Several posters on comp.arch are running their cpu designs on FPGAs.
I have several questions:
1. Which particular FPGA chip? (not just family but the particular SKU)
2. On what development board?
3. Using what tools?

Thanks,
brian

On 1/7/2024 3:07 PM, Brian G. Lucas wrote:
> Several posters on comp.arch are running their cpu designs on FPGAs.
> I have several questions:
> 1. Which particular FPGA chip? (not just family but the particular SKU)
> 2. On what development board?
> 3. Using what tools?
>

Main boards I am using:
Nexys A7, XC7A100T-1CSG324C
QMTECH XC7A200T, XC7A200T-1FBG484

A few other boards:
Arty S7-50, XC7S50
CMod S7, XC7S25

I am primarily using Vivado WebPack (AKA: the freeware version of Vivado).

I also technically have:
AVnet Ultra96
Got it from someone else in this group (which is cool).
But, never figured out how to make it actually do anything...

The Nexys A7 has the nice feature that one can stick the bitsteam onto
an SDcard, and just boot from the SDcard...

The Arty and CMod boards need to be booted via a USB cable, or have an
onboard Flash chip that can be configured to be able to boot the FPGA.

For the QMTech board, needed to get a JTAG cable, which cost almost as
much as the FPGA board itself...

Has the big annoyance that it needs to be booted via the JTAG cable
every time.

I have the "daughter board" for the QMTech board, which has an RP2040
chip and similar. Theoretically, the RP2040 was supposed to be able to
both serve the role of a JTAG interface and be able to boot the FPGA
from a bitstream loaded into the RP2040's Flash, but... Couldn't get
this part to work (the RP2040 itself seemed to work, just seemingly ).

As for cores I can fit on them:
XC7S25:
Small 1-wide cores, generally no FPU or MMU.
The CMod S7 could be more useful... if it had a RAM chip.
XC7S50:
Can sort-of fit a 2-wide core with an FPU and MMU.
Difficult to fit a full SIMD unit with 3-wide and 64 GPRs.
XC7A100T:
Could fit a full 3-wide BJX2 core, with full ISA the features.
XC7A200T:
Can fit two 3-wide BJX2 cores and a hardware rasterizer module.

> Thanks,
> brian

On 2024-01-07 4:07 p.m., Brian G. Lucas wrote:
> Several posters on comp.arch are running their cpu designs on FPGAs.
> I have several questions:
> 1. Which particular FPGA chip? (not just family but the particular SKU)
> 2. On what development board?
> 3. Using what tools?
>
> Thanks,
> brian

I primarily use a Nexys Video board which has a Xilinx/AMD
XC7A200T-SBG484-1 chip on it. I have several older boards decreasing in
density which I no longer use. They were great for working on things,
but I keep growing and am into really complex designs now. I am saving
up for a larger board as most of the stuff I work on now is quite
crammed into the FPGA. I also have a CMOD-A7 with a XC7A35 chip.

I am using Vivado Webpack (the free version). I also have been using
vasm, and vlink, to assemble and link software. Using a compiler I wrote
myself – cc64. Also using MS Visual Studio, community edition, to
maintain home grown tools. And I use MS Office 365 for documentation. I
use Ultra-edit as a source code editor. Files are being archive in Github.

On 1/8/2024 1:07 AM, Robert Finch wrote:
> On 2024-01-07 4:07 p.m., Brian G. Lucas wrote:
>> Several posters on comp.arch are running their cpu designs on FPGAs.
>> I have several questions:
>> 1. Which particular FPGA chip? (not just family but the particular SKU)
>> 2. On what development board?
>> 3. Using what tools?
>>
>> Thanks,
>> brian
>
> I primarily use a Nexys Video board which has a Xilinx/AMD
> XC7A200T-SBG484-1 chip on it. I have several older boards decreasing in
> density which I no longer use. They were great for working on things,
> but I keep growing and am into really complex designs now. I am saving
> up for a larger board as most of the stuff I work on now is quite
> crammed into the FPGA. I also have a CMOD-A7 with a XC7A35 chip.
>

Had been tempted by the Nexys Video, but, expensive...

Not sure about going bigger than this, as this seems like the upper
limit of what is supported by Vivado Webpack...

Had considered CMOD-A7, but the S7 was cheaper, and didn't notice until
later that the S7 lacked the RAM chip (which the A7 would have had).

Both the CMOD boards would be a bit small for what I want to do, since
as noted, even the Arty S7-50 is a bit cramped for my uses at present.

Both CMOD boards would likely be practically limited mostly to
microcontroller-like tasks.

Had at one point gotten a board with a Lattice ECP5 FPGA (the Orange
Crab), but then realized that despite having a moderately decent FPGA
(in terms of LUTs), its usable IO was actually a bit lacking (the number
of pads on the board is misleading, as many of them are not FPGA digital
IO pins, but rather fixed-purpose ADC input pins).

It is "somewhere around here"...

> I am using Vivado Webpack (the free version). I also have been using
> vasm, and vlink, to assemble and link software. Using a compiler I wrote
> myself – cc64. Also using MS Visual Studio, community edition, to
> maintain home grown tools. And I use MS Office 365 for documentation. I
> use Ultra-edit as a source code editor. Files are being archive in Github.
>

For Native-Windows coding:
Visual Studio Community 2022
Though mostly building via the command-line
I use the Visual Studio IDE mostly as a debugger.
I mostly use Notepad2 for code editing.

I have BGBCC, as my own C compiler, but currently this only really
targets my own BJX2 ISA. I generally try to keep my coding practices
portable enough that I can still use MSVC or GCC for native builds.

Main reason for targeting x86-64 or similar would be if I wanted to use
a variant of my own languages (BGBScript and BGBScript2) for native code
on Windows. But, for most native development I use C.

I could almost just go with C# instead if this is what I wanted, as
language-wise, BGBScript2 is fairly similar to C# just with a syntax
that is a little closer to a mix of Java and ActionScript (but using
manual memory management, more like C and C++).

But, for the cases where one usually uses C, C manages to work fairly
well (and neither BS2 nor C# is likely to out-compete C at doing "normal
C stuff"...).

Most of my documentation is plain text files using MediaWiki notation,
and admittedly not as much of a fan of MarkDown. Would be nice if there
were a native WYSIWYG viewer/editor for either notation though
(seemingly about the only WYSIWYG editors for these exist as web-apps).

At one point, had used HTML.

Technically have OpenOffice, but seemingly pretty much no one
understands ".ODT" files, and the closest one can get them to something
"generally viewable" is either to export them as HTML or "Print to PDF".

At least MediaWiki remains usable as plain text, even without all the
formatting.

On 2024-01-08 3:08 a.m., BGB wrote:
> On 1/8/2024 1:07 AM, Robert Finch wrote:
>> On 2024-01-07 4:07 p.m., Brian G. Lucas wrote:
>>> Several posters on comp.arch are running their cpu designs on FPGAs.
>>> I have several questions:
>>> 1. Which particular FPGA chip? (not just family but the particular SKU)
>>> 2. On what development board?
>>> 3. Using what tools?
>>>
>>> Thanks,
>>> brian
>>
>> I primarily use a Nexys Video board which has a Xilinx/AMD
>> XC7A200T-SBG484-1 chip on it. I have several older boards decreasing
>> in density which I no longer use. They were great for working on
>> things, but I keep growing and am into really complex designs now. I
>> am saving up for a larger board as most of the stuff I work on now is
>> quite crammed into the FPGA. I also have a CMOD-A7 with a XC7A35 chip.
>>
>
> Had been tempted by the Nexys Video, but, expensive...
>
> Not sure about going bigger than this, as this seems like the upper
> limit of what is supported by Vivado Webpack...
>
I do not think it is that expensive considering the amount of use I have
gotten out of it over the last five or six years. Cost divided by use is
pretty low. Expensive is somewhat relative, I think it is expensive now
because I am not a millionaire, yet. I have some qualms about spending a
lot of money on what in some ways is just an entertainment toy, but I am
likely to spend double the money at some point.

My issue with larger boards is the system build times. At some point it
is not practical as a hobby because builds take too long. It takes about
an hour or so to build the system for the current board. Not keen on the
idea of day long build times.

>
> Had considered CMOD-A7, but the S7 was cheaper, and didn't notice until
> later that the S7 lacked the RAM chip (which the A7 would have had).
>
> Both the CMOD boards would be a bit small for what I want to do, since
> as noted, even the Arty S7-50 is a bit cramped for my uses at present.
>
I think the CMODA7 has more I/O too. I had a RISCV project running on
the CMODA7.
>
> Both CMOD boards would likely be practically limited mostly to
> microcontroller-like tasks.
>
>
> Had at one point gotten a board with a Lattice ECP5 FPGA (the Orange
> Crab), but then realized that despite having a moderately decent FPGA
> (in terms of LUTs), its usable IO was actually a bit lacking (the number
> of pads on the board is misleading, as many of them are not FPGA digital
> IO pins, but rather fixed-purpose ADC input pins).
>
> It is "somewhere around here"...
>
>
>> I am using Vivado Webpack (the free version). I also have been using
>> vasm, and vlink, to assemble and link software. Using a compiler I
>> wrote myself – cc64. Also using MS Visual Studio, community edition,
>> to maintain home grown tools. And I use MS Office 365 for
>> documentation. I use Ultra-edit as a source code editor. Files are
>> being archive in Github.
>>
>
> For Native-Windows coding:
>   Visual Studio Community 2022
>     Though mostly building via the command-line
>     I use the Visual Studio IDE mostly as a debugger.
>   I mostly use Notepad2 for code editing.
>
>
> I have BGBCC, as my own C compiler, but currently this only really
> targets my own BJX2 ISA. I generally try to keep my coding practices
> portable enough that I can still use MSVC or GCC for native builds.
>
>
> Main reason for targeting x86-64 or similar would be if I wanted to use
> a variant of my own languages (BGBScript and BGBScript2) for native code
> on Windows. But, for most native development I use C.
>
> I could almost just go with C# instead if this is what I wanted, as
> language-wise, BGBScript2 is fairly similar to C# just with a syntax
> that is a little closer to a mix of Java and ActionScript (but using
> manual memory management, more like C and C++).
>
> But, for the cases where one usually uses C, C manages to work fairly
> well (and neither BS2 nor C# is likely to out-compete C at doing "normal
> C stuff"...).
>
>
>
> Most of my documentation is plain text files using MediaWiki notation,
> and admittedly not as much of a fan of MarkDown. Would be nice if there
> were a native WYSIWYG viewer/editor for either notation though
> (seemingly about the only WYSIWYG editors for these exist as web-apps).
>
> At one point, had used HTML.
>
> Technically have OpenOffice, but seemingly pretty much no one
> understands ".ODT" files, and the closest one can get them to something
> "generally viewable" is either to export them as HTML or "Print to PDF".
>
> At least MediaWiki remains usable as plain text, even without all the
> formatting.
>
>

Forgot to mention a tool: I am using a Windows workstation,
INTEL-i7-7770, with 32GB of RAM. It was advertised as a game machine. I
got it refurbished, and installed additional RAM myself. The hard disk
crashed on it, so I replaced that with an SSD. I used to buy the parts
and assemble the machines myself, but more recently I just try to buy a
suitable machine.

On 1/8/2024 5:24 AM, Robert Finch wrote:
> On 2024-01-08 3:08 a.m., BGB wrote:
>> On 1/8/2024 1:07 AM, Robert Finch wrote:
>>> On 2024-01-07 4:07 p.m., Brian G. Lucas wrote:
>>>> Several posters on comp.arch are running their cpu designs on FPGAs.
>>>> I have several questions:
>>>> 1. Which particular FPGA chip? (not just family but the particular SKU)
>>>> 2. On what development board?
>>>> 3. Using what tools?
>>>>
>>>> Thanks,
>>>> brian
>>>
>>> I primarily use a Nexys Video board which has a Xilinx/AMD
>>> XC7A200T-SBG484-1 chip on it. I have several older boards decreasing
>>> in density which I no longer use. They were great for working on
>>> things, but I keep growing and am into really complex designs now. I
>>> am saving up for a larger board as most of the stuff I work on now is
>>> quite crammed into the FPGA. I also have a CMOD-A7 with a XC7A35 chip.
>>>
>>
>> Had been tempted by the Nexys Video, but, expensive...
>>
>> Not sure about going bigger than this, as this seems like the upper
>> limit of what is supported by Vivado Webpack...
>>
> I do not think it is that expensive considering the amount of use I have
> gotten out of it over the last five or six years. Cost divided by use is
> pretty low. Expensive is somewhat relative, I think it is expensive now
> because I am not a millionaire, yet. I have some qualms about spending a
> lot of money on what in some ways is just an entertainment toy, but I am
> likely to spend double the money at some point.
>
> My issue with larger boards is the system build times. At some point it
> is not practical as a hobby because builds take too long. It takes about
> an hour or so to build the system for the current board. Not keen on the
> idea of day long build times.
>

As the moment, my build times (on the XC7A200T) are around 25 minutes.

Though, this is not using the full FPGA, and with a fair bit of timing
slack at the moment. Takes a lot longer if timing is tight.

>>
>> Had considered CMOD-A7, but the S7 was cheaper, and didn't notice
>> until later that the S7 lacked the RAM chip (which the A7 would have
>> had).
>>
>> Both the CMOD boards would be a bit small for what I want to do, since
>> as noted, even the Arty S7-50 is a bit cramped for my uses at present.
>>
> I think the CMODA7 has more I/O too. I had a RISCV project running on
> the CMODA7.

Yeah.

A RAM chip would make a big difference here...
As does more IO.

Could fit a basic RISC-V core or similar on it, but probably not
something with multiple execute lanes and SSE style FPU-SIMD (where the
4-wide SIMD unit was being a big pain trying to fit it into the XC7S50;
but for what I wanted to run on it, would kinda really want the fast
FPU-SIMD).

Though, like many other things, this sub-project stalled.

>>
>> Both CMOD boards would likely be practically limited mostly to
>> microcontroller-like tasks.
>>
>>
>> Had at one point gotten a board with a Lattice ECP5 FPGA (the Orange
>> Crab), but then realized that despite having a moderately decent FPGA
>> (in terms of LUTs), its usable IO was actually a bit lacking (the
>> number of pads on the board is misleading, as many of them are not
>> FPGA digital IO pins, but rather fixed-purpose ADC input pins).
>>
>> It is "somewhere around here"...
>>
>>
>>> I am using Vivado Webpack (the free version). I also have been using
>>> vasm, and vlink, to assemble and link software. Using a compiler I
>>> wrote myself – cc64. Also using MS Visual Studio, community edition,
>>> to maintain home grown tools. And I use MS Office 365 for
>>> documentation. I use Ultra-edit as a source code editor. Files are
>>> being archive in Github.
>>>
>>
>> For Native-Windows coding:
>>    Visual Studio Community 2022
>>      Though mostly building via the command-line
>>      I use the Visual Studio IDE mostly as a debugger.
>>    I mostly use Notepad2 for code editing.
>>
>>
>> I have BGBCC, as my own C compiler, but currently this only really
>> targets my own BJX2 ISA. I generally try to keep my coding practices
>> portable enough that I can still use MSVC or GCC for native builds.
>>
>>
>> Main reason for targeting x86-64 or similar would be if I wanted to
>> use a variant of my own languages (BGBScript and BGBScript2) for
>> native code on Windows. But, for most native development I use C.
>>
>> I could almost just go with C# instead if this is what I wanted, as
>> language-wise, BGBScript2 is fairly similar to C# just with a syntax
>> that is a little closer to a mix of Java and ActionScript (but using
>> manual memory management, more like C and C++).
>>
>> But, for the cases where one usually uses C, C manages to work fairly
>> well (and neither BS2 nor C# is likely to out-compete C at doing
>> "normal C stuff"...).
>>
>>
>>
>> Most of my documentation is plain text files using MediaWiki notation,
>> and admittedly not as much of a fan of MarkDown. Would be nice if
>> there were a native WYSIWYG viewer/editor for either notation though
>> (seemingly about the only WYSIWYG editors for these exist as web-apps).
>>
>> At one point, had used HTML.
>>
>> Technically have OpenOffice, but seemingly pretty much no one
>> understands ".ODT" files, and the closest one can get them to
>> something "generally viewable" is either to export them as HTML or
>> "Print to PDF".
>>
>> At least MediaWiki remains usable as plain text, even without all the
>> formatting.
>>
>>
>
> Forgot to mention a tool: I am using a Windows workstation,
> INTEL-i7-7770, with 32GB of RAM. It was advertised as a game machine. I
> got it refurbished, and installed additional RAM myself. The hard disk
> crashed on it, so I replaced that with an SSD. I used to buy the parts
> and assemble the machines myself, but more recently I just try to buy a
> suitable machine.
>

My case:
Ryzen 2700X, 3.7 GHz base-clock, 8 core / 16 thread;
Currently 112GB of RAM (was 48GB, upgraded recently);
ASUS ROG MOBO;
GeForce RTX 3060 (recently upgraded from GTX 980);
Still have a DVD-RW drive, still useful occasionally.

Had bought 128GB, but the ASUS ROG MOBO I am using, can't handle a full
128GB (freaks out and drops back down to 4GB).

Windows can see that the full 128GB exists, but identifies 124.5 GB of
it as "System Reserved".

So, stuck with a wonky configuration of 3x 32GB sticks and 1x 16GB
stick, which does at least seem to work.

Currently 12TB worth of HDD space, 1TB SSD (for OS).
Can't install any more HDDs, out of SATA ports.

CPU was left at base clock:

When I was running a Piledriver core, had down-clocked it from its stock
4.2 GHz to 3.6 GHz and disabled Turbo Boost, mostly because it was a lot
more reliable and well-behaved at the lower speed (didn't overhead under
load, wasn't constantly trying to melt).

Similar situation with the former Phenom II, where I had ended up
down-clocking it to 2.8 GHz from the stock 3.4, for similar reasons...

My Ryzen7 has a base-clock of 3.7 but occasionally turbos up, but
generally only under 1 or 2 cores; if trying to use all the cores, it
settles around the 3.7 GHz base clock anyways. Generally well behaved
enough that I have been able to leave it at the stock settings.

The RAM is IIRC running at 2733 MHz. Operation at much higher speeds
than this seems to be unreliable.

Can note that I also have an old Dell rack server (gotten as second-hand
/ they were throwing it out):
Dual Xeon E5410, 2.133 GHz IIRC
8GB of RAM as 16 sticks (so, like, 512MB each).

For a while, while its single-thread performance isn't that great, it
held up very well under multi-threaded workloads (against the Phenom and
Piledriver).

However, it seems roughly break-even vs my Ryzen. The Ryzen is
"obviously faster" at 4 threads, but too much past this and stuff slows
down to some extent and per-thread performance drops off (seemingly
leaving it pretty close to a tie under this situation).

However, running stuff like Verilator on my main PC is more convenient,
uses less power, and doesn't sound like a jet turbine...

....

BGB <cr88192@gmail.com> writes:
>Had bought 128GB, but the ASUS ROG MOBO I am using, can't handle a full
>128GB (freaks out and drops back down to 4GB).
>
>Windows can see that the full 128GB exists, but identifies 124.5 GB of
>it as "System Reserved".

This sounds to me like a Windows problem, not a problem of the board.
We have a number of systems, many of them with ASUS boards, with 128GB
RAM, and they all work fine, but of course none of them runs Windows.

- anton
--
'Anyone trying for "industrial quality" ISA should avoid undefined behavior.'
Mitch Alsup, <c17fcd89-f024-40e7-a594-88a85ac10d20o@googlegroups.com>

Anton Ertl <anton@mips.complang.tuwien.ac.at> schrieb:
> BGB <cr88192@gmail.com> writes:
>>Had bought 128GB, but the ASUS ROG MOBO I am using, can't handle a full
>>128GB (freaks out and drops back down to 4GB).
>>
>>Windows can see that the full 128GB exists, but identifies 124.5 GB of
>>it as "System Reserved".
>
> This sounds to me like a Windows problem, not a problem of the board.
> We have a number of systems, many of them with ASUS boards, with 128GB
> RAM, and they all work fine, but of course none of them runs Windows.

The workstation still under my desk at work (delivered 2015, a
replacement has finally arrived end of last year) has 512 GB of
main memory. It ran using full memory using Windows 8.1,
and currently runs Windows 10.

Needless to say I would prefer a Linux box, but the IT departments
of big corporations have their own priorities...

Thomas Koenig <tkoenig@netcologne.de> writes:
>Anton Ertl <anton@mips.complang.tuwien.ac.at> schrieb:
>> BGB <cr88192@gmail.com> writes:
>>>Had bought 128GB, but the ASUS ROG MOBO I am using, can't handle a full
>>>128GB (freaks out and drops back down to 4GB).
>>>
>>>Windows can see that the full 128GB exists, but identifies 124.5 GB of
>>>it as "System Reserved".
>>
>> This sounds to me like a Windows problem, not a problem of the board.
>> We have a number of systems, many of them with ASUS boards, with 128GB
>> RAM, and they all work fine, but of course none of them runs Windows.
>
>The workstation still under my desk at work (delivered 2015, a
>replacement has finally arrived end of last year) has 512 GB of
>main memory. It ran using full memory using Windows 8.1,
>and currently runs Windows 10.

Different variants of Windows support different amounts of memory.
See, e.g.,
<https://www.compuram.de/blog/wie-viel-ram-lasst-sich-unter-32-bit-und-64-bit-betriebssystemen-maximal-adressieren/>

This says that, e.g., Windows 11 Home only supports 128GB, and Windows
11 Enterprize 32-bit supports 4GB. Interesting entries in this table
are:

Windows 2003 Server Datacenter Edition 32-Bit 128GB
Windows 2008 Server Datacenter 32-Bit 64GB

Makes you wonder why they reduced the capabilities between 2003 and
2008.

- anton
--
'Anyone trying for "industrial quality" ISA should avoid undefined behavior.'
Mitch Alsup, <c17fcd89-f024-40e7-a594-88a85ac10d20o@googlegroups.com>

anton@mips.complang.tuwien.ac.at (Anton Ertl) writes:
>Thomas Koenig <tkoenig@netcologne.de> writes:
>>Anton Ertl <anton@mips.complang.tuwien.ac.at> schrieb:
>>> BGB <cr88192@gmail.com> writes:
>>>>Had bought 128GB, but the ASUS ROG MOBO I am using, can't handle a full
>>>>128GB (freaks out and drops back down to 4GB).
>>>>
>>>>Windows can see that the full 128GB exists, but identifies 124.5 GB of
>>>>it as "System Reserved".
>>>
>>> This sounds to me like a Windows problem, not a problem of the board.
>>> We have a number of systems, many of them with ASUS boards, with 128GB
>>> RAM, and they all work fine, but of course none of them runs Windows.
>>
>>The workstation still under my desk at work (delivered 2015, a
>>replacement has finally arrived end of last year) has 512 GB of
>>main memory. It ran using full memory using Windows 8.1,
>>and currently runs Windows 10.
>
>Different variants of Windows support different amounts of memory.
>See, e.g.,
><https://www.compuram.de/blog/wie-viel-ram-lasst-sich-unter-32-bit-und-64-bit-betriebssystemen-maximal-adressieren/>
>
>This says that, e.g., Windows 11 Home only supports 128GB, and Windows
>11 Enterprize 32-bit supports 4GB. Interesting entries in this table
>are:
>
>Windows 2003 Server Datacenter Edition 32-Bit 128GB
>Windows 2008 Server Datacenter 32-Bit 64GB
>
>Makes you wonder why they reduced the capabilities between 2003 and
>2008.

The 32-bit processors were limited to 36 bits, which is 64GB. Not sure
how the ever would have managed to fit 128GB in and use all of it.

On 1/9/2024 4:38 PM, Anton Ertl wrote:
> Thomas Koenig <tkoenig@netcologne.de> writes:
>> Anton Ertl <anton@mips.complang.tuwien.ac.at> schrieb:
>>> BGB <cr88192@gmail.com> writes:
>>>> Had bought 128GB, but the ASUS ROG MOBO I am using, can't handle a full
>>>> 128GB (freaks out and drops back down to 4GB).
>>>>
>>>> Windows can see that the full 128GB exists, but identifies 124.5 GB of
>>>> it as "System Reserved".
>>>
>>> This sounds to me like a Windows problem, not a problem of the board.
>>> We have a number of systems, many of them with ASUS boards, with 128GB
>>> RAM, and they all work fine, but of course none of them runs Windows.
>>
>> The workstation still under my desk at work (delivered 2015, a
>> replacement has finally arrived end of last year) has 512 GB of
>> main memory. It ran using full memory using Windows 8.1,
>> and currently runs Windows 10.
>
> Different variants of Windows support different amounts of memory.
> See, e.g.,
> <https://www.compuram.de/blog/wie-viel-ram-lasst-sich-unter-32-bit-und-64-bit-betriebssystemen-maximal-adressieren/>
>
> This says that, e.g., Windows 11 Home only supports 128GB, and Windows
> 11 Enterprize 32-bit supports 4GB. Interesting entries in this table
> are:
>
> Windows 2003 Server Datacenter Edition 32-Bit 128GB
> Windows 2008 Server Datacenter 32-Bit 64GB
>
> Makes you wonder why they reduced the capabilities between 2003 and
> 2008.
>

I am running Windows 10 Professional, it is supposed to be able to
support 2TB of RAM apparently...

But, whatever is going on, seemingly 112GB is fine, but 128GB is not...

> - anton

BGB <cr88192@gmail.com> writes:
>On 1/9/2024 4:38 PM, Anton Ertl wrote:
>> Thomas Koenig <tkoenig@netcologne.de> writes:
>>> Anton Ertl <anton@mips.complang.tuwien.ac.at> schrieb:
>>>> BGB <cr88192@gmail.com> writes:
>>>>> Had bought 128GB, but the ASUS ROG MOBO I am using, can't handle a full
>>>>> 128GB (freaks out and drops back down to 4GB).
>>>>>
>>>>> Windows can see that the full 128GB exists, but identifies 124.5 GB of
>>>>> it as "System Reserved".
>>>>
>>>> This sounds to me like a Windows problem, not a problem of the board.
>>>> We have a number of systems, many of them with ASUS boards, with 128GB
>>>> RAM, and they all work fine, but of course none of them runs Windows.
>>>
>>> The workstation still under my desk at work (delivered 2015, a
>>> replacement has finally arrived end of last year) has 512 GB of
>>> main memory. It ran using full memory using Windows 8.1,
>>> and currently runs Windows 10.
>>
>> Different variants of Windows support different amounts of memory.
>> See, e.g.,
>> <https://www.compuram.de/blog/wie-viel-ram-lasst-sich-unter-32-bit-und-64-bit-betriebssystemen-maximal-adressieren/>
>>
>> This says that, e.g., Windows 11 Home only supports 128GB, and Windows
>> 11 Enterprize 32-bit supports 4GB. Interesting entries in this table
>> are:
>>
>> Windows 2003 Server Datacenter Edition 32-Bit 128GB
>> Windows 2008 Server Datacenter 32-Bit 64GB
>>
>> Makes you wonder why they reduced the capabilities between 2003 and
>> 2008.
>>
>
>I am running Windows 10 Professional, it is supposed to be able to
>support 2TB of RAM apparently...
>
>But, whatever is going on, seemingly 112GB is fine, but 128GB is not...

Are the graphics on-chip? They may be sharing DRAM.

On 1/9/2024 6:23 PM, Scott Lurndal wrote:
> BGB <cr88192@gmail.com> writes:
>> On 1/9/2024 4:38 PM, Anton Ertl wrote:
>>> Thomas Koenig <tkoenig@netcologne.de> writes:
>>>> Anton Ertl <anton@mips.complang.tuwien.ac.at> schrieb:
>>>>> BGB <cr88192@gmail.com> writes:
>>>>>> Had bought 128GB, but the ASUS ROG MOBO I am using, can't handle a full
>>>>>> 128GB (freaks out and drops back down to 4GB).
>>>>>>
>>>>>> Windows can see that the full 128GB exists, but identifies 124.5 GB of
>>>>>> it as "System Reserved".
>>>>>
>>>>> This sounds to me like a Windows problem, not a problem of the board.
>>>>> We have a number of systems, many of them with ASUS boards, with 128GB
>>>>> RAM, and they all work fine, but of course none of them runs Windows.
>>>>
>>>> The workstation still under my desk at work (delivered 2015, a
>>>> replacement has finally arrived end of last year) has 512 GB of
>>>> main memory. It ran using full memory using Windows 8.1,
>>>> and currently runs Windows 10.
>>>
>>> Different variants of Windows support different amounts of memory.
>>> See, e.g.,
>>> <https://www.compuram.de/blog/wie-viel-ram-lasst-sich-unter-32-bit-und-64-bit-betriebssystemen-maximal-adressieren/>
>>>
>>> This says that, e.g., Windows 11 Home only supports 128GB, and Windows
>>> 11 Enterprize 32-bit supports 4GB. Interesting entries in this table
>>> are:
>>>
>>> Windows 2003 Server Datacenter Edition 32-Bit 128GB
>>> Windows 2008 Server Datacenter 32-Bit 64GB
>>>
>>> Makes you wonder why they reduced the capabilities between 2003 and
>>> 2008.
>>>
>>
>> I am running Windows 10 Professional, it is supposed to be able to
>> support 2TB of RAM apparently...
>>
>> But, whatever is going on, seemingly 112GB is fine, but 128GB is not...
>
> Are the graphics on-chip? They may be sharing DRAM.

Dedicated graphics card, RTX 3060, via PCIe.

CPU is a Ryzen 2700X, which does not have an integrated GPU.
Looks like only a subset of the low-end models have integrated GPU.

I guess, some stats for the CPU:
https://www.amd.com/en/product/7656

Closest looking motherboard:
https://rog.asus.com/us/motherboards/rog-strix/rog-strix-b450-f-gaming-model/spec/

https://rog.asus.com/us/motherboards/rog-strix/rog-strix-b450-f-gaming-model/

Not certain if the same motherboard, appears to be the closest match I
can find at the moment.

Looking elsewhere, it says that the AMD B450 chipset has a 128GB limit.

Not sure if just for RAM, or if this would also include MMIO for PCIe
devices and similar...

On 10/01/2024 01:10, BGB wrote:
> On 1/9/2024 4:38 PM, Anton Ertl wrote:
>> Thomas Koenig <tkoenig@netcologne.de> writes:
>>> Anton Ertl <anton@mips.complang.tuwien.ac.at> schrieb:
>>>> BGB <cr88192@gmail.com> writes:
>>>>> Had bought 128GB, but the ASUS ROG MOBO I am using, can't handle a
>>>>> full
>>>>> 128GB (freaks out and drops back down to 4GB).
>>>>>
>>>>> Windows can see that the full 128GB exists, but identifies 124.5 GB of
>>>>> it as "System Reserved".
>>>>
>>>> This sounds to me like a Windows problem, not a problem of the board.
>>>> We have a number of systems, many of them with ASUS boards, with 128GB
>>>> RAM, and they all work fine, but of course none of them runs Windows.
>>>
>>> The workstation still under my desk at work (delivered 2015, a
>>> replacement has finally arrived end of last year) has 512 GB of
>>> main memory.  It ran using full memory using Windows 8.1,
>>> and currently runs Windows 10.
>>
>> Different variants of Windows support different amounts of memory.
>> See, e.g.,
>> <https://www.compuram.de/blog/wie-viel-ram-lasst-sich-unter-32-bit-und-64-bit-betriebssystemen-maximal-adressieren/>
>>
>> This says that, e.g., Windows 11 Home only supports 128GB, and Windows
>> 11 Enterprize 32-bit supports 4GB.  Interesting entries in this table
>> are:
>>
>> Windows 2003 Server Datacenter Edition 32-Bit   128GB
>> Windows 2008 Server Datacenter         32-Bit   64GB
>>
>> Makes you wonder why they reduced the capabilities between 2003 and
>> 2008.
>>
>
> I am running Windows 10 Professional, it is supposed to be able to
> support 2TB of RAM apparently...
>
> But, whatever is going on, seemingly 112GB is fine, but 128GB is not...
>
>

Try booting it from a Linux live USB, with 128 GB in it. That should at
least let you know if you have a hardware problem (perhaps a faulty
DIMM) or a Windows problem.

Using mismatched DIMMs like this can significantly affect the speed of
the memory. You may find the system is using the DIMMs one at a time,
in series, because it can't match them up in sets for parallel access.
For some work, your system will then be a lot slower than if you had 4
16 GB DIMMs for a total of 64 GB.

And maybe you'll like Linux. It is not uncommon for tough jobs - such
as FPGA builds - to run significantly faster under Linux than Windows,
on the same hardware. Windows-only programs that you need to use can
often run fine under Wine (or Steam, if the Windows program you /need/
to run is a game :-) ), and there's always VirtualBox or KVM.

On 08/01/2024 18:38, BGB wrote:
> On 1/8/2024 5:24 AM, Robert Finch wrote:
>> On 2024-01-08 3:08 a.m., BGB wrote:
>>
>> My issue with larger boards is the system build times. At some point
>> it is not practical as a hobby because builds take too long. It takes
>> about an hour or so to build the system for the current board. Not
>> keen on the idea of day long build times.
>>
>
> As the moment, my build times (on the XC7A200T) are around 25 minutes.
>
> Though, this is not using the full FPGA, and with a fair bit of timing
> slack at the moment. Takes a lot longer if timing is tight.
>

I remember working with a fairly large CPLD in the late 90's, with a
16-bit Windows system. It could take 2 or 3 hours for the build, if it
fitted - and 6 to 10 hours if it failed. And being 16-bit Windows, the
system was useless for anything else while the build was running. That
was tedious development work!

BGB <cr88192@gmail.com> writes:
>On 1/9/2024 6:23 PM, Scott Lurndal wrote:

>
>Looking elsewhere, it says that the AMD B450 chipset has a 128GB limit.
>
>Not sure if just for RAM, or if this would also include MMIO for PCIe
>devices and similar...

Almost certainly the latter. Which explains the missing GB.

On 1/10/2024 1:08 AM, David Brown wrote:
> On 10/01/2024 01:10, BGB wrote:
>> On 1/9/2024 4:38 PM, Anton Ertl wrote:
>>> Thomas Koenig <tkoenig@netcologne.de> writes:
>>>> Anton Ertl <anton@mips.complang.tuwien.ac.at> schrieb:
>>>>> BGB <cr88192@gmail.com> writes:
>>>>>> Had bought 128GB, but the ASUS ROG MOBO I am using, can't handle a
>>>>>> full
>>>>>> 128GB (freaks out and drops back down to 4GB).
>>>>>>
>>>>>> Windows can see that the full 128GB exists, but identifies 124.5
>>>>>> GB of
>>>>>> it as "System Reserved".
>>>>>
>>>>> This sounds to me like a Windows problem, not a problem of the board.
>>>>> We have a number of systems, many of them with ASUS boards, with 128GB
>>>>> RAM, and they all work fine, but of course none of them runs Windows.
>>>>
>>>> The workstation still under my desk at work (delivered 2015, a
>>>> replacement has finally arrived end of last year) has 512 GB of
>>>> main memory.  It ran using full memory using Windows 8.1,
>>>> and currently runs Windows 10.
>>>
>>> Different variants of Windows support different amounts of memory.
>>> See, e.g.,
>>> <https://www.compuram.de/blog/wie-viel-ram-lasst-sich-unter-32-bit-und-64-bit-betriebssystemen-maximal-adressieren/>
>>>
>>> This says that, e.g., Windows 11 Home only supports 128GB, and Windows
>>> 11 Enterprize 32-bit supports 4GB.  Interesting entries in this table
>>> are:
>>>
>>> Windows 2003 Server Datacenter Edition 32-Bit   128GB
>>> Windows 2008 Server Datacenter         32-Bit   64GB
>>>
>>> Makes you wonder why they reduced the capabilities between 2003 and
>>> 2008.
>>>
>>
>> I am running Windows 10 Professional, it is supposed to be able to
>> support 2TB of RAM apparently...
>>
>> But, whatever is going on, seemingly 112GB is fine, but 128GB is not...
>>
>>
>
> Try booting it from a Linux live USB, with 128 GB in it.  That should at
> least let you know if you have a hardware problem (perhaps a faulty
> DIMM) or a Windows problem.
>

The DIMM itself works, just the problem is when there are 4 of them
installed.

From what I can gather, the most likely limitation at the moment is
that it has hit the limit of the AMD B450 chipset on my motherboard
(looking more, it seems the Zen+ may also have this limit as well).

Likely, the combination of 128GB RAM + PCIe MMIO and similar, exceeds
the 128GB limit of the chipset or similar.

Also putting in all of the RAM causes the PC to boot-cycle several times
before it actually boots.

> Using mismatched DIMMs like this can significantly affect the speed of
> the memory.  You may find the system is using the DIMMs one at a time,
> in series, because it can't match them up in sets for parallel access.
> For some work, your system will then be a lot slower than if you had 4
> 16 GB DIMMs for a total of 64 GB.
>

I hadn't actually noticed slowdowns in my own tests.
Granted, yes, I am aware with this being not ideal...

Had I realized sooner that there would have been an issue, probably
would have only ordered half as much RAM, and upgraded my PC to 96GB
(this could have matched sticks, but would have been less RAM).

>
> And maybe you'll like Linux.  It is not uncommon for tough jobs - such
> as FPGA builds - to run significantly faster under Linux than Windows,
> on the same hardware.  Windows-only programs that you need to use can
> often run fine under Wine (or Steam, if the Windows program you /need/
> to run is a game :-) ), and there's always VirtualBox or KVM.
>

I have run Linux before, just not as my main OS, in a fairly long time.
Usual drawbacks are things like software support and driver issues (GPU,
Sound, and Ethernet, being the main perennial issues).

Granted, I guess it is possible Linux could be like, "well the RAM is
there, gonna use it", and possibly just mouse-hole out the part that is
eaten up by PCIe (such as the graphics card), dunno, would need to test...

BGB <cr88192@gmail.com> writes:
>Looking elsewhere, it says that the AMD B450 chipset has a 128GB limit.

AM4 generally does not support more than 128GB RAM, because DDR4 can
only have 16GB/chip-select and channel, and AM4 only supports 4
chip-selects per channel and 2 channels. But B450 certainly does
support 128GB. We have 128GB in at least one machine with a B450
board:

# dmidecode|grep -B1 B450
Manufacturer: ASUSTeK COMPUTER INC.
Product Name: TUF B450M-PLUS GAMING

# dmesg
....
[ 3.859281] EDAC MC: UMC0 chip selects:
[ 3.859282] EDAC amd64: MC: 0: 16384MB 1: 16384MB
[ 3.860295] EDAC amd64: MC: 2: 16384MB 3: 16384MB
[ 3.861280] EDAC MC: UMC1 chip selects:
[ 3.861280] EDAC amd64: MC: 0: 16384MB 1: 16384MB
[ 3.862238] EDAC amd64: MC: 2: 16384MB 3: 16384MB
....
# free
total used free shared buff/cache available
Mem: 131832604 104094356 17986860 24772 11018444 27738248
Swap: 0 0 0

Apparently the kernel uses a little over 2GB for its own purposes (the
number of total KB shown is only 125.7GB).

- anton
--
'Anyone trying for "industrial quality" ISA should avoid undefined behavior.'
Mitch Alsup, <c17fcd89-f024-40e7-a594-88a85ac10d20o@googlegroups.com>

On 10/01/2024 15:32, Scott Lurndal wrote:
> BGB <cr88192@gmail.com> writes:
>> On 1/9/2024 6:23 PM, Scott Lurndal wrote:
>
>>
>> Looking elsewhere, it says that the AMD B450 chipset has a 128GB limit.
>>
>> Not sure if just for RAM, or if this would also include MMIO for PCIe
>> devices and similar...
>
> Almost certainly the latter. Which explains the missing GB.

If he could see 124.5 GB ram but was missing 3.5 GB, it would be a
perfect explanation. But he can only see 3.5 GB and is missing 124.5 GB.

Anton Ertl <anton@mips.complang.tuwien.ac.at> wrote:
>Different variants of Windows support different amounts of memory.
>See, e.g.,
><https://www.compuram.de/blog/wie-viel-ram-lasst-sich-unter-32-bit-und-64-bit-betriebssystemen-maximal-adressieren/>
>
>This says that, e.g., Windows 11 Home only supports 128GB, and Windows
>11 Enterprize 32-bit supports 4GB. Interesting entries in this table
>are:
>
>Windows 2003 Server Datacenter Edition 32-Bit 128GB
>Windows 2008 Server Datacenter 32-Bit 64GB
>
>Makes you wonder why they reduced the capabilities between 2003 and
>2008.

Microsoft's actual documentation[1] say:
Windows Server 2003 R2 Datacenter Edition, X86: 64 GB (16 GB with 4GT)

There's no 32-bit entries with more than 64GB supported.

The MS document does say that PAE support 37-bit (128GB) *addressing*,
and Googling shows that some early Windows PAE implementation such as
2003 (and "hacked" XP) allowed (but not supported) up to 128GB before
MS implemented limits. And people had "interesting" issues when they
did try it.

From my vague memories the default Windows virtual memory map (2GB
application, 2 GB system) is the reason why MS "only" allowed 64GB and
it's also why MS only supported 16GB on the same OS with the "4GT"
mode (3GB application, 1GB system). Basically, my theory is that it's
to avoid running out of system space in the virtual mapping (which is
still 32-bit).

I DO have a recollection that there was some kind of Windows edition
that allowed 128GB via PAE?, perhaps that WAS a variant? of 2003
Datacenter and MS later decided to try to erase history?

IIRC the trick is to completely split user-space and kernel-space
virtual mappings (IE 4/0+0/4 instead of the default of 2/2 or 4GT's
3/1) and just eat the MASSIVE performance penalty that result from
this when moving data between user-space and kernel-space.

You do get the "32-bit programs can used 4GB instead of 2GB (default)
or 3GB (4GT)" benefit with the "performance sucks" though.

But you'd need to be pretty damned desperate to do this once 64-bit OS
was well established. And I have to wonder how many hardware drivers
worked with this anyway.

1. https://learn.microsoft.com/en-us/windows/win32/memory/memory-limits-for-windows-releases

On 1/10/2024 12:26 PM, David Brown wrote:
> On 10/01/2024 15:32, Scott Lurndal wrote:
>> BGB <cr88192@gmail.com> writes:
>>> On 1/9/2024 6:23 PM, Scott Lurndal wrote:
>>
>>>
>>> Looking elsewhere, it says that the AMD B450 chipset has a 128GB limit.
>>>
>>> Not sure if just for RAM, or if this would also include MMIO for PCIe
>>> devices and similar...
>>
>> Almost certainly the latter.  Which explains the missing GB.
>
> If he could see 124.5 GB ram but was missing 3.5 GB, it would be a
> perfect explanation.  But he can only see 3.5 GB and is missing 124.5 GB.
>

Yes.

If it were, say, 116 GB usable, with 12GB or so quietly being cut off
for the graphics card and similar, this would be fine...

But, with it falling back to having only 3.5GB available for use by the
system, this makes the computer basically unusable...

In other random news, went to dentist to get dental work done (after
roughly 2 decades of not going to dentist). For now, they did several
fillings, I guess next they also want to do wisdom-teeth removal, grr...

While waiting:
Two females were sitting nearby, one kept looking in my direction for
some reason (general description: dyed hair, slightly overweight,
estimated age range, late 20s to mid 30s; was not anyone I recognized).
The other was thinner and younger (with significant facial piercings),
more interested in her phone; they seemed to know each other. Both were
wearing pajamas style clothing.

Further away in the room, there were a male and female with several
small children (and a baby in a plastic baby holder / carrying-case
thing, padded / cloth-lined interior with baby secured in place using
straps, with an external carrying handle).

Also another female with short hair (a "buzzcut" style), also slightly
overweight, and wearing black yoga pants; she was initially looking at
the small children but seemed to quickly lose interest and also started
messing around with her phone after that, ...

Etc...

Afterwards, the dentists used several types of pain suppressor (first a
paste, then injections). Now my face is numb, along with various other
sensory anomalies. Should theoretically clear up after a while.
Seems to interfere with general tactile perception and fine motor
control, had noted some visual and auditory anomalies but these have
since mostly cleared up.

Did not go to the machine shop today. Probably not a good idea to try to
work with tools or machines in this condition.

After initial drilling, they seemed to use some sort of light-cured
resin, with a fairly bright blue-colored light tool (sort of like a
curved dental-tool flashlight). Based on color, I am guessing it was
around 390nm to 400nm (was a similar color to some 400nm LEDs I had
gotten in the past, but a somewhat higher brightness; though, it looked
like there might have also been some ~ 465nm mixed in as well).

Then in the process, the dentist did make a random comment to the other
helper that I had "unusual dental anatomy" (the structure of the ridges
and grooves is apparently atypical).

Though, it is kind of funny that many of the times I have gone to
hospitals, doctors have often just ended up making comments along the
lines of "well, that is weird", about lots of random stuff.

Then again, maybe everyone gets this, just about different types of
features? Statistically, I wouldn't expect there to be any reason to
fall much outside the bell curve on most features.

Nor, in my case, to necessarily fall outside the realm of what would be
"typical" for someone of a mixed Celtic/Norse/Ashkenazim background.
Which, are not exactly rare demographics in the US.

....

David Brown <david.brown@hesbynett.no> writes:

> I remember working with a fairly large CPLD in the late 90's, with a
> 16-bit Windows system. It could take 2 or 3 hours for the build, if
> it fitted - and 6 to 10 hours if it failed. And being 16-bit Windows,
> the system was useless for anything else while the build was running.
> That was tedious development work!

I have a similar memory but I'm not sure what kind of Windows it
was. FPGA was something fairly large for the time, the FPGA tool was
probably early Quartus and there was no progress indication at all. When
you hit start, mouse cursor changed to a hourglass for hours and then it
was done, pass or fail. I think someone rigged the then new fangled
invention called a web camera to watch the display so we didn't have to
walk to the machine to see if it was finished or not.

On 1/10/2024 11:04 AM, Anton Ertl wrote:
> BGB <cr88192@gmail.com> writes:
>> Looking elsewhere, it says that the AMD B450 chipset has a 128GB limit.
>
> AM4 generally does not support more than 128GB RAM, because DDR4 can
> only have 16GB/chip-select and channel, and AM4 only supports 4
> chip-selects per channel and 2 channels. But B450 certainly does
> support 128GB. We have 128GB in at least one machine with a B450
> board:
>
> # dmidecode|grep -B1 B450
> Manufacturer: ASUSTeK COMPUTER INC.
> Product Name: TUF B450M-PLUS GAMING
>
> # dmesg
> ...
> [ 3.859281] EDAC MC: UMC0 chip selects:
> [ 3.859282] EDAC amd64: MC: 0: 16384MB 1: 16384MB
> [ 3.860295] EDAC amd64: MC: 2: 16384MB 3: 16384MB
> [ 3.861280] EDAC MC: UMC1 chip selects:
> [ 3.861280] EDAC amd64: MC: 0: 16384MB 1: 16384MB
> [ 3.862238] EDAC amd64: MC: 2: 16384MB 3: 16384MB
> ...
> # free
> total used free shared buff/cache available
> Mem: 131832604 104094356 17986860 24772 11018444 27738248
> Swap: 0 0 0
>
> Apparently the kernel uses a little over 2GB for its own purposes (the
> number of total KB shown is only 125.7GB).
>

Not sure, is this with an integrated or discrete GPU?...

But, yeah, as-is, at least for Win10 (Pro, X64), my current setup is the
one that has the most RAM available that I can get, and the performance
doesn't seem to have taken an obvious performance hit relative to the
96GB setup, so...

But, yeah, IIRC, the setup is currently:
1A, 1B, 2A: 32GB sticks.
2B: 16GB stick.

Currently 2733 MHz IIRC, as 2933 MHz seemed to have stability issues
(the new RAM modules, from Corsair, claimed to be good for 3200 MHz, but
I didn't see stable results much over 2733; was running 2933 with the
old modules, which IIRC claimed 3000 MHz...).

IIRC, I have Windows currently set up with around 384GB of swap space,
spread across several HDDs.

....

BGB-Alt <bohannonindustriesllc@gmail.com> writes:
>On 1/10/2024 11:04 AM, Anton Ertl wrote:
>> BGB <cr88192@gmail.com> writes:
>>> Looking elsewhere, it says that the AMD B450 chipset has a 128GB limit.
>>
>> AM4 generally does not support more than 128GB RAM, because DDR4 can
>> only have 16GB/chip-select and channel, and AM4 only supports 4
>> chip-selects per channel and 2 channels. But B450 certainly does
>> support 128GB. We have 128GB in at least one machine with a B450
>> board:
>>
>> # dmidecode|grep -B1 B450
>> Manufacturer: ASUSTeK COMPUTER INC.
>> Product Name: TUF B450M-PLUS GAMING
>>
>> # dmesg
>> ...
>> [ 3.859281] EDAC MC: UMC0 chip selects:
>> [ 3.859282] EDAC amd64: MC: 0: 16384MB 1: 16384MB
>> [ 3.860295] EDAC amd64: MC: 2: 16384MB 3: 16384MB
>> [ 3.861280] EDAC MC: UMC1 chip selects:
>> [ 3.861280] EDAC amd64: MC: 0: 16384MB 1: 16384MB
>> [ 3.862238] EDAC amd64: MC: 2: 16384MB 3: 16384MB
>> ...
>> # free
>> total used free shared buff/cache available
>> Mem: 131832604 104094356 17986860 24772 11018444 27738248
>> Swap: 0 0 0
>>
>> Apparently the kernel uses a little over 2GB for its own purposes (the
>> number of total KB shown is only 125.7GB).
>>
>
>Not sure, is this with an integrated or discrete GPU?...

Discrete. The CPU is a 3900X.

>Currently 2733 MHz IIRC, as 2933 MHz seemed to have stability issues
>(the new RAM modules, from Corsair, claimed to be good for 3200 MHz, but
>I didn't see stable results much over 2733; was running 2933 with the
>old modules, which IIRC claimed 3000 MHz...).

The machine reported above uses DIMMs rated at 2666MT/s, and runs them
at 2666MT/s. dmidecode outputs:

...
Speed: 2666 MT/s
...
Configured Memory Speed: 2666 MT/s
...

>IIRC, I have Windows currently set up with around 384GB of swap space,
>spread across several HDDs.

Have you computed how long it would take to page 384GB out to the HDDs
and to page them back in? IME paging to HDDs does not make sense
anymore (and paging to SSDs is questionable, too).

- anton
--
'Anyone trying for "industrial quality" ISA should avoid undefined behavior.'
Mitch Alsup, <c17fcd89-f024-40e7-a594-88a85ac10d20o@googlegroups.com>

On Thu, 11 Jan 2024 14:27:59 +0200
Anssi Saari <anssi.saari@usenet.mail.kapsi.fi> wrote:

> David Brown <david.brown@hesbynett.no> writes:
>
> > I remember working with a fairly large CPLD in the late 90's, with a
> > 16-bit Windows system. It could take 2 or 3 hours for the build, if
> > it fitted - and 6 to 10 hours if it failed. And being 16-bit
> > Windows, the system was useless for anything else while the build
> > was running. That was tedious development work!
>
> I have a similar memory but I'm not sure what kind of Windows it
> was. FPGA was something fairly large for the time, the FPGA tool was
> probably early Quartus and there was no progress indication at all.
> When you hit start, mouse cursor changed to a hourglass for hours and
> then it was done, pass or fail. I think someone rigged the then new
> fangled invention called a web camera to watch the display so we
> didn't have to walk to the machine to see if it was finished or not.

What David describes does not sound like Altera.
By late 90s in Altera world everybody were using MAX+Plus II. It was
32-bit.
Altera's biggest CPLD family back then was Max 9000 which was not
particularly big. May be, the biggest design could take 30 minutes to
compile, but I never encountered that. 5-7 minutes was more typical on
decent Pentium-II under Win NT4. And it didn't even need a lot of RAM.
64 MB was fully sufficient.
Now, the biggest contemporary Altera FPGAs (Flex 10K family) is a
different story. I never used biggest members of the family myself,
but heard that compilation could take several hours.

I still have project with ACEX FPGA to maintain. ACEX has exactly the
same architecture as Flex 10K, the only difference is fewer SKUs and
much lower price tag. But ACEX is supported by Quartus-II v.6 (circa
2006) so I have no need to somehow convince old MAX+Plus II software to
work on newer OSes. So on this front I have no war stories to share.

As for Qaurtus-II, I don't remember ever using very early versions.
Likely v.4 is the first I used. This one, of course, had progress bar
(unreliable, but then every progress bar I had ever seen in this sorts
of software was unreliable) and it could beep when it ends compilation.