Ok, duh - disabled auto config and tweaked. Am I tweaked at the right settings here, or can I get more out of it? Thanks.

3DBench 1.0C went from 81.x to 92.8 w/ this new config.

bp

Looks perfectly fine so far...

A few notes: You seem to use a different version of Speedsys (f.e.: it measures a "Memory speed index" instead of the usual "Memory bandwith").
I can't make out the version number in the top right corner because the image is not centered on the monitor, but the one we usually use is 4.78 (BTW, the program can save the results into a PCX image by pressing the R key).

There are still some things you could tweak for a bit more performance in some situations, but they come with the risk of an unstable system:
- Overclock the ISA bus even more: ISA Bus Clock: 1/3 PCLK (13.3 MHz instead of 10 MHz)
- Tighter ISA timings: 16-bit I/O Recovery Time: 2 BUSCLK, 8-bit I/O Recovery Time: 3 BUSCLK
- Spread out RAM refresh cycles: Slow DRAM Refresh: Enabled

Finally, I'd like to ask you for a favour: Could you run the Speedsys test with the L1 cache disabled, but L2 cache enabled (BIOS Features Setup -> CPU Internal cache: Disabled, External cache: Enabled) and post your speedsys results here? I'm getting strange results on my board when doing this...

Thanks, will do. Will check ver # also.

So I went from 7.xMHz ISA to 1/4 bus (10MHz) and in Win95 the EtherLink III card was no longer working. So, I rolled back the ISA clock to 7.xMHz and still issues w/ the Ethernet card. Coincidental w/ something else? Can most cards handle 10MHz ISA? (40MHz bus.)

ISA cards I have: 3Com EtherLink III, GUS, Adaptec 1542.

Thanks...

bp

My AM5x86 @ 160mhz gets 59.79. In real world performance, having tighter memory timings improve Quake heavily but turns soft resets into a freeze.

leileilol wrote:

My AM5x86 @ 160mhz gets 59.79.

Where? What metric is that? Tnx.

bp

Also my memory timings are extremely similar to yours. Notice that I lack L2 cache however. Once you disable the L2 you should see a similar dip after the peak.

blakespot wrote:

So I went from 7.xMHz ISA to 1/4 bus (10MHz) and in Win95 the EtherLink III card was no longer working. So, I rolled back the ISA clock to 7.xMHz and still issues w/ the Ethernet card. Coincidental w/ something else? Can most cards handle 10MHz ISA? (40MHz bus.)

May be coincidence, it should work with the fixed ISA clock. I've never used an ISA NIC on this board, however I had trouble finding a PCI NIC that would reliably work with the 40 MHz bus.

BTW, found some old 5x86-on-PVI-486SP3 results:

I find this tool useless for benchmarking. Supposedly slow computers are race horses in other benchmarks.

Stick to 3dbench2, Doom, PCPBENCH and Quake.

Mau1wurf1977 wrote:

I find this tool useless for benchmarking. Supposedly slow computers are race horses in other benchmarks.
Stick to 3dbench2, Doom, PCPBENCH and Quake.

Well, you could say the same about 3dbench and Doom. 😉 And PCPbench heavily depends on bus speeds. AMD 5x86@160 MHz scores the same as Intel DX4@120 MHz.

I agree that the SST CPU score is not very useful in some cases, but the cache/memory graphs are really cool, especially for noobs who don't know how to interpret cachechk readings for example.

Benchmarking applications themselves are mostly useless. Benchmarks can be heavily optimised for certain hardware and then not be a real representation of how that same hardware performs on real world.

I just see how well my hardware runs the programs i need. If i need something faster i'll buid something faster.