VOGONS


VIA Apollo Pro bad AGP performance?

Topic actions

First post, by link

User metadata
Rank Newbie
Rank
Newbie

Hi. I just completed my new "oldschool" setup. I wanted to recreate my first own PC from 2000.
Here is my current setup:
- Motherboard with VIA Apollo Pro (vt82c693a) chipset (I don't really know exact model, there is nothing written on this mobo, it has 1 ISA slot and 3 RAM slots)
- XFX GeForce 2 MX 400 64MB
- 3x 128MB RAM

I feel that something is very... wrong. I don't remember having any trouble with games like Hard Truck 2, Colin 2, Adventure Pinball Forgotten Island and few other. It was smooth before, I'm sure of it. Even my friend with TNT2 had no problem with running these games. Now these games are incredibly laggy.
3dMark 2000 results in ~2800-3000 points, so it's WAY below average for this GPU. I'm using nvidia drivers v56, all chipset drivers are installed too.
Any ideas?
Thanks in advance.

Reply 4 of 27, by dionb

User metadata
Rank l33t++
Rank
l33t++

Slight correction: the VT82C693A is the ApolloPro133, not the ApolloPro (which is the 691). Not that that is too relevant for drivers or performance, the chipsets are basically identical apart from the supported bus speed.

Last edited by dionb on 2018-09-11, 17:46. Edited 1 time in total.

Reply 5 of 27, by link

User metadata
Rank Newbie
Rank
Newbie
The Serpent Rider wrote:

It's just a bad chipset. Stick to 694X/694T.

This is what I thought, but 694X/694T are ridiculously expensive (like used hi end modern MOBOs).

appiah4 wrote:

Which version if VIA Hyperion 4 in 1 do you have installed?

I have read about this (broken agp drivers on anything newer). 4.17.

Deksor wrote:

I would recommend an older Nvidia driver as well

I started from 42.

dionb wrote:

Slight correction: the VT82C693A is the ApolloPro133, not the ApolloPro (which is the 693 non-A). Not that that is too relevant for drivers or performance, the chipsets are basically identical apart from the supported bus speed.

Aren't VIA drivers universal for this chipsets?

This MOBO has AGP x2 socket (there's even a switch in bios "Enable 2x AGP: Enabled/Disabled"), I think this could be the reason of terrible performance 😢

PS. I used Windows ME in the past. I don't think it could be the reason, but maybe?
And my CPU was Celeron 800, this one is Pentium III 800 (but as far I know PIII are far better?)

Reply 7 of 27, by link

User metadata
Rank Newbie
Rank
Newbie
okenido wrote:

Did you checked the video card driver settings ? Colour mode (try 16bits instead of 32bits), "Performance/Quality" settings, Antialiasing/Filters...

I tried it. Of course, all games works much better in 16bit mode, but it still pretty bad. I'm talking about CMC Rally 2.0 or even DX7 games with resolution set to 1024x768 (1280x1024 is completely unplayable in literally all 3d games but I used only 1024x768 in past so I can't compare it)
I remember playing NFS Underground on this card with reasonable speed (my friend with TNT2 was jealous because his card gave him weird textures).
I'm installing Windows ME on second partition now (it was much more stable for me back then too) and I'll report back.
PS. My CPU is identified as Pentium III 800EB

Edit:
Same on ME.

Reply 8 of 27, by ph4nt0m

User metadata
Rank Member
Rank
Member
link wrote:
appiah4 wrote:

Which version if VIA Hyperion 4 in 1 do you have installed?

I have read about this (broken agp drivers on anything newer). 4.17.

It's recommended to use either 4.35 or the latest 5.24a. Although if the AGP status is working and there are no crashes, it shouldn't be an issue.

My Active Sales on CPU-World

Reply 9 of 27, by dionb

User metadata
Rank l33t++
Rank
l33t++
link wrote:

This is what I thought, but 694X/694T are ridiculously expensive (like used hi end modern MOBOs).

Eh? 694T is a bit hard to find, but 694X (aka ApolloPro133A) is common as muck, probably the second commonest So370 chipset after the i810. Unfortunately, performance is hardly any better than the 693A (confusingly the ApolloPro133 non-A).

Aren't VIA drivers universal for this chipsets?

They are for all Via AGP chipsets. This was just correcting an error in nomenclature.

This MOBO has AGP x2 socket (there's even a switch in bios "Enable 2x AGP: Enabled/Disabled"), I think this could be the reason of terrible performance 😢

Highly unlikely.

Doc1_image002.gif

This is with a Gf2GTS, so significantly faster/more demanding than your MX

Almost no period-correct cards even came close to the limits of AGP 1.0 (2x). Your Gf2MX is somewhat newer than the rest of the system, but is also a relatively low-end card. Its GPU and video memory bandwidths are its bottlenecks, not the AGP bus.

PS. I used Windows ME in the past. I don't think it could be the reason, but maybe?
And my CPU was Celeron 800, this one is Pentium III 800 (but as far I know PIII are far better?)

"Better" is rather meaningless without adding "Better at" something. A Celeron 800 runs at 100MHz FSB and has 128kB L2 cache. A P3-800 runs at either 100MHz or 133MHz FSB and has 256kB L2 cache. Due to the extra L2 cache, a P3-800 will always outperform a Celeron 800, although only the 800EB will do so really significantly due to the 133MHz FSB. But the extra L2 cache does come at a cost: more power consumption so running hotter. That means more noise. So if you want a quiet system, the Celeron 800 might be better...

Reply 10 of 27, by link

User metadata
Rank Newbie
Rank
Newbie
dionb wrote:
Highly unlikely. […]
Show full quote

This MOBO has AGP x2 socket (there's even a switch in bios "Enable 2x AGP: Enabled/Disabled"), I think this could be the reason of terrible performance 😢

Highly unlikely.

Doc1_image002.gif

Now I suspect my particular GF2 model or my mistake somwehere. I just watched some benchmarks and Riva TNT2 has 2x more fps in many games. Riva TNT2 is able to pass 5000 points in 3dmark 2000 1024x768x16.
It lags even in Rally Championship menu (when you are choosing the car).
This is my board. Do you know the purpose of jumpers nest to CPU socket?
mlXmMQWl.jpg

dionb wrote:

So if you want a quiet system, the Celeron 800 might be better...

The loudest part in my current setup is actually a hard drive. It's way too loud and I will replace it as soon as possible.

Reply 11 of 27, by dionb

User metadata
Rank l33t++
Rank
l33t++
link wrote:

Now I suspect my particular GF2 model or my mistake somwehere. I just watched some benchmarks and Riva TNT2 has 2x more fps in many games. Riva TNT2 is able to pass 5000 points in 3dmark 2000 1024x768x16.
It lags even in Rally Championship menu (when you are choosing the car).

A Gf2MX isn't as fast as a Gf2GTS, but it should beat the pants of any TNT2:
image018.gif
(the MX400 is basically a later rebrand of the old MX, so the figures for MX should be representative of what you could expect)

This is my board. Do you know the purpose of jumpers nest to CPU socket? https://i.imgur.com/mlXmMQWl.jpg […]
Show full quote

This is my board. Do you know the purpose of jumpers nest to CPU socket?
mlXmMQWl.jpg

With no identifying features visible/legible: no. Quite likely something to do with CPU and/or RAM bus speeds.

But that sticker between northbridge and DIMMs might have a name, or otherwise the sticker on the BIOS EEPROM, or the sticker between first and second PCI slots.

Names & jumpers aside, several capacitors on that board are bulging. It's unlikely to directly impact performance, but if you want that board to stay stable in the longer term, you need to replace all the capacitors of the types that are bulging.

The loudest part in my current setup is actually a hard drive. It's way too loud and I will replace it as soon as possible.

Ah, those very 'authentic' details. Yes, with a hard disk like that any difference in CPU fan noise won't make too much difference.

Last edited by dionb on 2018-09-11, 17:44. Edited 1 time in total.

Reply 12 of 27, by shamino

User metadata
Rank l33t
Rank
l33t

I suggest you try experimenting with older nVidia drivers.
I've never experimented with this chipset, but I have had experience with running a Geforce2 MX on the VIA MVP3 (super socket 7) chipset. There's a huge variation in performance between different nVidia drivers.

On a good setup, the K6-3 450MHz running a Geforce2 MX scores about 3000pts in 3DMark2000 at 1024x768x16 and about 2300-2500pts at 1024x768x32. These are with 16-bit Z buffer and hardware T&L.
With a P3-800EB yours should be faster, so I agree something is wrong.

For whatever it's worth, nVidia drivers that have worked for me on the MVP3 are v5.32, 8.05, 12.41. The popular v45.23, among others, ran slowly.
Whatever the issue is might apply to your chipset also. I can only guess that nVidia's attention to performance with older VIA AGP chipsets was intermittent in older versions, then completely forgotten in later versions.

AGP 1X/2X/4X hardly matters. I had to switch my MVP3 from AGP 2X to 1X in order to fix a stability issue, and the impact on benchmarks was less than 1%. On a P3-800, maybe it would be more than 1%, but still not much. And that's regarding AGP1X vs 2X. The jump from AGP 2X->4X would be even less.

Reply 14 of 27, by PARKE

User metadata
Rank Oldbie
Rank
Oldbie

Here are some 3DMark 2000 default benchmarks that I collected over time - I think(?) that the marriage VIA Apollo Pro + Gf2 MX 400 + 3DMark 2000 is just a very unhappy one.
The best score of the VIA Apollo PRO 133 below here is even some 1.000 points lower than the one that started this thread.
It may be worthwhile to figure out the best VIA 4-in-1 chipset drivers for this kind of board but they are +/-20% slower than 440BX and Apollo Pro "A" to start with, so....

ooooooooooooo
VIA Apollo PRO 133
MSI Geforce2 MX 400 64Mb // Jetway 993AN // 933/133 MHz PIII SECC // 1936 // driver 30.82
MSI Geforce2 MX 400 64Mb // Jetway 993AN // 933/133 MHz PIII SECC // 1424 // driver 45.32

ooooooooooooo
440BX
MSI Geforce2 MX 400 64Mb // ASUS P2B-S // 1 GHz Celeron FCPGA /100 // 3878
MSI Geforce2 MX 400 64Mb // Chaintech 6BJM // 850/100 MHz FCPGA // 4535

ooooooooooooo
Aladdin V
MSI Geforce2 MX 400 64Mb // ASUS P5A // AMD K6-3+ 400@550 // 3649
MSI Geforce2 MX 400 64Mb // ASUS P5A // AMD K6-3+ 400@600 // 3881

ooooooooooooo
440BX
Powercolor MX 400 32MB // Chaintech 6BJM // 850/100 MHz FCPGA // 4414

ooo
AMD Athlon
Powercolor MX 400 32MB // MSI 6167 // Athlon K7 500 MHz // 3450
ooooooooooooo

Reply 15 of 27, by okenido

User metadata
Rank Member
Rank
Member
appiah4 wrote:

The MX card may be 64bit?

I tested a GF2MX 64bit on my Pentium II 450Mhz (440bx chipset) and score was approx 3000 in 3DMark2000. I'm pretty sure it's heavily CPU limited, the op *should* get way better results with his PIII 800...

Maybe try various driver / DirectX version combinations ?

@PARKE omg those scores with VIA Apollo PRO 133 😕 😲 Would be interesting to discover what is eating MORE THAN HALF of the performance with this chipset/graphics card combination...

Reply 16 of 27, by PARKE

User metadata
Rank Oldbie
Rank
Oldbie
okenido wrote:
appiah4 wrote:

The MX card may be 64bit?

@PARKE omg those scores with VIA Apollo PRO 133 😕 😲 Would be interesting to discover what is eating MORE THAN HALF of the performance with this chipset/graphics card combination...

>>
>>
Omg, indeed 😉. But it is the combination with 3DMark2000 that seems to be the main hurdle. The pc with the 939AN+Via Apollo Pro chipset / 933Mhz/133 cpu / MSI Gf2 MX400 is standing next to me right now and the 3DMark99 outcome is much more in line with what to expect - see below.
Also noteworthy is that the cpu benchmark outcomes in my tests with 3DMark2000 are also -way- below what they should be - like value +/-100 instead of +/-400 for a 933Mhz chip.

VIAtest.jpg
Filename
VIAtest.jpg
File size
165.66 KiB
Views
2795 views
File license
Fair use/fair dealing exception

Reply 17 of 27, by link

User metadata
Rank Newbie
Rank
Newbie
shamino wrote:

12.41.

Tried this and it is A BIT better, but still terrible. All my games are crashing pretty often additionally. Same on 21.83.

appiah4 wrote:

The MX card may be 64bit?

I don't really know how to check it. There is only 1 sticker on this GPU, XFX hologram.
Some snaps from Sandra (now I know who was the manufacturer at least - Shuttle HOT-687V):
https://imgur.com/a/MbMFpvN

PS.
Am I missing something, or my board is missing one capacitor? Take a look on my photo above and:
https://www.shuttle.eu/_archive/older/de/687v.htm
(above CPU)
Could it be the reason?

Reply 18 of 27, by ph4nt0m

User metadata
Rank Member
Rank
Member
link wrote:
PS. Am I missing something, or my board is missing one capacitor? Take a look on my photo above and: https://www.shuttle.eu/_arc […]
Show full quote

PS.
Am I missing something, or my board is missing one capacitor? Take a look on my photo above and:
https://www.shuttle.eu/_archive/older/de/687v.htm
(above CPU)
Could it be the reason?

Sigh, your board has issues with many more than one capacitor in fact. I see at least 5 of them bulged. Must have a recapping job.

My Active Sales on CPU-World

Reply 19 of 27, by okenido

User metadata
Rank Member
Rank
Member

Find the reference of the memory chips on your card (by just looking at them). Search google to find if they are 16 or 32 bit each (most of time there is a "16" or "32" in the reference so you don't even need searching)

Multiply by the number of chips on your card and you have the total.

Most of time cards with 8 memory chips are 128 bit and those with 4 are 64bit if they don't have the big chips with legs on the four sides. Most low-profile cards are 64 bit too.