^ OK, thanks a lot for the info. I didn't realise there was a Windows version of PcpBench.

Yeah, it's useful to test that early 3D cards that will crawl in 3DMark or Final Reality benchmarks. It'll also let you test Direct3D software modes.

Recently when I tried to benchmark my 486 boards, I tested also a pentium to compare it. And while an 75MHz cpu scored about 79fps, and the 133MHz 124fps, later I thought to try my main pentium pc. It's pentium 233MMX with a soyo at board, having eteq chipset. I was very surprised to see that it scored only 72fps, using the 1.0c version. This is the same machine that some time ago had scored 147,something, and is mentioned here. I didn't know why that happened. I tried to change BIOS settings, change RAM/CPU/Video card, but nothing changed it's performance. So I decided to get rid of that board. I never liked the eteq chipset, anyway. I replaced it with my old soyo 5VA2, with the Intel 82430VX chipset. However this board has an issue. It can recognize mmx cpus, but they can only work @ 166MHz. The 200Mhz one works @166, and the 233 with the multiplier set to 1,5x works at 133MHz.... I really don't know why , and I've tried every bios upgrade I've ever found, but still can't get to work with an MMX cpu faster that 166. Now, my new system specs and score are the following :
Intel-Pentium_MMX-66-2.5x-166-Soyo-5VA2-i82430VX-256kB-PCI-Diamond Stealth 3D 2000-4MB-Markk-141.6
Now, I think it is a little remarkable that, having a cpu slower by 66MHz, and 256kB cache instead of 512, it is only 6.5 fps slower. Maybe if I had 512kB cache in my current board, it would come very close. And also in general, I think that the faster MMX cpus, have better to be used with a little more advanced and newer ATX boards, rather than older AT ones, even though if they support them... 😀

MMX Pentiums are a lot faster!

I did a quick test on my SS7 Iwill board with a Pentium 200 (non MMX):

66 x 2 = 133 MHz: 114.9
66 x 2.5 = 166: 126.7
66 x 3 = 200: 136.2
100 x 2 = 200: 150

Markk, you seem to have the same problem as me: I have a Soyo 5BT board (i430TX) which I love, but unfortunately, using an Intel MMX CPU, I can't pass the 166 Mhz mark. Even if I set the multipliers to 3x or 3.5x, it stays at 2.5x (166MHz). At least I can OC the FSB to get 190Mhz in that CPU (75Mhzx2.5x). I would not have imagined that it seems to be common on Soyo motherboards.
Also, it only happens with MMX CPUs. Normal P54C ones works ok, all multipliers.

Yeap, that's right, if you use a P54 cpu it works fine. I don't know if it's happening on all soyo boards, I guess it has to be on earlier models. I had another soyo with intel chipset, that worked fine with the 233 cpu, but it wasn't in good condition, due to battery leaking (very uncommon for a board of this era, as most of them had either the dallas chip or 2032 battery...). I've been searching for months for a solution for the faster mmx cpus in that board I use now. Everyone has the same problem. I've only found one url that says it has the solution, but you have to pay to view it.... And the other thing is that a friend of mine remembers having the same board but with 512kB on board cache (that was 5VA5, mine is 5VA2 with 256kB, and a socket for a coast module) working @ 200MHz. As a matter of fact, he remembers that he had a 166MMX, overclocked to 200. Something that is just impossible with my board....

Mau1wurf1977 wrote:

MMX Pentiums are a lot faster! […]
Show full quote

MMX Pentiums are a lot faster!

I did a quick test on my SS7 Iwill board with a Pentium 200 (non MMX):

66 x 2 = 133 MHz: 114.9
66 x 2.5 = 166: 126.7
66 x 3 = 200: 136.2
100 x 2 = 200: 150

So, I suppose it is right what people say, that the 166 mmx is slighty faster than the 200 non-mmx.

Markk wrote:

So, I suppose it is right what people say, that the 166 mmx is slighty faster than the 200 non-mmx.

Looks like it! The results would suggest so and I have a very modern SS7 board with lots of cache and a fast MX 440...

One more benchmark for the chart...

Intel-486DX2-33-2-66-ECS-UM4980-UMC UM8498F-256KB-VLB-Cirrus Logic GD5428-1MB-sprcorreia-43.4

And just to show that graphic card does matter here is same setup this time with an OAK ISA card....

Intel-486DX2-33-2-66-ECS-UM4980-UMC UM8498F-256KB-ISA-Oak Technology OTI 087-1MB-sprcorreia-31.2

Thanks!

Database has been updated...

@Mau1wurf1977 You have done great job! (:

This is my result:
AMD-5x86-60-3-180-Shuttle-Hot-433-UMC8881-512kb-PCI-Tseng ET6000-4MB-udam_u-102.1
3dbench 1.0c.

unmei220 wrote:

Markk, you seem to have the same problem as me: I have a Soyo 5BT board (i430TX) which I love, but unfortunately, using an Intel MMX CPU, I can't pass the 166 Mhz mark. Even if I set the multipliers to 3x or 3.5x, it stays at 2.5x (166MHz). At least I can OC the FSB to get 190Mhz in that CPU (75Mhzx2.5x). I would not have imagined that it seems to be common on Soyo motherboards.
Also, it only happens with MMX CPUs. Normal P54C ones works ok, all multipliers.

This is really strange (and sorry maulwurf for going off topic a bit!). Have you measured the speed with a program, or benchmarked it?
Perhaps theres something wrong with the way the board 'sends' the multipliers to the cpu?

You could opt to set the multi directly on the cpu by shorting a couple pins. I don't know the details about it but at least you can bypass the motherboard this way.

Yes, I measured it with a lot of programs, and all say the CPU is at 166Mhz.
Didn't know about shorting the pins on the CPU, will try it someday.
At first I thought the dip switch was broken, but it worked perfectly when I put a non-MMX CPU. Also tested it with a multimeter.
I already forgot now, but I remember after searching a little using google, it seems this were not the only Soyo model with that kind of problem.
As I'm stuck at 2.5x multiplier, at least the CPU is running happily now at 210Mhz (83x2.5x).

I've benchmarked my soyo with various cpus, and the the speed it shows is indeed the actual. That was the first thought I had, that maybe it shows for example 166MHz while it runs at 233, but the benchmarks showed equal performance to the 166MHz cpu. It would be interesting to work on that matter of the "forced" multiplier setting, but I'm more or less convinced that when the MMX cpus where made more common, it was the time of ATX motherboards that would work fine with them. And also I believe that the performance gain isn't that huge, when comparing the 166 MMX with the faster ones. For something quicker, it's better to have a pentium 2 or 3.

Markk wrote:

I've benchmarked my soyo with various cpus, and the the speed it shows is indeed the actual. That was the first thought I had, that maybe it shows for example 166MHz while it runs at 233, but the benchmarks showed equal performance to the 166MHz cpu. It would be interesting to work on that matter of the "forced" multiplier setting, but I'm more or less convinced that when the MMX cpus where made more common, it was the time of ATX motherboards that would work fine with them. And also I believe that the performance gain isn't that huge, when comparing the 166 MMX with the faster ones. For something quicker, it's better to have a pentium 2 or 3.

Just to be sure, are you using an actual 233Mhz MMX chip or trying to overclock a 166Mhz MMX chip?

Edit: Has anyone run a benchmark on the Cx486DLC by any chance? This chip should basically be a 486SX with a 386 pinout.

Yep, I've got 2 or 3 MMX cpus of each different frequency, so I've tried 166,200,233 and only the first works fine. I've got another 200MHz which is different from the others(ceramic, like the older/slower Pentiums), but doesn't make a difference. And I also have two almost identical motherboards. Both are Soyo 5VA2, but one of them is slightly older, and is missing a few jumbers, but works exactly the same. So far I can't make it work correctly using the faster MMX. I've been googling it from time to time amost for a year now, but I haven't found a solution, besides just one url that claims to have the way to solve it, but they need to get paid to tell you how.... And about the 486DLC, If I remember well, it has just 1KB of L1 cache instead of 8 that the 486s have. I don't have such a cpu, but what I can tell you is the result of a Greek magazine that tested it in 1993. It was when AMD released the 40MHz 486, and they had 3 systems using that cpu, and one with the cx486DLC. The Amd 486s had a score of 1,75-1,79. I think that is how faster it is than the reference pc of the benchmark which was a compaq 386dx/33(but I'm not completely sure about that). That score they wrote is almost the 40/33 of the i486DX/33. So by doing the math, I think the 33MHz scores about 1,45. And the cx486DLC they measured scored 1,41. The also said that for that time it was a alternative more to the 486SX that to the 486DX, but also to the 386DX/40, and in general, the motherboard they tested was about 16% faster than the fastest 386 they had ever measured.

I've found one reference of someone running your board @ 233Mhz here:
http://old.benchmark.pl/archiwum/kwie98.htm

Which cpu's -exactly- do you own?
Edit2: I'm asking this as many of the slower MMX chips had the higher multipliers disabled.

Edit: both the AM386-40 and the Cx486DLC don't have the fpu integrated in the cpu like the 486DX does. Both these cpu's should work with a 387 so it's the integer unit that's the only real difference between them.

Well, right now I have 4 . All of them have the code "FV80503233/SL27S" and the are different only by date codes. I've got also one 200Mhz like them, which is "BP80503200/SL23W", and has glued heatsink/fan on it. Also another same looking as that, but without heatsink, and the code is "FV80503200/SL26J". And finally the ceramic 200MHz, which is labeled "A80503200/SL2RY". The rest I have are 166s. When I try the 200/233@ 200Mhz, the run @ 166, and when I try the 1,5xmulti for the 233, it runs @ 133... I've also tried them on other boards that supported MMX, and they run fine. The mystery is how to make them run correctly on those specific soyo boards. And the confusing part is, as mentioned before, that the non-mmx cpus, run fine....

edit: I've got a couple of identical 4C87DLC-40 IIT fpus. I suppose they work with 486DLCs, but I use them for my 386/40 and work fine. I wonder if the opposite is possible. I mean, to use a 387/40 on a 486DLC. The reason why I wonder about that, is because I see that cyrix had both 83D87-40 for the 386DX and Cx87DLC-40 for the 486DLC.

Markk wrote:

edit: I've got a couple of identical 4C87DLC-40 IIT fpus. I suppose they work with 486DLCs, but I use them for my 386/40 and work fine. I wonder if the opposite is possible. I mean, to use a 387/40 on a 486DLC. The reason why I wonder about that, is because I see that cyrix had both 83D87-40 for the 386DX and Cx87DLC-40 for the 486DLC.

I suppose Cyrix first made the 83D87-40 and later started making an improved (?) fpu and named it the Cx87DLC-40?

I'm not savvy when it comes to the 386, but I'd say those should work in any combination