You are right according to the reviews of the time, however in my own personal testing - just preliminary - things are looking different! This is an idea for a future project I had in mind:
Athlon 1400C tested on Abit KT7A - VIA KT133A with SDRAM PC133; Abit KG7-raid - AMD760 with DDR and Abit KR7A - VIA KT266A with DDR266. The main problem is finding a compatible graphic card for all 3 systems!

wouldn't the radeon 8500 and gf3 work fine on all 3?

Yes, they would. I do not have a radeon 8500 however I tested Abit geforce3 ti500 & ti4600 and they work fine, however like in the socket 462 comparative period correct graphic cards are outclassed in games and benchmarks, even in period correct ones, at maximum settings! The scope is to have a platform bottleneck and not a GPU bottleneck, all others variables being equals! For example in 3d mark 2001 geforce4 gets crushed and all 3 platforms get the same score at max settings despite great differences in 3d mark 2000! Geforce 7800gs, 6600gt and any bridged cards I tested refuse to run on any of the 3 boards!

Well, isn't that a good thing? If with a GF3 Ti500 all 3 boards perform identically, then it doesn't matter which board is fastest and you would rather pick the one with the nicest features 😀 That would likely be the KT7A because it has ISA

nd22 wrote on 2021-03-16, 16:48:

You are right according to the reviews of the time, however in my own personal testing - just preliminary - things are looking different! This is an idea for a future project I had in mind:
Athlon 1400C tested on Abit KT7A - VIA KT133A with SDRAM PC133; Abit KG7-raid - AMD760 with DDR and Abit KR7A - VIA KT266A with DDR266. The main problem is finding a compatible graphic card for all 3 systems!

For my first AMD Athlon DDR system I used a MSI K7T266 Pro2 motherboard with VIA KT266. I had some issues with it. My ASUS V8200 Deluxe GeForce3 did not run at AGP 4X, only AGP 2X. I later upgraded it to the ASUS A7V266-E with VIA KT266A and the performance increase was noticeable with the same CPU, AMD Athlon Thunderbird 1200C. AGP 4X was also no problem anymore. The VIA KT266 was not very good, but the KT266A was a good chipset.

God Of Gaming wrote on 2020-06-04, 10:29:

btw, mobo trays are awesome, you use them as a test bench, and when happy just slide them in from the back and you're done, so convenient! Why dont they make such cases anymore?

Some business-oriented machines still have them. I'm just glad many consumer cases still have rear cutouts behind the CPU to allow changing of a cooler's mounting hardware without having to take the motherboard out of the case.

About this time 11 years ago I decided to build a KT7A nostalgia system, and I wanted to max it out. So I bought 1.5 gigs of SDRAM, a PCI SATA card, and a Radeon HD 4650 AGP. I was a bit disappointed in the result, and ended up parting it out, but the HD 4650 worked fine in the KT7A I had.

God Of Gaming wrote on 2021-03-17, 08:23:

Well, isn't that a good thing? If with a GF3 Ti500 all 3 boards perform identically, then it doesn't matter which board is fastest and you would rather pick the one with the nicest features 😀 That would likely be the KT7A because it has ISA

IMHO it is not a good thing because you would not know that AMD760 is faster than VIA KT133A for example - in 3d mark 2000 KT266A + AMD760 far outscores KT133A even with geforce4 ti4600. As I said my purpose is to have a platform bottleneck and not a GPU bottleneck, all others variables being equals! However if you like having an ISA sound card in your system than by all means, please use KT133A motherboard!

Of course, having a platform bottleneck is a good way to find out how they compare, but then for actual use in a 2001 build with a 2001 card, if the bottleneck is on the gpu side and all 3 perform the same, all Im saying the ultimate board to use may not necessarily be the fastest board, but the one with the most features. With all that said, 3dmark is not necessarily representitive of all gaming. It might be gpu bottlenecked in 3dmark, but might be platform bottlenecked in some actual game, maybe perhaps some rts with more AI calculations or whatever

So a bit of an update from me. I lost my interest in doing yearly builds a while ago. My plan was to make a PC for each calendar year with components released within the calendar year. However calendar years did not match very well hardware generation release dates, and as a result some of the planned builds ended up not being very interesting to do or keep. My new plan that I can find a lot more enjoyment with, is to do builds per hardware generations, of sorts... not the easiest thing to define but I have done some thinking on it.

With that said. I am renaming and repurposing this "2001 Dream PC" project into a "DX8 (generation) PC" project, and trying to re-name the thread to that now. I would define that with components released between the first dx8 gpu (the geforce3, released on february 2001) and the first dx9 gpu (the radeon 9700 released on august 2002). Actually I noticed that nvidia detonator driver version 6.36 dated 24 october 2000 has a line entry for the "nvidia nv20" that is the gf3, for some reason. Maybe pre-release beta driver used for testing engineering samples or something, I don't know. But this is where I begin the journely, I am building this machine into a geforce3 pc with components and software matching october 2000, and will slowly upgrade it towards august 2002 later. I will try all the compatible driver versions to see which one(s) have the best game compatibility and performance.

So here is where I start: Asus CUSL2-C motherboard (~oct 2000), Intel Celeron 700 (june 2000, $192 msrp - core i5 money today, PIIIs were expensive back then) and the Cooler Master ATC-201 case (january 2001 I think? maybe a bit earlier). Waiting for a few other components including the geforce3 to arrive from ebay, currently.

God Of Gaming wrote on 2024-03-25, 13:10:

With that said. I am renaming and repurposing this "2001 Dream PC" project into a "DX8 (generation) PC" project, and trying to re-name the thread to that now. I would define that with components released between the first dx8 gpu (the geforce3, released on february 2001) and the first dx9 gpu (the radeon 9700 released on august 2002).

With the change from yearly releases, what is your new goal? You mention a DX8 build, but many games released at that time didn't perform their best. Moving to 2003 you'd get access to the Jan 2003 GeForce FX which is listed as a DX9a card, but has terrible DX9 performance, but great DX8 with older features like 8-bit palleted textures and fog table still supported. As well as getting access to Sept 2003 Socket 754 which now-a-days are slow, lower end XP systems, but fly with 98.

The generational take is how I did my builds. I have a Pentium MMX /w Voodoo 1 for speed sensitive 386 -> Win95 with early 3D accelerated gaming. It can run DOS, Win 3.1, and Win95 for very early DX games. Then an Athlon 64 with GeForce FX 5500 running Win98 for anything that needs more power than that early build, but is pre-DX9 as well as early EAX and A3D supported games. And finally a i5 3550 with a GeForce 750ti for DirectX9. Most games released after that can still run on modern systems today.

So I guess the question is do you want to showcase the games, see them over time. Or are you interesting in how the hardware changes over time and using games to see the differences?

Of course I will also use 2003 hardware like the geforce FX, but not on this build. Even though the FX was a bad dx9 card, it was still technically a dx9 card nonetheless, so its not strictly dx8 hardware that Im targeting with this build

The FX can run DX8 games with additional things like anti aliasing enabled and get better performance over the GeForce 4 cards. This wasn't quite the era of 3D accelerators double in performance in nine months. But I had an original GeForce 3, and upgraded to the more budget GeForce 4 4200 which was a good performance increase.

With the way DX works, there were many cards that "supported" a version of D3D. But in reality, they just used compatibility modes, so features of DX8 could run on this card, but it was dog slow because it's falling back on performing some things in software. While a card that's a version of DX newer handles all of the previous features fully in hardware.

That's why I questioned this being a DX8 build versus a 2002 build. Having a date of release limit for hardware in a build is fine, but it doesn't allow all of the DX8 games to be their best. PCs don't have easy dividers between generations like consoles, so we have to come up with our own sets of goals, and any self imposed limitations.

SScorpio wrote on 2024-03-25, 16:02:

Having a date of release limit for hardware in a build is fine, but it doesn't allow all of the DX8 games to be their best.

And this is why Im gonna have a myriad of builds, if any game works better on another hardware combination, Im gonna have that as well. And its not like Im gonna run dx8 games exclusively on this dx8 build. I will use each build to run all the games that it could run, in fact that would be the criterium I use to compare the components and driver versions I test for each of the builds - which combination can run the most amount of games with the minimum amount of issues and the best possible performance, and Im gonna do extensive testing to find that out myself, rather than relying on just common knowledge

Update on the DX8 build, finally found a reasonably-priced GF3, an Elsa Gladiac 920, as well as some pc133 cas-2 ram, 3x128mb (micron?)

Starting to take shape now

After being on pause for half an year till I was able to re-cap the power supply I chose (Enermax EG451P-VE), I'm finally getting close to getting this machine in an usable state to start performing game compatibility and performance testing. I added the infamous IBM Deskstar 75GXP hard drive, a Sony MPF920-E floppy drive, and a Pioneer DVD-105S slot-loader optical drive.

I tested the oldest driver I could find that lists the geforce3, the detonator 6.36 dated 24 october 2000 (the reason all the components are period-correct for october 2000), and established that it does not work at all. On windows Me, an odd thing happens where as soon as driver loads, PC does a hard shutdown. Freaky. On 98SE, it goes into a boot loop instead. I only found a win9x version of this driver, and I couldnt be bothered to try it on win95 after seeing what it did in Me and 98se, so I decided to scratch it off the list and move on to next version to test.

So, next one is detonator 7.16 dated 31 october 2000, and I have more luck with this one, it appers to work so far, in both 98se and Me, it has a win2k version too. I will go with Me first, as it was the main home version of windows at the time. Card is detected as a "NVIDIA NV20" in device manager, and as a 64mb Riva TNT in the nvidia control panel. I was able to select 32bit color and raise the resolution, however it wont let me select display refresh rate, and 60hz is painful on CRT, so I suppose I will do a period-incorrect thing here and use a LCD instead

So I finally got it up and running with some mostly period-correct software. Remaining bits I added are Creative Sound Blaster Live! 5.1 SB0060, Promise Ultra100 TX2 ide controller, Intel Pro-1000GT network card, IBM Deathstar 75GXP hard drive as system drive, and a couple other period-correct drives I had around as extra storage (a Quantum Fireball lct10 and a Seagate U-series 5). Still need a few more addons to fully pimp it out, most importantly a cd burner to fill the empty gap as I dont have a 5.25 cover for it.

Detonator 7.16 wont let me change display refresh rate, and stuck on 60 I decided to use a LCD instead of CRT while testing the current config. For other peripherals I went with an unbranded multimedia keyboard with buckling spring switches, and a logitech ifeel mouse.

A few preliminary benchmarks: 3.99 score in final reality, 3dmark 99max at 1024x768x32 scores 3799 3d marks and 8333 cpu marks, and 3dmark 2000 at 1024x768x32 scored 3529 3d marks.

I started playing Serious Sam. Avg fps seems to vary between about 25 and 40 depending on the scene, so its not great but it is about playable. Seems to be a cpu limitation as fps does not change between 640x480 and 1024x768, and only chnages by about 2-3 fps by changing graphics quality settings. Kind of odd, the og xbox port runs better than this, despite having more or less the same hardware specs. Also has a few minor rendering errors, so I can already tell detonator 7.16 will not be the best driver for the gf3, but I will try out all the games it is able to run and log data about them before I move on to the next version

Time for a little upgrade, can you believe it I actually found one more IBM Deathstar 75gxp that still works 😀 WIth that all four 3.5" mounts are finally populated 😉

Small update on the dx8 build, I already had a sb live 5.1 sb0060 mounted inside, but I was able to find a full platinum set so I mounted that. This finally closed the gaping hole in the front due to missing 5.25 cover. Now I will finally be able to enjoy Serious Sam the way it was meant to be heard 😉

Lovely build. Well done! I too am opting to do builds by generation. Though in some cases I am not pairing the GPU with a period correct CPU. The generations I am building are:

DirectX 3 - Pentium MMX @300Mhz + Voodoo 1
DirectX 5 + Glide 2 - Pentium II @300Mhz + Riva 128 4mb + Voodoo 2
DirectX 6 + Glide 2 - Pentium III @550Mhz - TNT2 + Voodoo 2 SLI
DirectX 6 + Glide 3 - Pentium III @700Mhz + Voodoo 3500 (August 1999)
Directx 7 - Pentium III @933Mhz - GeForce 2 GTS 64MB
DirectX 8.0a - Pentium 4 @3.4Ghz - Geforce 4600
DirectX 9.0c - Core 2 Duo E8600 @3.33Ghz -
2x Geforce 7900 512Mb in SLI
DirectX 10.0 - i5 3570 @3.8Ghz + Geforce 285 GTX 1024Mb