FPS Column is not correctly calculated if there is no data in Quake field (or other field). Divider needs to be always 3. Now you have a 286 on the 6. place 😉 I recommend to use 0 where there is no data. Maybe add a new column "Misc notes" where additional information about the system can be added - like no FPU, crashes at X benchmark...).

keropi wrote:

Well, it also craps out with the 486slc2 without cache enabled.
I don't know, I don't trust it even on slower machines after that... it's too unpredictable 😐

What do you mean by that?

kixs wrote:

keropi wrote:

Well, it also craps out with the 486slc2 without cache enabled.
I don't know, I don't trust it even on slower machines after that... it's too unpredictable 😐

What do you mean by that?

Phil's Ultimate VGA Benchmark Database Project

Aha... I didn't see that post. It is interesting... does the same happen with the original 3dbench 1.0?

Also... does it really run so fast or it's just the wrong FPS number at the end?

It would be good to not include the 3Dbench2 numbers in the avarage FPS score if that lowers the avarge FPS

I got the exact same "low" score in 3Dbench2 using the Radeon 5770 with the CPU @ 3.33 and @ 4.5 GHz
Also the Geforce 7900 GTX that scored otherwise poorly was too fast in that test...

3Dbench2 as I can see it's not included in AVG FPS. Only PCPBench, Doom and Quake.

kixs wrote:

3Dbench2 as I can see it's not included in AVG FPS. Only PCPBench, Doom and Quake.

Good to know 😀
How do you access the table for overclocked results?

Like in Excel... on the bottom of the window you have Tabs/sheets (this is one document with different sheets).

AH 😀

How did I manage to miss that
I am not allowed to move my overclocked results.
Something about that the stuff I am trying to paste overlaps protected cells so Phil will have to do it.

Copy cell by cell. Or even better all the cells till you get to the protected ones. There are just two protected cells you can't paste (DOOM FPS and AVG FPS).

kixs wrote:

Copy cell by cell. Or even better all the cells till you get to the protected ones. There are just two protected cells you can't paste (DOOM FPS and AVG FPS).

I guess I cant be lazy and let someone else do it when you provide step by step instructions 😉

Its moved.

Just before 5 AM, I get up to check what's going on and find that the 1000 FPS in Quake has been cracked 🤣

Well done!

Yes I'll move any OC results when I spot them.

Some tips:

- You can switch pages down the botton
- Please read the notes, it has valuable tips and explains FAQ like how the average FPS is calculated 😀 I will work on a decent formula that checks what's inside a cell first
- Please always do a stock run first, so we have something for the main table, especially if it's an under-represented CPU 😀

Here in Europe its still sunday.
I am a bit slow today, perhaps I have been drinking to much beer 😀
To get a high avarage FPS you have to submit a high PCPBench score and a high Quake score but no Doom score 😜
Make it so you need a Doom score to get an avarge FPS !

Need your input, guys...

Let's calm everyone down 😀

So looking at the data we can see that machines, which can't complete all tests, end up appearing more powerful on paper because they don't have the lower scores of the more demanding benchmarks lower their average.

I can think of a few options but there are likely more and it has to be fair as well:

- We can take entries out that don't complete PCPBENCH, DOOM and Quake. (I really don't like that option)
- We can interpolate / guess an approximate figure (not hard with so much surrounding results) and use that in the formula behind the scenes
- We can calculate 0 fps for benchmarks not completed and by doing so "punishing" systems that don't complete all tests
- We can leave it as it is

Other ideas? What's fair?

@ Skyscraper

That 0 fps for 3dbench2 for one of your coppermines, I changed that to 1000 fps because that's likely what's going on 😀

"We can calculate 0 fps for benchmarks not completed and by doing so "punishing" systems that don't complete all tests" is my vote.

And "too fast" at 3dbench for the contemporary machines would be 500.0 (seems to be the upper limit on my amd phe ii)

The systems that didn't/don't complete all the benchmarks should be listed/moved to a separate sheet.

Geoman:
3Dbench isn't used in AVG FPS, so it doesn't matter what is the score - it is only for the reference.

kixs wrote:

The systems that didn't/don't complete all the benchmarks should be listed/moved to a separate sheet.

That would mean no 386, no 286, no 486-SX in the main table. It's a choice, but a quite radical one. Because for gaming a 486SX-33 is no worse than a 486DX-33 in the vast majority of cases.

kixs wrote:

3Dbench isn't used in AVG FPS, so it doesn't matter what is the score - it is only for the reference.

It IS for systems that ONLY have a 3DBENCH score, such as 286 computers 😀

Perhaps we use the formula of the pros at SpecOPC:

http://spec.it.miami.edu/gwpg/pastissues/Oct1 … /geometric.html and rank the individual influences of the benchmarks, quake should have the highest influence, since it's the most complex piece of software...

There are plenty 386 with FPU that complete all the benchmarks.

I didn't says that 3Dbench isn't important as a benchmark. But it isn't important for calculation of AVG FPS.

The other option is to give score 0 to non complete benchmarks and the formula for calculation of AVG should stay the same for all systems = (PCPBench+Doom+Quake)/3 in this case the non complete benchmarks wouldn't artificially increase their AVG score.

Geoman wrote:

Perhaps we use the formula of the pros at SpecOPC:

http://spec.it.miami.edu/gwpg/pastissues/Oct1 … /geometric.html and rank the individual influences of the benchmarks, quake should have the highest influence, since it's the most complex piece of software...

At the moment there is more basic problem. What to include and how to calculate the AVG FPS.