VOGONS


EGA DOS benchmarks?

Topic actions

First post, by Ozzuneoj

User metadata
Rank l33t
Rank
l33t

I know this is pretty obscure, but I have an IBM 5150 which I have upgraded with an EGA card to get 16 color support on the IBM 5153 CGA monitor (this works as long as the resolution isn't too high). The system also has a Tiny Turbo 286 7.26Mhz upgrade card installed along with 640K memory.

I wasn't all the interested in the limited 4 color CGA games that the system could run natively (our first PC, a Tandy 1000 at least had 16 colors), and I basically already had everything I needed to upgrade it to be far more capable while still using the original monitor, so the quest began to squeeze as much as possible out of the original IBM PC.

I have noticed that some EGA games like Galaxian tend to bog down a bit and I'm considering trying a newer video card to see if it is a video bottleneck, rather than CPU. The system currently has an Everex EV659 Micro Enhancer Deluxe (8bit ISA, EGA, plus parallel). I also have another Everex EV-653 EGA card without a parallel port. I have an OAK OTIVGA 16bit ISA card that has 8bit support and has an EGA port a well. I'd like a way to compare the performance of these cards.

Anyone know of any video benchmarks with EGA support? I guess it'd also need to support running on an IBM PC, or at least a 286 with 640K.

So far the only things I've found that will run and give a video benchmark result are Checkit 3.0, and Landmark System Speed Test 6.00, but neither really show much on the display for these tests, so I don't know how well these reflect 2D gaming performance.

I guess I should also ask if anyone has tested EGA or VGA+EGA cards before and found any that perform significantly better than others when attached to an EGA display?

Now for some blitting from the back buffer.

Reply 1 of 28, by vlask

User metadata
Rank Member
Rank
Member

Try Diag, it might work (tested on 286 by developer, might work on older too)....

http://web.archive.org/web/20041216133952/htt … ei=diagbeta.zip

Not only mine graphics cards collection at http://www.vgamuseum.info

Reply 2 of 28, by Ozzuneoj

User metadata
Rank l33t
Rank
l33t

Thanks! I'll give DIAG a try. 😀

In the mean time, here are my results with the cards I have that will connect to my 5153 (by the way, the 5153 is technically a CGA monitor, but for reasons that are beyond my technical knowledge, it can display most 16 color EGA modes, aside from the highest resolutions).

Landmark Video Test:
Everex EV-659 = 413 chr/ms
Everex EV-653 = 380 chr/ms
Oak OTIVGA (OTI-037C, set to CGA) = 408 chr/ms

Checkit 3.0 Video BIOS CPS:
EV659 = 462 Char./Sec
EV653 = 1488
OTIVGA = 860

Checkit 3.0 Direct Video CPS:
EV659 = 15708 Char./Sec
EV653 = 15708
OTIVGA = 14740

I'm really not sure how to interpret these results. Looking at other benchmarks for OAK cards, they tend to be extremely slow, so this probably isn't the best representation of a later card that might improve performance. Most likely the 7Mhz 286 add on CPU is bottlenecking things, but I'd probably need more cards and more benchmarks to know for sure. I wish I could find something that at least looked like a 2D game that had a benchmark feature.

Anyone know why the BIOS speed would vary so much? What does that even mean anyway?

Now for some blitting from the back buffer.

Reply 3 of 28, by Jo22

User metadata
Rank l33t++
Rank
l33t++
Ozzuneoj wrote:

Anyone know why the BIOS speed would vary so much? What does that even mean anyway?

Hi there! while CGA is handled by the PC BIOS itself,
EGA and VGA cards have their own BIOS extension each.

Usually, it is stored on EPROM chips on the cards themselves.
During the benchmark, CheckIt! accesses graphics cards both directly and via BIOS ROM routines.

Without Shadow Memory, ROM code is read from these EPROMs each time a graphics command is issued.
Depending on the access or response time of each EPROM and how it is connected (interleaved or not),
speed may vary. For example, there are EPROMs with an access time of 45 to 200ns.

"Time, it seems, doesn't flow. For some it's fast, for some it's slow.
In what to one race is no time at all, another race can rise and fall..." - The Minstrel

//My video channel//

Reply 4 of 28, by dr.zeissler

User metadata
Rank l33t
Rank
l33t

https://www.flickr.com/photos/94839221@N05/al … 157660715046016

Some "Benches" of my FAST 😀 TowerAT (EGA)

Last edited by dr.zeissler on 2023-02-21, 09:07. Edited 1 time in total.

Retro-Gamer 😀 ...on different machines

Reply 5 of 28, by derSammler

User metadata
Rank l33t
Rank
l33t
Ozzuneoj wrote:

(by the way, the 5153 is technically a CGA monitor, but for reasons that are beyond my technical knowledge, it can display most 16 color EGA modes, aside from the highest resolutions)

Strictly speaking, there is no such thing as a CGA monitor. The 5153 is a color monitor with digital RGBI input. Restrictions of CGA don't apply, as those come from the video memory limitations of the CGA card, not from the monitor. EGA works, since it is using digital RGBI as well. Only modes with a higher scan rate (>15 KHz) won't work, as the 5153 can not sync to anything but 15 KHz.

Reply 6 of 28, by Scali

User metadata
Rank l33t
Rank
l33t
derSammler wrote:

Strictly speaking, there is no such thing as a CGA monitor. The 5153 is a color monitor with digital RGBI input. Restrictions of CGA don't apply, as those come from the video memory limitations of the CGA card, not from the monitor. EGA works, since it is using digital RGBI as well. Only modes with a higher scan rate (>15 KHz) won't work, as the 5153 can not sync to anything but 15 KHz.

Yes, and to expand on that:
The EGA card was specifically designed to be backwards compatible with RGBI monitors for CGA in 200-line modes (320x200 and 640x200).
EGA actually has an RrGgBb interface rather than RGBI (that is, r, g and b are separate intensity signals for the 3 basic colours). This allows for 64 indivdual colours on a real EGA monitor (you can select 16 of these 64 colours into the EGA palette).
But because of the CGA compatibility, this only works in the 640x350 mode (EGA monitors are also hardwired to be RGBI-compatible in 200-line modes, so they simply use one global intensity and ignore the other intensity signals). This is a shame, since this resolution is too high for any kind of action game. So most games are stuck to 320x200 and the fixed 16 colours known from CGA textmode, PCjr, Tandy and Plantronics.

http://scalibq.wordpress.com/just-keeping-it- … ro-programming/

Reply 7 of 28, by Ozzuneoj

User metadata
Rank l33t
Rank
l33t

Thanks for the info guys! One thing I discovered is that using the OTIVGA card with the 5153 requires the card to be set to CGA mode, otherwise the display is garbled (probably because it is trying to send a higher refresh rate text mode during bootup?). This works for some things that use EGA graphics, like display mode tests in Checkit, but other games and programs detect only CGA graphics, and when you force them to use EGA graphics the display is blank. Switching to CGA graphics in those games simply gives 4-color CGA.

When I use one of my "native" EGA Everex cards set to CGA (Color 80x25) via their dip switch, programs and games usually detect EGA graphics capability and display the correct graphics automatically, or when told to use EGA modes.

So, this does limit my ability to compare performance between the newer OAK VGA\EGA card and the older Everex EGA cards. I would assume that a more capable EGA monitor, like the 5154 wouldn't have any of these problems.

Now for some blitting from the back buffer.

Reply 8 of 28, by dr.zeissler

User metadata
Rank l33t
Rank
l33t

That thread brings me to another question:
- Is the fastest EGA graphics always a VGA Card in EGA mode?
- There are some games that do not work on EGA with a VGA card, even if the software for the card is set to "ega-mode"

Retro-Gamer 😀 ...on different machines

Reply 9 of 28, by Ozzuneoj

User metadata
Rank l33t
Rank
l33t

Wow, this is blast from the past.

Funny timing though... Here we are a little over 3 years since I posted this, and the day before you posted this I sorted through my now-much-larger video card collection. I now have so many dedicated EGA\CGA cards and VGA+EGA cards that I must have a pretty large percentage of the different EGA-capable chipsets in existence. I would love to compare these in an XT and a 386.

*sigh* Yet another benchmarking test I need to do, right after ISA and VLB cards on 386, 486 and 586 systems.

Now for some blitting from the back buffer.

Reply 10 of 28, by dr.zeissler

User metadata
Rank l33t
Rank
l33t

would be interesting though.

I just thought about putting an ega card in my amiga2000 instead of a vga card.
with the mce adapter I still can use my dual tft-setup nec71vm.

Last edited by dr.zeissler on 2021-12-23, 20:14. Edited 1 time in total.

Retro-Gamer 😀 ...on different machines

Reply 11 of 28, by mogwaay

User metadata
Rank Newbie
Rank
Newbie

I'd be really interested to see how EGA cards perform in 16bit, 286/386/486 computers that the CPU runs much faster than the bus to see how they perform. I built my own 10MHz 8088 pc and was a little disappointed with the performance of games using the EGA card and assumed it must be the CPU, but as the games would run really smooth on CGA modes, I started to think it must be the video, which led me to wonder if all games that support EGA were sluggish regardless of CPU platform?

For example of you play Prince of Persia on a 386 with an EGA card in EGA mode, is it still a bit slow compared with CGA mode?

Reply 13 of 28, by digger

User metadata
Rank Oldbie
Rank
Oldbie
Ozzuneoj wrote on 2021-02-26, 06:20:

Funny timing though... Here we are a little over 3 years since I posted this, and the day before you posted this I sorted through my now-much-larger video card collection. I now have so many dedicated EGA\CGA cards and VGA+EGA cards that I must have a pretty large percentage of the different EGA-capable chipsets in existence. I would love to compare these in an XT and a 386.

It would indeed be nice to have some benchmarks done comparing performance between EGA (and EGA-compatible) cards.

It would be especially interesting to include the Graphics Gremlin in these benchmarks, once it has gained EGA compatibility.

Reply 14 of 28, by mkarcher

User metadata
Rank l33t
Rank
l33t
Plasma wrote on 2021-12-23, 20:28:

Well there's twice as much data to move around with 4-color CGA (2bpp) vs 16-color EGA (4bpp). On a slow system with an 8-bit bus that makes a big difference.

Exactly that. The limiting factor is the bandwidth of data transfers into the EGA video memory. The processor and the screen refresh are contending for access to the EGA memory, so the EGA card is typically adding wait states to video memory writes, limiting the performance even more than required by the 8-bit bus interface. The available write bandwidth in CGA and EGA graphics modes is around the same, so the double amount of data means the double amount of time spent just to update the sceen.

Reply 15 of 28, by mogwaay

User metadata
Rank Newbie
Rank
Newbie

This is what I'm thinking, 8bit EGA is always slow, regardless of CPU speed, EGA Prince of Persia on an 8bit EGA is a slow game, regardless of whether it's on a 8088 or a 386. My thinking is that high Res EGA games was just very compromised games and never a great experience - you probably have better gameplay experience using CGA or VGA. Bit sad as when making my retro PC I'd focused on EGA as it's the most pleasing retro PC aesthetic to me. I did get a 8bit compatible VGA card too that needs fixing, so interested to try EGA modes on that as I think it has a higher vRAM bandwidth so should have less wait states and maybe better performance...

Thanks for the replies, really love the VOGONS community 😄

Reply 16 of 28, by mkarcher

User metadata
Rank l33t
Rank
l33t
mogwaay wrote on 2021-12-24, 10:25:

This is what I'm thinking, 8bit EGA is always slow, regardless of CPU speed, EGA Prince of Persia on an 8bit EGA is a slow game, regardless of whether it's on a 8088 or a 386. My thinking is that high Res EGA games was just very compromised games and never a great experience - you probably have better gameplay experience using CGA or VGA. Bit sad as when making my retro PC I'd focused on EGA as it's the most pleasing retro PC aesthetic to me. I did get a 8bit compatible VGA card too that needs fixing, so interested to try EGA modes on that as I think it has a higher vRAM bandwidth so should have less wait states and maybe better performance...

It's normal to get better performance from EGA cards in CGA mode than from EGA cards in EGA mode, but I am unsure about the comparison between EGA games on an EGA card and CGA games on a CGA card. In EGA-optimized games that use the hardware capabilities of the EGA, e.g. for smooth scrolling, you get a nice high-performance experience even with 8-bit cards. Take a look at games like the original Duke Nukem trilogy or Crystal Caves.

Don't get your hops up too high with VGA cards - unless they use a buffer to accept a write from the CPU and forward it to the video RAM as soon as it is possible, while the CPU is already busy doing other stuff. Cards known to have a buffer like that are the Tseng ET4000 series (of course I have to mention them - no retro post without someone praising the ET4000...), or the Trident 8900D. Possibly the 8900CL also has a write buffer. The cheaper Trident models do not and have mediocre performance.

I tried a couple of VGA cards in my 10MHz Turbo-XT, and the only cards that enabled the V20 processor to deliver 2MB/s into the video RAM (i.e. the bus started being the bottleneck) were two of my three ET4000 cards and a mach32 card. The third ET4000 card refused to work in an 8-bit slot.

Reply 17 of 28, by rmay635703

User metadata
Rank Oldbie
Rank
Oldbie
dr.zeissler wrote on 2021-02-24, 09:33:

That thread brings me to another question:
- Is the fastest EGA graphics always a VGA Card in EGA mode?
- There are some games that do not work on EGA with a VGA card, even if the software for the card is set to "ega-mode"

Nope, I can list several horrifying “VGA” cards that are outrun by better examples of EGA.

Card age correlates to its speed and capabilities alongside high end / low end of the market.

This is shown by the fact some systems with “Tandy graphics “ will outperform some of the older EGA cards, despite 16 color Tandy being the slowest of them all by spec.

EGA as a distinct card was very short lived compared to CGA/vga as the 3rd parties mostly ignored it until VGA was a thing which then begat the hybrid cards which were sort of VGA compatible but fully EGA compatible.

My personal experience was before anybody really bought a VGA system there were folks with very fast (for the time) EGA graphics that outran CGA and other EGA systems comparatively.
Then a few years later some folks got systems with horrifyingly slow buggy VGA that was slower than other friends EGA systems

If we were to look at a Candy cane graph for EGA VRS VGA I’m gonna guess there is lots of overlap based on the age of the card, it’s simply a case that when faster VGA cards were being made nobody was designing new EGA cards, also rarity and cost plays a role in what people actually bought regardless of the speed, some very late cards were very slow being sold into a shrinking market with only cost / availability as a factor giving a big f U to late bloomers.

Hard to win without a horse in the race.

Reply 18 of 28, by BitWrangler

User metadata
Rank l33t++
Rank
l33t++

Back when I was a n00b to PC hardware, and thus just trying to get things working any how, not into how and how fast they work, I had an EGA card that "spoiled" me, it was one of those multi cards, Herc/CGA/EGA switchable on a single chip, 8 bit card, 64kB RAM.. I wanna say the chip was by Paradise but I could be wrong. Well it spoiled me because any EGA card I tried after that one seemed slower, though I've only occasionally had access to working EGA capable monitors.

Unicorn herding operations are proceeding, but all the totes of hens teeth and barrels of rocking horse poop give them plenty of hiding spots.

Reply 19 of 28, by kdr

User metadata
Rank Member
Rank
Member

So far the only objective benchmarks suggested are the video tests from Check It and Landmark, which seem oriented towards text mode performance. Any others out there?

I don't think EGA 4bpp has to be any slower than CGA 2bpp, because one of the big features of EGA graphics is the capability to modify all four colour planes with a single 8-bit memory write. The main performance limitation is the wait states that the EGA imposes on the CPU when reading/writing video RAM.

The original EGA chipset from IBM (along with the first clone EGA chipset from Chips & Technologies) clocks the video RAM at the dot clock frequency, which is 14.318Mhz for the 320x200 and 640x200 modes. The chipset sequencer coordinates the CRT and CPU accesses to video RAM. Over the course of 32 clock cycles the sequencer provides 5 opportunities to access the video RAM. The CRT needs 2 out of 5 accesses in 320 pixel modes and 4 out of 5 accesses in 640 pixel modes. When the CPU wants to read/write to video RAM, the video chipset inserts wait states until an access opportunity becomes available. (This will be quite a few wait states in the 640 pixel modes!)

So I think on these early EGA cards, the video RAM would actually be *faster* in 640x350 mode (since it uses a 16.257Mhz clock) as compared to 640x200 mode (which is using the standard 14.318Mhz clock). There's a "bandwidth" bit in the SR01 register that determines whether the CPU gets 1 access or 3 accesses during each 32 cycle group. I wonder what happens if you program the sequencer to give the CPU 3 accesses in a 640 pixel mode (which should starve the CRT of its required accesses). It might be safe to do so during the vertical blanking, providing a brief window of high performance access to the video RAM...

Anyhow, at some point the video chipsets began to incorporate a FIFO buffer, so that the CPU writes into the buffer and can be released immediately without any wait states, and also switched to clocking the video RAM with a faster clock that was decoupled from the video dot clock. Did any EGA chipsets use these techniques? The early VGA chipsets used essentially the same design as the original EGA, adding a whole bunch of wait states to synchronize the CPU writes with the gaps in the CRT reads.