.
.
CPU Benchmark:

Landmark 2.0:

Landmark 6.0:

Checkit 3.0

Norton SysInfo 6.0:

286 = 19.4 XTs.
386 = 27.1 XTs. Best.

PC-Tools SysInfo 9.0:

286 = 15.1 XTs. Best.
386 = 13.3 XTs.

NSSI 60

DoctorHard 3.7:

PC Doctor 1.7:

286 = 6.9 MIPS and Read 16bit = 11mb/s. Best
386 = 6.3 MIPS and Read 16bit = 8mb/s

PC_info 4.04:


286 = 18.1 XTs.
386 = 21.3 XTs. Best.

K_Speed4:

286 = 101%.
386 = 105%. Best.

Power Meter:

286 = 6.7k dhrystones.
386 = 7.7k dhrystones. Best.

FPU Benchmark:

FBench:

287XL =
IIT387 = Best

CABT:

287XL = 6.26 sec.
IIT387 = 2.58 sec. Best

FLOPS:

287XL = 0.04 MFLOPS.
IIT387 = 0.14 MFLOPS. Best

NSSI 60:

287XL =
IIT387 = Best

RAM Benchmark:

CCT:

286: TIME = 898. Calculating: 1 / 898 * 10000 = 11 mb/s RAM
386: TIME = 1531. Calculating: 1 / 1531 * 10000 = 6,5 mb/s RAM ( x*2 for 32bit ?)
386: TIME = 1123. Calculating: 1 / 1123 * 10000 = 9 mb/s Cache ( x*2 for 32bit ?)

SST (sorry only 386+)

286: Memory________Failed
386: Memory________R=7.6 MB/s______W=18.5 MB/s_____Move=9 MB/s
386: Cache_________R=16 MB/s_______W=18.5 MB/s_____Move=24 MB/s

SysTest:

286 = 12mb/s, 0WS.
386 = 9mb/s, 1WS

CompTest 2.6:

286 = 12mb/s, 0WS.
386 = 9mb/s, 3WS ....... Cache=24mb/s

agSI 1.2.3:

286 = 12mb/s, ?WS.
386 = 9mb/s, 4WS........Cache = 24mb/s

HDD <-> ISA Benchmark:

4_Speed:

286 = 2873 kb/s. Best.
386 = 1797 kb/s.

CheckIt 3.0:

286 = 2774 kb/s. Best.
386 = 1681 kb/s.

DiskMeter:

286 = 2979 kb/s. Best.
386 = 1807 kb/s.

CoreTest:

286 = 3018 kb/s. Best.
386 = 1845 kb/s.

VGA <-> ISA Benchmark:. All test are run only in VGA mode 320x200x256color

VGAFPS:

286 = 8236kb/s, 129fps. Best.
386 = 4770kb/s, 75fps.

VidSpeed 3.2:

286 = 8192kb/s. Best.
386 = 4886kb/s.

VMAX256 1.11:

286 = 125.6 fps. Best.
386 = 77.5 fps.

VideoSpeed 0.42:

286 = 8324kb/s. Best.
386 = 4884kb/s.

MIPS Benchmark:

MIPS:

286 = 3.57 MIPS. Best.
386 = 3.53 MIPS.

PMIPS (comparing between each other)

286 = 4.0 MIPS.
386 = 4.5 MIPS. Best.

Speed-XT:

286 = 18 sec. 15.86 XTs. Best.
386 = 19 sec. 14.94 XTs.

QuickTECH PRO 1.28:

286 = 16.51 XTs. Best.
386 = 15.55 XTs.

Index:

286 =
386 = Best.

DMA Benchmark:

DMAspeed:

286 = 1559kb/s. Best.
386 = 1042kb/s

T8237DMA:

286 = 5.67mhz. Best.
386 = 3.89mhz

OTHER Benchmark:

TopBench:

286 = Score 73. Best.
386 = Score 68

This made a LOT more sense when I realized the 386 was clocked at 25mhz.

You know, I read that spiel over at Red Hill about them always suspecting a good 286 was faster than the 386SXs. And here you came, showing them that they could hold their own against the full blown DX-class machines provided all was equal other than the chip and motherboard.

And I never would have believed it if I hadn't been shown the benchmarks all the way down. As I was reading, I couldn't believe my eyes.

You, sir, are a legend.

Very interesting results. Obviously the lower VGA throughput confirmed by VGAFPS/VidSpeed/VMAX256/VideoSpeed will hold the 386 back, otherwise it would be faster in the graphical benchmarks. (I'm surprised the 286 can push 8MB/sec to the card, that's about the maximum ISA can do.) Unfortunately I don't know how good the UMC chipset is.
Since you mention you underclocked the 386: Could you check the ISA clock? Maybe there is a BIOS setting for the divider? Alternatively you could check the VGA throughput with the original 40MHz. If VGA throughput gets better, than it really might be the ISA clock. A 386DX25 should be able to saturate the ISA bus.

idspispopd wrote:

Very interesting results. Obviously the lower VGA throughput confirmed by VGAFPS/VidSpeed/VMAX256/VideoSpeed will hold the 386 back, otherwise it would be faster in the graphical benchmarks. (I'm surprised the 286 can push 8MB/sec to the card, that's about the maximum ISA can do.) Unfortunately I don't know how good the UMC chipset is.
Since you mention you underclocked the 386: Could you check the ISA clock? Maybe there is a BIOS setting for the divider? Alternatively you could check the VGA throughput with the original 40MHz. If VGA throughput gets better, than it really might be the ISA clock. A 386DX25 should be able to saturate the ISA bus.

Actually not, bandwidth needed for the performance they're yielding, like 19.5 fps in wolf3d is lower than the throughput that can be achieved, even on the slower ncr on the slower 386. 320x200x19.5 is only 1218.5k.

awgamer wrote:

Actually not, bandwidth needed for the performance they're yielding, like 19.5 fps in wolf3d is lower than the throughput that can be achieved, even on the slower ncr on the slower 386. 320x200x19.5 is only 1218.5k.

You are correct that the available bandwidth is enough for 19.5 fps assuming a fast enough CPU. But with lower available bandwidth the CPU will stay busy longer transferring the image to the VGA. With (not the exact numbers) 8192MB/s the CPU will be busy about 15% of the time for this example. With 4096MB/s the CPU will be busy transferring for about 30% of the time which is quite significant.
(The pure VGA benchmarks I referred to use 100% of the time for transferring.)
Unfortunately transferring the data to the VGA is not being done using DMA, otherwise you would be right.

idspispopd wrote:

Since you mention you underclocked the 386: Could you check the ISA clock? Maybe there is a BIOS setting for the divider?

if auto is disable, then you can change the parameter "AT clock select" on:
1) "CPUCLK/2" for high througput ISA. In this overclocked ISA->VGA from 4888kb/s to 5450kb/s. ISA limit 5450kb/s.
2) fixed "7.2mhz" - system halted on POST

As you can see that it will not give superiority over 286 with 8200kb/s for VGA 320x200x256. ISA limit >10200kb/s

OK, CPUCLK/3 would be correct for 25MHz, giving 8.33MHz. CPUCLK/2 is indeed an overclock.
I'm not sure what "IO Recovery Time" exactly refers to. If it only refers to I/O port operations it doesn't matter, if it also refers to memory mapped I/O it would.
Do you know at which speed the 286 board is running the ISA bus?
Anyway, what I'm saying is if the 386 could manage the same VGA throughput as the 286 it would score higher in the graphical benchmarks. Perhaps the UMC chipset just doesn't have good ISA performance, the HDD benchmarks also say that the 286 has better throughput. I searched for UMC82C491F, it seems this is supposed to be a 486 VLB chipset, which is quite confusing. Maybe it doesn't work as fast with a 386?
In my experience it is definitely possible to have a throughput of > 8MB/s with a 386, more if overclocking the ISA bus. (386DX40, OPTi chipset, WD90C31 VGA which should be a good as your CL.)
I think you have a very well tuned 286 system while the 386 system is not quite as good, for whatever reason.

Two identical 386 baby-mobo, but with different chipsets: UMC and BIOteq.
I checked and both run almost identical to 40MHz (3Dbench 16,8..16,9 fps).
Performance is the same on VGA card. Hence the same for ISA-bus.
But Bioteq no jumpers to 25MHz. A minimum of only 33MHz. 🙁

You can test on your 386 (downclock to 25mhz) ?

idspispopd wrote:

Do you know at which speed the 286 board is running the ISA bus?

look carefully as the spread of the table:

It is seen that the 386 limit on 4888kb/s, the maximum values for all video modes.
While for 286 maximal 10200kb/s for CGA mode. And this is not the limit.
You can still pick up overclock videocard CS5420 (new bios with program "mclock").
Therefore, ISA able to pass over 10200 kb/s for CGA-mode.

I still suspect that the ISA clock on the 286 is higher than 8.3MHz, but I don't have a good idea how to check this, except using an oscilloscope.
I don't have the 386 board anymore, IIRC I gave it to kixs (and even before it wasn't in a working system for a long time which is why I gave it away). I don't even know if the board could be underclocked. (IIRC it was a Shuttle HOT-307 which can only be switched from 40MHz to 33MHz).

I also had a look at your TOPBENCH screenshots and compared some values to the TOPBENCH database. Your 286 had a VidMem result of 131 µsec. No 286 or 386 reaches this value in the database.

I don't know what's going on here. Please understand that I don't accuse you of cheating (after all, why should you?), I'm just trying to figure out the reason behind this numbers.

IIRC I gave it to kixs

Still have it 😉 Back then I didn't have much old hardware and still hope to make my own round of HW benchmarks.

As for ISA speed. It's obvious the ISA speed on a 286 is higher.

@Ekb:
Why did you leave the AUTO bios settings? 386 would gain around 10% if set to fastest.

idspispopd wrote:

I don't know what's going on here. Please understand that I don't accuse you of cheating (after all, why should you?), I'm just trying to figure out the reason behind this numbers.

See pic (click to large). This is original mainboard with oscilator at 48mhz. 48/4=12mhz AMD CPU.
i'm replaced this oscilator on new oscilator at 100mhz. 100/4=25mhz Harris CPU.

apparently ISA bus was NOT tied to secondary quartz at 14.31818mhz ? (14.31818 / 2 = 7.15mhz ISA) 🙁
It was attached to main oscilator for processor (through a divider: 25mhz CPU / 2 = 12mhz ISA ?)..

kixs wrote:

Why did you leave the AUTO bios settings? 386 would gain around 10% if set to fastest.

I would compare 286 with classics 386DX's 😀

idspispopd wrote:

I think you have a very well tuned 286 system while the 386 system is not quite as good, for whatever reason.

I agree to this. The 386 seems to run on rather unoptimized settings.
Isn't it possible to reduce the Cache and RAM waitstates?
Is it possible to decrease the divider for ISA clock further to get closer to the 286 ISA bus clock?

I would not have to read bench results if I know that the competitors are a very well tuned 286 vs. a default 386 at the same clock to know that the 286 will end up faster.

kixs wrote:

IIRC I gave it to kixs

Still have it 😉 Back then I didn't have much old hardware and still hope to make my own round of HW benchmarks.

As for ISA speed. It's obvious the ISA speed on a 286 is higher.

Nice to hear!
About the ISA speed: I don't see why that's obvious. (Of course it's obvious in this comparison judging from the numbers, but why is it obvious in general? I'd say maybe a 286 system where the CPU and the ISA bus are running in sync might have an advantage, but asides of that?)

Ekb wrote:

i'm replaced this oscilator on new oscilator at 100mhz. 100/4=25mhz Harris CPU.

apparently ISA bus was NOT tied to secondary quartz at 14mhz ? 🙁
It was attached to main oscilator for processor (through a divider: 25mhz CPU / 2 = 12mhz ISA ?)..

12.5MHz for the ISA bus, that would explain something. So to make things fair you'd have to use "CPUCLK/2" for the 386.

I don't see why that's obvious.

Maybe not so obvious. But I like to see HDD transfer speed and then deduct the ISA speed from it.

Yesterday I tested one 386DX-40 with UMC chipset (like the one in this thread). It can do ISA/2 at 40MHz - HDD transfer rate is 3500KB/s (Checkit) and VGA speed 8800KB/s (Speedsys). It also gets 18.1 in 3DBench, 6.0 in PCPBench. These are the best scores I ever seen on a 386DX-40.

idspispopd wrote:

12.5MHz for the ISA bus, that would explain something. So to make things fair you'd have to use "CPUCLK/2" for the 386.

It is my guess, I do not insist on the truth.
especially as I wrote above about what 386-25mhz with ISA=CPUCLK/2 can not catch up to 286-25mhz.

Ekb wrote:

if auto is disable, then you can change the parameter "AT clock select" on: 1) "CPUCLK/2" for high througput ISA. In this overcl […]
Show full quote

if auto is disable, then you can change the parameter "AT clock select" on:
1) "CPUCLK/2" for high througput ISA. In this overclocked ISA->VGA from 4888kb/s to 5450kb/s. ISA limit 5450kb/s.
2) fixed "7.2mhz" - system halted on POST

As you can see that it will not give superiority over 286 with 8200kb/s for VGA 320x200x256. ISA limit >10200kb/s

***

kixs wrote:

Yesterday I tested one 386DX-40 with UMC chipset (like the one in this thread). It can do ISA/2 at 40MHz - HDD transfer rate is 3500KB/s (Checkit) and VGA speed 8800KB/s (Speedsys). It also gets 18.1 in 3DBench, 6.0 in PCPBench. These are the best scores I ever seen on a 386DX-40.

I'm agree you!
It is one of the favorite motherboards 386DX-40
some used a videocard?

I think it is interesting:

DMA Benchmark:

DMAspeed:

286 = 1559kb/s. Best.
386 = 1042kb/s

T8237DMA:

286 = 5.67mhz. Best.
386 = 3.89mhz