UPDATE: Right - after hours of testing, taking apart, inspecting (and finding nothing wrong) then researching, I figured out what went wrong:

*clears through: The K8M880 doesn't really like some video cards... witch explains why the rig would randomly crash / hang with the Radeon 7000 / 7200 / 8500, the TNT2 Ultra and the TNT2 PRO, as well as the Geforce 256 (occasionally) and the 6200A (bsod upon installing any driver). It's not defective, there's nothing wrong with it, apart it not liking some video cards. I did manage to match performance of my 939 rig with the 754 test bench by dropping the HT speed from 1000 to 800MHz, as well as using single channel ram as opposed to dual channel - aida64 synthetics are identical, and 3dmark scores using the x800xt are almost identical (within a 1% margin). Funny thing is hyper threading and dual channel ram makes a difference with fast video cards - at least in quake at 640x480 and in 3dmark 2001.

So, to summarize, I will be redoing the Riva TNT2 Ultra test (since it would not run 3dmark 2k on the testbench but runs fine on this machine), as well as half of the ATi card tests. Will keep you posted.

Added ATi card benchmaks - check the 2nd post!

kanecvr wrote:

well not really... it equals the Ti4200 in some tests, and the MX 440 in others (namely 3dmark). Not really a bad card for DX8 gaming, but if I wanted to run at 1600x1200 I'd go for a Ti 4600, or the FX 5700, or the 5900XT. as for overall performance ***SPOILERS*** the 9800PRO and the X800XT beat the crap out of everything else tested so far. Maybe the 6800LE @ 16pp will fair better.

Back then we were expecting upgrades, not downgrades when a new series of cards was released. After GF4 MX was out you had to be careful not to step in the same shit twice. It's amazing how poorly this (so called) 5-series card performed against many older models. There really was no point in releasing it in the first place, who would ever want to buy something like that and for what purpose? Older, faster cards could be had for less and it didn't really cater to the gaming needs either.

sf78 wrote:

Back then we were expecting upgrades, not downgrades when a new series of cards was released. After GF4 MX was out you had to be careful not to step in the same shit twice. It's amazing how poorly this (so called) 5-series card performed against many older models. There really was no point in releasing it in the first place, who would ever want to buy something like that and for what purpose? Older, faster cards could be had for less and it didn't really cater to the gaming needs either.

The difference is in perspective. You're looking at it from the point of buying it back thenwhen it was new, then being disappointed it wasn't as fast as what you were upgrading. I am looking at it from a retro gamer's perspective. And that perspective is: a quieter, cooler, less expensive, easier to find substitute for a GF3 Ti200. 😀

clueless1 wrote:

sf78 wrote:

Back then we were expecting upgrades, not downgrades when a new series of cards was released. After GF4 MX was out you had to be careful not to step in the same shit twice. It's amazing how poorly this (so called) 5-series card performed against many older models. There really was no point in releasing it in the first place, who would ever want to buy something like that and for what purpose? Older, faster cards could be had for less and it didn't really cater to the gaming needs either.

The difference is in perspective. You're looking at it from the point of buying it back thenwhen it was new, then being disappointed it wasn't as fast as what you were upgrading. I am looking at it from a retro gamer's perspective. And that perspective is: a quieter, cooler, less expensive, easier to find substitute for a GF3 Ti200. 😀

I added the 6200 to the nvidia charts. It's kind of on par with the 5200 in DX7 and 8. I also added the 6800LE.

My benchmark results confound me. ATi cards are faster then they are in old period correct reviews and initially I couldn't figure out why - but I think it has to do with drivers. I expected testing ATi cards to be driver hell, but it turns out it was the other way around - all ATi cards except for the rage128 were tested with Catalyst 6.2 and everything worked perfectly. On the nvidia side, you have to carefully match the driver to the platform / video card / games you need to play, otherwise you will get compatibility issues or mediocre performace.

The forceware 66 series stands out as offering the best performance for all cards, while the 43.xx driver offers best compatibility, and performance improvements for some cards (GF 4 titanium performs a little better).

Overall the ATi cards also proved faster - with the x800xt being on top. ATi cards had a notable advantage in unreal and quake 3 @ 640x480 as well - at least the newer cards. This is not important of you play at 1024x768, but if you play at 1600x1200 like me, it makes quite a difference. I'm actually contemplating installing the 9700 in my tualatin rig because of this - yes, it will be severly CPU bound, but DK2 at 1600x1200 will see a notable improvement.

The ATi RV360 cards and the X800 series are AGP 8x only, witch restricts them to newer platforms - but you can easily built a win98 socket A / 754/ 939 / 775 machine and use one of these cards w/o hassle. For older machines the FX5900XT and the R300/RV350 series stand out, but 9700/9800 cards are getting rarer and rarer, not to mention expensive. Reliability-wise I'd say the FX5900XT is your best bet, since the 9800 series gets pretty hot and is not very well cooled. They also use more power then the 59xx series from nvidia, so a beefy PSU with a strong 12v rail is required. In contrast the FX5900XT will happily chug along on a good brand name 300W PSU, while the 9800 wants at least 350w, 400 in some case, 450 for socket 478 builds.

clueless1 wrote:

The difference is in perspective. You're looking at it from the point of buying it back thenwhen it was new, then being disappointed it wasn't as fast as what you were upgrading. I am looking at it from a retro gamer's perspective. And that perspective is: a quieter, cooler, less expensive, easier to find substitute for a GF3 Ti200. 😀

Yes, because back then you could only afford to upgrade your hardware every 1-2 years and had to make sure you got the best bang for your buck. These days, you can find most GPU's from that era for pennies so you can experiment and use them however you like. You could get one of the many Radeon cards (or all of them for that matter) instead and it still wouldn't cost you more than pocket change.

kanecvr wrote:

The chinese cooler works well. They come in two sizes - one fits the Riva TNT / Geforce 2/3/4MX, the other fits the 5900/6600 and so on. They also fit equivalent ATi cards. Both coolers are very very quiet and are quite a bit better then the stock solution. They are of surprisingly good build quality too, from the heatsink to the ball bearing fan. I use them on cards with insufficient (radeon 9800XT) or defective (Club3D FX5900 / Hercules GF 3 Ti500) coolers. With the large version chinese cooler the 5900xt does about 65-67C tops. The mounting systems differ as well - the small ones use classic plastic spring backed clips, while the large one uses a bolt-trough metal system it comes with.

thanks for the answer, it gives me more confidence now about using it with the 9500, since it's the only affordable option I can find.

and thanks for the ATI results, one thing, is your 9500 running with the stock configuration (4 pipelines) or is it running with the 8 unlocked (like a 9500PRO)?

@kanecvr - mate look for a P4 platform which has a AGP 2x slot, the performance for 3dfx card will be much improved compared with AMD platforms.

Here is the P4 with V5 5500 3dmark 2001 score that I did some time back 😁

rodarkone wrote:

@kanecvr - mate look for a P4 platform which has a AGP 2x slot, the performance for 3dfx card will be much improved compared with AMD platforms.

Here is the P4 with V5 5500 3dmark 2001 score that I did some time back 😁

That's about what I get on my KT333 + Athlon XP 3200+ - a little less then that. Socket 478 only gets faster then socket A from 3.2GHz onward, and I don't own any 478 CPU's faster then that, or VIA / SiS based 478 boards with 3V AGP slots. Frankly right now the only working 478 boards I have are an Asus P4P800-VM and a couple of old SD-RAM boards that only support 400MHz FSB CPUs. All intel chipset based.

Good luck getting a Voodoo 1 to work with the cpu at 2333mhz. Mine only reliably works at under 600mhz, just blocks on the screen above that. That's on a KT133 motherboard.
You will likely have to underclock the cpu, but then that will mess up the results.

Imperious wrote:

Good luck getting a Voodoo 1 to work with the cpu at 2333mhz. Mine only reliably works at under 600mhz, just blocks on the screen above that. That's on a KT133 motherboard.
You will likely have to underclock the cpu, but then that will mess up the results.

Didn't know that. You just saved me a lot of grief, thanks 😁

Whew Radeon 7000 gets murdered. I got a bunch of them still new in box and need to find good uses for them.

Rhuwyn wrote:

Whew Radeon 7000 gets murdered. I got a bunch of them still new in box and need to find good uses for them.

Yeah it does. But keep in mind there's two versions of the radeon 7000 - one based on the radeon VE, and one based on the radeon LE - and both come with either a 64 bit or 128 bit bus. The card I tested is a simple, crippled 64 bit radeon VE witch has only half of the full R100 chip activated. Some radeon 7000 cards use a full R100 or a rv200 chip and use a full 128 bit bus - some even come with DDR - those cards would perform as well as a Radeon DDR (Radeon 7500) witch is about as fast as a 128bit Radeon 9250 - a very usable card. The SDR 128 bit bus Radeon 7000 RV200 cards are slightly faster then the radeon 7200 I tested - witch is not a bad card. Not brilliant, but not terrible either - so it's a good idea to stick one in a PC and fire up GPU-Z before making a decision. ATi cards can be very confusing.

tl;dr - Radeon 7000 types:

- radeon VE based - crippled RV100 chip, half the pipelines of a R100 or RV200 radeon core - slowest of the lot - what I tested.
- radeon LE based - uses a R100 chip - full radeon core - comes in 64 and 128 bit versions, with either DDR or SDR - two times faster then what I tested - even more if it's 128 bit / DDR
- radeon 7500 based - uses full RV200 core - full radeon core - comes in 64 and 128 bit versions, with either DDR or SDR - fastest radeon 7000 card.

Personally i've only seen Radeon 7000 cards based on the RV200 core with either DDR and a 64 bit memory bus or SDR with a 128 bit memory bus, but some people actually have and posted GPU-Z screens. What I do know is that newest radeon 7000 cards are either RV100 or RV200 based - R100 cards are pretty rare, and when they do pop up they are 128 bit and use DDR, usually clocked at 183 MHz for the core and 366 or 400MHz for the vRAM. On-board radeon 7000 cards are almost always RV100 based with the exception of radeon mobility 7000 witch are always R100 cards.

I would be curious to see exactly how a Radeon 7500 stacks up against the other ATi's. I know we can make good estimates/approximations, but I *just* built a Tualatin system with a 64MB 7500 DDR as the primary card (will be doing a Voodoo2 SLI setup with it down the road as well). Back in the day, I built a P4 Northwood system for my dad with an AIW 7500, so it seemed appropriate for the P3. Originally I had planned to put in a 9800pro, but considering I have a 9800pro in my 478 Prescott box, doing the same in the P3 didn't feel right.

ultimate386 wrote:

I would be curious to see exactly how a Radeon 7500 stacks up against the other ATi's. I know we can make good estimates/approximations, but I *just* built a Tualatin system with a 64MB 7500 DDR as the primary card (will be doing a Voodoo2 SLI setup with it down the road as well). Back in the day, I built a P4 Northwood system for my dad with an AIW 7500, so it seemed appropriate for the P3. Originally I had planned to put in a 9800pro, but considering I have a 9800pro in my 478 Prescott box, doing the same in the P3 didn't feel right.

Unfortunately I don't own a 7500... had one back in the day and it was great. Performance-wise I suppose it would go in between the radeon 7200 and the radeon 9200 (not in my tests) - the latter being ~10-15% slower (depending on clocks) then the 9000 PRO card in my tests - so - not much slower then a 9200/9250 then, and a great match for a fast pIII system.

sf78 wrote:

kanecvr wrote:

well not really... it equals the Ti4200 in some tests, and the MX 440 in others (namely 3dmark). Not really a bad card for DX8 gaming, but if I wanted to run at 1600x1200 I'd go for a Ti 4600, or the FX 5700, or the 5900XT. as for overall performance ***SPOILERS*** the 9800PRO and the X800XT beat the crap out of everything else tested so far. Maybe the 6800LE @ 16pp will fair better.

Back then we were expecting upgrades, not downgrades when a new series of cards was released. After GF4 MX was out you had to be careful not to step in the same shit twice. It's amazing how poorly this (so called) 5-series card performed against many older models. There really was no point in releasing it in the first place, who would ever want to buy something like that and for what purpose? Older, faster cards could be had for less and it didn't really cater to the gaming needs either.

the biggest reason it existed was OEM sales. Most Geforce 4 MX cards where sold in Dells, HP's, E-Machines ect, it allowed them to check yes it has a stand alone gpu, yes it can play some games, and yes its DX8 compatabile, just not compliant. The FX5200 existed for that same reason, most of those where like wise OEM cards.

candle_86 wrote:

sf78 wrote:

kanecvr wrote:

well not really... it equals the Ti4200 in some tests, and the MX 440 in others (namely 3dmark). Not really a bad card for DX8 gaming, but if I wanted to run at 1600x1200 I'd go for a Ti 4600, or the FX 5700, or the 5900XT. as for overall performance ***SPOILERS*** the 9800PRO and the X800XT beat the crap out of everything else tested so far. Maybe the 6800LE @ 16pp will fair better.

Back then we were expecting upgrades, not downgrades when a new series of cards was released. After GF4 MX was out you had to be careful not to step in the same shit twice. It's amazing how poorly this (so called) 5-series card performed against many older models. There really was no point in releasing it in the first place, who would ever want to buy something like that and for what purpose? Older, faster cards could be had for less and it didn't really cater to the gaming needs either.

the biggest reason it existed was OEM sales. Most Geforce 4 MX cards where sold in Dells, HP's, E-Machines ect, it allowed them to check yes it has a stand alone gpu, yes it can play some games, and yes its DX8 compatabile, just not compliant. The FX5200 existed for that same reason, most of those where like wise OEM cards.

In my country geforce 4 MX cards sold really well. As did FX5200 cards. Mostly because they were cheap, and back then there were very few games you could not run on the two. The 5200 was a bit worse because when it was on sale, games like NFS Underground 2 and Far Cry were launched, and frankly the 5200 could only run them or reduced resolutions / details. Particularly farcry...

NVIDIA, being evil since forever!
I actually found about the 2003 story with 44.03 drivers and cheating in 3DMark03 not too long ago. Amazing!
And who can forget 3.5GB GTX 970...and gimping older cards with newer 98SE drivers...and making the GTX 1060 in 3GB variant while cheaper 1050Ti has 4GB...and GameWorks...bleh!
I wish PC market was just like it was in 2002-2003 when ATI ruled the world!

MrMateczko wrote:

NVIDIA, being evil since forever! I actually found about the 2003 story with 44.03 drivers and cheating in 3DMark03 not too long […]
Show full quote

NVIDIA, being evil since forever!
I actually found about the 2003 story with 44.03 drivers and cheating in 3DMark03 not too long ago. Amazing!
And who can forget 3.5GB GTX 970...and gimping older cards with newer 98SE drivers...and making the GTX 1060 in 3GB variant while cheaper 1050Ti has 4GB...and GameWorks...bleh!
I wish PC market was just like it was in 2002-2003 when ATI ruled the world!

They still cripple previous generation cards with new drivers. I recently got a hold of a GTX 760 and benchmarked it with several drivers - worse performance is with newest drivers, followed by launch date drivers. The best drivers for the 7xx series cards are those from 1.5 years ago it seems. In Contrast a 7950 gains as much as 35% performance with most recent drivers over release, and performs significantly better in newer games.

As for the 970 - had one until autum last year - great card acutally. I does indeed have 4GB or usable ram - you can write 4096MB of data into it's buffer - the problem is that when going over 3.5GB the read/write speeds plummet - almost like there's a problem with the memory controller or it's not really 256 bit wide.

As for ATi... I REALLY think AMD should sell their graphics division to some other company and focus on CPUs. They don't have the R&D budget to do both... 🙁 I don't mind that AMD bought ATi, but I don't think they can make both great GPUs and CPU's.

I guess the crippling performance is only with the GTX 7xx series?

As for the AMD, they have one last shot with Ryzen/Vega. If it fails, it will be catastrophic for them.