Jump to content
Search In
  • More options...
Find results that contain...
Find results in...
D2ultima

The video RAM information guide

Recommended Posts

Posted · Original PosterOP
On 3/8/2019 at 8:25 AM, Sett said:

Excellent (if quite old) guide. But There is a slight issue, and that's the "SLI memory bandwidth doubling".

No, technically it IS true, but..
 

The total number doubles, because now there are two GPUs accessing two sets of VRAM chips over two sets of copper lines.. But each GPU only has access to it's own copy of VRAM so the respective GPUs don't benefit at all (this is different from some Quadro/NVlink applications AFAIK),.

From a different point of view it means that there is double bandwidth, but also double demand for bandwidth, so you're breaking even, not doubling your assets.

The actual benefit is that each card only has to render half of the frames, so in total you can do more with the same ... times two. Therefore even if you are bandwidth limited for some reason, you should see some improvement from SLI.

For the 970 that means, that you WILL get SLI scaled performance when it comes to the last 1/2GB. So 13FPS instead of 8. Yay.

Read it a few times, I think we're talking about the same thing. If you're running 150fps single GPU then you run it SLI each GPU is doing (roughly) half the work, so each GPU only needs half the memory bandwidth demands due to the extra downtime. It does mean you get more of a memory bandwidth break because if one card was bandwidth bottlenecked at say 100fps and you were only using... 60% GPU, SLI would probably get you around 190fps and each card would be sitting around 60-70% uitlization (assuming good scaling).

 

NVLink or normal bridge, it doesn't change how SLI works. The change is the connection between cards, and how much data can be passed between each card (which is a HUGE benefit to SLI functionality (if Nvidia would get their drivers working that is). When you experience bandwidth choking with SLI (and generally poor scaling) you are in a situation where the game probably needs more data from the slave to the primary (and vice versa) than can be provided. It can't send completed frames AND the frame buffer data across without widening the choke. That was the entire point of the High Bandwidth bridge, and now NVLink SLI. It only even exists because Nvidia knows the BFGD it wants to put out simply cannot be powered by single GPU, but even PCI/e 3.0 x16/x16 + HB bridge isn't enough for 4K high refresh + gsync, so NVLink to the rescue.

 

I have a big stink about it in my SLI guide (you can see in my signature). I honestly would love to update these guides but for whatever reason, the guides are broken. They were written before this forum went to IPB v4, and after updating to IPB v4 they encountered... problems. I have had to simply copy and paste the ENTIRE guide without context and re-write them, re-add links, re-make tables, etc literally every single time I want to make more than a single small change near the bottom. If I could do this from scratch on a new page and have them re-stickied after they were done, where editing them would cause no problems with formatting or colour for night mode or light theme etc, then I would be happy to oblige. I haven't stopped learning it's just become extremely chore-ish to update.

 


Clevo P870DM3 (Eurocom) | i7-7700K | 32GB DDR4 2400MHz | GTX 1080N SLI | 850 Pro 256GB | 850 EVO 500GB M.2 | Samsung PM961 256GB NVMe | Crucial M4 512GB | Intel 8265ac | 120Hz Matte screen | 780W PSU

 

THE INFORMATION GUIDES: SLI INFORMATION || vRAM INFORMATION || MOBILE i7 CPU INFORMATION || Maybe more someday

Link to post
Share on other sites
On 8/16/2014 at 11:35 AM, D2ultima said:

It's 5:35am and I have no idea if you're serious or joking xD.

hey, I was wondering how to add your pc specs under your comment?

Link to post
Share on other sites
On 3/10/2019 at 4:30 AM, D2ultima said:

Read it a few times, I think we're talking about the same thing.

 

Well, like I said, It's technically correct.

The issue I have is that it might be somewhat misleading, just adding up the bandwidth - kind of like just adding up VRAM. You can do neighter, that's not how it works. A cards performance is a cards performance. You can sort of add up the results in SLI, but if a card has some issue, SLI won't fix it.
Anyways, that's just nitpicking on my part. I learned quite a bit here, and will definitely be redirecting people with questions this way. 👍

Link to post
Share on other sites
Posted · Original PosterOP
18 hours ago, Sett said:

Well, like I said, It's technically correct.

The issue I have is that it might be somewhat misleading, just adding up the bandwidth - kind of like just adding up VRAM. You can do neighter, that's not how it works. A cards performance is a cards performance. You can sort of add up the results in SLI, but if a card has some issue, SLI won't fix it.
Anyways, that's just nitpicking on my part. I learned quite a bit here, and will definitely be redirecting people with questions this way. 👍

Haha that's all fine, I'm good with the technicalities. There's often a line where you simply CANNOT explain something in more depth or people just won't get it, this would be one of those things. But yeah, it doesn't make one card faster at all.


Clevo P870DM3 (Eurocom) | i7-7700K | 32GB DDR4 2400MHz | GTX 1080N SLI | 850 Pro 256GB | 850 EVO 500GB M.2 | Samsung PM961 256GB NVMe | Crucial M4 512GB | Intel 8265ac | 120Hz Matte screen | 780W PSU

 

THE INFORMATION GUIDES: SLI INFORMATION || vRAM INFORMATION || MOBILE i7 CPU INFORMATION || Maybe more someday

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now


×