Using The K6-III on Older Boards

Pages: 1 2 3 4 5 6 7

The M Tech R581A

UPDATED! Rise Computer, Inc – the actual manufacturer of this motherboard, has provided an updated BIOS file to support the CXT core, and the K6-III. Using this new BIOS, we retested the R581A, and found the performance improved as much as 6% in Winstone 99! Read on for the latest info

The M Tech R581A motherboard is similar to the PA-2007 in that it sports 1MB cache, supports SDRAM and is limited to a 10A maximum current. One possible advantage with this motherboard over the PA-2007 is its official 83MHz bus speed support, courtesy of the SiS 5591 chipset. The 83MHz bus speed runs the PCI/AGP bus asynchronously (32MHz/64MHz) for more stable operation. The 75MHz bus speed can be set synchronously or asynchronously, so we tried both and found that there was almost no difference in the test results. For this reason, all test results listed are for the asynchronous setting. The new R008 BIOS file provides support for all the new features of the CXT core, and recognizes the K6-III processors.

Just as with the PA-2007, the fastest K6-2 speeds we can run without exceeding 10A is 366MHz (5.5 x 66MHz = 9.45A) and 375MHz (5.0 x 75MHz = 9.7A), but we also have the option of also using 373.5MHz (4.5 x 83MHz). With the K6-III, we must run at a the slower speeds of either 333MHz (5.0 x 66MHz = 9.3A), 337.5MHz (4.5 x 75MHz = 9.4A) or 332MHz (4.0 x 83MHz). While the motherboard does offer 90MHz and 100MHz jumper settings, these are not very stable. The only voltage setting available that will work with both the K6-2 and K6-III is 2.31v, by jumpering both VID0 and VID1. The multipliers are obtained by jumpering JP8, JP9 and JP10 as follows: 4.0x = COC, 4.5x = CCC, 5.0x = OCC, 5.5x = OOC where C=closed and O=open.

The test setup included the R581A rev B3 board with 1MB onboard cache (BIOS Level R008), 64MB PC66 SDRAM, Western Digital 1.2GB (PIO 4) HDD, Diamond FireGL 1000 Pro w/ 8MB (BIOS Level 1.54), Windows 98 (release 1), DirectX 6.1 and Diamond 4.10.01.2359 video driver. The BIOS settings were set to ‘Setup Defaults’ for this run except for one comparison test where the SDRAM CAS Latency was set to 2T, WR Retire Rate was set to X-1-1-1 and Wait State Control was set to 0WS. The display properties were set to 1024×768 and 16 bit color. The benchmarks run were Winstone 99, 3DMark99 Max Pro and Rage’s Incoming game benchmark.

CPUSettingWinstone 993DMark99 Max Pro
CPUMarks
3DMark99 Max Pro
3DMarks
Incoming
Worst FPS
Incoming
Best FPS
Incoming
Avg FPS
K6-2 3665.5 x 66MHz15.243926747.0831.8518.29
K6-2 3755.0 x 75MHz15.748286869.5631.8918.47
K6-2 373.54.5 x 83MHz16.2499572110.6231.8918.80
K6-III 3335.0 x 66MHz17.5476471410.1231.9218.99
K6-III 337.54.5 x 75MHz17.9500572010.4431.8518.96
K6-III 3324.0 x 83MHz18.2503074210.4631.9219.19

CAS Latency = 2, WR Retire Rate = X-1-1-1, Wait State Control = 0WS
CPUSettingWinstone 993DMark99 Max Pro
CPUMarks
3DMark99 Max Pro
3DMarks
Incoming
Worst FPS
Incoming
Best FPS
Incoming
Avg FPS
K6-III 3335.0 x 66MHz17.8498872710.2931.9019.05

With the new R008 BIOS, the R581A appears to be the fastest of all motherboards tested so far in Winstone 99, but lags behind the others in the 3D Games department. Using the 83MHz bus speed, the performance of the K6-III @ 332MHz in Winstone 99 is very near what we have gotten with a Celeron 400 on a ZX chipset motherboard. We did try to run the K6-III at 4.0 x 90MHz (360MHz), which would be just under 10A with the board running at 2.3v, but we could not get it to complete any of the tests.

As would be expected, the difference in performance between the 66MHz bus and 83MHz bus is much less noticable with the K6-III. In fact, the K6-III at 83MHz is only 12% faster than the K6-2 at 83MHz. While the K6-III is running slower than the K6-2, running either one any faster would potentially cause problems by exceeding the motherboard specifications.


Pages: « Prev   1 2 3 4 5 6 7   Next »

Be the first to discuss this article!