Am386DX-40 wrote on 2023-11-03, 22:52:

Bump! Looks like now it's really gone...

I have noticed that about a month ago.

At least archived versions can be accessed thru Wayback Machine: https://web.archive.org/web/20230823172523/ht … r.hu/DOS_TESTS/

Gona's website still works fine from the EU.

Possibly some US provider issue?

It works fine for me, too (in USA). I've never noticed an issue to date.

Am386DX-40 wrote on 2023-11-03, 22:52:

Bump! Looks like now it's really gone...

My provider said that their servers were getting too many attacks from China, so he blocked all of China. He may have added other countries to the blacklist too. I don't think this is a good solution, but I get the hosting and service for free, so I can't really interfere/complain about this. But I can make a copy, a 2nd site.
Am386DX-40 could you recomnend a free website for 2nd copy?
Or contact me in private, and might your IP range could be on a white list.

Site still works for me (Canada).

Gona wrote on 2023-11-04, 17:31:

My provider said that their servers were getting too many attacks from China, so he blocked all of China. He may have added oth […]
Show full quote

Am386DX-40 wrote on 2023-11-03, 22:52:

Bump! Looks like now it's really gone...

My provider said that their servers were getting too many attacks from China, so he blocked all of China. He may have added other countries to the blacklist too. I don't think this is a good solution, but I get the hosting and service for free, so I can't really interfere/complain about this. But I can make a copy, a 2nd site.
Am386DX-40 could you recomnend a free website for 2nd copy?
Or contact me in private, and might your IP range could be on a white list.

Oh I have no idea, maybe migrating to a spreadsheet in google docs?

I'm in Argentina, my current IP address is 181.166.xxx.xxx (but it changes from time to time)

Now should be OK from there too. My provider has removed that blocking.

Gona wrote on 2023-11-12, 07:33:

Now should be OK from there too. My provider has removed that blocking.

Thanks! I (in Latin America) can now access the website just fine.

As a test, I set up a ChatGPT interface to the data: xxxx. You need the paid ChatGPT subscription to use it.

To answer questions the bot has access to the chart's data harvested around 2019, and has been told to stick to the facts as much as it reasonably can. It can do some numerical analysis as well etc., but I don't know how people would typically use the data so I haven't tested it much. Give it a go and see whether it can do something interesting.

For example:

1You: Are 3dfx or s3 cards better for DOS compatibility?
2
3Video chip DOS compatibility chat:
4
5Based on the DOS compatibility test results from my knowledge source, both 3dfx and S3 video cards show varied results in terms of DOS compatibility. Here's a summary of the average compatibility scores for these brands (higher percentages indicate better compatibility):
6
73Dfx Cards:
8- Voodoo Banshee: 87.5%
9- Voodoo3: 87.5%
10- Voodoo4 4500 PCI: 89.02%
11- Voodoo5 5500 AGP: 89.39%
12
13S3 Cards:
14- The scores for S3 cards range from as low as 31.82% (Savage/IX) to as high as 95.83% (ViRGE/GX).
15- Notably, several S3 cards like Trio64V+, Trio64UV+, ViRGE/GX, and Savage4 Xtreme show high compatibility scores above 90%.
16
17In general, S3 cards exhibit a wider range of compatibility scores, with some models achieving higher scores than 3dfx cards. However, 3dfx cards consistently show good compatibility, with scores mostly above 87%.
18
19It's important to note that specific model performance can vary, and these scores are averages of the tests conducted on each model.​

vvbee wrote on 2023-11-13, 08:51:

[…]
Show full quote

Copy code to clipboard

1Here's a summary of the average compatibility scores for these brands (higher percentages indicate better compatibility):
2
33Dfx Cards:
4- Voodoo Banshee: 87.5%
5- Voodoo3: 87.5%
6- Voodoo4 4500 PCI: 89.02%
7- Voodoo5 5500 AGP: 89.39%

My table is not really suitable for evaluating the compatibility results as a percentage, so these are false percentage results, since I bring particularly difficult and problematic cases only in my table. For example both 3Dfx (Banshee or later) and S3 Trio/Virge family average compatibility scores are more than 99%. This is a big problem, very misleading % results.
The statement would be true like this:

1Here's a summary of these brands' compatibility scores tested on more problematic games (higher percentages indicate better compatibility):

Furthermore a minor problem but also could be interesting what value are given to colors. For example green "OK" vould be 100% of the value and red are 0%, this clear, value but there are 5 other colors in the table and white also, and white are a big "range": depends by it's BIOS, or even "1MB memory not enough", or simply "not tested".

Pretty much any benchmark or test would be an incomplete sampling of a domain, so as you say it comes down to expressing the uncertainty in some way. If the tests are with problematic software then they all have that in common and it's a problem of wording still. It should be fairly easy to tweak the way the bot words things, as far as my limited understanding goes. Since the bot is able to run analysis on the data, I assume as a user you could ask it to massage or skew it in some way too.

In the data the bot was fed the scores go linearly from 100%/green to 0%/red in seven bins, white is ignored.

Some examples of visualizing the data (looks plausible, but I didn't verify by hand):

The bot seems to do worse when what you're asking requires information from other sources (the original data doesn't list release dates):

I never had issues in loading dos on any GPU , is this exclusevly for gaming(chart?)

ediflorianUS wrote on 2023-11-14, 15:10:

I never had issues in loading dos on any GPU , is this exclusevly for gaming(chart?)

None of the video cards are 100% compatible with all DOS games as some people think when it comes to VESA support, tweaked EGA/VGA modes and scrolling. If that was the case, then Gona's list wouldn't exist.

As a heads up, I set up the bot as a feature test and to see if people do something interesting with it. I don't plan to keep it around as a service, and as a test it's already served its purpose. Don't rely on it being available tomorrow or a week from now or whatever, if you plan on setting one up yourself then now's the time to take notes.

Some more use cases of the bot. As an abstraction layer/UI I think it's pretty good. Not perfect and knowing the dataset to begin with helps, but one could build a workable frontend this way.

Extract specific data as a CSV file. (If you're making your own bot, better give it instructions about case insensitive string matching.)

Find the most compatible resolution of Quake.

Find the best card for a specific selection of games.

Let the bot come up with some numbers on its own.