Can I increase the Video Memory of an integrated HD card to improve FPS/performance? Is there a maximum?

mickael28

Reputable
Aug 7, 2014
26
0
4,530
Hi,

I bought a new mini-tower PC that should arrive next week. It's got an i5 6th generation, integrated intel HD 530 graphics card and 32GB RAM.

I've never done it but I was reading today that there's a way to assign more RAM to an integrated HD graphics card that could help with the performance on games.

Do you guys know if this is indeed possible and there's an actual difference in the performance and FPS when assigning more video memory?

And is there a limit on the maximum RAM one can assign. Eg, could I assign 24GB of Video Memory and keep 8GB of normal RAM? and how would that compare to let's say a dedicated graphics card with only 1GB or 2GB of dedicated memory? ie, would the dedicated be better than an integrated with a much higher assigned memory (maybe x12 or x24 times higher in this case)?

Cheers!
 
All extra memory does is give the GPU more space to queue up textures for scenes. You don't need to manually assign it it automatically takes what it needs from your system ram that is not being used. I guess in theory if you could find a way to force it to dedicate more so its not limited by what is in memory already it may net you a couple fps, but honestly with 32gb, that integrated gpu will have all the ram it could ever need.

Basically, especially for you, this is a waste of time.
 
I see, I won't do anything then and see how things run. I am not much of a PC game player, and I mainly bought it for business purposes, but thinking about it once I ordered it, I saw the oculus rift and I was thinking about getting one for giving it a go, but not quite sure it'd be able to run Virtual Reality stuff or if I could improve it somehow.

The pc in question is the 'hp elitedesk mini 800 g2', with a 256GB SSD drive.

Any view about what's required for this VR stuff to run kind of smoothly?
 
The main problem is not memory so much as it is the sharing of that memory, AND the processing power of integrated graphics. Having to access shared system memory is slow, this is why videocards come with their own memory. Sharing memory is very cost effective though, which is why it's the goto choice for business use. Processing power is the other concern. Having access to 32gb of ram is worthless unless the integrated graphics could actually process that amount of data fast enough to make it useful. Needless to say, it can't.

You'll want a fast videocard for VR. An AMD RX 480 8gb or Nvidia 1060 6gb is where I'd start to look if I was putting together a basic VR system.
 
Since it has no dedicated amount of vram, it uses system ram as vram. Setting a pre-allocated amount makes no difference in performance as nothing changes in how much it can use. By default, windows 10 limits dvmt to half your total ram. http://www.intel.com/content/www/us/en/support/graphics-drivers/000020962.html This is a dynamic amount so it can be used by the system or the igpu as needed.

More vram does not mean more performance as it will only use as much as it needs but being so weak, it'll probably never go above 2gb in games. The amount of vram does not give more performance on a dedicated card either. As you can see from the chart, all that really matters is what gpu model it is.

Any of the vr stuff has requirements listed. You are way under on what's required as far as gpu goes.
 
The amount of video memory isn't really the issue.

Computer RAM has a bandwidth of about 10-20 GB/s. This is due to the width of the data channel (64 bits transferred per request), and distance from the CPU to the RAM (light only travels 7.5 cm during one clock cycle of a 4 GHz system). Integrated graphics cards use the computer's RAM as their video memory, so are limited to these memory speeds.

A dedicated graphics card has its own built-in memory. They're designed to respond faster (DDR5 vs. DDR3 and DDR4 for system RAM), have a wider data channel (256-384 bits transferred per request on higher end GPUs), and are located right next to the GPU so the electrical impulses do not have to travel very far. This allows them to hit 200-500 GB/s.

So yeah you could increase the amount of memory reserved for your integrated graphics, but it wouldn't really help speed things up because it's being bottlenecked by the memory bandwidth, not the amount of memory. Video memory is primarily used to hold textures, and almost any modern game run at 1080p will be just fine with 2-3 GB of video memory (most will run in 1 GB or less). Additional video memory is only needed to hold larger textures (2k or 4k). But to see an improved image from those larger textures, you need to be running the game at higher resolutions than 1080p. And an integrated video card is so slow it's not worth running them at resolutions higher than 1080p. So there is no need to assign them more video memory.
 
Thanks for the explanations, I see it won't make much difference now.

Is the list in the link provided above, http://www.tomshardware.com/reviews/gpu-hierarchy,4388.html?_ga=1.257786504.204844781.1483751292, sorted by performance then? ie, the higher the better in that list?

And is there a way to plug any of the good cards for VR into either an usb port-c or usb 3.0? (as this is a mini pc and I think it doesn't have slots inside for a full size card)...
 
The motherboard's BIOS determines the maximum amount of memory that you can allocate for the IGP.

For example the ASUS H170-PRO/USB 3.1 motherboard only allows a maximum of 1024MB to be allocated for the IGP. The ASUS Z170-PRO only allows a maximum of 512MB to be allocated for the IGP.
 
As I mentioned, you have m.2. It is similar to that video but keep in mind your ssd is in the slot you need. It probably has the os on it too. Something more neat will never come out simply because these are all internal connections and components and you want an external device. There's no way around that except having a cable sticking out of it and no cable management.
 
@k1114, but what could you install in that m.2 slot?

Is there any graphic card that could go directly there without additional external power and which would be powerful enough to run VR applications?
 

Some USB-C implementations include an alternate mode which will support an external graphics card. You will have to check with your motherboard's specs to see if one of these alternate modes is supported. e.g. If the board supports Thunderbird alternate mode, then you would (in theory) be able to use a Thunderbird external video card adapter.

https://en.wikipedia.org/wiki/USB_Type-C#Alternate_Mode_partner_specifications
http://www.pcworld.com/article/2984716/laptop-computers/how-to-transform-your-laptop-into-a-gaming-powerhouse-with-an-external-graphics-card.html

If the motherboard includes a slot for a WiFi card, that is a mini PCIe interface. Some people have hacked it to connect an external GPU to the system (only PCIe x4 speed though).

http://forum.notebookreview.com/threads/diy-egpu-experiences.418851/
http://superuser.com/questions/999490/egpu-connected-through-wifi-card-mini-pcie
http://amfeltec.com/products/flexible-minipci-express-to-pci-express-adapter/
http://www.hwtools.net/adapter/pe4h.html
http://www.gearbest.com/laptop-accessories/pp_229101.html

There's a web page out there by someone who did this with their laptop, so it's definitely possible. It has lots of pics and benchmarks, but I can't seem to find it. You are constrained by the PCIe x4 limit, and the need for an external enclosure + power supply usually means you're better off just replacing the motherboard and case for a different one which lets you plug in a full-size PCIe card. But if you really want to do it, you can.

Be forewarned that the mini PCIe slot is physically identical to a mSATA slot (manufacturers tried to be cheap and saved money by using the same physical slot for two different functions). So make sure it's for a WiFi card, not for a mSATA drive.
 
There's a reason why gpus are so big. We just don't have the technology to have that performance so small and sip power. You'll have to wait years like how phones are as powerful as top pcs from 10 years ago.

M.2 is typically used for ssds, like what you currently have in it. It uses pcie protocols which is why a gpu can be connected to it but it's not made for gpus and you run into the diy realm. All gpus are normal pcie size which is too big for a mini pc.

@Solandri, that is thunderbolt which is thunderbolt not usb. Thunderbolt combines usb and pcie. Usb does not have pcie. Usb is usb only. These ports are clearly marked because of the differences.

I can go in depth about diy and commercial products but all he has is mpcie and m.2 so there is no need. These 2 are similar connections but I did not mention mpcie because it's x1 while m.2 is x4. Mpcie is being used for the wireless card here but because of the bottleneck, I would not recommend it for an egpu when the m.2 is right next to it.

Msata being the same is not about saving money. That's like saying thunderbolt 3 uses usb type c to save money. The real reason was for easier adaptation. M.2 is msata's replacement so you won't find both on a single device.
 
So if I understood it correctly, there might be a workaround to install an eGPU via the m.2 slot (after adding a standar GPU to a case, power it up, etc)?

I couldn't find an article, DIY info for this setup. Could you please see if you manage to see something useful to have a further look (I don't quite know what I should start researching).

I need to go at the moment. But I'll have another look and research further over the weekend.

Thanks for the ideas...
 

That was true up to USB 3.0. OP states he has USB-C (USB 3.1). USB-C has an alternate mode built into the spec which allows transmission of non-USB signals over the USB-C connector. Not all USB-C implementations take advantage of this, but of the ones that do the common alternate modes supported are Displayport, MHL, HDMI, and Thunderbolt. OP would have to refer to the manual or motherboard specs to determine which, if any, alternate modes his motherboard supports.

https://en.wikipedia.org/wiki/USB_Type-C#Alternate_Mode
 
You are stil confusing usb vs the use of the port in other implementations. Right from your own link: "The USB Implementers Forum is working with its Alternate Mode partners to make sure that ports are properly labelled with respective logos." Like I said, it will be clearly marked. A usb port labeled as a usb port is only usb. There's the usb logo right above the port and you can google the g2 specs. If it supported alternate modes, it will have the other logo/s and list it in the specs.
 
@k1114. I got what you said now, thanks.

Yes, it seems the G2 has the usb logo (unfortunately, so I guess it won't support thunderbolt speed there, which I guess make it already incompatible for the new external cases to fix this problem).

Regarding a graphic card to stick on the m.2 slot. Could you please have a quick look when you've got the chance to see if you see any shop where they sell them? I don't really know how they look or what I am looking for, I was trying to google by 'm2 graphics card' but the images that I see don't look like a graphic card with a m2 connector, plus I've not seen any shop were to get one (or an adaptor for this solution)
 
Thanks, I had not understand that connection type (PCIE x16). I'll have a look.

I noticed that the motherboard has two m.2 slots, one for a wifi card and another for an ssd drive. If such an m.2 to pcie x16 adapter exists, could I plug it in any of those 2 slots? Or I would need to make sure that it's just plugged into one of those, which one?
 
Cool, I need to re-read all the info now that I understand a little bit more. At the beginning I thought x1 and x4 referred to the number of slots available, but it seems it refers to the speed instead so it was difficult to follow the content.

Many thanks for the help. I'll see if some kind of m.2 to PCIE x16 adapter exists, now or in the near future.

Cheers!