. @OpenGL4ever It's 4 years ago, so I don't recall the details at this point. But I can't think of any reason why I would have even had different versions, let alone used different versions.
One common mistake about PCI cards: the voltage selection changes the signal voltage, not the power supply voltage. The slot provides BOTH 5V and 3.3V no matter what variant.
Would be interesting which early vga-cards give stutterfree scroller on menace or other bitmap-brother sidescrollers. my schneider tower at 286/10/ega stutters, but A2286 with ET4000 is fine...not stuttering.
Yes, that's possible. The problem with early games is that many don't have framerate counters and many games are locked to a given speed. So one is limited to the very small number of games where one can measure framerate, that aren't locked and that are not entirely dependent on CPU rather than graphics. For this reason, it is much easier to measure graphics performance with benchmarks designed for that purpose.
@@PCRetroTech I have another idea. Most games barely achieved more than 20 fps under DOS when they were new because the developers immediately put any additional performance of the CPUs available at the time into better 3D features. That's why most games rarely ran faster than 20 fps. If it was 23 fps, the CPU was already great. That's why instead of relying on the fps information from benchmarks, you could choose a different approach and simply film the screen with a camera. Most cameras record at least 25 fps, so any jerking should be noticeable. The screen and the camera would have to run synchronously and then at the end you either have to compare the video material directly or evaluate it using some method to draw conclusions about the fps. As a CPU I would suggest a 486DX 33 MHz and a 486DX2 66 MHz. When it comes to games that have a built-in benchmark mode, Wing Commander 3 comes to mind. For Doom, the engine was released, so an fps display could have been added later. That would make it two games with an integrated benchmark function that were on the market around 1994 when VLB was an issue. Other games with inbuilt fps counter are Indycar Racing II and Archimedean Dynasty.
I wonder if it would be possible to create a video card for one of these old systems like with an FPGA or some discrete logic that could outperform all these old cards. The kind of thing to create a high watermark for the test bench to know if the card is being limited by the platform or not.
That sort of thing should be possible. You won't get it too much faster though. You'd drop some wait states for sure, but you'd still be limited by the bus speed. You could probably create such a card in era-correct chips. I've thought about doing something like this one day.
@@PCRetroTech Being limited by the bus speed would be the point. If you had this theoretical Uber card and it still generated 62.5 fps in the benchmark you'd know that you haven't tapped even these card's full potential yet.
@@Jackpkmn You could run a test with a Voodoo 3 2000 PCI. But not because the Voodoo 3 has advanced 2D and 3D accelerator functions that would be useless under normal VGA, but because it was one of the last chips where they had to do a completely new implementation of VGA. The predecessors were actually pure 3D accelerator cards and did not have a 2D VGA mode. This means with a Voodoo 3 you have a VGA-capable chip without the legacy of other manufacturers, and thus modern technology like SDRAM and it also should properly supports the VESA standards including all errata.
Performance is not so much important for me on early ISA cards. Image quality is a big issue (jailbars!), compatibility to old software (ega-special tricks (BeverleyHillsCop), scrolling issues), shifted image on TFT between vga-textmode, ega/vga 16 color lowres modes and vga 256color lowres-mode. all et4000 shift vga 256 to the right compared to vga textmode.
.. The speed of the processor matters a lot for superscape.. Superscape uses the processor to calculate the 3d rotations, so if the processor is slower you'll get lower results on the same bus, so unless you were comparing the VLB at 40 and the PCI bus on one of those rare cards that actually had BOTH those alternatives using the same processor, and the same briges, etc.. the bus-speed will only be a small part of it.
@rogerlundstrom6926 In that time all 3d calculations were done by the CPU. The graphic chips were passive. VLB might give a slight performance win in higher SVGA modes. In a standard VGA mode (320x200x8) it would be interesting to find out how fast the CPU has to be in order to see advantages there too.
i think in the 90s up to 1997 video cards made no difference on frames as it was all done by the CPU the video card just game you the image buffer and colours... my ISA 8900C trident card could play everything from 1991 to 1996, even red alert ran great with a 486 dx4 100 i had no idea later on i learn that the trident cards were bare budget min cards at the time that were common, i could even play duke nukem 3D the cpu and ram was the biggest difference.... later on in 2006 or so the video card was very important
Uh PCI IS running at the FSB speed (or a fraction of it). On a pentium system, if you have a 75MHz FSB, you get 37.5MHz PCI. During your last test, PCI was also running at 40MHz, hence why they had the same performance.
This does not seem to be what the references I consulted say. They say that PCI version 1 called for a maximum frequency of 33MHz. In this particular machine, there is a jumper which either runs the PCI bus at the PCI clock, or at a set fraction (1/2) of the CPU clock. Naturally I had it set on the former for this test, as the latter would have presumably been slower (20MHz). There could be a misprint in the motherboard manual of course. They could mean that the PCI bus is set to the CPU clock (rather than the PCI clock) with the jumper setting I used. This would make it a lousy PCI implementation. But the timings would certainly support this. And of course there may be motherboards which support overclocking the PCI bus. But many Pentiums have PCI version 2.1 (1995), which allows up to 66 MHz anyway. Others may just be cheap implementations or may just allow overclocking of the PCI bus beyond what the spec allowed. Of course the P54C was 1994, so I could easily imagine some motherboard manufacturers cheaping out and just using 1/2 CPU clock for PCI. But I don't think you can rely on that.
@@PCRetroTech Well yeah PCI is designed to run at 33MHz max, but it doesn't prevent you from overclocking it. To my knowledge, every PC with PCI implementation have 33MHz PCI, even when it's PCI 2.1. On the other hand, macs have 66MHz PCI. Proof of that is the i440BX chipset (that's pentium II era as you've probably guessed) which has a 1/2 and 1/3 divider for PCI, made for 66MHz FSB and 100MHz FSB. Later revisions even have a 1/4 divider for PCI so you can safely run your board at 133MHz FSB (the i440BX isn't designed for that, but when you have that revision, you're pretty much safe except for the AGP which gets overclocked to 83MHz) They didn't cheap out, it's pretty much how intel, and the other chipset manufacturers implemented PCI 2.x.
@@DxDeksor There were some raid cards that ran at 66MHz, but you are right in that most consumer PCs ran everything on the PCI bus at 33MHz. Mostly it was server boards that supported 66MHz cards. There were boards that ran PCI at a fraction of CPU and there were boards that independently clocked PCI at 33MHz. So we are both right in some sense. It seems the spec was abused quite a bit. I had simply assumed that my board independently clocked it, as FSB can go up to 50MHz on this board. But maybe that is not correct and it was intended that you run it at CPU/2 when using that FSB.
@@PCRetroTech yeah probably, because 50MHz PCI would be pretty unsustainable for many PCI cards designed for 33MHz. I'm not even sure that all PCI 2.1 cards can really handle 66MHz. Some sure do (I've read that 3Dfx Voodoo cards can, and since PC 3dfx cards can run in macs, I trust that statement). But anyways, unless you're using a super high-end server, you'll never see PCI 66 in a PC. (same for 64 bit PCI) so I think you can simplify and say that 99% of PCs have 32 bit PCI made for 33MHz max.
"Retro" is something styled to look older than it is. The PT Cruiser was a retro design but is now vintage. A NES mini is a retro system that plays vintage games. From wiki. en.wikipedia.org/wiki/Vintage_(design) There is debate over what determines if an item is vintage. Some rely on the definition of anything old and of value. Others create a more limited bracket dictating only objects old enough to have been "used", but not more than two generations old.[4] The terms vintage, retro, and antique are oftentimes used interchangeably and have some overlay, however the words possess different meanings. Retro refers to styles of a previous era, while vintage refers to an older object that contains important value, and lastly antique refers an item of the previous era or at least 100 years old.[5][1][6] A related term is antiquity, which indicates something of past eras, or simply put, ancient.
@@extrameatsammich Retrocomputing is not computing using parts styled to look older than they are [1]. I think the distinction you are trying to make is pedantic and doesn't reflect common usage. If you are an antiques appraiser, there might be a usefulness in these definitions. But for the ordinary person, retro can refer to 1) a time period from 20-40 years ago (the time period probably varies depending on what type of item it is and any significant periods in the history of such items), 2) an item which imitates OR which actually comes from such a period (but is used in a later period) or 3) solely restricted to a modern imitation of an item from that era. For me, vintage computing evokes computers from before the IBM PC, though I wouldn't complain if someone called a 30 year old part vintage. The way I use the word, these parts are retro (though not retro-styled, as that would probably be more specific to imitation) in the same way that my hobby is retro. They are parts for a hobby in a later era that imitates computer usage in a former era. If I had said the parts were retrocomputing parts you would not have complained. My usage is merely an abbreviation of this for common speech. [1] en.wikipedia.org/wiki/Retrocomputing
@@extrameatsammich As I was bored I decided to look a bit more into this vintage/retro distinction, as it is the first I've heard of it. Apparently the word comes directly from Latin and simply means backwards. The Miriam-Webster Dictionary defines retro as "relating to, reviving, or being the styles and especially the fashions of the past : fashionably nostalgic or old-fashioned." Thus imitation is not an essential part of the definition. I accept that it is my use of these parts in the present that makes them retro. But I see no reason to call these vintage components instead of retro ones. The term is widely used as an adjective to describe early PC parts. And I am absolutely certain it does not cause any confusion. Naturally I have no problem with antiques appraisers using the word vintage if it means something special to them, and likewise no objection if the prefer retro in other contexts, though it seems to me that to some of them retro means items from 20-40 years in the past, rather than an imitative style.
@@extrameatsammichVintage is special podcathegory of retro. where you specify, it is specially old, so you don't expect like only 10 year old stuff, but more like 40-100 year old stuff. You are not obligated to call it vintage, when you don't need to specify it is "very old". You can still call it Retro Hardware. So all Vintage retro hardware belongs to cathegory Retro hardware. That's it, why he didn't make any mistake call it Retro hardware.
Just one thing to point out, the 16 bit isa came with the IBM AT, not with the 286 since some 286 where XT systems.
Yes that's true. I glossed over that a bit. I am sure there are other inaccuracies as well. Thanks for pointing this one out.
Try to shift the Roms on the pci cards.. if they are the same.. maybe you get same results but shifted..
Nice idea.
@@PCRetroTechDid you use the same version of the software and VESA VBE drivers in the first and last test?
. @OpenGL4ever It's 4 years ago, so I don't recall the details at this point. But I can't think of any reason why I would have even had different versions, let alone used different versions.
One common mistake about PCI cards: the voltage selection changes the signal voltage, not the power supply voltage. The slot provides BOTH 5V and 3.3V no matter what variant.
Would be interesting which early vga-cards give stutterfree scroller on menace or other bitmap-brother sidescrollers. my schneider tower at 286/10/ega stutters, but A2286 with ET4000 is fine...not stuttering.
I have an ISA Video 7 that does 15 fps in Quake.... I have no idea why
@PCRetroTech
Could you make a comparison between a VLB and ISA card and show in which games the VLB is faster due to its higher bandwidth?
Yes, that's possible. The problem with early games is that many don't have framerate counters and many games are locked to a given speed. So one is limited to the very small number of games where one can measure framerate, that aren't locked and that are not entirely dependent on CPU rather than graphics. For this reason, it is much easier to measure graphics performance with benchmarks designed for that purpose.
@@PCRetroTech I have another idea. Most games barely achieved more than 20 fps under DOS when they were new because the developers immediately put any additional performance of the CPUs available at the time into better 3D features. That's why most games rarely ran faster than 20 fps. If it was 23 fps, the CPU was already great.
That's why instead of relying on the fps information from benchmarks, you could choose a different approach and simply film the screen with a camera. Most cameras record at least 25 fps, so any jerking should be noticeable.
The screen and the camera would have to run synchronously and then at the end you either have to compare the video material directly or evaluate it using some method to draw conclusions about the fps.
As a CPU I would suggest a 486DX 33 MHz and a 486DX2 66 MHz.
When it comes to games that have a built-in benchmark mode, Wing Commander 3 comes to mind. For Doom, the engine was released, so an fps display could have been added later.
That would make it two games with an integrated benchmark function that were on the market around 1994 when VLB was an issue. Other games with inbuilt fps counter are Indycar Racing II and Archimedean Dynasty.
I wonder if it would be possible to create a video card for one of these old systems like with an FPGA or some discrete logic that could outperform all these old cards. The kind of thing to create a high watermark for the test bench to know if the card is being limited by the platform or not.
That sort of thing should be possible. You won't get it too much faster though. You'd drop some wait states for sure, but you'd still be limited by the bus speed. You could probably create such a card in era-correct chips. I've thought about doing something like this one day.
@@PCRetroTech Being limited by the bus speed would be the point. If you had this theoretical Uber card and it still generated 62.5 fps in the benchmark you'd know that you haven't tapped even these card's full potential yet.
@@Jackpkmn I guess so.
@@PCRetroTech
Im doing something wildly different. Maybe it would be interesting. Its a lot of work and taking all my free time.
@@Jackpkmn You could run a test with a Voodoo 3 2000 PCI. But not because the Voodoo 3 has advanced 2D and 3D accelerator functions that would be useless under normal VGA, but because it was one of the last chips where they had to do a completely new implementation of VGA. The predecessors were actually pure 3D accelerator cards and did not have a 2D VGA mode. This means with a Voodoo 3 you have a VGA-capable chip without the legacy of other manufacturers, and thus modern technology like SDRAM and it also should properly supports the VESA standards including all errata.
chip quality ram differences board quality and trace layout would be my assumption as to why there would be any difference.
The clock rate is probably far too low to make trace route play a big role. The use of cache might be a viable reason.
Performance is not so much important for me on early ISA cards. Image quality is a big issue (jailbars!), compatibility to old software (ega-special tricks (BeverleyHillsCop), scrolling issues), shifted image on TFT between vga-textmode, ega/vga 16 color lowres modes and vga 256color lowres-mode. all et4000 shift vga 256 to the right compared to vga textmode.
I'm surprised you haven't made a video looking into Hercules yet, planning to do?
Not many PCI2.0 Cards can do 66MHz and also 66MHz capable Mainboards are rare and mostly to be found in Workstations and Servers
Tseng-6000 video card, one of the fastest 2D cards. thanks to fast memory with 128bit bus.
.. The speed of the processor matters a lot for superscape.. Superscape uses the processor to calculate the 3d rotations, so if the processor is slower you'll get lower results on the same bus, so unless you were comparing the VLB at 40 and the PCI bus on one of those rare cards that actually had BOTH those alternatives using the same processor, and the same briges, etc.. the bus-speed will only be a small part of it.
Well it's a while since I made this video, so I don't remember the details, but I should hope I used the same processor!
@rogerlundstrom6926
In that time all 3d calculations were done by the CPU. The graphic chips were passive. VLB might give a slight performance win in higher SVGA modes. In a standard VGA mode (320x200x8) it would be interesting to find out how fast the CPU has to be in order to see advantages there too.
i think in the 90s up to 1997 video cards made no difference on frames as it was all done by the CPU the video card just game you the image buffer and colours... my ISA 8900C trident card could play everything from 1991 to 1996, even red alert ran great with a 486 dx4 100 i had no idea later on i learn that the trident cards were bare budget min cards at the time that were common, i could even play duke nukem 3D the cpu and ram was the biggest difference.... later on in 2006 or so the video card was very important
Uh PCI IS running at the FSB speed (or a fraction of it). On a pentium system, if you have a 75MHz FSB, you get 37.5MHz PCI.
During your last test, PCI was also running at 40MHz, hence why they had the same performance.
This does not seem to be what the references I consulted say. They say that PCI version 1 called for a maximum frequency of 33MHz. In this particular machine, there is a jumper which either runs the PCI bus at the PCI clock, or at a set fraction (1/2) of the CPU clock. Naturally I had it set on the former for this test, as the latter would have presumably been slower (20MHz). There could be a misprint in the motherboard manual of course. They could mean that the PCI bus is set to the CPU clock (rather than the PCI clock) with the jumper setting I used. This would make it a lousy PCI implementation. But the timings would certainly support this.
And of course there may be motherboards which support overclocking the PCI bus. But many Pentiums have PCI version 2.1 (1995), which allows up to 66 MHz anyway. Others may just be cheap implementations or may just allow overclocking of the PCI bus beyond what the spec allowed.
Of course the P54C was 1994, so I could easily imagine some motherboard manufacturers cheaping out and just using 1/2 CPU clock for PCI. But I don't think you can rely on that.
@@PCRetroTech Well yeah PCI is designed to run at 33MHz max, but it doesn't prevent you from overclocking it. To my knowledge, every PC with PCI implementation have 33MHz PCI, even when it's PCI 2.1. On the other hand, macs have 66MHz PCI.
Proof of that is the i440BX chipset (that's pentium II era as you've probably guessed) which has a 1/2 and 1/3 divider for PCI, made for 66MHz FSB and 100MHz FSB. Later revisions even have a 1/4 divider for PCI so you can safely run your board at 133MHz FSB (the i440BX isn't designed for that, but when you have that revision, you're pretty much safe except for the AGP which gets overclocked to 83MHz)
They didn't cheap out, it's pretty much how intel, and the other chipset manufacturers implemented PCI 2.x.
@@DxDeksor There were some raid cards that ran at 66MHz, but you are right in that most consumer PCs ran everything on the PCI bus at 33MHz. Mostly it was server boards that supported 66MHz cards. There were boards that ran PCI at a fraction of CPU and there were boards that independently clocked PCI at 33MHz. So we are both right in some sense. It seems the spec was abused quite a bit. I had simply assumed that my board independently clocked it, as FSB can go up to 50MHz on this board. But maybe that is not correct and it was intended that you run it at CPU/2 when using that FSB.
@@PCRetroTech yeah probably, because 50MHz PCI would be pretty unsustainable for many PCI cards designed for 33MHz. I'm not even sure that all PCI 2.1 cards can really handle 66MHz. Some sure do (I've read that 3Dfx Voodoo cards can, and since PC 3dfx cards can run in macs, I trust that statement). But anyways, unless you're using a super high-end server, you'll never see PCI 66 in a PC. (same for 64 bit PCI) so I think you can simplify and say that 99% of PCs have 32 bit PCI made for 33MHz max.
There were no Pentiums with 75 FSB, just 50, 60 or 66,67.
These are vintage cards, not retro.
They were made in the 1990's, so not yet 40 years old but more than 20 years ago, so they are retro AND vintage.
"Retro" is something styled to look older than it is. The PT Cruiser was a retro design but is now vintage. A NES mini is a retro system that plays vintage games.
From wiki.
en.wikipedia.org/wiki/Vintage_(design)
There is debate over what determines if an item is vintage. Some rely on the definition of anything old and of value. Others create a more limited bracket dictating only objects old enough to have been "used", but not more than two generations old.[4]
The terms vintage, retro, and antique are oftentimes used interchangeably and have some overlay, however the words possess different meanings. Retro refers to styles of a previous era, while vintage refers to an older object that contains important value, and lastly antique refers an item of the previous era or at least 100 years old.[5][1][6] A related term is antiquity, which indicates something of past eras, or simply put, ancient.
@@extrameatsammich Retrocomputing is not computing using parts styled to look older than they are [1]. I think the distinction you are trying to make is pedantic and doesn't reflect common usage. If you are an antiques appraiser, there might be a usefulness in these definitions. But for the ordinary person, retro can refer to 1) a time period from 20-40 years ago (the time period probably varies depending on what type of item it is and any significant periods in the history of such items), 2) an item which imitates OR which actually comes from such a period (but is used in a later period) or 3) solely restricted to a modern imitation of an item from that era.
For me, vintage computing evokes computers from before the IBM PC, though I wouldn't complain if someone called a 30 year old part vintage. The way I use the word, these parts are retro (though not retro-styled, as that would probably be more specific to imitation) in the same way that my hobby is retro. They are parts for a hobby in a later era that imitates computer usage in a former era. If I had said the parts were retrocomputing parts you would not have complained. My usage is merely an abbreviation of this for common speech.
[1] en.wikipedia.org/wiki/Retrocomputing
@@extrameatsammich As I was bored I decided to look a bit more into this vintage/retro distinction, as it is the first I've heard of it. Apparently the word comes directly from Latin and simply means backwards. The Miriam-Webster Dictionary defines retro as "relating to, reviving, or being the styles and especially the fashions of the past : fashionably nostalgic or old-fashioned." Thus imitation is not an essential part of the definition.
I accept that it is my use of these parts in the present that makes them retro. But I see no reason to call these vintage components instead of retro ones. The term is widely used as an adjective to describe early PC parts. And I am absolutely certain it does not cause any confusion.
Naturally I have no problem with antiques appraisers using the word vintage if it means something special to them, and likewise no objection if the prefer retro in other contexts, though it seems to me that to some of them retro means items from 20-40 years in the past, rather than an imitative style.
@@extrameatsammichVintage is special podcathegory of retro.
where you specify, it is specially old, so you don't expect like only 10 year old stuff, but more like 40-100 year old stuff.
You are not obligated to call it vintage, when you don't need to specify it is "very old".
You can still call it Retro Hardware.
So all Vintage retro hardware belongs to cathegory Retro hardware.
That's it, why he didn't make any mistake call it Retro hardware.