VOGONS


Trident Blade3D AGP: opinions?

Topic actions

Reply 40 of 54, by 386SX

User metadata
Rank l33t
Rank
l33t
Joseph_Joestar wrote on 2022-01-14, 09:33:
That sounds promising. FF8 will give you the definitive answer since it has a paletted texture check in its configuration utilit […]
Show full quote
386SX wrote on 2022-01-14, 09:10:

Also while I don't know if it's the same texture format it also says it supports the "8-bit RGB PAL8" texture.

That sounds promising. FF8 will give you the definitive answer since it has a paletted texture check in its configuration utility.

file.php?id=123813&mode=view

Additionally, it uses paletted textures for rendering the in-game menu, which will have washed out colors on cards that don't support this feature.

file.php?id=123814&mode=view

Tested the card with Final Fantasy 8: the config tool notice the "Pass" result for the 8-bit Paletted Textures! And I can confirm the menu colors rendering is mostly similar to the Geforce2 result above posted, certainly not like the other two. The grey shades before the center brighter part is divided in four sort of darker columns until the sides. In-game backgrounds also looks quite good in colors even if I've never played this game.
Soon I'll post some game screenshots.

Reply 41 of 54, by Joseph_Joestar

User metadata
Rank l33t
Rank
l33t
386SX wrote on 2022-01-14, 12:19:

Tested the card with Final Fantasy 8: the config tool notice the "Pass" result for the 8-bit Paletted Textures! And I can confirm the menu colors rendering is mostly similar to the Geforce2 result above posted, certainly not like the other two. The grey shades before the center brighter part is divided in four sort of darker columns until the sides. In-game backgrounds also looks quite good in colors even if I've never played this game.

That's awesome! If you want to check what other texture formats the Blade 3D supports, Forsaken is also a good test. Just go to Options > Visuals > Select Texture Format.

file.php?id=123601&mode=view

BTW, if anyone's curious, here's a period correct Blade 3D review on Anandtech. Their initial impressions of the card were favorable, considering the value it provided while being quite affordable at the time.

PC#1: Pentium MMX 166 / Soyo SY-5BT / S3 Trio64V+ / Voodoo1 / OPTi 82C930 / AWE64 Gold / SC-155
PC#2: AthlonXP 1700+ / Abit KT7A / Voodoo3 / SBLive / Vortex2
PC#3: Athlon64 3400+ / Asus K8V-MX / 5900XT / Audigy1
PC#4: i5-3550P / MSI Z77A-G43 / GTX 650Ti / X-Fi

Reply 42 of 54, by 386SX

User metadata
Rank l33t
Rank
l33t

I'll try Forsaken demo. It's interesting to read those time correct review (obviously also the Vogons users reviews on youtube or their websites) to see how obvious should have been to choose and keep those cheap cards until possible and skip some video card period instead of usual expensive faster middle-end cards that most of the time would have run on low resolution CRT monitors or played at 640x480.

It's interesting anyway that the Anandtech review scored around 1400 points on 3DMark99 with a Pentium II 400Mhz while I get around 1000 points more with the Pentium III 500Mhz. I'm sure the better architecture, the SSE and the faster clock help here but it's quite a lot compared to the Rage Pro where I got 1400 points with the same P3-500. So at the end the Blade3D could still give more speed with a faster cpu and not already at its limit.

Reply 43 of 54, by Joseph_Joestar

User metadata
Rank l33t
Rank
l33t
386SX wrote on 2022-01-14, 13:27:

It's interesting anyway that the Anandtech review scored around 1400 points on 3DMark99 with a Pentium II 400Mhz while I get around 1000 points more with the Pentium III 500Mhz. I'm sure the better architecture, the SSE and the faster clock help here but it's quite a lot compared to the Rage Pro where I got 1400 points with the same P3-500. So at the end the Blade3D could still give more speed with a faster cpu and not already at its limit.

Well, in this video (around 29:20) @swaaye got almost 2500 in 3D Mark 99 Max.

It seems that he was using a much faster CPU though.

PC#1: Pentium MMX 166 / Soyo SY-5BT / S3 Trio64V+ / Voodoo1 / OPTi 82C930 / AWE64 Gold / SC-155
PC#2: AthlonXP 1700+ / Abit KT7A / Voodoo3 / SBLive / Vortex2
PC#3: Athlon64 3400+ / Asus K8V-MX / 5900XT / Audigy1
PC#4: i5-3550P / MSI Z77A-G43 / GTX 650Ti / X-Fi

Reply 44 of 54, by 386SX

User metadata
Rank l33t
Rank
l33t
Joseph_Joestar wrote on 2022-01-14, 14:35:
386SX wrote on 2022-01-14, 13:27:

It's interesting anyway that the Anandtech review scored around 1400 points on 3DMark99 with a Pentium II 400Mhz while I get around 1000 points more with the Pentium III 500Mhz. I'm sure the better architecture, the SSE and the faster clock help here but it's quite a lot compared to the Rage Pro where I got 1400 points with the same P3-500. So at the end the Blade3D could still give more speed with a faster cpu and not already at its limit.

Well, in this video (around 29:20) @swaaye got almost 2500 in 3D Mark 99 Max.

It seems that he was using a much faster CPU though.

So I suppose it reached the limit at the early Pentium III cpus.. 😁 (anyway I've read some internet modern reviews where someone reached even 2800 points but I suppose it was the Turbo version or the few models that had higher clocks, I think in the Anand review they talked about 120Mhz while my card seems clocked at 100Mhz). Anyway I was looking also his youtube game test of Quake3 with the same card and seems like it runs (even with a much powerful cpu) more or less similar with a much variable frame rate. It looks OpenGL driver remained in a sort of beta status. The driver file seems anyway updated like the others into the 2000 year but it had to be polished much more imho if the Direct3D path of the same games ran (much) faster or at least better with a stable frame rate. Too bad after the 2000, even searching in the old homepage archived driver pages, the Blade3D has been left with this drivers and not more. At least another year of updates would have been a good thing for what has been a good chip after all.

Reply 45 of 54, by LubieCipy

User metadata
Rank Newbie
Rank
Newbie

My Blade3D Turbo @135 MHz scores 3331 points in 3Dmark99 and 989 points in 3Dmark2000. Everything at standard resolution and settings using tualatin CPU

Reply 46 of 54, by 386SX

User metadata
Rank l33t
Rank
l33t

Impressive. It probably means it could get more speed increasing the main clock freq but I suppose on 250nm and that type of package the Turbo might have already been the few selected lucky cores but it might have needed probably a BGA package for more speed. While instead I suppose my card might have the not lucky cores and cheap sdram modules. Unfortunately there're not many technical information on these chips with more details.

Last edited by 386SX on 2022-01-17, 09:30. Edited 1 time in total.

Reply 48 of 54, by 386SX

User metadata
Rank l33t
Rank
l33t

Yes AGP texturing enabled but I suppose it's a right score for a time correct machine using PC100 ram and a 500Mhz Pentium3. Around 2390 points almost at 3DMark99 and more than 790 in 3DMark2000 and 4,55 in Final Reality for 100Mhz clock of the card. Here is the 3DMark2000 results.
Soon I'll post a comparison of the Rage IIC and Rage Pro with 3DMark99 and it looks like the Blade3D win in the games frame rate and fill rate but texture memory speed is much faster on the Rage Pro, I suppose cause has SGRAM (I don't remember exactly its memories clock).

Attachments

  • 3dmark2000_Blade3D.jpg
    Filename
    3dmark2000_Blade3D.jpg
    File size
    68.77 KiB
    Views
    406 views
    File license
    Fair use/fair dealing exception

Reply 49 of 54, by 386SX

User metadata
Rank l33t
Rank
l33t

What a day.. turned the computer on today and protection error at boot, tried system.ini fix and still something is wrong. I switch back to the Pentium II (considering the i440BX) and I probably reinstall the o.s... Good old Win 98 times... 🙁

EDIT: reinstalled the o.s. but with the Pentium II 400Mhz even with its plastic clips on the socket it randomly see or not see the L2 cache.. I have to detach the original heatsink/fan to check that cpu so I went for another Pentium III, the early 600Mhz and last cpu supported by this mainboard. Reinstalled the Blade3D and the Vibra 128 PCI sound card. Testing right now the various benchmarks to see everythings is stable.

Interesting, the Blade3D score with 3DMark99 didn't increase (2394 points) so I suppose the video card clocks are the limit here while the CPU score increased to 9061 (along with the power watt requirement, the computer reach 72 watts only with the vga and the sound card, the disk and the drives). The 600Mhz early P3 seems quite watt demanding I hope the mainboard will be stable with this.
The 3DMark2000 scores is even a bit lower, sign that the CPU was already too much for this video chip with these clocks as suspected. It'd be interesting to find detailed specifications of the chip and the ram modules.

Reply 50 of 54, by 386SX

User metadata
Rank l33t
Rank
l33t

What I don't understand are the fill rate numbers that make me think about the core clock of the video chip might be lower than the theorical 100MT/s. Most benchmark I've run at lower resolution result in mostly 2MPolygons/s at best that I think was the supposed maximun geometry limit while the fill rate seems to be 75MT/s both single and obviously multi too. Shouldn't that value be close to the chip freq theorically in these early accelerators? Might it be the video chip freqs lower than 100Mhz? PowerStrip gives only the mem clocks and that's strange too.

From the old Trident press release: "The 3D-triangle peak rate is 2.5 million triangles per second (with backface culling); the maximum fill rate is 110 Million pixels per second."

Last edited by 386SX on 2022-01-15, 17:11. Edited 1 time in total.

Reply 51 of 54, by 386SX

User metadata
Rank l33t
Rank
l33t

Meanwhile, I'm testing the card with the game Max Payne.. I'm stressing the card with the impossible tests but I'm surprised it can render at almost high level details 640x480-16bit quite ok. Maybe at this resolution high level details for most options is too much from 15fps to 25fps I'd say, probably need medium levels for some of the options I'm going to test this. Just some few rare strange textures. Instead 3DMark2001 runs (not the Demo that gives a strange requirement error) the benchmark of course at low frames (650 points at 640x480 16bit) but insterestin it runs also with the DXT1 texture format as supposed to from the features list.

Also I tested again the old Powerdvd that seems clearly detecting and make use of the vga Trident acceleration and with the faster 600Mhz P-III the CPU usage is lower than before, with Motion Compensation and this sw engine and without any real time graph running, it's running at 35% CPU usage during a good quality MPEG2 DVD (8Mbit/s of bitrate) and for 58-60 watts of total system power during playback (and DVD drive running). I'll try eventually with a slower cpu, I was thinking to something more time correct like a P-II 350.

Reply 53 of 54, by 386SX

User metadata
Rank l33t
Rank
l33t
Putas wrote on 2022-01-16, 07:10:

The fillrate is in line with my results. The spec numbers are usually theoretical maximums hard to reach in the real world.

Thanks. 😀
Indeed theorical numbers were often much more theorical than real and sometimes also quite optimistic.. reading some press release info from your great video cards webpage is interesting to read how much they tried to sell these solutions. For example I was reading about the hardware video accelerations, lines like Trident's highly acclaimed Trident Hardware Assisted etc.. and I was wondering acclaimed by who? I couldn't find much info about it or remember someone talking about it (while the ATi video acceleration was indeed well received and discussed not only for its mostly complete acceleration but the general quality).
I can say that from my early DVD test it does a very good job this Trident chip but I suspect only Powerdvd were optimized for it not finding the others time correct trial version of the player to see if they worked with it too. And I think I've read somewhere the Powerdvd was suggested at first. And also the sw engine always was a factor for the final quality before the acceleration itself. I'd like to test those 1999/2000 demo version of sw like Cinemaster or Windvd but I can't find them.

For the game part, Max Payne which of course is a much later heavy game, impressively runs quite well and with acceptable quality at medium/high mixed video options.

Reply 54 of 54, by 386SX

User metadata
Rank l33t
Rank
l33t

Temporary I switched back to the Rage Pro Turbo and installed an old style Pentium II 350Mhz the early one with the white printed brand and name with some OEM big passive heatsink already installed on it, to have a more time correct machine also for future tests. The CPU is lighter on power usage and stays quite cold even with a passive heatsink. Compared to the Blade3D the desktop quality improves indeed but that's imho for the PCB quality difference besides ramdac or filters.
The cpu speed difference from the Pentium III is quite heavy and the DVD test clearly notice that impact. The Rage Pro Motion Compensation acceleration helps but not as often expected considering with that cpu the usage is around 80% with acceleration enabled and detected (but something is strange while sw decoding is around 95% CPU usage checking with the processes reading app in background).
Also cause for example the Blade3D press release said "the Blade 3D easily meets the film industry's full-motion video requirement for 30 frames per second and still leaves up to 50% CPU headroom on a PII 300MHz only system." This would mean the Blade3D video acceleration might do a faster job then the well known ATi one with the same player (sound strange). I still need to test again the Blade3D on this cpu to double check cause the previous tests did benefit from SSE feature and 150/250Mhz more. On the video final pixel quality I'd say maybe the ATi results in a sharper upscaling and noise stability compared to the Blade3D that did a great job indeed but lines and details seems sharper and colors vibrant with default settings on the Rage Pro. It's difficult to estimate how much but seems a bit better.
Now I try some games to see how they run with the ATi cards and after testing later the Blade3D with such slower cpu.

EDIT: To add this It's interesting with the ATi Player the same MPEG2 DVD with the same CPU runs @ 55% usage. That's quite low and what I was expecting at first (around 60%). Need to understand what's wrong with Powerdvd and the Rage Pro while the acceleration is noticed as "In Use" and lower of 10-15% CPU usage compared to sw decoding. ATi player seems to do a very good, I always thought that engine was one of the best but I think with the Rage Pro was more optimized for speed more than quality beside is still great, probably at best with the Rage 128 Pro.