Babasha wrote on 2025-08-06, 11:51:Hi!
With little delay my BIOS speed comparison table is ready
https://docs.google.com/spreadsheets/d/1Set4g … dit?usp=sharing […]
Show full quote
Chkcpu wrote on 2025-08-01, 10:02:Hi Babasha, […]
Show full quote
Babasha wrote on 2025-07-30, 06:13:
Hi!
I already try all BIOS'es for my mobo inc. eSupport 2004 version.
Its sad eSupport 2004 is rich-function one while been one of the slowest in DOS and W98 too with AMD K6-2/450 and AMD K6-2/500. I Try to repatch it too but its gain intermediate performane boost compared to "my" patched version.
Hi Babasha,
Weird, the 2004 eSupport BIOS is the most advanced and should give you the best performance, not the lowest.
I got curious to where these performance differences come from, so in the 86Box emulator I experimented a bit with the original 0619 BIOS and the BIOS Patcher versions, both with and without the New Koeffs support patch.
However, I didn’t find any difference in chipset programming that could explain a difference in timing settings for L2 cache or RAM.
Can you explain how you measured these performance differences?
I’m also curious about the board setup:
- Did you use the Setup defauts in the BIOS, or Manually optimized settings?
- How much RAM was installed?
- Which memory sticks were used; FPM, EDO, or SDRAM?
- What was the CPU setting? Was it 6x75 for the K6-2/450 and 6x83 for the K6-2/500?
Thanks in advance for answering these question.
Regards, Jan
Hi!
With little delay my BIOS speed comparison table is ready
https://docs.google.com/spreadsheets/d/1Set4g … dit?usp=sharing
So on 500MHZ CPU - everything are OK and works as expected
If I try it with 450MHz:
1) eSupport 2004 is slowest BIOS (Q2DOS and Q2WIN shows it and there is strange video memory speed slowdown in SST 4.78)
2) 80619-L pathed with BP 4.23 without "NEW KOEFFS SUPPORT" is fastest BIOS but without proper FREQ detection on POST screen
3) 80619-L patched witt BP 4.23 and with "NEW KOEFFS SUPPORT" is intermediate performance, correct FREQ detection on POST screen
I try BIOSes from ZIDA 5SVA - they are with same symptoms((( You can get correct FREQ detection on POST screen but not so fast performance or vice versa(((
Hi Babasha,
Thanks for the nice benchmark table. This makes your oberservations quite clear.
Here are some possible explanations for the differences at 500 and 450MHz.
-----
The performance figures with the original 80619-L BIOS are obviously lower than with the eSupport and patched BIOSes, both at 450 and 500MHz.
This is caused by the lack of Write-Allocation. All K6-2/400 and faster CPUs have the improved CXT core with x6 multiplier support. The original BIOS doesn’t support Write-Allocation for the CXT core, but the patched versions and the eSupport BIOS do.
A clear indication of missing Write-Allocation support is lower performance figures in SpeedSys, except for the MEM figure which is higher. This is a curious side effect of how Write-Allocation works.
-----
The patched ‘PTC’ BIOSes, without the New Koeffs support, have the best performance at both 450 and 500MHz, although the BIOS POST reports 400MHz.
This is expected, because the 400MHz display limit in these 1998 BIOSes is just cosmetic and doesn’t influence performance. The table for the CPU speed display in the BIOS simply doesn’t go any further than 400MHz.
-----
The attachment Monitor_1_20250813-162915-331.png is no longer available
As shown in the above picture, the BIOS sets the “DRAM Auto Configuration” to Disabled and “DRAM Timing Control” to Normal as defaults. These settings give you a fixed but conservative DRAM timing. Did you use these settings for your tests?
With these settings I can’t explain the strange low performance of the eSupport and fully patched BIOSes at 450MHz.
However, if you used “DRAM Auto Configuration” at 60ns or 70ns and “DRAM Timing Control” at Auto, the eSupport and fully patched BIOSes can have influenced the DRAM timing here.
My hypothesis that this only happens at 75MHz FSB, and not at 83MHz, lays in the fact that the VPX chipset has native 75MHz support and the BIOS supports this, causing relaxed the DRAM Timings when 75MHz FSB is detected. However, at 83MHz FSB, the BIOS is unaware of this setting and simply uses the faster DRAM timings for 66MHz FSB.
You can test this hypothesis by setting “DRAM Timing Control” on Fast or Turbo instead of Auto. If I’m correct, you should get the same performance with, or without, New Koeffs support at 450MHz, just like at 500MHz.
-----
To add another BIOS to the mix, here is a Unicore BIOS upgrade for your VPX board.
This 03/15/2000-580VPX-PRIME3-2A5LDV3AC-00 BIOS supports all socket 7 CPUs ever made and supports IDE drive up to 128GB.
The attachment 2A5LDV3A.zip is no longer available
Happy testing,
Jan
