One error (maybe).

Ti486DLC-40BGA / IIT4C87DLC-40 is on a 386 platform not 486 (Socket 1-3). Otherwise you have to change Cyrix 486slc2-40 from 386 to 486 class.

jxhicks:
use the model name that NSSI or CHKCPU report.

kixs wrote:

One error (maybe).

Ti486DLC-40BGA / IIT4C87DLC-40 is on a 386 platform not 486 (Socket 1-3). Otherwise you have to change Cyrix 486slc2-40 from 386 to 486 class.

Yea these chips confuse me 😀

Does this chip go into a 386 socket? That's really what determines it. I thought it was a Intel 486 competitor.

EDIT: Found an image and counted the pins 😀 I'll fix it ASAP.

About DLC: (SLC is the same but goes into 386SX board)

Introduced in June 1992, like the later and more famous Cyrix Cx5x86 it was a hybrid CPU, incorporating features of a new CPU (in this case the Intel 80486) while plugging into its predecessor's (386DX) socket. It ran at speeds of 25, 33, and 40 MHz.

The 486DLC can be described as a 386DX with the 486 instruction set and 1 KB of on-board L1 cache added.

If you determine it by a socket then it is a 386. But if you determine it by an instruction set then it is a 486(sx).

Where would you put Intel RapidCAD? It's a 486DX upgrade for 386DX? Socket 386 and 486 class 😉

Yea the socket rules. If it goes into a 386 socket, it goes in that class. That's the whole point about these upgrade chips right? And it shows how they compare to a regular 386.

I'm doing the same with the POD and leaving it with the 486.

Anyone who wants to compare such an upgrade chip with the next class up, well there will be a graph for the entire machines, but there will others showing each class so we can bring out the detail. Well that's the plan at least.

Anyway, less talking, more BENCHING! We haven't cracked 200 submissions yet... 🤣

If I had more time to dig out and setup the machines and then tested them with different configurations... you'd have had more then 200 by now. Right now I'm benchmarking a machine a day (on average).

No pressure 🤣

I appreciate your contribution!

I've done some test with 486DLC-33 with and without FPU. It is interesing that PCPBenchmark results are slower without FPU. On this configuration with FPU 5.4, without 5.0.

On 486DLC-40 with FPU 5.8, without 5.3.
On 386DX-40 with FPU 4.3, without 4.0

All with fastest possible BIOS settings and ISA CL 5429 2MB (fastest ISA VGA card I've got).

Awesome, thanks!

Good work everyone!

We got 200 entries so far 😀

Tomorrow (Australian time) I will sit down and create some charts. This gives everyone in the other timezones time to add anything this Sunday.

Mau1wurf1977 wrote:

Good work everyone!

We got 200 entries so far 😀

Tomorrow (Australian time) I will sit down and create some charts. This gives everyone in the other timezones time to add anything this Sunday.

Nooo we didn't 🤣

First, my 386SX board turned out to have 20, 25, 30 MHz CPU clock jumpers that do nothing. Second, my 486DX2 refused to boot for unknown reason. I started typing the configs early, as doom on 386 takes so much time 😵

No problem 😀

After looking at the results, I believe underclocked results should be moved from stock settings to either the overclocked tab (which will be renamed) or a separate tab. They are not "stock" CPU settings.

^ agreed.

Do you mean running a Pentium MMX 233 at 66 x 2 to submit a MMX 133 result?

Or do you mean "odd" values of CPUs that didn't really exist?

So I submittet a Bochs score - but to configure this thing is painful!

Quake won't even start:

1FindFile: can't find gfx/pop.lmp

😵

Mau1wurf1977 wrote:

Do you mean running a Pentium MMX 233 at 66 x 2 to submit a MMX 133 result?

The problem I see with this is that it is easy to produce skewed results without an encyclopedic knowledge about CPUs and their features. Your example is already a borderline case, since the MMX-133 is a rather obscure mobile variant, but you can't really compare the results with a "mainstream" non-MMX Pentium 133, because that one only has half the L1 cache.
If you look at the downclocked K6-III scores, they seem similar to Pentiums in the same clock ranges, but these machines are in some aspects much faster than their older counterparts.

Sorting results in a "stock" and "non-stock" tab would make it easier to work with the results and minimize the risk of drawing false conclusions.

I agree with this. Default sheet should have only real stock speeds.

I just added my laptop's results on both 133MHz and "compatible mode". CHKCPU says that's 8.4MHz so I entered that. Seems to make a nice 386 at that speed. Would this be appropriate in the "stock" tab or should I move it to "overclocked"?

Charts added

Quite a bit of work. Will update them once we have a significant number of new entries 😀

Really strange results on the SS7 boards!

Ali V seems to be generally faster than MVP3, except in Quake were it seems to be a bit quicker. There seems to be significant differences on motherboards as well. This is different compared to Intel based Socket 7 motherboards which generally behave very similar if based on the same chipset. Motherboard brand doesnt matter much if you for instance have a board with the Intel 430HX chipset.

I have an Epox MVP3 board in addition to my Aopen AX59-PRO. I'll bench it today.