Max resolution of an integrated graphics processor?

On 7/10/2020 3:44 PM, Andy Burns wrote:
Chris wrote:

On here it says it supports DP 1.2 and HDMI 1.4a. So that means it can do
4K.

reports on some Dell machines using that APU say it can do 4K over DP
but only 2K over HDMI

Should still be enough for my purposes, the person who wants it has an
ultrawide 1440p monitor: 3440 X 1440.

Yousuf Khan

On Fri, 10 Jul 2020 at 20:07:59, Paul wrote:
J. P. Gilliver (John) wrote:
On Fri, 10 Jul 2020 at 18:19:44, Paul wrote:
J. P. Gilliver (John) wrote:
On Fri, 10 Jul 2020 at 14:37:42, VanguardLH wrote:
[]
mostly just the monitor and keyboard. Probably the biggest performance
degrade for CPU-embedded graphics is it uses shared memory; see
https://en.wikipedia.org/wiki/Shared_graphics_memory. System memory
(SDRAM) is slower but cheaper than the more expensive video RAM (that
you get with video cards). If your mobo doesn't allowed dedicated
shared memory for onboard video, for example, gaming assets get
moved to
local storage when out of system memory, so even s-l-o-w-e-r. If
dedicated shared memory is available, it takes a permanent bite out of
system memory to lower what is available to the OS and your programs.

Unless you want a just-okay system, you don't want to use onboard
video.
Get a video card. Then you'll also be able to get more specs. Onboard
video is for a workhouse setup doing typical desktop apps, but bad if
you want to do gaming, video processing, or other high-demand GPU
tasks.
That _is_ true today. It doesn't _have_ to be: I remember building
for my brother a system based around a motherboard that had
on-board video, but _not_ shared memory: there were dedicated
video memory chips - they just happened to be located on the
motherboard. (And not a shared bus to them, either.) That was a
very long time ago, though - I think it might have been in 486 or

I don't think any mobo that has on-board video these days has
separate video memory, or at least, has other compromises that
degrade performance (video, general processing, or both). It
doesn't have to be so - for example, the majority of mobos these
days have on-board sound, and ditto disc controllers, and ports (I
remember when _all_ of these were on plug-in cards!) - but for
some reason, on-board video is considered an excuse to cut
corners. (Arguably, justifiably, for the majority of users.)

AMD put out a single generation of designs, with a 32-bit wide memory
chip off the side. The memory was soldered to the motherboard. It was
[]
Here's a picture, where the pink "performance cache" on the right hand
side, was the side-port memory.

https://img.hexus.net/v2/motherboard...785G/Block.jpg
[]
I see that diagramme mentions DVI, HDMI, PCI Express, 6 SATA and 12
USB. The board I'm thinking of was far before any of those first
three; I'm not even sure it had USB, but if it did, I'm pretty sure
it didn't have anywhere near 12 of them. It was certainly no later
than XP - might have even been in '9x days.

If you go back far enough in the history of
computing, the CPU didn't have a cache.

Instead, they had some sort of cache DIMM. There was
a socket for the cache DIMM, and the user was expected
to pony up for this "missing bit of their CPU" :-)

And then there are the various levels of cache: L1, L2, L3. I remember
when one (and maybe then two) levels were inside the CPU (at least
inside its package), with the last one outside (the inner ones were a
smaller amount of memory). I don't even know if that's still the case -
I stopped paying close attention to chipset architecture about that
point, if it isn't.

Also, caching schemes back then, only "covered" a
limited address range. Maybe you had 512MB of main memory,
and the cache only covered the first 256MB. Programs
running in the lower part of memory "go like snot".
Programs in the upper part of memory would be
quite slow by comparison (no cache to hit on).
So not only was cache an "option", it didn't even
match the arch it was plugged into :-)

I hadn't come across that particular wrinkle! Sounds vaguely connected
with the 640K, paging, HIMEM, and so on stuff I've forgotten all about
.... (-:

*******

This motherboard covers a few concepts. See page 11,
[]
The L2 consists of two chips, soldered to the motherboard.
(No cache DIMM in this case.) There's an L2 cache chip
(which should really be 64 bits wide or so). There's
a TAG RAM, and that might be the Content Addressable Memory
that controls the cache static RAM chip.

The board has an SIS Northbridge with integrated graphics.
The graphics were so gutless (a basic frame buffer), the
chip only draws 2W and doesn't need a heatsink. But next
to the Northbridge, is video RAM for the frame buffer.
Four chips, totaling 8MB. The width of the memory array,
would be a function of what room they had for pins for the
purpose, on the Northbridge. While you would prefer the
four chips to be 16 bits wide each, they could do anything
they wanted for a frame buffer. Maybe that RAM interface
only worked, if the board features no AGP slot.

I have been trying to remember whether the board I'm thinking of had a
GPU, as such; it might have just been a frame buffer, as you describe.
It definitely had separate video RAM (at the time I'm thinking of, the
main criticism of on-board video was [a] that it ate into system RAM,
and [b] it slowed things down a lot as it was addressed over the same
bus as main memory [as it would have to if it _was_ main memory). The
advantages being plugged were that it _didn't_ use up some of the main
memory and it was addressed via a _separate_ bus. (But what was
addressing it, I can't remember.)

In those days, most on-board video (with _shared_ memory), you had to
take into account what resolution you were running at: if a high one,
the amount of RAM available for general use was lower. This makes me
think there _was_ no GPU as such, and the main CPU was generating all
the graphics - though presumably with a frame buffer at least. I think
this was from the period when both system memory and video memory were
in the single numbers of MB, not GB (and thus very OT for W7 and W10 [or
even XP!] 'groups! But I'm continuing it for historical curiosity).

(I remember being puzzled why "graphics cards" _needed_ so much memory;
surely they only needed enough to store the screen! Or maybe two or
three buffers' worth, anyway. That was before I understood the concept
of the separate graphic _processor_! [I was never much of a gamer. I
remember being impressed with the wall "texturing" in DOOM. I don't
think I've played any graphic-intensive game since those days!])

Static RAM never really ran all that fast. But maybe two
years ago, I found a static RAM chip that runs at 2GHz.
Only trouble is, each chip is $500, which means there's
likely only one (desperate) customer for it :-) If the
chip was $4, I could "see some interesting home projects".
But at $500, the item goes into my "what were they thinking"
file. It might cost $20K to do a home project.

Indeed! (It's a lot easier to design with though - no need to worry
about refresh. But I haven't played with it since its capacity - and in
DIL packages, mostly! - was measured in K. [And often less than 8 bits
wide: 1 bit not unknown.])

Paul
John
--
J. P. Gilliver. UMRA: 1960/1985 MB++G()AL-IS-Ch++(p)Ar@T+H+Sh0!:`)DNAf

"Bother," said Pooh, as Windows crashed into piglet.