For us it's easy to churn out one result after the other and posting them here, but for Maulwurf it means editing the table, as well as converting and uploading it each time.
What we really need is a wiki or a shared database.
Yea that would be a lot easier 🤣
wrote:For us it's easy to churn out one result after the other and posting them here, but for Maulwurf it means editing the table, as well as converting and uploading it each time.
What we really need is a wiki or a shared database.
That would be a good idea! Thing is, someone will need to maintain it.
But I think that for a shared database, Maulwurf's database would be a good starting point!
I wouldn't mind continuing Maulwurf's work, but (this may sound complicated) I would need to set up a "digital database manufacturing system" for ease of maintaining, expanding and exporting into an easy readable table, the results.
Perhaps a Google Docs spreadsheet would suffice? You wouldn't need to convert it to a picture and host it, you could just link to the spreadsheet and people can view it in their browser
The thing is, Maulwurf used a system that made it very easy to insert data into the database and it's easy to manage. I don't know the 1st thing about google docs.
Having to insert 10 different words/numbers is 10 times as much work as being able to insert the entire text string at once.
If you guys can live without the "fancy lines", it would be much easier / faster for me, despite me using auto format...
I do like the idea of google docs however. Think of excel, just on a webpage. But if everyone has to sign up for google (which I guess is the case), that might be an issue...
I'll keep adding results, so don't worry about this. Maybe we can focus on the systems with little results so far and missing CPUs.
Another though is making a reference database. E.g. DK has a 386DX and could provide 25, 33 and 40 MHz results with his system.
Tetrium could do a 486 VLB system with 25, 33, 40, DX2-50, DX2-66, DX4-75, DX4-100 and DX4-133 results (for VLB and ISA), so we see the effect. Maybe there is a board that can do PCI, VLB and ISA. That would be awesome...
Just something that is simples / less lines / easier to use as a reference AND is using the same system for each class, so the CPU get's isolated.
Just a thought...
wrote:If you guys can live without the "fancy lines", it would be much easier / faster for me, despite me using auto format... […]
If you guys can live without the "fancy lines", it would be much easier / faster for me, despite me using auto format...
I do like the idea of google docs however. Think of excel, just on a webpage. But if everyone has to sign up for google (which I guess is the case), that might be an issue...
I'll keep adding results, so don't worry about this. Maybe we can focus on the systems with little results so far and missing CPUs.
Another though is making a reference database. E.g. DK has a 386DX and could provide 25, 33 and 40 MHz results with his system.
Tetrium could do a 486 VLB system with 25, 33, 40, DX2-50, DX2-66, DX4-75, DX4-100 and DX4-133 results (for VLB and ISA), so we see the effect. Maybe there is a board that can do PCI, VLB and ISA. That would be awesome...
Just something that is simples / less lines / easier to use as a reference AND is using the same system for each class, so the CPU get's isolated.
Just a thought...
I agree. One platform only changing CPU and maintaining bios settings would be optimal. I still have lots of cpus and could test them per platform.
I can contribute with several benchmarks. It would not take long after platform is put together.
That would be awesome!
What benchmarks could we add? That Russian benchmark is a given, maybe Norton SI 8.0 could be used and also the PC Player Bench in VGA mode (SVGA scores are too low on a lot of machines)?
wrote:That would be awesome!
What benchmarks could we add? That Russian benchmark is a given, maybe Norton SI 8.0 could be used and also the PC Player Bench in VGA mode (SVGA scores are too low on a lot of machines)?
This page as a few programs that might be good (most links are dead, but a google search may help).
http://active-hardware.com/english/benchmarks/benchmarks.htm
wrote:If you guys can live without the "fancy lines", it would be much easier / faster for me, despite me using auto format... […]
If you guys can live without the "fancy lines", it would be much easier / faster for me, despite me using auto format...
I do like the idea of google docs however. Think of excel, just on a webpage. But if everyone has to sign up for google (which I guess is the case), that might be an issue...
I'll keep adding results, so don't worry about this. Maybe we can focus on the systems with little results so far and missing CPUs.
Another though is making a reference database. E.g. DK has a 386DX and could provide 25, 33 and 40 MHz results with his system.
Tetrium could do a 486 VLB system with 25, 33, 40, DX2-50, DX2-66, DX4-75, DX4-100 and DX4-133 results (for VLB and ISA), so we see the effect. Maybe there is a board that can do PCI, VLB and ISA. That would be awesome...
Just something that is simples / less lines / easier to use as a reference AND is using the same system for each class, so the CPU get's isolated.
Just a thought...
That would be a good idea 😉
I would vastly prefer to have a benchmark that's bootable from floppy though, so I wouldn't need to do total windows installs. Having the boot media has the added benefit that other hardware doesn't influence the results. When I did my SuperPi benchmarks, I got the vague idea that the speed of the harddrive was influencing the results, which kinda bugged me. For me, ideal would be that doing a CPU benchmark should be about the CPU exclusively.
However, maybe it would be a nicer idea if anyone who is willingly, could do a whole class of CPU's per platform, as when 1 person only does all 386's, someone else does all 486's etc, theres no guarantee the results are 'objective'.
It would be nice if people could contribute a large range of results per test, for instance doing all CPU's from, say, 486DX-25 all the way to the 5x86, preferably all on the same day. This would also prevent results from getting screwed because someone accidentally benches using slightly different settings by accident (I know I would 😜).
So basically, instead of people posting 1 result at a time, people would post their complete range of results from 1 platform 😉.
But whoever manages the 3dbench results, for me this would always be Maulwurf's project, so I'd have no problem with continuing for a little while and having Maulwurf take over again at some point 😉.
Bootable is not an issue. 3dbench, speedsys and sysinfo should easily fit on a floppy. PC Player bench might be pushing it, but haven't tried yet.
Regarding results, just need to have some trust 🤣 Most of us kinda know what we are doing, so I'm not worried to be honest. If something is out of the ordinary, I'm sure we would notice.
Reason I want to spread the load, is because I also want to spread the love and have several people take pride in a document, that who knows what / where it will end up...
In 10 years people might be looking at that document for reference 🤣
I do like the idea of google docs however. Think of excel, just on a webpage. But if everyone has to sign up for google (which I guess is the case), that might be an issue...
I'm pretty sure you can have a public access doc without signing up for it - I'm sure I've seen that before. Will have to look into it
One potential way to get around the "one line of input" thing by storing the master copy as a CSV or TSV file, and simply adding each new entry to the bottom in the correct format, (using commas or tabs instead of the hyphens used in this OP) then uploading to google docs as a new spreadsheet and overwrite the old one.
DK has a 386DX and could provide 25, 33 and 40 MHz results with his system.
Unfortunately no, I don't have jumpers for clock speed, only ISA bus speed. I need to actually swap crystals to change the clock speed- and I've only got one crystal 😜 RG100 will be able to help you with this though.
Ok, but I'll need to upgrade the harddrive in my working computer if I'd want to go further. "400GB, you'll NEVER be able to fill it all up!"...lmao haha! 😜
wrote:Ok, but I'll need to upgrade the harddrive in my working computer if I'd want to go further. "400GB, you'll NEVER be able to fill it all up!"...lmao haha! 😜
That's what i thought when i had my 286@12MHZ with a 5.25" 20MB MFM Seagate HDD.
Loved seeing the drive work without it's cover... It was built like a tank! Flawlessly!
Fake cache!
Intel-486dx2-33-2-66-PC Chips-M919-UMC 8881F-9Kb-PCI-S3 Trio 64V2+-Unknown-Lazy1-47.6
[Edit]
Wait, actually the board isn't trying to "fake" L2 - it simply has none.
The _OTHER_ M919 I have does have fake chips though.
Bad press early on?
Intel-286-20-1-20-Octek-HT1X-Unknown-Unknown-ISA-Tseng ET4000-1MB-SquallStrife-6.2
Intel-486DX2-33-2-66-DataExpert-EXP4044-OPTI-256KB-ISA-Diamond Stealth 24-1MB-SquallStrife-27.7
Intel-Pentium-66-2-133-PC Partner-VXB830DS-Intel 430VX-512KB-PCI-S3 Virge DX-4MB-SquallStrife-123.8
Performance on the 486 system seems a bit crap for the specs?
Also, what's a good util I can run to find out the unknowns on a DOS box?
Edit: Sorry, had to amend the video card in the Pentium box.
Edit: Argh, missed the VRAM size!! 😖
That 486 is held back by the ISA video card...
Ah righto.
wrote:Ah righto.
Yea, pop in a VLB ET4000 and you should get a much better score!
Hey, I just noticed something, my 286 is faster than the first two 386s on your leaderboard! 🤣
