I suppose VLB is considered faster because of the specification. PCI above 33MHz is out of spec, while VLB goes up to 40 or 50 MHz.

firage wrote:

I suppose VLB is considered faster because of the specification. PCI above 33MHz is out of spec, while VLB goes up to 40 or 50 MHz.

Well, correct me if I'm wrong but if your vlb cards are stable enough @40 Mhz bus with no WS added (for memory and its timings as well) it's indeed the case but @50 Mhz very few cards would support that and with WS from both mb and cards added there will be no benefit at all.

firage wrote:

I suppose VLB is considered faster because of the specification. PCI above 33MHz is out of spec, while VLB goes up to 40 or 50 MHz.

I think it's also because VLB is a direct interface to the 486 socket. PCI requires a PCI controller between the CPU and the cards.
Especially on 486 systems, this introduced some overhead, so performance wasn't always that great. Depends a bit on the chipset and card used as well, I suppose. Bad PCI chipsets gave PCI a bad rap on 486.

What 486 motherboards are PCI only? I've only ever seen them with ISA/PCI, ISA/VLB or ISA/VLB/PCI.

sliderider wrote:

What 486 motherboards are PCI only? I've only ever seen them with ISA/PCI, ISA/VLB or ISA/VLB/PCI.

I doubt there are any. ISA support was very important even into the Pentium II era.

sliderider wrote:

What 486 motherboards are PCI only? I've only ever seen them with ISA/PCI, ISA/VLB or ISA/VLB/PCI.

I don't think any are. I am using "PCI-only" and "VLB-only" to shorten the names. They all have ISA.

vetz wrote:

It would be interesting if you could get hold of a 964 or 968 PCI card. I believe the differences you're seeing between SIS471 and 496 is mainly due to the VRAM and DRAM of the graphic cards and that Trio64 is newer and faster than the Vision series. I've gotten similar performance differences on my two VLB cards running in my Asus VL/I-486SV2GX4 rev. 2.0 with 1024kb cache, the DRAM based S3 Vision864 (Number Nine GXE#64 with 2MB) and the VRAM based Diamond Stealth64 VRAM - S3 Vision968 4 MB (the same card you have). I can post the benchmarks when I get home from work late tonight 😀

I can also switch out the CPU to the AMD 5x86 @ 160mhz

Another member who has the VLB Trio64 and VLB S3 968 tested the two with an Am5x86-160 and it was determined that the Trio64 was faster in DOS (~4-8% faster) and the 968 was faster in Windows (~20-80% faster).

What was the result of your 864-DRAM vs. 968-VRAM comparison on the X5-160?

Seems like I have my X5-160 inserted in my UMC PCI board which is packed away. I'll need to find it in the shed another day. Anyway, did some benches on the POD.

Questions:
1. Did you use Phil's VGA benchmark settings for Doom and Quake? These settings run the benchmarks in fullscreen. I always bench according to these settings as they are becoming the norm.
2. PCPBench non-vga, is it mode 100 or 101?
3. Did you use synchronize or transparent on the Local Bus Ready setting?

Turns out I don't have a Vision968 from Diamond after all, it's an earlier Diamond Stealth 64 Vision964 card! Like this one:
http://www.amoretro.de/wp-content/uploads/dia … 3_vision964.jpg

Here are my BIOS settings (which are the fastest on the POD). For this test to get a comparable result I had to switch Local Bus Ready to Synchronize. The Vision964 card could not complete the tests if it was set to Transparent:

If you're going to run an overclocked POD in this ASUS motherboard, then L1 cache must be set to Write-through, or else Soundblaster cards won't work for some strange reason. It also provides the best performance :)

SYSTEM:
Asus VL/I-486SV2GX4 rev. 2.0 with 1024kb cache
Pentium 83@100
1x32MB of FPM 60ns:

Number Nine 9#GXE (S3 Vision 864) 2MB DRAM (Pic)
Local Bus Ready: Synchronize (Transparent)
3DBench: 91.1 (96.7)
PCPBench (VGA):: 27.6 (27.8)
PCPBench (640x400 mode 100):: 11.4 (11.4)
PCPBench (640x480 mode 101):: 10 (10)
DOOM with Phil's settings: 1421 realticks (52,6 FPS)) (1325 realticks (56,4 FPS))
DOOM with UI: 1351 realticks (55,3 FPS) (1264 realticks (59 FPS))
Quake: 23.3 FPS (23.6 FPS)

Diamond Stealth64 (S3 Vision964) 2MB VRAM
3DBench: 91.1
PCPBench (VGA):: 27.1
PCPBench (640x400 mode 100):: Could not run without TSR
PCPBench (640x480 mode 101):: 10.3
DOOM: 1420 realticks (52,6 FPS)
Quake: 23.1

All in all, very small differences between the DRAM and VRAM, but there is a small gain to the DRAM card.

feipoa wrote:

Very intersting. Thank you for this. My Trio64 VX2/DX PCI results were better than the VLB results. I wonder if the VX2/DX improved upton the original Trio inasmuch as 3DBench, Topbench, and Doom are concerned. DO you have a Trio VX2/DX card? If so, could you run these benchmarks again with the Trio64 VX2/DX card?

I don't. I have been using a Virge while I wait to repair the TSeng board this machine is meant to use though, and it doesn't yield much difference. I also have a Trio 64 V+ which is the same as the Virge, briefly tested with something else - can't remember what, maybe a G220 - and whilst it did skew results more towards 3DBench / Doom, it still felt the same overall.

One thing I want to know is how you guys get such high scores, given I've used some of the same boards and could never make them do that. I suppose it could be that I tweak for different applications instead of for benchmarks. The only time I ever caught up (and indeed, beat everything else to death) was on my VLB system, I seem to have a strange affinity with that.

Vetz, I did not use Phil's benchmark settings. You are comparing your own graphics card to another of your own, so it shouldn't be necessary to be consistent with my settings provided that you are consistent with your own tests. I ran Doom and Quake as they are with whatever given defaults are in place. I am using the demo versions of DOOM and Quake as it seems that later revision non-demo versions yield different scores. I used PCPBench Vesa mode 100. I needed to use "Synchronize" for the 968; "Transparent" w/968 would not boot. My Mach64 card, on the other hand, works fine with "Transparent".

I gave up on the POD on this motherboard because it does not allow for L1:WB mode when using a SCSI VLB controller card. I will stick with the X5-160. The POD100 runs beautifully in an MB-8433UUD version 2+.

Vetz: Could you test your cards in Windows to see if the VRAM card came out ahead of the DRAM card? In Windows 3.11, I have been using WinTune 2.0, Windows Speed, Speedy, and WinBench96 - Graphics Winmark, with the S968 VRAM card squashing the Trio64 DRAM card.

feipoa wrote:

Vetz, I did not use Phil's benchmark settings. You are comparing your own graphics card to another of your own, so it shouldn't be necessary to be consistent with my settings provided that you are consistent with your own tests. I ran Doom and Quake as they are with whatever given defaults are in place. I am using the demo versions of DOOM and Quake as it seems that later revision non-demo versions yield different scores. I used PCPBench Vesa mode 100. I needed to use "Synchronize" for the 968; "Transparent" w/968 would not boot. My Mach64 card, on the other hand, works fine with "Transparent".

That explains your high Doom scores on the X5-160. I was comparing them to Phil's settings. As I recall the default settings in Doom and Quake is with UI and "window frame", both which increase framerate. I use the same versions as you do. Interesting that your 968 card would not boot with Transparent setting. Could it be tied to the VRAM? Is your Mach64 VRAM or DRAM?

feipoa wrote:

I gave up on the POD on this motherboard because it does not allow for L1:WB mode when using a SCSI VLB controller card. I will stick with the X5-160. The POD100 runs beautifully in an MB-8433UUD version 2+.

Yes, that is the drawback with the POD and this board. With a modern IDE drive it doesnt matter much for gaming. I only keep this system the way it is to maximize performance with the Creative 3D Blaster VLB. The games that the card supports runs better with the POD than the X5-160.

feipoa wrote:

Vetz: Could you test your cards in Windows to see if the VRAM card came out ahead of the DRAM card? In Windows 3.11, I have been using WinTune 2.0, Windows Speed, Speedy, and WinBench96 - Graphics Winmark, with the S968 VRAM card squashing the Trio64 DRAM card.

Yup, but let me switch the CPU first.

I beleive my Mach64 is DRAM. It is shown here, Benchmarks for high-end VLB graphic cards - ET4000W32P/ARK1000VL/Trio64/S3 Vision968/Mach64

Are you able to run the POD in WB mode with the Creative 3D Blaster VLB when using an IDE HDD?

feipoa wrote:

I beleive my Mach64 is DRAM. It is shown here, Benchmarks for high-end VLB graphic cards - ET4000W32P/ARK1000VL/Trio64/S3 Vision968/Mach64

That is the DRAM version, or the ATI Graphics Xpression as it was marketed.

feipoa wrote:

Are you able to run the POD in WB mode with the Creative 3D Blaster VLB when using an IDE HDD?

Yeah, no problem. With a newer Samsung IDE drive I don't notice any difference compared to the older SCSI drive I had with the X5-160. The games load just as fast. PIO mode is 4 times slower than the VL-bus in theory, but I don't notice it in practice. The type of drive plays a bigger role.

The 3D Blaster is not very picky, it initiates no matter what. 40mhz FSB is also not a problem, haven't tested 50mhz though.

vetz wrote:

feipoa wrote:

Vetz: Could you test your cards in Windows to see if the VRAM card came out ahead of the DRAM card? In Windows 3.11, I have been using WinTune 2.0, Windows Speed, Speedy, and WinBench96 - Graphics Winmark, with the S968 VRAM card squashing the Trio64 DRAM card.

Yup, but let me switch the CPU first.

Found this benchmark from PC Mag in December 1994. It features most of the cards mentioned here:
https://books.google.no/books?id=PITtFPwTaWwC … diamond&f=false (page 268)
The Diamond Stealth 64 VRAM came out on top in the VLB test for Windows, even faster than boards featuring the same chipset.

I own the Orchid Kelvin 64, and the Diamond Stealth 32 and some other VLB cards (CD5428, ET4000/W32p and W32i), so I'll bench them as well for the heck of it.

Unfortunately, most of those benchmarks are for Windows. For Windows-based benchmarks on a VLB/ISA system, the Diamond Stealth 64 Video VRAM (S3 Vision 968) was the card to have. I wish it was a little faster in DOS. Not that it is slow, but the fact that it is not the fastest in DOS as well bothers me from an 'ultimate perfection' standpoint.

feipoa wrote:

Unfortunately, most of those benchmarks are for Windows. For Windows-based benchmarks on a VLB/ISA system, the Diamond Stealth 64 Video VRAM (S3 Vision 968) was the card to have. I wish it was a little faster in DOS. Not that it is slow, but the fact that it is not the fastest in DOS as well bothers me from an 'ultimate perfection' standpoint.

Hehe, can't have it all 😉

Found the CPU in the shed, and I've prepared a Windows for Workgroups bootable CD (no way I'm installing through all the floppies again!). Going to try keropi's patch so I can run WfW on DOS 7.1 as the computer already have Win98 installed.

Here are some numbers with a POD @100 running on a PC Chips M921 (UMC chipset, PCI only) with 16MB FPM memory and 256KB cache.

The memory speed in Speedsys and Cachechk is OKish but not great.

The PCPbench mode100 and VGAmode numbers seem close to what feipoa got.

I will see if I can get Quake and Doom installed on this system later.

Skyscraper wrote:

I will see if I can get Quake and Doom installed on this system later.

You can download Phil's benchmark and get it all in one nice package: Phil's Ultimate VGA Benchmark Database Project

Additionally, mine uses the same metric but has a nicer menu (as well as a few silly easter eggs.) because I was bored. Might still cause the screen to jump when making selections though and I currently cannot test.

Currently stored here; https://www.mediafire.com/?bzo4pdyv2ch3lqh

Looks like this;

The keys are a bit weird and earlier I did remap the arrows and return, but I figured that would be bad if something crashed as you would then have to "un-map" them manually or reset the machine. Plus some ANSI implementations don't allow remapping. Otherwise it needs a few more tweaks before it gets a proper release, but it is functional now, only really minor tidying up left to do.

vetz wrote:

Skyscraper wrote:

I will see if I can get Quake and Doom installed on this system later.

You can download Phil's benchmark and get it all in one nice package: Phil's Ultimate VGA Benchmark Database Project

Im working on it! 😀

The system diddnt have any means for transfering files as I only had a test drive connected to it but I have now added a CD-ROM drive, a floppy drive and even a horrible mouse.

The new more or less complete test system, Im just going to add a SB16.

Here is my first Quake result, this is just a quick test for now as I need to do some other non computer related things.

I had to run Quake 1.08 (from one of my bench discs) inside Windows 95 as I got page faults in DOS (Quake 1.01 works fine), this is strangly enough often solved by installing a sound card. Running inside Windows 95 with only 16MB memory without a sound card do probably not produce a better score than running Quake in DOS with sound. The score seems OK for an UMC chipset board with only 256KB cache.

I will try to run Phils full benchmark package in DOS later when I have installed a sound card and created a CD with the benchmarks.

Quake 1.08 demo1 running inside Windows 95, POD@100, PC-Chips M921, S3 Trio64.: 21.7 FPS