I have at last one 386 motherboard which has a small amount of cache built into the chipset for use when no L2 cache is installed. I forgot the a mount of cache it provided, but 1 KB or less.

konc wrote:

Anonymous Coward wrote:

Konc, are you certain your system has no cache? Your speedsys graph almost looks as if your chipset has 16kb cache.

Positive, I was referring to L2 cache on the m/b. 16KBs of on-chip L1 are obviously there and correctly detected.
In fact I was puzzled by SpeedSys's incorrect detection of this SX. I did open the case and visually confirmed it's an SX@20MHz and, just to be 100% sure, also ran a couple of other "system information" utilities that did identify it as such.

What 386SX has 16kb on-chip cache? 😕 In your case I though the cache is on the motherboard. A few 386SX boards have that implemented.

Anonymous Coward wrote:

Both sxl 40s and sxl2 50s default to 1x clock mode, and both can enable clock doubling by toggling the direct mapped/2 way associated cache register in the Cyrix dlc software util. Both chips are essentially the same exept for the silkscreen label.

Just so that my understanding is correct.
Regardless of which of the above CPU's you have, if you default it to 1x clock, it will run at 40 MHz? If you enable the clock doubling, it will run at 50 MHz (25 x2)?.
How does the motherboard crystal play a role? For example, if your motherboard only supports 33 & 40 MHz 386 DX CPU's, will the clock doubling work?

kixs wrote:

konc wrote:

Anonymous Coward wrote:

Konc, are you certain your system has no cache? Your speedsys graph almost looks as if your chipset has 16kb cache.

Positive, I was referring to L2 cache on the m/b. 16KBs of on-chip L1 are obviously there and correctly detected.
In fact I was puzzled by SpeedSys's incorrect detection of this SX. I did open the case and visually confirmed it's an SX@20MHz and, just to be 100% sure, also ran a couple of other "system information" utilities that did identify it as such.

What 386SX has 16kb on-chip cache? 😕 In your case I though the cache is on the motherboard. A few 386SX boards have that implemented.

That was my question initially, I only knew of some IBM chips that had on chip cache, but it's detected as L1 everywhere (and no L2). I could post some CacheCheck & CTCM screenshots if they are of any help.

Let's learn something new today: You're saying that some motherboards implement a "L1" cache on their own that gets detected as such? That would explain its speed, seems too slow to be on the chip. I didn't know they could do that , I always thought m/b cache = L2. I'll open the case anyway soon to replace the VGA so I'll try to properly identify the CPU.

About L1 and L2. It doesn't automatically mean the L1 is on-chip. It usually means it's the first one. And before 486 there were only on the motherboard. As you see with 386DX tests - L1 is on-board cache. I'd say it's the same with your SX board - unless you have some 486slc chip but then the L1 speed should be like double of main memory (by my tests).

i put in cyrix fastmath fpu.

result is here.

top is cyrix 486drx2 66Mhz, bottom is ti 486dlc 40mhz

cyrix 486drx2 speedsys youtube : https://youtu.be/A8jqm6-wp-M
ti 486dlc 40MHz speedsys youtube : https://youtu.be/W4AHeZaj_g4

and i run doom timedemo demo3 benchmarks.

result is here.

AMD 386DX 40MHz + CL-GD5428 VLB 2MB 9,124 (8.19 fps) youtube : https://youtu.be/EAPnbZjOOxQ
AMD 386DX 40MHz + CL-GD5434 VLB 1MB 8,986 (8.31 fps) youtube : https://youtu.be/iFKv0DFVooA
AMD 386DX 40MHz + CL-GD5434 ISA 1MB 10,154 (7.35 fps) youtube : https://youtu.be/8-rJcTLY8GE

TI 486DLC 40MHz + CL-GD5428 VLB 2MB 5,838 (12.79 fps) youtube : https://youtu.be/EzJfuCAVfdQ
TI 486DLC 40MHz + CL-GD5434 VLB 1MB 5,813 (12.84 fps) youtube : https://youtu.be/NvVS9KtNr10
TI 486DLC 40MHz + CL-GD5434 ISA 1MB 6,621 (11.28 fps) youtube : https://youtu.be/O1m-Q7XhSVY

Cyrix 486DRx2 66MHz (without 486drx utility) + CL-GD5428 VLB 2MB 4,950 (15.08 fps) youtube : https://youtu.be/dW1OpcHMdEA
Cyrix 486DRx2 66MHz (without 486drx utility) + CL-GD5434 VLB 1MB 4,767 (15.66 fps) youtube : https://youtu.be/wDdfEkWKYz0
Cyrix 486DRx2 66MHz (without 486drx utility) + CL-GD5434 ISA 1MB 6,212 (12.02 fps) youtube : https://youtu.be/S_EbIwLshFc

Cyrix 486DRx2 66MHz (with 486drx utility) + CL-GD5428 VLB 2MB 4,705 (15.87 fps) youtube : https://youtu.be/kTkd8FZjOUY
Cyrix 486DRx2 66MHz (with 486drx utility) + CL-GD5434 VLB 1MB 4,556 (16.39 fps) youtube : https://youtu.be/IPp316s2TCc
Cyrix 486DRx2 66MHz (with 486drx utility) + CL-GD5434 ISA 1MB 5,938 (12.57 fps) youtube : https://youtu.be/vmZN3jwlbus

CL-GD5434 VLB 1MB is Orchid Kelvin64 VLB
CL-GD5434 ISA 1MB is Diamond Speedstar 64 ISA

CL-GD5434 have 64-bit memory bus. but it requires 2mb memory. if have 1mb memory, CL-GD5434 running only 32-bit memory bus mode.
my cards have 1mb memory.

kixs wrote:

About L1 and L2. It doesn't automatically mean the L1 is on-chip. It usually means it's the first one. And before 486 there were only on the motherboard. As you see with 386DX tests - L1 is on-board cache. I'd say it's the same with your SX board - unless you have some 486slc chip but then the L1 speed should be like double of main memory (by my tests).

Ah, thanks for the explanation. I always thought L1=on chip, L2=m/b. Now it totally makes sense, as well as the measured speed.

i put in ULSI Math-Co DX-2 66MHz US83D87-C 387 FPU.

and running speedsys on Cyrix 486DRx2 66MHz 386 CPU.

youtube : https://youtu.be/paSpmw30itY

CPU : Cyrix 486DRx2 66MHz
FPU : ULSI Math-Co DX-2 66MHz (US83D87-C)
RAM : 32MB FPM 60ns (4MB x4)
M/B : Aquaris MB-4D33/50 3/486 VLB Motherboard (OPTI 495SX)
I/O : IWill SIDE Jr. Pro VLB
HDD : IBM Travelstar DARA-212000 12GB E-IDE
VGA : Orchid Kelvin64 VLB CirrusLogic CL-GD5434 1MB 70ns

Nice 😀

Can you also run 3dbench, Pcpbench, Doom and Quake 😉

Quake needs at least a pentium.

RoyBatty wrote:

Quake needs at least a pentium.

It runs on a 386 with FPU - even 386sx. Of course it's not playable, but as a benchmark it's OK 😉

Interesting. Where did you find the Math-Co DX2? It's a pretty rare part.

Are you as equally unimpressed as I am with the performance compared to a non clock doubled FPU? It almost makes me wonder if there is a software utility to enable the doubling, or if doubling on an FPU is just worthless.

Are you as equally unimpressed as I am with the performance compared to a non clock doubled FPU? It almost makes me wonder if there is a software utility to enable the doubling, or if doubling on an FPU is just worthless.

I think you should see the screenshot again 😉 It's more than double of normal 387 FPU. It's in the range of 486DX fpu performance.

EDIT:
My mistake. Didn't see his previous posts. Yeah, it doesn't seem that fast for 66MHz operation.

Went ahead and watched his Youtube videos. By NSSI benchmark even DRx-66 doesn't seem any better then DLC-40. But with DOOMS there is big difference ~ 4800 vs 5800 realtics.

Would be interested in Autocad and Quake benchmarks between Cyrix Fasmath-40 and ULSI-66.

kixs wrote:

Went ahead and watched his Youtube videos. By NSSI benchmark even DRx-66 doesn't seem any better then DLC-40. But with DOOMS there is big difference ~ 4800 vs 5800 realtics.

Would be interested in Autocad and Quake benchmarks between Cyrix Fasmath-40 and ULSI-66.

i will download pcpbench later because i shoot film yesterday.

i bought ULSI Math-CO DX2 in germany seller.

I tested quake timedemo demo1.

result is here.

Phil's benchmark

1. Landmark System Speed Test Ver 2.0
2. Landmark System Speed Test Ver 6.0
3. FBench DOS
4. FLops Ver 1.31
5. Circuit Analysis Benchmark Test
6. SuperScape VGA Benchmark v1.0 (3DBench)
7. SuperScape VGA Benchmark v1.0c (3DBench2)
8. NSSI CPU
9. NSSI FPU

TI 486DLC 40MHz + IIT 4C87DLC + CL-GD5434 1MB VLB - youtube : https://youtu.be/BHGDxjyDTJY
TI 486DLC 40MHz + Cyrix FasMath + CL-GD5434 1MB VLB - youtube : https://youtu.be/suab7uyiAwQ
TI 486DLC 40MHz + ULSI Math-co DX2 + CL-GD5434 1MB VLB - youtube : https://youtu.be/XQUPdniVAoc
Cyrix 486DRx2 66MHz + IIT 4C87DLC + CL-GD5434 1MB VLB - youtube : https://youtu.be/w3e7YNMnSBI
Cyrix 486DRx2 66MHz + Cyrix FasMath + CL-GD5434 1MB VLB - youtube : https://youtu.be/cJUyCRQfFMY
Cyrix 486DRx2 66MHz + ULSI Math-co DX2 + CL-GD5434 1MB VLB - youtube : https://youtu.be/3ReJKaJW0l8

Quake timedemo demo1 benchmark

Command is quake -nosound -noipx -nojoy -nocdaudio +stopdemo +timedemo demo1

TI 486DLC 40MHz + IIT 4C87DLC + CL-GD5434 1MB VLB - youtube : https://youtu.be/C3gCFzQJClo
TI 486DLC 40MHz + Cyrix FasMath + CL-GD5434 1MB VLB - youtube : https://youtu.be/tg_39kGQwtI
TI 486DLC 40MHz + ULSI Math-co DX2 + CL-GD5434 1MB VLB - youtube : https://youtu.be/ts7gVhcZAcM
Cyrix 486DRx2 66MHz + IIT 4C87DLC + CL-GD5434 1MB VLB - youtube : https://youtu.be/wrcvoiV1540
Cyrix 486DRx2 66MHz + Cyrix FasMath + CL-GD5434 1MB VLB - youtube : https://youtu.be/uIzGzapeEkY
Cyrix 486DRx2 66MHz + ULSI Math-co DX2 + CL-GD5434 1MB VLB - youtube : https://youtu.be/3rg_Vt74lgA

during 486DLC 40MHz's quake benchmark, i set Cache timing to 3-1-1-1 because My W24257AK-10 cache sram is can't allow 2-1-1-1 or 2-2-2-2 timing on 40MHz FSB.
other 486DLC 40MHz's benchmark set cache timing is 2-1-1-1.

ULSI Math-co DX2 is 80MHz operate with 486DLC 40MHz because use 40MHz system bus.

Continue.

skv400 wrote:

i bought ULSI Math-CO DX2 in germany seller.

He is very expensive... I put a couple of offers in for that same chip and he declined them all.

He has another at around 120 euros I think... I hope there is a decent performance boost for the price (not that I would pay of course).

Edit; I just saw your FPU results... there is a little bit of a boost after all, not a significant amount... say 10% to 20%. It would be intersting to experience if in practice there is a difference also.

Thanks for posting those btw!

386_junkie wrote:

He is very expensive... I put a couple of offers in for that same chip and he declined them all. […]
Show full quote

skv400 wrote:

i bought ULSI Math-CO DX2 in germany seller.

He is very expensive... I put a couple of offers in for that same chip and he declined them all.

He has another at around 120 euros I think... I hope there is a decent performance boost for the price (not that I would pay of course).

Edit; I just saw your FPU results... there is a little bit of a boost after all, not a significant amount... say 10% to 20%. It would be intersting to experience if in practice there is a difference also.

Thanks for posting those btw!

i boutght 100 euro without shipping price.

this fpu's performance is not good. 486drx2 66mhz and ulsi 387dx2 66mhz is waste of money.

this item's only one advantage is highest clock of 386 cpu and fpus.. it is all

overall performance is can't reach 486dx-33.

486dlc-40 + cyrix fasmath 40gp is so cheap. but reach 75~85% performace of this 66mhz cpu&fpus.

ah.. one advantage is added. 33MHz system bus useful for 33mhz vlb cards. (ex:s3 864/868, e-ide cards)

don't use iit fpus. it is sucks.

i think iit's 487DLX2 50MHz fpu is lower than Cyrix fasmath 40MHz. because see iit's 40mhz performance..

add : i added Cyrix 486DRx2 66MHz's cachechk data.

Thank you for posting all the results skv400.
Would be interesting to see how the TI 486SXL 40 MHz CPU stacks up against the TI 486DLC 40MHz & the Cyrix 486DRx2 66MHz on these benchmark tests.
I must be honest, I didn't even know there was an ULSI Math-co DX2, but then again, I also didn't investigate this much further.

The first PC I bought for myself was a Cyrix 486DLC 40MHz with the Cyrix FastMath 40MHz FPU (I still have it).
For its time, it was a very cheap alternative and was able to stand up to (almost) the speed of an Intel 486DX 33MHz.

But I agree, the speed increase of the Cyrix 486DRx2 66MHz over the Cyrix/TI 486DLC 40MHz simply doesn't justify the price premium.
If Cyrix had brought out the Cyrix 486DRx2 66MHz much earlier, it might have been worthwhile buying one and perhaps paying that price premium, but by the time that CPU hit the market, your (true) 486 CPU's were already affordable to most users.
I can still remember that, when I bought my Cyrix 486DLC 40 MHz back in late 1993, an equivalent Intel 486DX 33MHz rig cost only slightly more. But, back then, every cent counted and I was very intrigued by the Cyrix DLC CPU's and therefore decided to go for that one.