for specific things let me
know and I'll do it.
I asked him to run some of the apps and benchmark tests I noted on the G4/733 w/GeForce3 performance page which used a 2001 digital audio system (w/1MB L3). Even though he doesn't have a Geforce3, the results of CPU bound tests would be interesting to compare with the 733 w/o L3.
Benchmarks: (these are linked in my G4/733 review also, as well as notes on application tests you may want to run)
Test results from Quicksilver system owners are welcome - email me.
Last week Oliver sent results of Quake3 in OS 9.2 and OS X with a G4/867 noted on this page.