Tertz wrote:

Would be interesting also to see the comparision of 286 on same clock with 386SX. We had a discussion previously about is 286 faster than 386SX.

The 286 should win that. As said, putting the 386 on a 16-bit bus results in inefficient bus usage.
As we've seen here, even a true 386DX *with* cache can barely outperform a 286. 386SX systems don't have cache, and no 32-bit interface to memory. They will be marginally slower than a 286 at the same speed.

Scali wrote:

Tertz wrote:

Would be interesting also to see the comparision of 286 on same clock with 386SX. We had a discussion previously about is 286 faster than 386SX.

The 286 should win that. As said, putting the 386 on a 16-bit bus results in inefficient bus usage.
As we've seen here, even a true 386DX *with* cache can barely outperform a 286. 386SX systems don't have cache, and no 32-bit interface to memory. They will be marginally slower than a 286 at the same speed.

When used with "equal" chipsets that is, many 286 chipsets are dog slow but some later ones seem to compete with the 386SX chipsets just fine.

Some musing, a card that can do 5250k has the bandwidth to do VGA at full speed, that's the bandwidth needed for 320x240@70fps, and 320x200@70 is 4375k, so from your vspeed results, on the 286 the ncr and cirrus card are capable of full speed for mode x, and 386 only the cirrus and only for VGA, though neither of these full & bottlenecked speeds are realized since neither CPU is strong enough to run games that fast as shown by 3dbench and wolf3d results. Results also show that the ncr's poor 386 handling drops game performance even though with its throughput, hobbled as it is, should still be able to match the cirrus performance, wolf3d@19.5fps only needs 1218k.

If Software is coded for 386 the 286 lose.

To get a real result whe need a Software that has a 16 and a 32bit Version for Benchmark.

A 386 with a 16bit ISA Bus is also not optimal.
I can imagine that a 386 with a 32bit Bus Interface will run much quicker.
Does someone have a 386 EISA Board ?

Skyscraper wrote:

When used with "equal" chipsets that is, many 286 chipsets are dog slow but some later ones seem to compete with the 386SX chipsets just fine.

True. I have 2 286 motherboards, both have totally different performance figures, different chipsets. 286 systems based on Headland and VLSI seem to really fly. But i really doubt that back in the day OEM 286 systems would have this kind of performance.

386 EISA systems are pretty rare, but the most common ones are probably those made by AST. Unfortunately, they use the CUPID bus, which accepts the weird 64-pin CUPID memory. I believe Compaq sold a number of EISA systems as well. I have never seen a 3rd party EISA 386 motherboard in person, but they do exist.

Chipsets have quite an impact on the system performance and it was true for 286 and it's true today.

Problem is finding which ones are fast for each platform. It's much more difficult with old chipsets as not much old hardware exists - at least I only have 3 working 286 motherboards.

someone has 386sx?
try to test and put the results here?

Scali wrote:

Tertz wrote:

Would be interesting also to see the comparision of 286 on same clock with 386SX. We had a discussion previously about is 286 faster than 386SX.

The 286 should win that.

I suppose same. But it would be good to have concrete experimental data in different tests and applications, - how much is the difference. It's expected near 15% slower than 286.

!
I suggest to Ekb reformat this theme or create a separate theme as "XT-386 benchmark theme" in the way similar to Phil's benchmark theme. All tests in this theme should work on XT up to 486DX 33 MHz (ideally up to modern PCs or at least late DOS machines like Pentium 3). Benchmark's compatibility may be checked on real hardware and PCem. Part of tests will be obligate, while others - matter of personal choice. The charts table may to be on google with open access, as Phil did.
It's good to make all tests wich work from XT to Pentium 3 as obligate, an example of wich is Speedtst.

to Tertz

Good idea! but I bad know English. 🙁

Unfortunately, these game not work on XTs:
1) Wolf3d - optimized opcodes for 286
2) Formula One - 286+

While it's obvious that XT with 300-700 dhrystones is extremely slow. See on youtube: Cracked Wolf-3D on XT. It is very small window!

Statistics benchmarks XT-286-386 in GoogleDOC: https://docs.google.com/spreadsheets/d/1tPId4 … dit?usp=sharing
It can be based on it to expand and make the most convenient?

I'd keep it 286 only.

kixs wrote:

I'd keep it 286 only.

I'd extend to 386 as well, as Phil's VGA benchmark runs very poorly on those machines and it's mainly for 486 and faster.

Ekb wrote:

I have bad English.

It's ok for this purpose as you've made this theme. Google translator may be used too.

these game not work on XTs:
1) Wolf3d - optimized opcodes for 286
2) Formula One - 286+

If a test runs and gives more than 0, it may be used. If the test works only on higher CPU or testing time is too long (>15 min), there are 2 ways:
1) Obligate part may to have CPU specific lists of tests. Those who have XT - have to do one tests. Those who have 286 - XT obligate part + some tests wich need 286+. In similar way for 386. Tests wich need more than 1 mb should not be used as obligate for XT-386.
2) To place such tests in non-obligate part with subsections for 286+, 386+. Those who want - will do them.

Cracked Wolf-3D on XT

Normal XT will not run it as it needs "V20/V30/80186". I'd relate it to 286+.

It can be based on it to expand and make the most convenient?

Something like in Phil's theme. Your spreadsheet has vertical subsections, instead it's better to add CPU type, - who needs grouping by CPU will make sorting by that field. Less colors, no bright colors. Tester name is redundant. Results of same test are better to place close, not in distant subsections (fields: test1-APU, test1-FPU, test2-APU, test2-FPU not like test1-APU, test2-APU, test1-FPU, test2-FPU).
Possible fields: CPU class (386SX), CPU freq, cpu model (Intel 386SX), fpu, video (chip name+model+mem ISA), chipset+MB model, cache, mem size and speed, tweaks/settings: BIOS ..., PC manufacture-model (for brands), additonal comments.

Ekb, I have a 386SX and a 50MHz Quartz to complete youe research! What videocard would you expect me to use? I've got Trident 8900 and WD 90C11-LR.

Привет из Омска!

to RacoonRider
hello! 😀

I have a Trident 8900С (1mb), I would be if that could compare with your results 😀
But it is also interesting even your WD 90C11-LR. I do not know how high-speed card?
If no trouble, then check both? first Trident, after WD.
The archive is the program CAPTURE - to save the screen (not only works on NSSI60).
Thank 😉

Russian:
У меня есть 8900С Я если что, мог бы сравнить результаты с Вашими:)
Также интересно вашего WD. Я беспонятия, что это за карта и какая производительность?
Если не затруднит, то попробуйте оба? Сначало Trident, затем WD.
В архиве есть программка CAPTURE - чтобы сграбить экран (не работает только на NSSI60).
Потом можно в архивчик и я разложу здесь по красивее, чтобы было нагляднее.
Возможно, что даже сделаю тоже на своем Trident

Here's the WD. I have doubts concerning speed 😀

Scali wrote:

The 286 should win that.

There are interesting stats. Seems early 286 and 386SX have close performance per clock. Later models (286 16 MHz from ~1990, 386SX 20-25 MHz from 1992-summer) are 12% faster and are close too.

New test:

3DSpace:

286 = 15%.
386 = 19%. Best.

I redid new archive, added more new programs. See first post.

Besides pictures, to make the reading easier you need a table with all results gouped in it.
I used Speed Test in my benchmark theme. I'd run it in long mode, not medium to get more accuracy.

Performance comparison of CPU: SuperFast-286 and AMD-386SX

System 1: 286-25mhz

Clocker in 25mhz (100/4=25mhz)
Mobo = Noname with five chips VLSI
CPU = Harris 25mhz.
FPU = Intel 287XL
RAM = only DIP20 Siemens 60ns(super faster) + Aluminium cooler
Cache = No
VGA = Trident 8900C, 1mb
Sound = ESS1868
Multi I/O = Chips "TANS" (china?) for fast speed ISA.
HDD = CompactFlash 1GB, with Ontrack

For more information see here: Kixs's 286 to the Max

286 BIOS default:

System 2: 386SX-25mhz

Clocker in 25mhz (50/2=25mhz)
Mobo = 303n1
CPU = AMD 386SX-40
FPU = IIT 3c87SX-33
RAM = 2x4mb=8mb, 60ns SIMM30
Cache = No
VGA = Trident 8900C, 1mb
Sound = No
Multi I/O = Chip GoldStar Prime 2 (not "2C")
HDD = Quantum Fireball AT 6.4, with Ontrack

386 BIOS is Auto.

WARNING: Left only 286. Right only 386SX.

3D Benchmark:

3Dbench 1.0:

286 = 9.7 fps. Best.
386 = 6.4 fps.

3Dbench 2.0:

286 = 9.6 fps. Best.
386 = Error calculating. Value < 7.0 = overflow.

Formula One Benchmark 1:

286 = 108% load. Best.
386 = 172% load.

Formula One Benchmark 2:

286 = 316% load. Best.
386 = 525% load.

Wolf 3D. PC Speaker (NOT SoundBlaster) + Mouse + FullScreen

286 = 14.7 fps.
386 = - fps. (wait..)

3DSpace:

286 = 14%.
386 = -. (wait..)

Morph3D:

286 = 77.7 fps. Best.
386 = 59.1 fps.

.
.
.
.
CPU Benchmark:

Landmark 2.0:

Landmark 6.0:

Checkit 3.0

Norton SysInfo 6.0:

286 = 19.4 XTs.
386 = - XTs.

PC-Tools SysInfo 9.0:

286 = 15.1 XTs. Best.
386 = 10.0 XTs.

NSSI 60

DoctorHard 3.7:

PC Doctor 1.7:

286 = 6.9 MIPS and Read 16bit = 11mb/s.
386 = (wait..)

PC_info 4.04:

286 = 18.1 XTs.
386 = (wait..)

SpeedTest 1.14:

286 = 47740 scores, 17.7 XTs.
386 = (wait..)

FPU Benchmark:

Fbench:

287XL =
IIT387 = Best

CABT:

287XL = 6.26 sec.
IIT387 = 3.24 sec. Best

FLOPS:

287XL = 0.04 MFLOPS.
IIT387 = 0.11 MFLOPS. Best

NSSI 60:

287XL =
IIT387 = Best

RAM Benchmark:

CCT:

286: TIME = 898. Calculating: 1 / 898 * 10000 = 11 mb/s RAM
386: (wait..)

SST (sorry only 386+)

286: Memory________Failed
386: Memory________R=10(?) MB/s______W=18.5 MB/s_____Move=10 MB/s

SysTest:

286 = 12mb/s, 0WS.
386 = 4,3mb/s, 1WS

CompTest 2.6:

286 = 12mb/s, 0WS.
386 = (wait..)

agSI 1.2.3:

286 = 12mb/s, ?WS.
386 = (wait..)

HDD <-> ISA Benchmark:

4_Speed:

286 = 2873 kb/s.
386 = (wait..)

CheckIt 3.0:

286 = 2774 kb/s. Best.
386 = 2090 kb/s.

DiskMeter:

286 = 2979 kb/s.
386 = (wait..)

CoreTest:

286 = 3018 kb/s.
386 = (wait..)

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

MIPS Benchmark:

MIPS:

286 = 3.57 MIPS.
386 = (wait..)

PMIPS (comparing between each other)

286 = 4.0 MIPS.
386 = (wait..)

Speed-XT:

286 = 18 sec. 15.86 XTs. Best.
386 = (wait..)

QuickTECH PRO 1.28:

286 = 16.51 XTs.
386 = (wait..)

Index:

286 =
386 = (wait..)

DMA Benchmark:

DMAspeed:

286 = 1559kb/s.
386 = (wait..)

T8237DMA:

286 = 5.67mhz.
386 = (wait..)

OTHER Benchmark:

TopBench:

286 = Score 64. Best.
386 = Score 44