Onboard Graphics -- Onboard Memory!! (noob question really)

[jaykishankrk]

Hello all,

i would like to know whether,

there exists any OnBoard Memory for OnBoard Garphics??

i do know that it shares the memory from the RAM installed on the system but apart from this doest it have any onboard memory??

this might be a stupid question but i am really not getting any answer for this. Googled results did not fetch me an answer

 

[jaykishankrk]

Come on guys i need one word answer just say yes or no

 

[Stumps]

Old motherboards used to have a seperate frame buffer for the onboard video, these days it's all shared with the system ram.

 

[Bill Brasky]

AFAIK, no. Integrated graphics only use system memory.

 

[jaykishankrk]

Originally posted by: Stumps
Old motherboards used to have a seperate frame buffer for the onboard video, these days it's all shared with the system ram.

can u give me an Chipset Model number if u have any??

 

[memories2002]

Nope.

 

[Peter]

SiS has been offering the option for ages - like, the 620 and 530 chipsets (Pentium-II and Pentium, respectively) as well as the 630 P-III chipset let you have it. Not sure about the inbetweens, but even today, on the M760, dedicated RAM is supported.

http://www.sis.com/products/sis760.htm

To enhance system performance, the SiS760 also supports the local frame buffer solution and can support SDRAM and SGRAM. SiS760 can support up to 128MB display memory with shared memory and/or local frame buffer.

ATI appear to be introducing the feature in the 690G, apparently.

 

[Stumps]

Originally posted by: jaykishankrk

Originally posted by: Stumps
Old motherboards used to have a seperate frame buffer for the onboard video, these days it's all shared with the system ram.

can u give me an Chipset Model number if u have any??

IIRC the old Intel 810 came with a seperate frame buffer on some mobo's, usuallly 4mb. some Packard bell's and HP's had a Voodoo 3 with 8mb buit in, many all in one Super7 boards had a trident (not blade 3D) with 1 or 2 mb onboard.

There are heaps....but it pretty much stopped once the MVP4 came out and introduced the UMA setup of sharing system ram.

 

[jaykishankrk]

Originally posted by: Peter
ATI appear to be introducing the feature in the 690G, apparently.

it appears that 690G will have 128 to 256mb frame buffer built in with the board. this is really kind of amuses me as to why we need to back to OLD setup if the present shared memory concept has been proved to be quite efficient. is there any advantages for doing so??

 

[Stumps]

Originally posted by: jaykishankrk

Originally posted by: Peter
ATI appear to be introducing the feature in the 690G, apparently.

it appears that 690G will have 128 to 256mb frame buffer built in with the board. this is really kind of amuses me as to why we need to back to OLD setup if the present shared memory concept has been proved to be quite efficient. is there any advantages for doing so??

yes, it allows the VGA Frame buffer(VRAM) to run much faster than the main system ram, and allowing much better performance of the intergrated VGA....but I suspect that boards with 128mb or 256mb built into them would be pretty pricey for all in one types.

 

[Peter]

The word is "integrated", thank you.

AMD64 chipsets have another reason beside performance: Power consumption. With the graphics unit using main RAM, which is on the CPU, it has to keep its bus interface and RAM controller alive during CPU powersave states, so that the graphics unit can still show a picture. With local graphics RAM, the CPU can put all its functional components into halted state.

 

[hans007]

There were some radeon xpress 200 boards (i think they renamed it xpress 1100) that had it. the rs480 chipset. I had one from jetway that had like 32mb onboard 32bitframe buffer. most companies did not bother with it since it did not help much. it was basically like the x300 hypermemory but for integrated video.

 

[Peter]

Originally posted by: StumpsIIRC the old Intel 810 came with a seperate frame buffer on some mobo's, usuallly 4mb. some Packard bell's and HP's had a Voodoo 3 with 8mb buit in, many all in one Super7 boards had a trident (not blade 3D) with 1 or 2 mb onboard.

There are heaps....but it pretty much stopped once the MVP4 came out and introduced the UMA setup of sharing system ram.

The 810 didn't use local RAM for the framebuffer, just as a 3D z-buffer cache of sorts.

UMA by far preceded the VIA MVP4. It started out as a sideband interface on chipsets that didn't have a VGA inside but let a discrete VGA chip borrow system RAM - as demonstrated by the SiS 5510 northbridge in combination with the SiS 6201 VGA chip.

The first commercially relevant chipset to have an integrated shared-memory VGA was the SiS 5596, first presented at Computex 1996. Its followup a year later, the 5597/5598, then was the first all-in-one chipset, integrating north, south, and VGA functions. (Dear NVidia marketing dept., you are nine years late for "presenting the first single-chip chipset with graphics function" as you claimed for the 6100.)

 

[jaykishankrk]

Originally posted by: Stumps

Originally posted by: jaykishankrk

Originally posted by: Peter
ATI appear to be introducing the feature in the 690G, apparently.

it appears that 690G will have 128 to 256mb frame buffer built in with the board. this is really kind of amuses me as to why we need to back to OLD setup if the present shared memory concept has been proved to be quite efficient. is there any advantages for doing so??

yes, it allows the VGA Frame buffer(VRAM) to run much faster than the main system ram, and allowing much better performance of the intergrated VGA....but I suspect that boards with 128mb or 256mb built into them would be pretty pricey for all in one types.

r u suggesting that the frame buffer used will be much faster in terms of their clock speed than the system ram speed ??

 

[bigsnyder]

Yes, frame buffers will generally be much faster for 3D functions vs. system memory.

C Snyder

 

[Schadenfroh]

Some old S3 chips that were on motherboards used to have video memory separate from system memory, that that would be Pentium 1 days at the latest.

 

[Steve]

Integrated video = shares system RAM
Onboard video = brings its own RAM

My first computer, an IBM Aptiva with a K6-2 and SiS 530 had 8MB of separate RAM for video. My first build had a PC Chips M847LU with a built-in Xabre 200 GPU (yes, GPU) and 64MB of its own RAM. And my HTPC with an nForce2 board has an integrated GeForce2 which I don't use.

 

[hans007]

like i said earlier.

theres , onboard video which is basically just a real video card soldered onto the board. such as those old s3 cards, or the voodoo3 1000 that a few boards had before or the xabre 200.

there is integrated video which is part of the chipset.

most integrated video does not have its own memory. the only ones that ever did were the rs480 ATI (which was optional... and worked in the same way as hypermemory on their discrete cards like the x300) and the only board i know of that actually had it was some variants of the jetway brand motherboards using that chipset (i know I HAD one, newegg used to sell them, but then stopped selling the ones with onboard memory). the i810 also had optional 4mb zbuffer memory and that is the other chipset that had one.

i believe one of the reasons that the rs480 supported real memory for itself, was that on athlon 64s the latency of getting system memory through the cpu was more than it would have been on say a chipset that also had a memory controller (like an intel chipst)

 

[Steve]

Um, that Xabre 200 board I mentioned is a step above onboard video. Onboard video generally does not feature a discrete GPU and all the hardware of an AGP video card built into the board

 

[Golgatha]

Originally posted by: sm8000
Integrated video = shares system RAM
Onboard video = brings its own RAM

My first computer, an IBM Aptiva with a K6-2 and SiS 530 had 8MB of separate RAM for video. My first build had a PC Chips M847LU with a built-in Xabre 200 GPU (yes, GPU) and 64MB of its own RAM. And my HTPC with an nForce2 board has an integrated GeForce2 which I don't use.

It's a matter of semantics.

Integrated video = shares system RAM
Onboard video = brings its own RAM

I like to think of onboard video as being essentially an add-in card that's hard soldered onto the motherboard to save space.

 

[Steve]

Originally posted by: Golgatha

Originally posted by: sm8000
Integrated video = shares system RAM
Onboard video = brings its own RAM

My first computer, an IBM Aptiva with a K6-2 and SiS 530 had 8MB of separate RAM for video. My first build had a PC Chips M847LU with a built-in Xabre 200 GPU (yes, GPU) and 64MB of its own RAM. And my HTPC with an nForce2 board has an integrated GeForce2 which I don't use.

It's a matter of semantics.

Integrated video = shares system RAM
Onboard video = brings its own RAM

I like to think of onboard video as being essentially an add-in card that's hard soldered onto the motherboard to save space.

Integrated video = shares system RAM
Onboard video = brings its own RAM

That's exactly what I said earlier. Point is, onboard GPU is more than onboard video, even if they feature dedicated RAM in common

 

[Peter]

Nope, sm8000, there are (and have been for a long time) chipset-integrated VGA units that /do/ have dedicated RAM, just as well as there have been discrete VGA chips that didn't.

It's about technology, not semantics. Just realize that we have two different aspects mangled into one: (1) is the VGA engine integrated into the chipset or is it a discrete chip, (2) does the VGA engine have dedicated RAM attached to it or does it share system RAM. These topics are typically tied together as integrated=shared, discrete=dedicated, but not necessarily - which is the OP's question.

btw, (2) gets even more complicated when we look at technologies like AGP, PCIE HyperMemory and friends, and SiS/NVidia chipsets that could run their integrated VGA engines in half-dedicated half-shared mode.

 

[hans007]

Originally posted by: Peter
Nope, sm8000, there are (and have been for a long time) chipset-integrated VGA units that /do/ have dedicated RAM, just as well as there have been discrete VGA chips that didn't.

It's about technology, not semantics. Just realize that we have two different aspects mangled into one: (1) is the VGA engine integrated into the chipset or is it a discrete chip, (2) does the VGA engine have dedicated RAM attached to it or does it share system RAM.

(2) gets even more complicated when we look at technologies like AGP, PCIE HyperMemory and friends, and SiS/NVidia chipsets that could run their integrated VGA engines in half-dedicated half-shared mode.

there are no discrete chips integrated onto the board, that ever had access to system ram.

 

[Peter]

Hans, please bother read what I said above. UMA started out as a sideband interface on the north bridge, for the sole purpose of letting discrete graphics chips use system RAM. Early Pentium days that was, and there have actually been mainboards that used it. (SiS 5510 NB and 620x graphics chips, mostly. VIA had the UMA interface on their early NBs too.) This was a VESA standard mechanism, northbridges and graphics chips not having to be from the same vendor.

Pulling the VGA engine into the NB was the next step, also pioneered by SiS, but it wasn't how it started.

 

[hans007]

Originally posted by: Peter
Hans, please bother read what I said above. UMA started out as a sideband interface on the north bridge, for the sole purpose of letting discrete graphics chips use system RAM. Early Pentium days that was, and there have actually been mainboards that used it. (SiS 5510 NB and 620x graphics chips, mostly. VIA had the UMA interface on their early NBs too.) This was a VESA standard mechanism, northbridges and graphics chips not having to be from the same vendor.

Pulling the VGA engine into the NB was the next step, also pioneered by SiS, but it wasn't how it started.

ah thanks for the clarification.