At 08:25 �� 19/4/2002, you wrote: >Phoebus Dokos wrote: > > Is that possible? > >Hard to say, my database is not based in GCC compiled binaries, but >once can take a rough factor of 1.3 to 1.33 between C68/GCC and >according to that it should be 25000-25500 i/s for an Athlon 750 >(Thunderbird, that is). > >For some reason I don't have any QXL values in my database, but >according to http://sol.spaceports.com/~vader/text/bench.htm it's > >QXL 20MHz >Dhrystone 6578 Dhrystones/Sekunde >Bogomips 9,58 Bogomips > >QXL2 25MHz >Dhrystone 8620 Dhrystones/Sekunde >Bogomips 13.42 Bogomips
It's @ 25 MHz Also the Dhrystone figures (Me thinks) are with the previous version of Dhrystone. This is the 2.1 one :-) (Thierry says IIRC that's it's 33% faster) >This is a lot less from your figures, which by the way almost resemble >the ones of a Q40 (36443) and that shouldn't really be the case. Mind >you, don't do any tests in a Windows DOS box, the timer rate is all >wrong there. Also on the QXL there quite some factors that can affect >the speed pretty much, like the DISP_UPDATE rate (which should be set >to IIRC 10,10 if you want a fair comparison). I run it starting with DR-DOS 7.3, however I am going to run some extra tests and change the DISP_UPDATE too I want to be accurate... however it did stun me. The QSBB benchmark shows the difference in the I/O (visible with the naked eye too). QPC of course beats the QXL (any QXL unless a PCI version materialises, -Knowing Dave that's not entirely impossible!-) by leaps and bounds in disk access and graphics. It would be extremely interesting to see a QPC version handling all the screen, disk I/O and assisting with the FPU and the 68040 doing the raw processing :-)Kinda like the QL emulator on the Atari :-) >According to the figures on the page a K6-III 400 is comparable to the >QXL. But generally one can say that the QXL has decent power but even >a much slower QPC usually feels faster because of the better IO >response. I agree. Even with a Celeron 400 and only (!) 64Mb QPC does feel faster (is indeed though slower in real number crunching operations). I am going to try and find a full blown 040 and also to push it up a tad to see what gives :-)) (Yeah I know! Speed freak!). One thing I did not try though is to test with QSBB and QL colours option. Also it would be interesting to measure QPC 2 v.3... Off I am to download the demo :-) Phoebus >Marcel > > > >--- >Incoming mail is certified Virus Free. �� ����������� ������ ����� >�������� ���. >Checked by AVG anti-virus system (http://www.grisoft.com). >Version: 6.0.350 / Virus Database: 196 - Release Date: 17/4/2002
