Speed testing

[russ Boehle]

Has anyone done any testing of load speed of A5 on a network optimized app with graphics and without?
I have an app developed on an Athlon 1.6 w. 512 meg ram. On this machine it runs fine with what I would call snappy performance. When I run this on slower machines on the network, ie. 450 mhz, it really slows down. The graphics (jpgs and bitmaps) are on the client machines. If it made a big difference I would strip all the graphics as they are cosmetic only. I don't want to strip them though if it doesn't make any difference. Any ideas?

Russ

[JerryBrightbill]

RE: Speed testing

Russ

Graphics can slow down any system, regardless of the program being used. There are a few ways to improve the situation. First, use the lowest resolution graphics you can. If the graphic is just for screen use, try 72 dpi and set the size to the size you want to display. For example a full screen display graphic for a 800 x 600 monitor should be 800 x 600 at 72 dpi, which creates a file size of about 1.4 meg. One thing that helps is to use bmp graphics since they don't have to decompress to view. This will be usually be slightly faster than using jpg. Another thing that may improve speed is to set most graphics to absolute size, rather than stretch, since this doesn't require any manipulation by the program to display.

I have used a number of graphics in applications, including full screen backgrounds. These open quite quickly, even with a 300 mhz processor. I try to follow the rules above and try to keep large grahics as simple as possible and no one has mentioned any speed issues, even on very slow computers.

Jerry

[JerryBrightbill]

RE: Speed testing

Russ

By the way, since the speed of a terminal running optimized is primarily determined by the terminal hardware since everything running locally, a 450 mhz machine will be significantly slower that a 1.6 ghz in all actions. With A5V5, a 450 mhz computer is at the very lower end of acceptable, although I have run A5V5 on a 300 mhz machine. Most people would say that a 1.4 to 1.6 ghz processor is a more normal lower limit to get good performance.

Jerry

[russ Boehle]

RE: Speed testing

thanks Jerry, Yeah, I know faster would be better, but unfortunately thats not in the cards at the moment. We have to do with what we have for the most part. What you say is pretty much what I figured, I hadn't thought of the jpg decompression though, so that's a good idea. I have a lot of bitmaps on buttons etc. I will probably leave these but take a look at the jpg's since they don't do anything anyway.

Hopefully by the end of the year we can bump everyone with slow machines up to something new-- lets hope the economy takes off, I think it's time for it to.

Russ

[JerryBrightbill]

RE: Speed testing

Russ

I haven't found that bitmaps on buttons have any real effect on speed. I usually stick to the built in bitmaps which load very fast. If you are specifying external bitmaps, I have found the best course is to embed them. If you link them, it is very easy to lose the link, plus it does seem slower than an embedded bitmap.

Jerry

[Raymond Lyons]

RE: Speed testing

Yes but. For most apps performing most functions you should not notice all that much difference between 450 mhz and, say, 1.6 ghz. If you do there is most likely some other explanation such as a mucked up windows setup (win98 especially seems to get sluggish after a year or so of loading this and that program) or something having to do with the network (protocols, cables, a slow hub instead of a fast switch, etc.) or a slow hard drive. I have 500 mhz win98se networked machines that seem remarkably fast while others are slow as can be. In one real slow case all we did is replace the hard drive ($75), load w2k instead of win98se and just that quick it became, relatively speaking, a speed demon.

But then too, I rarely use graphics in my apps, so I have no experience with the difference that might make. But I don't think it would make much difference.

One last thing. One of my clients recently switched from a 10BaseT hub to a 100BaseTX switch. As the numbers 10 mps vs. 100 mps indicate, it is a fantastic increase in speed for things that have to flow over the network. The only hitch was that in the wall cables that worked at 10 mps would not work at all at 100 mps. Regular Cat 5 cables should have worked but many did not. We had to replace everything with Cat 5 Enhanced (Cat 5e) cables and pay more attention to going over things like florescent lights. This of course has nothing to do with loading forms from the local workstation. It's just an aside that may be of interest to some.

Ray Lyons

[russ Boehle]

RE: Speed testing

Switching to win2k from 98se is on the list also but 98 is bought and paid for 2k isn't so that's waiting too. Our network is all switched 100 meg, so we're OK there but I also have seen that there are fast win 98 machines and slow ones. I also have seen what appear to be vast differences between motherboards (probably due to chipset) as well. One area where I do notice a difference with A5 is opening and positioning forms. On The fast machines you really cant see the locating. On the slow ones its obvious. Not a big deal but it's noticeable. Also some of the screens paint kind of funny (as in slowly) I assume this is the video card, but we don't buy gamer video cards for office machines.

Russ

[Raymond Lyons]

RE: Speed testing

FYI, I have found that ATI still makes some comparatively inexpensive ($30 to $80) video cards that work well with A5. Good warranty and support. Other even less expensive brands I have tried were a mixed bag and not worth fooling with. On my development machine I have an ATI AGP card with 128MB of DDR memory, analog, digital and TV out capable of running two monitors at once and that can be had for about $115. Too much for all the workstations but nice for a development machine.

I have also had good luck with the integrated video on Intel motherboards (what many Dell machines use) that I build into workstations myself and that only costs $7 - $15 more than an Intel board without integrated video.

Ray Lyons