How to tell if Windows 7's info on GPU indicates it's integrated or dedicated?

I have access to a cheap HP dv6 laptop and would like to figure out if it's GPU is integrated or dedicated.

When I go into Control Panel>Appearance and Personalization>Display>Screen Resolution, then click on "Advanced Settings," a window pops up. Under the "Adapter" tab it shows 2 bits of info that appears conflicting: that it has 64MB of "Dedicated Video Memory" and 1631MB of "Shared System Memory." The latter part sounds like it's an integrated GPU. Based on that info, can I determine what kind of GPU it is? Or should I be looking elsewhere?

Additionally, it says it has 1695MB of "Total Available Graphics Memory." Could someone help me understand what that means?

Thank you for any help.

 

It sounds like you have an INTEL 4500MHD which is a integrated gpu in the motherboard chipset, it can have 64mb of dedicated memory according to some websites, and the rest of the memory is stolen from your notebook system memory.

 

It's like you're looking over my shoulder! Yes, it's a 4500MHD. The first several Google results I had checked said it's integrated, but that 64MB of "Dedicated Video Memory" is what confuses me.

How can it have 64MB of dedicated memory, but still be integrated? Isn't that conflicting? I thought to be integrated it can't have any dedicated memory at all.

Thanks.

 

It`s all very confusing , some websites say it have 64mb of dedicated memory, so say it`s all system memory, if you run gpu-z it will say 64mb , but it might be dynamically allocated and 1695MB is the maximum amount.

 

What does dynamically allocated mean? That it could be a hybrid, with 64MB as the dedicated and using up to 1631 of system memory?

And is Windows 7 often this conflicting with info or is it just this particular GPU the odd one out? So if an Nvidia 330M with 512MB was the GPU, Windows would show 0MB as "Shared System Memory" and if there was another integrated GPU, it would show 0MB "Dedicated Video Memory"?

 

If you run dxdiag save save the info it will say it has 64mb of dedicated memory, but i think it is wrong and the 4500MHD really has only shared memory and no dedicated

 

It means once the dedicated GPU memory is filled it will go on your RAM, and consume as much it needs UP TO the memory amount mentioned.

Yes. Sadly 64MB is not even enough to manage Windows Aero at anything above 1280x1024... so it uses your RAM like almost all the time. If it was at least 128MB it would be much better As it would give you a chance to keep your RAM free when you don't run anything that really uses the GPU.

This is a complicated issues, as it could be many factor. It depends if the panel system, information that you look at, shows basic detailed or very deep details of your system as different approaches are used. As most people don't have 2 different kind of GPU in one system (or at least have the other graphic card disabled in the BIOS), it's usually of no concern. Also the information provided can also be due to the BIOS, all depending on how it gets the info.

 

When I asked above whether the Intel GPU behavior was like a hybrid GPU, I wondered that because I thought maybe this was unique to this Intel GPU.

But then I just recalled something I read a few weeks back, that in fact, this behavior is common to all dedicated GPUs, is that correct? So if a dedicated ATI/Nvidia's 1GB of VRAM is used up, the OS, Windows or OSX or Linux, will move on to RAM. So all dedicated GPU's dynamically allocate or are "hybrids" in my likley incorrect terminology.

So if I extend that further, that would mean integrated graphics never dynamically allocate? As they're using regular RAM to begin with.

Thank you!

 

Discrete GPUs these days have their own discrete set of memory. They do not go to the system memory if it runs out. Some older GPUs used to have a split between dedicated and shared memory (ex: HyperMemory - Wikipedia, the free encyclopedia) This isn't the case anymore. Hypermemory/TurboCache are the "hybrid" GPU memory systems you're thinking of.

In the case of integrated GPUs, they set aside a minimum amount of memory for graphics usage which the system cannot get back. If more memory is needed, more will be grabbed, but it's rare that an integrated GPU will request more memory from the system. However, it can, so it's technically dynamic.

 

Correct,

Now your System RAM is much slower than your typical GPU dedicated memory (if any). Hence why there is a big performance reduction. If Intel GPU used 128/256MB of own memory, it would be much much faster GPU's (but also much more expensive, rendering Nvidia/ATi even a better deal).
BUT, that is not all, You see, in order for the GPU to access your main memory, it needs access to your system BUS. And that is only 1 at a time (per cycle) access. So basically it reduce CPU and GPU performance as they both battle access to the memory.

Now you may ask, "But why not have 2 buses?!", The answer is: We wish.
It would be hyperly complicated to implement, and current memory technology doesn't allow this. For starters, your own memory would need a controller and processor system to manage all the data, and use a database approach system so that multiple access/read/write can be performed without any conflicts, all by preventing 2 critical things: starvation (a process or device or hardware waits for ever to get access), and potential deadlock problem ( http://en.wikipedia.org/wiki/Deadlock), which up to today, doesn't have a real solution to it.

 

Thanks for confirming. I was just taking a guess by extrapolating on something I had read.

The last sentence seems to be the opposite of what I proposed and GoodBytes confirmed. Have I misinterpreted what you're saying?

And when you say that discrete GPUs now have their own discrete memory, are you saying there's now another set/layer of memory in the discrete GPU besides just the VRAM portion? The way you've worded it, it sounds like you're saying that discrete GPUs in the past did not have discrete memory.

Based on what I shared in my initial post with the Windows 7 info on the HP's GPU, that 64MB of "Dedicated Video Memory" sounds like it's the equivalent of Hypermemory/TurboCache. Just Intel's version. It would be acting as the "small local memory" described in the Wiki you linked. Would you agree?

And when you say that "this isn't the case anymore," that GPUs no longer use dedicated and shared memory, that seems counter to what GoodBytes said about all dedicated GPUs eventually using system RAM when GPU RAM is all used up. Or am I missing some nuances in both your wording?

Thank you both!

 

Integrated Graphics Processor borrows memory from your system, and uses the CPU to do or help all the processing required, depending on what is being demanded. New models, have a small quantity of it's own dedicated memory.

Discrete GPU, has it's own full graphical processor, but uses your system RAM as memory. However, some models has it's own memory, where it may or may not extend to your RAM when filled. This blurs the line from dedicated and discrete. You can identify them by having: a. Low power GPU, b. Have 2 models or make the OEM decide if they want to put dedicated memory or not.

Dedicated GPU, has it's own full graphical processor and it's won dedicated memory.

 

My laptop has a ATI Mobility HD 5870 display adapter with 1GB DDR5 and no on-board GFX. In the Adapter Properties I see the following:

Code:

Total Available Graphic Memory: 4091MB
Dedicated Video Memory:         1024MB
System Video Memory:               0MB
Shared System Memory:           3067MB

I hope this clarifies it better.

 

Do not go around trying to distinguish discrete from dedicated GPUs on laptops. Few people if any use the terms distinctly - the GPU can be integrated or discrete, the memory can be shared or dedicated - this is the typical terminology.

For the most part, modern laptops can have discrete GPUs, their own set of dedicated (unavailable to the system) memory and the ability to access shared system memory. However, I've never seen one take from system memory under normal circumstances. It's mostly safe to assume that discrete GPUs will not go to system memory for anything, especially considering the low clock speeds of mobile graphics and the fact that 256mb is already sufficient for the vast bulk of users. Going beyond 512mb of dedicated graphics memory is usually called a waste since it doesn't improve gaming performance (the typical reason for choosing a discrete GPU). Going to system memory for graphics purposes is slower and hampers performance, because data jumps through so many hoops, as previously noted.

When you reference 64MB of video memory, that's a 64mb piece of your system RAM that's been reserved for video purposes, it's not physically distinct from the sticks of general purpose RAM. The piece that's reserved for video memory can flex somewhat, but that flex takes up more of your system ram. The total available memory is the either the most that can be called or the most that's in your system.

It's somewhat of a lie to call most of the discrete notebook GPUs discrete anyway - many are built onto the same piece of silicon as the motherboard even if the circuitry and the dedicated memory are mostly wholly separate (this is the case on my 1558). Some still exist as independent boards that can be connected to the motherboard, but at least on my Dell, it's one large board. My e1505 used a separate riser card and was a hypermemory system.

 

Firstly, thanks to GoodBytes and woofer00 for explaining discrete and dedicated so thoroughly. Indeed, I had thought the terms synonymous, as with shared and integrated. Apparently I was wrong on both.

So if there's a discrete GPU with some of its own memory, shouldn't it then just be called a dedicated GPU?

You and GoodBytes are saying much the same here, except that you both categorize graphics cards slightly differently. GoodBytes breaks it into 3 groups anchored mostly by GPU, then memory as the secondary factor; while you break it down into 4 distinct groups, clearly separating GPU and memory, which feels easier to comprehend. However, is one approach more accurate or more commonly used when people talk about graphics cards.

woofer is saying that the graphics card in the HP laptop is an integrated GPU with shared VRAM. However, based on the info I gave on the HP--64MB of "Dedicated Video Memory" and 1631MB of "Shared System Memory"--doesn't it seem like it could also be an integrated GPU with dedicated memory, in accordance with the last sentence of GoodByte's quote above, with that 64MB serving as the dedicated memory?

And going by GoodByte's definition of "Discrete GPU" (" Discrete GPU, has it's own full graphical processor... However, some models has it's own memory"), couldn't it possibly be a discrete GPU with dedicated memory?

This is one area where I'm still uncertain about since my first post. How do you reconcile the presence of listed memory in both the "Dedicated Video Memory" and "Shared System Memory" section in Windows 7's Adapter info?

How is the "Total Available Graphic Memory" determined? In my instance, the HP's total is 1695MB, while the system has 4GB RAM. In Kalim's example, his system can use a total of 4GB, while he has 8GB of RAM, assuming it hasn't changed since purchase (I googled his laptop, ASUS G73JH-X1). So his system used 50% of system RAM, while the HP only uses about 40%.

Thank you again. This is very educational. And a brain killer.

 

I'm not familiar with the specific models, but my guess is there's a limit either to the amount of memory the GPU can address or to the amount of memory the controllers on the board can make available. The more I've been reading on laptop specs, the more I suspect that many laptop's with discrete GPUs use dedicated and can access shared memory as well. However, with regard to the 4500mhd, I'm pretty sure the dedicated memory is just the minimum that's been blocked off from the main memory, not a physically distinct chip that would fall into "dedicated" the way I defined it.

I'm on my netbook at present (on linux), but I'll check the numbers on my Studio 1558 and post em here once I get home to see if it can help make sense of this.

This link may help, particularly the first paragraph:
Notebookcheck: Intel Graphics Media Accelerator 4500MHD (GMA X4500MHD)


update:

just checked on my 1558, looks like my adapter has 1024 mb dedicated ram, and i showing 2782mb of hypermemory (not that it's actually in use at the moment).

 

Ah, so it may just be Windows' static wording unable to accommodate for all the graphics card variations and nuances. Thanks.

 

Seams like it.
This is one of the small problem when you try to make an OS to support a few xillion hardware configuration possibilities of a system and models which keep popping up continuously.

 

Well, it's probably just reporting what the Intel GPU driver tell it.