home *** CD-ROM | disk | FTP | other *** search
-
- QDRAM -- Decreases refresh cycle timer for the dRAM refresh rate to speed
- DOC up CPU performance. Normally, a HEX value of 12h is in the I/O
- register location 41h of the 8253 timer controlling refresh rate.
- To find the right for your machine, start increasing the value of "12" in
- line 5 to a maximum of "FF" (these are HEX values) and run TEST.BAT after
- each increment. If "FF" is reached without the CPU stopping due to parity
- errors (you'll have to COLDboot the machine), then increase the value "00"
- in line 7 (this is the high byte value for the timer) to "01" and restart
- the incrementing in line 5 from "00" to "FF" while running the benchmarks
- in file TEST.BAT. You'll notice that the increase in CPU performance will
- rapidly reach a plateau and eventually you'll get a parity error (in my AT
- with an Inboard 386/AT, an error is obtained when the low byte [line 5] =
- "0C" to "0F" and the high byte [line 7] = "0F," but was obtained at lower
- values in an IBM XT).
-
- Use the lowest value providing the best performace increment for the final
- version of QDRAM. This should increase the equivalent CPU speed by about
- 5% to 10%, depending on your machine. I use a low byte value of "8F" and
- a high byte value of "00" in the 386ed AT to increase the equivalent speed
- to 24 Mhz from its original of 22 Mhz (Landmark test in CPUSPEED.COM), so
- far without any parity error problems.
-
- Based on PC Magazine July 1988 (v7 n13) PC Lab Notes (Brian K. Roemmele).
-
-
-
-
- TEST -- This is a batch file to automatically create QDRAM.COM via DEBUG
- DOC (you must supply your own version) and to test the results of the
- change in dRAM refresh rate with a variety of benchmarks:
-
- 1. Landmark Test in CPUSPEED.COM (provides equivalent CPU speed)
- 2. Processor Speed test in PC Magazine's BENCH21.EXE and BENCH21A.EXE
- 3. Instruction Mix Speed test in PC Magazine's BENCH28.EXE
- 4. Video Display times in PC Magazine's BENCH24.EXE
-
- TEST.BAT also invokes the performance tests of VBENCH.EXE from Golden Bow
- Bow Systems (usually distributed with the VOPT utilities), and of SI.EXE
- from the Peter Norton Computing, Inc (usually distributed with the NORTON
- utilities). These two programs are not included in this ARChive, so you
- should delete such calls if you do not have the programs.