Rawit wrote:

Didn't think the Savage4 would do well compared to the GeForce 2 MX400.

Neither did I!

Rawit wrote:

You might get to squeeze a tid bit more performance out of the Savage4 by using S3VBEFIX /B+.

Are these utilities for use in Windows? I thought they were mostly for DOS purposes. S3 Speed Up, for example, mentions it gets automatically disabled when booting Windows. According to the Vogons post, S3VBEFIX hasn't been tested with a Savage card.

Rawit wrote:

I would have expected the Matrox to pull ahead here because of its memory speed and my own experience with Matrox/Build engine games. Did you see any tearing when benching it?

I don't recall anymore. Too many benchmarks, too many drivers, too many tests.

The Serpent Rider wrote:

graphics core/memory = 120 / 120 MHz

That's too low for GeForce 2 MX400. It must be 200/166(200).

I had to say some Nvidia results are quite low and not any better than Pentium MMX 233 with 430TX chipset. These tests were done without sound?

For the 120/120 speed, I am only quoting what NVTweak and Powerstrip reported. Could they be wrong? As I was already finished using the GF2 in the PPRO system, I had put the GF2MX into my 430TX system, which normally uses a GF4, for the purposes of checking the core/memory speeds using NVTweak and Powertrip.

I ran GLQuake and Quake2 tests without sound. I followed the instructions you provided for Expendable/Shogo. For the other games, I left sound at default, so most likely enabled.

The Serpent Rider wrote:

Detonator 45.23 is too new.

Which version do you recommend? When I tried out a dozen drivers for my AMD K6-III, 430TX build, I determined that 45.x was suitable.

luckybob wrote:

That's why I covet my overdrive CPUs.

Interestingly, I stuck in a PII overdrive and noticed almost no improvement on the results. I have them penciled in on my rough datasheet if anyone is interested.

that was the selling point for MMX, as for reality I'm not sure it was ever utilized to anywhere near that level of benefit

Also Pentium MMX can't do MMX and FPU simultaneously.

Neither did I!

All hail Savage 4! Good thing I bought bunch of them before the "hype" 🤣

Which version do you recommend? When I tried out a dozen drivers for my AMD K6-III, 430TX build, I determined that 45.x was suitable.

Detonator 28.32 or probably even 12.10 (GF3 era drivers).

For the 120/120 speed, I am only quoting what NVTweak and Powerstrip reported.

For anything related to Nvidia and TNT+ cards, RivaTuner should be better.

I left sound at default, so most likely enabled.

I see. Althought it's still strange that Incoming has only 21fps. I need to test my Pentium MMX 233 with sound (28fps without sound and with Quadro4 PCI).

feipoa wrote:

Are these utilities for use in Windows? I thought they were mostly for DOS purposes. S3 Speed Up, for example, mentions it gets automatically disabled when booting Windows. According to the Vogons post, S3VBEFIX hasn't been tested with a Savage card.

I'm using S3VBEFIX with my Savage4 under DOS and it seems to work. Benchmarks show an increase in speed for banked VESA modes. Some games that give me "out of range" problems like Canon Fodder work with S3VBEFIX, although there are some games that refuse to run. Haven't tried it under Windows, my machine is DOS/WfW3.11.

I've been trying to find a cheap Savage4 PCI forever.. I guess I can give up now that the secret is out. Everyone knows it's a beast.

The Serpent Rider wrote:

Detonator 28.32 or probably even 12.10

I think I tested these and didn't see any speed up in the games I tested at the time with the K6-III. Perhaps I needed a broader spectrum of test games.

The Serpent Rider wrote:

For anything related to Nvidia and TNT+ cards, RivaTuner should be better.

I have located RivaTuner 2.24c, which they call "RivaTuner v2.24 MSI Master Overclocking Arena 2009 edition" If you have any older revisions that you think are better to run, could you provide a link for them?

Using RivaTuner 2.24c shows that the GF2 runs at 100 MHz core, 120 MHz memory. PowerStrip shows 120/120. NVTweak shows 120/120. So what's the truth? RivaTuner says the GF2 is SDR. Everest Home Edition says the GF2 has 120 MHz GPU, 350 MHz RAMDAC, and 120 MHz effective memory (1x120 MHz, SDR).

RivaTuner 2.24 shows that my GF4 runs at 100 MHz core, 300 MHz memory. PowerStrip shows 275/300. NVTweak shows 275/300. RivaTuner indicates that the GF4 uses DDR. Everest Home Edition states the GF4 has a 146 MHz core clock, 350 MHz RAMDAC, and 159 MHz effective memory (79MHz x 2, DDR).

The Serpent Rider wrote:

I see. Althought it's still strange that Incoming has only 21fps. I need to test my Pentium MMX 233 with sound (28fps without sound and with Quadro4 PCI).

What benchmark strategy do you use for this game? For me, I set the command to enable the instantaneous and average frame rate displays. Then I enter the game and hold down the right turn key (arrow) for exactly 60 seconds (turns your ship in circles), then record the average. Some whomever enemies are shooting at me, but the first 60 seconds is the same every time. I find the average value doesn't fluctuate a whole lot at this point.

Rawit wrote:

feipoa wrote:

Are these utilities for use in Windows? I thought they were mostly for DOS purposes. S3 Speed Up, for example, mentions it gets automatically disabled when booting Windows. According to the Vogons post, S3VBEFIX hasn't been tested with a Savage card.

I'm using S3VBEFIX with my Savage4 under DOS and it seems to work. Benchmarks show an increase in speed for banked VESA modes. Some games that give me "out of range" problems like Canon Fodder work with S3VBEFIX, although there are some games that refuse to run. Haven't tried it under Windows, my machine is DOS/WfW3.11.

Anyone know if these utilities will benefit the Savage4 in 3D mode in Windows 98se?

appiah4 wrote:

I've been trying to find a cheap Savage4 PCI forever.. I guess I can give up now that the secret is out. Everyone knows it's a beast.

There appears to be a russian seller on eBay who has a seemingly endless supply of these, but it is missing the bracket that attaches the card to the computer chassis. With any luck, one of your existing VGA cards will have the VGA hole and bracket mounts in the correct place.

What benchmark strategy do you use for this game?

Full lenght of the first demo, which barely if at all fluctuate near the end.

I have located RivaTuner 2.24c, which they call "RivaTuner v2.24 MSI Master Overclocking Arena 2009 edition"

Should not be an issue, I use it too.

Using RivaTuner 2.24c shows that the GF2 runs at 100 MHz core, 120 MHz memory. PowerStrip shows 120/120. NVTweak shows 120/120. So what's the truth? RivaTuner says the GF2 is SDR. Everest Home Edition says the GF2 has 120 MHz GPU, 350 MHz RAMDAC, and 120 MHz effective memory (1x120 MHz, SDR).

RivaTuner 2.24 shows that my GF4 runs at 100 MHz core, 300 MHz memory. PowerStrip shows 275/300. NVTweak shows 275/300. RivaTuner indicates that the GF4 uses DDR. Everest Home Edition states the GF4 has a 146 MHz core clock, 350 MHz RAMDAC, and 159 MHz effective memory (79MHz x 2, DDR).

Thats very weird, but apparently your GF2 MX400 is MX200 (64-bit bus), which is further underclocked by whatever reason.

P.S.
I think you have a terrific talent for causing unexplored bugs 🤣

That's a bummer about the GF2 not being the MX400, but you sure? The model number on the barcode on the PCB says 118PCI-64LTV, which according to their datasheet is an MX400.

When you load up Incoming and type in the command to view the average frame rate, you continue to let it autoplay? It seemed to me that it just kept on going forever, and the average frame rate kept dropping steadily. How long does this run for?

I don't type anything. There's a built-in benchmark via command file.

Jaton_Video-118PCI-64LTV_GF2MX400.pdf

I think it's 64-bit.

The Serpent Rider wrote:

I don't type anything. There's a built-in benchmark via command file.

Could you let me know how to run it?

http://alt.3dcenter.org/userbench/incoming.php

The Serpent Rider wrote:

http://alt.3dcenter.org/userbench/incoming.php

Great. I will try this out and compare the frame rate result with the method I used.

1Unfortunately, the game recalculates the average frames per second ( average fps) with every new timedemo, and unfortunately there is no such thing as a correct ending with a statistic of the achieved values. Therefore please read the average pictures per second at picture no. 4550 (the moment your own helicopter is hit) in the first demo before the second timedemo is loaded. This timedemo can be ended by the pause button. Alternatively: By pressing the Windows key (alternatively ALT + ESC ) and then ending the task in Task Manager ( CTRL + ALT + DEL ).

It is unfortunate that the timedemo doesn't stop upon completion. This sounds very similar to the autoplay that occurs when you first load the game and tell it to display the average frame rates.

I took some shots of the case while messing around with the PIIOD. My main gripe with the case is not having a dedicated 3.5" floppy bay. I ordered the Pentium case badge from eBay. Maybe one day a dual PPRO board will make its way into this case. The system boots Win98SE, NT4, and W2K.

The USB door opened. I wired the USB ports to a PCI USB 2.0 card by making cable headers from solder points on the PCB (not shown).

Inside with the PPRO 200 512K chip running at 233 MHz. Unfortunately, the 1M chip would not run reliably at 233 MHz on this motherboard. I tried two such chips. You can see the four IDE drives mounted, which are setup in RAID 10 mode using an Adaptec AAR-2400 controller with a 256 MB SDRAM stick.

The PIIOD installed. Benchmark results were surprising (discussed in the next post).

I had an extra 25 mm fan, so I added it to the Savage 4's heatsink. The heatsink is an odd size, so standard 40 mm 486 fans won't fit properly. The heatsink measures 35mm x 35mm. Digikey has several 35mm fans, but I had an extra 25 mm fan, so.. The keen eye might notice a Dreamblaster S2 on an ESS 1688 sound card. I personally think the S2 sounds just as good as the X2 for most situations, so I have installed them in my systems for which the wavetable header is not utilised.

For running Incoming in timedemo mode, one must add -gameindex -nocd -screenmode to the shortcut target. One must watch carefully to note the average FPS rate down just before the demo automatically moves on to demo2. Using this method, I record 34.7 fps when using the Savage4 card with the PPRO233. Using my previous method, I would get 29.8 fps. That should offer some idea as to how to translate between the two if you are intending to run similar benchmarks.

What I found most interesting is when I used the PIIOD CPU. I have attached a CPU-Z from instances when each CPU was installed, that is, PPRO 200 and PIIOD. CPU-Z correctly identifies them. The oddity occurs when running the benchmarks.

In GLQuake, I get 44.2 fps when using both the PPRO233 and the PIIOD-333. In Incoming, I get 34.7 fps with PPRO233 and 34.8 fps with the PIIOD. Why is this? I would have expected the PIIOD to yield significantly higher results. Should I be running Fastvid, or is this only for DOS? The latest BIOS for this motherboard correctly identifies the PIIOD.

It means the cpu isn't a bottleneck. I'd play with ram timings, if it were me.

luckybob wrote:

It means the cpu isn't a bottleneck. I'd play with ram timings, if it were me.

There aren't any. BIOS tweaking is very limited.

Intel settings for EDO aren't that bad though.

same id expect better, could be that your tests are erroneous. Id say run some other tests besides those so you can be for sure whats going on and perhaps shed some light on the problem.

Which tests would you suggest? I can run some synthetic CPU-only tests to see if it makes a differences. Perhaps 3DMark99Max as well

if not memory, the next culprit i'd look into is the PCI bus. is there an option for 30mhz?