🤣.
Do we have first of April, because you are complaining that the system is too fast?

SI is a synthetic benchmark, and does not say too much.
I guess you are running the L1 in WB, so you have high values.

AMD Am5x86@160MHz gives more than 300, so your values are not too high, don't worry.
Use other benchmarks to get more real world values (quake, doom, wolf3d, ....)

CoffeeOne wrote on 2024-01-04, 08:11:

LOL. Do we have first of April, because you are complaining that the system is too fast? […]
Show full quote

🤣.
Do we have first of April, because you are complaining that the system is too fast?

SI is a synthetic benchmark, and does not say too much.
I guess you are running the L1 in WB, so you have high values.

AMD Am5x86@160MHz gives more than 300, so your values are not too high, don't worry.
Use other benchmarks to get more real world values (quake, doom, wolf3d, ....)

If it's real, I'm not complaining 😉. LGR seemed to point out he couldn't get the CPU any faster, so I thought there must have been an issue with my benchmark.

I just ran a Quake benchmark and got 11.2 fps. LGR got 8.8 fps so the benchmark results must be real.

Very happy!

Um, do you have access to a copy of Central Point PC-Tools 7.x, by any chance?
It used to be the arch rival to Symantec Norton Utilities.
And it also has a "System Information" utility. Maybe it's more reliable, not sure.

Alternatively, there's CINFO, which looks similar. It can be found here.

zuldan wrote on 2024-01-04, 08:15:

If it's real, I'm not complaining ;-). LGR seemed to point out he couldn't get the CPU any faster, so I thought there must have […]
Show full quote

CoffeeOne wrote on 2024-01-04, 08:11:

LOL. Do we have first of April, because you are complaining that the system is too fast? […]
Show full quote

🤣.
Do we have first of April, because you are complaining that the system is too fast?

SI is a synthetic benchmark, and does not say too much.
I guess you are running the L1 in WB, so you have high values.

AMD Am5x86@160MHz gives more than 300, so your values are not too high, don't worry.
Use other benchmarks to get more real world values (quake, doom, wolf3d, ....)

If it's real, I'm not complaining 😉. LGR seemed to point out he couldn't get the CPU any faster, so I thought there must have been an issue with my benchmark.

I just ran a Quake benchmark and got 11.2 fps. LGR got 8.8 fps so the benchmark results must be real.

Very happy!

If you want to check what is possible with a 486:
There is a huge thread from pshipkov 3+3 retro .... and the worlds fastest 486 thread from Feipoa
My current values with Am 5x86 @ 160MHz
Re: 3 (+3 more) retro battle stations

I remember getting ~195 as well with my DX4-100 ODPR. Write-through and Write back L1.

Actually, I find Norton System Info to be a very good indicator of a particular CPU behaving correctly or not.

If I remember correctly, a DX2-66 should score exactly 144.0 in this benchmark if properly configured on a decent motherboard, which makes sense given the result for DX-33.
This would put the DX4-100 around 200 points. In any case, LGR's jumpers were either improperly configured or the motherboard is just bad.

I don't give too much on a single number.
Since a 486 has so many settings which will influence performance, I use that tools:
SpeedSys 4.78 --> the cache chart
ctcm from heise --> the values of the memory/cache access cases

Some random benchmark of AMD 5x86-133 @180. So 100Mhz should be ~216 points.

Also Intel 486DX4-100 has at least two versions. One with 8KB L1 cache and other wih 16KB L1 cache. So there should be some difference from get go.

EDIT:
Only now seen it's an Overdrive CPU. Not sure how much L1 have these.

Well, here is my result. Standard Intel DX4-100 (S-Spec SK051, write-through) on Soyo SY-4SAW2. It is identical to yours.
Therefore, I think we can conlude the "correct" score for this CPU is exactly 197.1.

Norton's SI will react to things like memory timing, so it's not like it should show exactly 197.1, but something close to that if settings are optimized. I get 197.0!
Also this is another case of LGR being not too... techy (to at least identify the issue).

LGR was using a ACER VL15g motherboard with a AMI BIOS. That BIOS is holding the board and processor back.

TheMobRules wrote on 2023-07-30, 01:25:

I noticed that the AMI BIOS for SiS471 boards suck ass (both on this board and MSI MS-4132G for example). For some reason they ignore the fastest timings for DRAM and cache set in the BIOS, and other settings are messed up (local bus T2/T3 and SYNC/TRANSPARENT are inverted). I wonder why AMI fucked up these in such a way and never bothered to fix them.

I've found that Norton's SI CPU benchmarks can vary wildly depending on motherboard and BIOS, and don't necessarily indicate true performance.

I have both a 486DX4-100 overdrive and a 486DX2-66 that both give a Norton SI CPU score of ~132. Yet the DX4-100 scores much faster in other benchmarks like PC Player, Doom, and Quake.

Something else to check with the DX4-100 overdrive is if L2 cache is enabled and functional. On my 486 DX4-100 overdrive setup using the stock BIOS results in no L2 cache. I had to switch to a MR BIOS in order for L2 cache to work.

Norton is CPU benchmark. While PCplayer, Doom, Quake are all around benchmarks with major impact on graphics performance and other subsets.

kixs wrote on 2024-01-04, 16:41:

Norton is CPU benchmark. While PCplayer, Doom, Quake are all around benchmarks with major impact on graphics performance and other subsets.

Norton's SI CPU benchmark doesn't seem to properly report the performance of the DX4-100 overdrive chip. I just tested a DX2-66 versus the DX4-100 overdrive in the same system, and both report identical Norton SI CPU scores of 132.

I'm going to compile some benchmarks and post those once I'm done.

I just did some benchmarks comparing a 486 DX2-66 versus the 486 DX4-100 overdrive in the same system. I get identical Norton SI CPU scores of 132 with both processors, but the DX4-100 is faster in every other benchmark.

System Specs

Motherboard: ECS UM486V (UMC UM82C480 chipset)
BIOS: MR BIOS v1.65 (default settings)
L2 Cache: 256KB
RAM: 16 MB
Video Card: Cirrus Logic GD5428 (1 MB)

486 DX2-66 Benchmarks

Norton SI CPU: 132.0
Speedsys processor score: 25.03
3D Bench: 43.4
PC Player 320x200: 9.4
Doom (high detail): 25.3 (2957 realticks)
Quake 320x200: 6.9

486 DX4-100 Benchmarks

Norton SI CPU: 132.0
Speedsys processor score: 39.40
3D Bench: 55.5
PC Player 320x200: 13.6
Doom (high detail): 31.4 (2377 realticks)
Quake 320x200: 9.2

Despite Norton's SI CPU score being the same in both cases, all other benchmarks show higher performance for the DX4-100 overdrive compared to the DX2-66. Performance differences range from 24% faster (Doom) up to 57% faster (Speedsys) for the DX4-100 overdrive.

I just did some further testing and experimenting with BIOS settings and the DX4-100 OD processor.

It seems that the Norton SI CPU score is specifically sensitive to the cache write timing setting. Changing it from 1 WS (default) to 0 WS results in Norton SI reporting a CPU benchmark of 198 instead of 132.

I did further benchmarks with results as follows:

Norton SI CPU: 198.0
Speedsys processor score: 42.37
3D Bench: 62.5
PC Player 320x200: 14.1
Doom (high detail): 33.1 (2258 realticks)
Quake 320x200: 9.5

Performance difference ranged from 3.3% faster (Quake) to 12.6% faster (3D Bench) based on this one BIOS setting.

zuldan wrote on 2024-01-04, 08:00:

Is there some bug that can distort these numbers? Or something I can turn off in the BIOS to get the real benchmark speed?

Just based on my own testing, changing the cache write timing from 1 WS to 0 WS resulted in changing the Norton SI CPU benchmark from 132 to 198.

It also resulted in a general performance improvement, but not nearly as dramatic as that Norton SI benchmark would indicate. 😆

kixs wrote on 2024-01-04, 13:44:

Only now seen it's an Overdrive CPU. Not sure how much L1 have these.

Mine has 16kB of L1 cache.

Your memory/cache performance seems to be poor.
At 33 MHz FSB you should not have a problem to get 2111 cache timings at all.
And your DRAM performance also seems to be weak. Try to remove waitstates.

You also may try ctcm.exe from heise: https://ftp.heise.de/ct/ctsi/ctcm17a.zip
On a 486 use ctcm /NOP