ryanfox81 wrote on 2022-06-05, 17:09:

What is happing here? 🤔

I dunno for sure. Maybe the SRAM gets mapped into UMA, were the BIOS code used go.

Edit: Some games can use UMBs. Commodore Keen IV,for example.
But that's old news to you, I assume. 😅

Oh, and if the whole UMA was flooded with SRAM..
How would this affect VGA performance? VGA cards didn't use SRAM, did they?
VGA's 128KB framebuffer window is in A segment, isn't it?
MCGA/VGA mode 13h games are requiring 64KB of memory, merely, I vaguely remember? So they'd need no bank-switching etc.?

ryanfox81 wrote on 2022-06-05, 17:09:

[..] basically anything with scrolling graphics runs significantly smoothly.

Awesome! Kudos to you! 😁

I don't dare to imagine what what
awesome XTs and ATs we could make with SRAM.
And interleaved, maybe, with 16-Bit wide memory i/o!
The 8086, V30 ou 80186 could use them directly.

Edit: PS: Have you ever heard of the Siemens PC-D?
It was an MS-DOS compatible based on an 80186.
An hot-rod XT-class system, so to say. It had a VG96 bus and a super quick serial port (10 MBits)..
Machines like this really show what XTs (pre- 286s) could be capable of! 🙂

http://oldcomputer.info/pc/SiemensPCD/index.htm

Those ROM sockets are intended for Basic interpreter, and mapped at F4000..FDFFF, right?
Strange... I wouldn't expect any software using that range as RAM...

You've added 64KB of SRAM 1x32KB, 4x8KB to the rom sockets, and you feel like the games run smoother.

The memory locations for those sockets tend to be pretty much fixed, and also tend to lack any signals which would allow writing to those locations.
Now due to circumstances, it might be possible that you could write data into those locations, (assigned pins in the sockets, address overlap etc) but that shouldn't have any effect on the actual processing of the bios.
I doubt that a normal write to those memory locations which are occupied by the srams you've inserted will actually succeed.

remove the chips and run your testprograms again and post the results.

placebo. As weedeewee mentioned there is no write signal, plus addresses decoded to rom sockets also trigger additional wait states
you can test write ability using dos DEBUG

It's a miracle that this setup even boots and isn't completely clobbering the data and address buses. It's definitely placebo and I wouldn't suggest running it like this for long. It's not acting as cache because not only does your 8088 system not have cache control circuitry (None of them do), I don't even think an 8088 would benefit that greatly from faster RAM if it didn't also mean increased clock speed, aside from the fact of lacking the DRAM refresh cycles meaning the CPU gets a few more cycles in per second.

Error 0x7CF wrote on 2022-06-05, 21:39:

It's a miracle that this setup even boots and isn't completely clobbering the data and address buses. It's definitely placebo and I wouldn't suggest running it like this for long. It's not acting as cache because not only does your 8088 system not have cache control circuitry (None of them do),

Well, the 62256 (?) seems to be pin compatible to 27256 EPROMs.
Have to double check, though.
So in worst case, they appear as empty EEPROMs.
I also wonder if the motherboard has the write enable lines for the ROM sockets.
Normally, if we assume logical thinking of the designers, they should be missing of course.
But after all these years, I've stopped thinking like that.
Everything seems possible nowadays. 🤷‍♂️

Error 0x7CF wrote on 2022-06-05, 21:39:

I don't even think an 8088 would benefit that greatly from faster RAM if it didn't also mean increased clock speed, aside from the fact of lacking the DRAM refresh cycles meaning the CPU gets a few more cycles in per second.

I'm curious, though. An experiment with real hardware would be worth a try.
I do know the difference between both RAM types thanks to my vintage books from the 1990s,
but my experience with XT hardware is limited.
I wonder if it would be possible to cut the traces for the DRAM refresh circuit on an XT board and use an 1MB UMB card with SRAM instead. With the 640KB DILs removed on the motherboard, of course.

There was that FASTV20.COM program to slow down RAM refresh on an XT, making it run a little faster.
So, it should be possible to replace DRAM chips with SRAM, and use that program to the extreme...

There are 8088 systems with SRAM already, the IBM 5140 Convertible used it.

That said, I would love to see what a V20/V30 system would look like when pushed to the limit with SRAM and frequency increase.

Error 0x7CF wrote on 2022-06-07, 18:14:

There are 8088 systems with SRAM already, the IBM 5140 Convertible used it.

That said, I would love to see what a V20/V30 system would look like when pushed to the limit with SRAM and frequency increase.

Very few, most were battery backed ram scenarios.
there were people who would remove dram and replace with SRAM in the PC/XT/JR era for a speed boost.
One of my friends fathers was extremely into the PCJR in ancient times and had it expanded using SRAM .

How this was done or what cards were used I have no idea.

What was strange is everyone talked about it so there must have been a commercial offering somewhere.

Error 0x7CF wrote on 2022-06-07, 18:14:

There are 8088 systems with SRAM already, the IBM 5140 Convertible used it.

That said, I would love to see what a V20/V30 system would look like when pushed to the limit with SRAM and frequency increase.

Add lotech 1mb ram board, remove all onboard ram.
Voila..

matze79 wrote on 2022-06-07, 19:07:

Error 0x7CF wrote on 2022-06-07, 18:14:

There are 8088 systems with SRAM already, the IBM 5140 Convertible used it.

That said, I would love to see what a V20/V30 system would look like when pushed to the limit with SRAM and frequency increase.

Add lotech 1mb ram board, remove all onboard ram.
Voila..

I’m not sure 0k dram is possible on all boards (I got errors with a low tec and no dram)
Maybe 16k (or whatever the lowest amount you can install is?)
That is why I always wondered how it worked historically