I really don't understand why you folk all insist on using Quake as a benchmark for non Intel CPU's when we all know it was specifically hand coded for Pentium architecture and it's pipelined FPU. To me it seems like trying to run a racing car on diesel instead of petrol.

I did use Wing Commander 3, Wintune and Sisoft Sandra as well. Quake is easy to bench, and vQuake is easy to install.

Wing Commander 3 has a massive performance drop for some mysterious reason when WA is enabled. It might run 50% slower. Both video speed and CPU speed are slower. It does seem like WA might be conflicting with the video frame buffer but I don't know how to determine or remedy that.

BSA Starfire wrote:

I really don't understand why you folk all insist on using Quake as a benchmark for non Intel CPU's when we all know it was specifically hand coded for Pentium architecture and it's pipelined FPU. To me it seems like trying to run a racing car on diesel instead of petrol.

What are some better non-synthetic real-world alternatives that are suitable for the 1996-1997 era and which display numerical performance results?

feipoa wrote:

What are some better non-synthetic real-world alternatives that are suitable for the 1996-1997 era and which display numerical performance results?

This is the reason I also used Wing Commander 3. It has CPU, video and CDROM data rate tests and gives a numerical score for each. It has SVGA support and is quite demanding. I think the engine is based on integer arithmetic instead of being floating point orientated. There is no "timedemo" gameplay benchmark functionality though as far as I know.

The WC3 demo includes the testing functionality.
http://www.wcnews.com/wcpedia/Category:Downlo … rt_of_the_Tiger

feipoa wrote:

BSA Starfire wrote:

I really don't understand why you folk all insist on using Quake as a benchmark for non Intel CPU's when we all know it was specifically hand coded for Pentium architecture and it's pipelined FPU. To me it seems like trying to run a racing car on diesel instead of petrol.

What are some better non-synthetic real-world alternatives that are suitable for the 1996-1997 era and which display numerical performance results?

Sure I uderstand your point, but I don't think that is really relevant, Quake was written for the Pentium, no getting away from that.

swaaye wrote:

feipoa wrote:

What are some better non-synthetic real-world alternatives that are suitable for the 1996-1997 era and which display numerical performance results?

This is the reason I also used Wing Commander 3. It has CPU, video and CDROM data rate tests and gives a numerical score for each. It has SVGA support and is quite demanding. I think the engine is based on integer arithmetic instead of being floating point orientated. There is no "timedemo" gameplay benchmark functionality though as far as I know.

And very commendable to try something different, perhaps we should try WC3 against all the different 5th gen chips, might be fun and pretty sure it's not been done before. I don't know a thing about the history of the that game beyond "Mark Hamill & the porn star", but if it's not specifically coded to the Pentium like Quake then it sounds like a really good option.
Sorry if i'm sounding difficult, I'm just sick of all the Cyrix 6x86, K5,K6 sucks purely on the basis of Quake. It's one game, & one we can play great on modern hardware far better too, so not that relevant today for retro machines
. I'm sure in the day reviewers latched onto it because it was super popular and easy to bench but it's a tradition I think we should drop now for non intel CPU, it doesn't give anything like an accurate idea how they run run other contemporay games.
Best,
Chris

BSA Starfire wrote:

And very commendable to try something different, perhaps we should try WC3 against all the different 5th gen chips, might be fun […]
Show full quote

And very commendable to try something different, perhaps we should try WC3 against all the different 5th gen chips, might be fun and pretty sure it's not been done before. I don't know a thing about the history of the that game beyond "Mark Hamill & the porn star", but if it's not specifically coded to the Pentium like Quake then it sounds like a really good option.
Sorry if i'm sounding difficult, I'm just sick of all the Cyrix 6x86, K5,K6 sucks purely on the basis of Quake. It's one game, & one we can play great on modern hardware far better too, so not that relevant today for retro machines
. I'm sure in the day reviewers latched onto it because it was super popular and easy to bench but it's a tradition I think we should drop now for non intel CPU, it doesn't give anything like an accurate idea how they run run other contemporay games.
Best,
Chris

WC3 was a bleeding edge game in 94. The movies certainly, because it's one of the more interesting SVGA FMV games, but also the SVGA, texture mapped 3D graphics. People complained about how demanding it was, just like Crysis or Unreal. TIE Fighter came out in 1994 too and it was only Goraud shaded VGA. WC3 runs acceptably on a DX2 50, but you want something more like a Pentium 200 to run it fast in SVGA. The Pentium was barely out at the time so I doubt it is as Pentium targeted as Quake.

Quake may be written for Pentiums, but it was still an easy thing to get up and running to see if write allocate was doing anything useful. And since I am playing with the Verite too, Vquake was another idea.

Doom can be benched as well. But I don't have any other game ideas. I really wish I could bench a RTS like Age of Empires or Total Annihilation somehow. Maybe even a turn-based strategy game. AI and pathfinding calculations are just as hardcore as 3D graphics.

I merely mentioned Quake because it is the de facto test and really easy to set it up. I played through Wing Commander 3 on my K6-III+ 550 and I'm pretty sure there were some points with a lot of action in which I saw the framerate take a slight dive.

Swaaye, you could give MDK a try, it has some weird built-in performance meter. You could also try it it out in normal gameplay. Can't really of anything else that would allow for any performance differences to be substantiated!

I recall the MDK performance monitor being fairly sensitive. Worth a run through.

I have used Write Allocate Monitor to enable write allocate with AMD K5 in W9x, it works with AMD K6 as well. Also it is possible to disable CPU internal chache and CPU L2 chache probably in k6-II+,k6-III, k6-III+ and has nice GUI.

For me it did work with Apollo MVP3 chipset motherboard and did not with ALI Aladdin V chipset as i remember, but it may had do to something with bios or windows settings.

Download it here:
https://www.upload.ee/files/8175174/wamo2057.rar.html

Maybe try Blood - it's a very stressful application and awesome stability test. Similar to Doom it's highly dependent on memory speed but the gains seem more substantial when compared to formerly mentioned.

What's up with all the necroposting as of late ? 🤣

F2bnp wrote:

What's up with all the necroposting as of late ? 🤣

^^This 🤣

At first I thought it was to do with the wiki page I kinda maintain, but this thread isn't listed there I think.

F2bnp wrote:

What's up with all the necroposting as of late ? 🤣

Some people like it, some people hate it. For me, if it fits very well with an old post which contains very specific, non-broad spectrum information, then I prefer a "necropost" if it isn't too too old (e.g. 10+ years). This isn't my forum, so if there are specific owner rules against this, then they should be followed.

Did you compare minimum FPS ?

Ah 2015..

I did a quick translate of the text files for the K5.sys driver from Japanese to English using Google translate. Here is the complete file set with those translated text files.