A few weeks ago, I was mentioning that I was considering a CPU upgrade. I have now done that, and I wanted to mention the following single benchmark that I made. I went from a 233Mh Pentium 1 to a 2Gh Pentium 4. I used DOS 4DOS under DOS 6.22 on both machines. (Sorry, Rich!) I was told that the P4 emulates 16-bit applications slower than the original Pentium does, so I didn't know what increase in speed I'd get. There was an 858% increase in CPU processor speed for the two machines, but what would be the increase in Fractint calculation speed? I ran an arbitrary precision mandelbrot on both: Pentium 233Mh: 0:37:44.42 (2264.42 sec) Pentium 4 2Gh: 0:06:04.33 ( 364.33 sec) or a 621.53% increase speed for the Fractint CPU time. The ratio of the Fractint speed increase to the CPU speed increase is .7244, the penalty for running a 16-bit application on the P4. This ratio is higher than I expected, and I do get more than a 6-fold increase in Fractint calculation speed. But what if I had upgraded to an AMD Athlon? I would like to know what this ratio would have been for that! Lee Skinner
Lee Skinner wrote:
A few weeks ago, I was mentioning that I was considering a CPU upgrade. I have now done that, and I wanted to mention the following single benchmark that I made. I went from a 233Mh Pentium 1 to a 2Gh Pentium 4. I used DOS 4DOS under DOS 6.22 on both machines. (Sorry, Rich!) I was told that the P4 emulates 16-bit applications slower than the original Pentium does, so I didn't know what increase in speed I'd get. There was an 858% increase in CPU processor speed for the two machines, but what would be the increase in Fractint calculation speed? I ran an arbitrary precision mandelbrot on both:
Pentium 233Mh: 0:37:44.42 (2264.42 sec) Pentium 4 2Gh: 0:06:04.33 ( 364.33 sec)
or a 621.53% increase speed for the Fractint CPU time.
The ratio of the Fractint speed increase to the CPU speed increase is .7244, the penalty for running a 16-bit application on the P4. This ratio is
Er, it's a lot more complicated than that. You can't really compare clock speeds between different processor models. The only meaningful comparison is just running a benchmark and comparing times (which is what you did, of course :)
higher than I expected, and I do get more than a 6-fold increase in Fractint calculation speed. But what if I had upgraded to an AMD Athlon? I would like to know what this ratio would have been for that!
I have a 1.4GHz Athlon. Please post the PAR you used for your benchmark (and the video mode) and I'll be glad to give it a run and post the times. I'm curious about the result myself. Pedro
and i'll throw more petrol on the fire: it's more likely that a lot of what's going on with display is directly attached to the amount of video memory you have, how fast the card can display, and the quality of its dos drivers... i think, lee, a better comparison would be done amongst machines with absolutely identical hardware, then changing out video and cpu as you can... i would not be surprised to see vast differences in speed for same CPU when video cards are different... Pedro Lopes <paol1976@yahoo.com> wrote: Lee Skinner wrote:
A few weeks ago, I was mentioning that I was considering a CPU upgrade. I have now done that, and I wanted to mention the following single benchmark that I made. I went from a 233Mh Pentium 1 to a 2Gh Pentium 4. I used DOS 4DOS under DOS 6.22 on both machines. (Sorry, Rich!) I was told that the P4 emulates 16-bit applications slower than the original Pentium does, so I didn't know what increase in speed I'd get. There was an 858% increase in CPU processor speed for the two machines, but what would be the increase in Fractint calculation speed? I ran an arbitrary precision mandelbrot on both:
Pentium 233Mh: 0:37:44.42 (2264.42 sec) Pentium 4 2Gh: 0:06:04.33 ( 364.33 sec)
or a 621.53% increase speed for the Fractint CPU time.
The ratio of the Fractint speed increase to the CPU speed increase is .7244, the penalty for running a 16-bit application on the P4. This ratio is
Er, it's a lot more complicated than that. You can't really compare clock speeds between different processor models. The only meaningful comparison is just running a benchmark and comparing times (which is what you did, of course :)
higher than I expected, and I do get more than a 6-fold increase in Fractint calculation speed. But what if I had upgraded to an AMD Athlon? I would like to know what this ratio would have been for that!
I have a 1.4GHz Athlon. Please post the PAR you used for your benchmark (and the video mode) and I'll be glad to give it a run and post the times. I'm curious about the result myself. Pedro _______________________________________________ Fractint mailing list Fractint@mailman.xmission.com http://mailman.xmission.com/cgi-bin/mailman/listinfo/fractint --------------------------------- Do You Yahoo!? Get personalised at My Yahoo!.
At 10:49 06/03/02 +0000, you wrote:
Lee Skinner wrote:
A few weeks ago, I was mentioning that I was considering a CPU upgrade. I
I have a 1.4GHz Athlon. Please post the PAR you used for your benchmark (and the video mode) and I'll be glad to give it a run and post the times. I'm curious about the result myself.
Pedro
Hi Pedro, a few years ago I took this .par (on a floppy-disk with fractint V.18.2) to buy my now used PC, an Athlon K7-600. The time for the fractal was 2.35 Sec. (under Fractint Version 18.2, Video=f3 and viewwindows=1.8) Can you please test it, as it is, on your machine? Now, under the Ver. 20.02.3 it runs in only 1.42"... here is my old original speedtst.par (again): ******************************************** speedtst { reset=1820 type=mandel center-mag=-0.74909880336762710/+0.05084338424079404/1.304918e+012 params=0/0 float=y maxiter=9000 inside=0 logmap=2178 periodicity=0 viewwindows=1.8/0.75/yes/0/0 colors=000zI0<2>z00<3>m06j07g09<3>W0ET0FP0HM0II0K<3>809506303000<6>21H21\ J31L<3>42V42Y52_<3>63i69i5Gh4Mf<3>2Zb2ab2ca<3>1k_1mZ0oZ0qY0rX0tX0vX0vX<1\ 6>8KG9HF9FE9CDAAC<2>B39C08C09<3>G0GH0II0KL0JO0HS2GV0E<3>h08l06o05<2>z00<\ 4>zT0zY0zc0<2>zt0yy0zu0<3>zZ0zT0zN0<3>z00<3>m06j07g09<3>W0ET0FP0HM0II0K<\ 3>90B708506303000<2>01612913B14D14F<6>39V3AX3BZ<2>4De5Eh4Kg4Of3Sd<3>2ba2\ e`1g`<3>1oZ0pY0rY0tX0vW0vX<16>0KB0H90F80C70A6<2>032000202<3>B0CD0FG0HI0K\ <4>Z0Da0Ce0A<2>o05s04w01y00<3>yN0zS0zY0<3>zt0yy0zu0<5>zN0 } ********************************* Cheers, Guy
Hi Lee, At 20:13 05/03/02 -0500, you wrote: ....
But what if I had upgraded to an AMD Athlon? I would like to know what this ratio would have been for that!
Lee Skinner
sorry, I was unable to run most of the new machines running now Windows-XP with my old DOS 6.22 bootable floppy-disk (1.44MB) without being detected by vendors.. I had to set the primary boot to floppy-disk... and so on.. When I got more infos's about your question, I will tell ya.. cheers, Guy
I ran Lee's benchmark on my home machine (1.4GHz Athlon, Win98, Fractint 20.2.03). The result was a cool 4m43s, even better than I expected. I actually had to time a run in batch mode because fractint goes weird when I try to leave 1600x1200 graphics mode, so that time includes Fractint startup time and image loading & saving. Back when I bought my machine the Athlon 1.4 and the P4 2.0 were the fastest chips available. Most benchmarks I read gave similar performance for both, with the P4 pulling ahead only when running code optimized for it's new architecture. Also, at the time the price of CPU+Motherboard+RAM for the P4 2.0 system was something like 3x more (!) than the Athlon 1.4 system, so the choice was easy :) It would be interesting to see the same benchmark run on a P4 under Xfractint (the non-ASM version). I would expect it to perform a bit better than the DOS version. Anyone? Now for Guy's speedtst parameter, it finished in 0.86 seconds. That's actually too fast for a useful benchmark I think. Maybe we should pick one of the PARs distributed with Fractint as the (semi-)official benchmark. It would make it easier for people to run tests and compare results. Pedro
Hi Pedro,
Now for Guy's speedtst parameter, it finished in 0.86 seconds.
thanks for this info..
That's actually too fast for a useful benchmark I think.
right, but for my under cover actions, about 2-3 years ago, it was ok..
Maybe we should pick one of the PARs distributed with Fractint as the (semi-)official benchmark. It would make it easier for people to run tests and compare results.
yeahh! let's take one for the Fractint version-optimized "Type=Mandel" (that's a problem when it has to be comparable) and one without optimized code..
Pedro
_______________________________________________ Fractint mailing list Fractint@mailman.xmission.com http://mailman.xmission.com/cgi-bin/mailman/listinfo/fractint
participants (4)
-
caren -
Guy Marson -
Lee Skinner -
Pedro Lopes