mao@eden.Berkeley.EDU (Mike Olson) (12/10/90)
straying just a little from the original subject of this thread: we're running mit's R4 server on our decstation cluster at berkeley. we have noticed the following curious fact: identical X11 programs running under the identical X11R4 server binary do substantially faster screen drawing on our decstation 3100's than on our decstation 5000's. we have none of the interesting graphics options on either our 3100's or our 5000's. worse performance on the newer hardware is counter-intuitive. to be fair, this is the only instance of performance degradation we've seen. can anyone offer a plausible explanation for this behavior? mike olson postgres research group uc berkeley mao@postgres.berkeley.edu
jg@crl.dec.com (Jim Gettys) (12/11/90)
In article <9606@pasteur.Berkeley.EDU>, mao@eden.Berkeley.EDU (Mike Olson) writes: > straying just a little from the original subject of this thread: we're > running mit's R4 server on our decstation cluster at berkeley. we have > noticed the following curious fact: identical X11 programs running under > the identical X11R4 server binary do substantially faster screen drawing > on our decstation 3100's than on our decstation 5000's. we have none of > the interesting graphics options on either our 3100's or our 5000's. > worse performance on the newer hardware is counter-intuitive. to be fair, > this is the only instance of performance degradation we've seen. > > can anyone offer a plausible explanation for this behavior? I presume you are using DS5000/CX's... The CX does not have a plane mask. The DS3100/2100 does. So for programs which use plane masks .ne. all ones, it is possible a DS5000 could run more slowly than a DS3100. (since the server has to do read/modify/writes rather than just writes, for such operations). The other graphics options on the DS5000 do have plane mask hardware. For most applications, the DS5000/CX is substantially faster than the 3100. - Jim