• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

970's having performance issues using 4GB Vram - Nvidia investigating

Status
Not open for further replies.
I know I created this thread but the content of them always is the same and this is textbook lol.

You have those in denial.

The purchase justifiers try to explain why the problem wouldn't be that bad if confirmed.

Then there's the armchair software engineers doubting the method of measurement.

The law suit guys.

The impatient refund guys.

And the buying a competing product guys who enjoy the situation a bit too much.

Well I'm none of those. I'm currently unaffected and waiting with interest to see what happens.

My concern is that if it's better to only use 3.5gb, can you cap memory usage?
 
I just realised something...

If the 970's memory problem was due to some reason of how it was cut down from the 980...what about the new 960? I mean if they were cut down using the same method, would they also potentially have the problem of memory bandwidth taking a dive, when memory usage exeed 87.5% of the total capacity (1.75GB)?

The 960 is a different chip, not a cut down GM204, the upcoming 960ti on the other hand who knows.
 
ok but we dont know what the usual fps is?
the issue is it tanks after 3.5?

I had SOM sitting at 3.7gb and fps was over 100. I haven't had time to test properly, but it looks like there have to be other factors in place as not everyone can say for sure they are affected, which explains why this issue hasn't been raised much sooner.
 
ok but we dont know what the usual fps is?
the issue is it tanks after 3.5?


well, dropping down to 2560x1080 drops the ram using to 3.5gb and increases the fps from averaging 30fps to averaging 50fps, it that area of the map anyway.

8waLz1ih.png.jpg

So there is quite a drop there however i dont know if that's because of the ram usage or because i'm hitting bandwidth limits or something else - it is ac:u after all and i'm running everything up full for these screenshots. im trying to find something that allow me to fill the ram without loading the gpu up so much, get a better idea of what's going on.

Kitch9 said:
There's always one denying their denial too lol.

IPhone dude? Really?

I'm not in denial, i'm just trying to be factually correct here.
 
Last edited:
well, dropping down to 2560x1080 drops the ram using to 3.5gb and increases the fps from averaging 30fps to averaging 50fps, it that area of the map anyway.

8waLz1ih.png.jpg

So there is quite a drop there however i dont know if that's because of the ram usage or because i'm hitting bandwidth limits or something else - it is ac:u after all and i'm running everything up full for these screenshots. im trying to find something that allow me to fill the ram without loading the gpu up so much, get a better idea of what's going on.



I'm not in denial, i'm just trying to be factually correct here.

Why are we comparing graphics cards to phones instead of other graphics cards then?
 
well, dropping down to 2560x1080 drops the ram using to 3.5gb and increases the fps from averaging 30fps to averaging 50fps, it that area of the map anyway.

So there is quite a drop there however i dont know if that's because of the ram usage or because i'm hitting bandwidth limits or something else - it is ac:u after all and i'm running everything up full for these screenshots. im trying to find something that allow me to fill the ram without loading the gpu up so much, get a better idea of what's going on.

yeh you need a better way of testing that, way too many variables
isnt there any vram speed tests out there ><

you could open a ton of browsers im sitting here in windows using 500mb of vram with 20 browsers open, tho thats just as lame i guess
 
I had SOM sitting at 3.7gb and fps was over 100. I haven't had time to test properly, but it looks like there have to be other factors in place as not everyone can say for sure they are affected, which explains why this issue hasn't been raised much sooner.

I don't think the issue is whether the memory is in use or not, but whether that memory is actually USED. If the GPU isn't currently using that part of the memory for anything (just assets on standby, ready for use) your performance wouldn't drop. And I think it would be very difficult to ensure you get the GPU to use that part of the memory, so it's more likely it shows as a sporadic problem.
 
Why are we comparing graphics cards to phones instead of other graphics cards then?

you dont need to quote the screenshots every time you respond to me :p


I'll explain once more. I compared Legend's argument to the situation we had with the release of the latest galaxy phones and iPhones because people in this thread (Legend being one of them) are claiming that the last 500mb or so of ram is simply unavailable. This was the case with the phones where the OS took up so much of the advertised storage (in the case of the 16gb models). So very similar situation. Phones got released and people got angry, but did apple/samsung do anything about it? nope. No lawsuits, class action or any of that rubbish. Now, never mind the fact the phones lost far more storage that what was being claimed here; the real difference here is that the ram IS very much much available, as the last couple of screenshots in this thread show. So, again, that argument is silly and needs to be dropped.

The performance issue, the issue you brought up in the first post, is very real however, as i have maintained so no- again, I am not in denial of anything. Perhaps i'm just trying to be too rationing in a forum full of hotheads.
 

i know when i use to play some unreal engine games you can edit the ini files to change the size of the textures that are displayed, that affected the used vram a lot! that might help reach the magic number without changing res and all the rest of it

just a thought
 
Well, I tried some BF4 maxed at 4K with about 10% extra resolution scaling.

Was filling up to just under 3.8GB but maintaining around 55/60 fps, and the only major stuttering I noticed was when dying, for a sec. Didn't seem to be running too badly by any means.

Until it simply crashed. Whether crash caused by this I could not say. I don't normally play with such high settings though, so not normally maxing out the vram to such an extent.

Will maybe try another couple of games through the day, but I want to enjoy myself a bit so not going to spent a lot of time on it.
 
i know when i use to play some unreal engine games you can edit the ini files to change the size of the textures that are displayed, that affected the used vram a lot! that might help reach the magic number without changing res and all the rest of it

just a thought

I'll see what i can find :)


3641mb used, 59fps

k4ph69lh.png.jpg

Shiari said:
I don't think the issue is whether the memory is in use or not, but whether that memory is actually USED. If the GPU isn't currently using that part of the memory for anything (just assets on standby, ready for use) your performance wouldn't drop. And I think it would be very difficult to ensure you get the GPU to use that part of the memory, so it's more likely it shows as a sporadic problem.

That seems to be the case!
 
Just played BF4 @4k with 4xAA and High AA post process enabled, over 3.7GB VRAM used:

Lxmb.png


Frame rate was perfectly playable 40-50fps with no frame rate plunges or anything you wouldn't expect @4k 4AA.
 
All the naysayers, deniers and anyone who's basically saying "shut up, the card has 4GB, stop moaning" (which is a surprising number of people)... Nvidia have already said they are aware of the issue (not that this definitively admits a fault I know). Now, unless they turn around and say the same as what you're saying (i.e "stop complaining and enjoy your perfectly functional cards")... you're all going to look a bit silly and have egg all over your faces, because it is quite obvious to MOST people that there is a problem here. Once Nvidia acknowledge that and outline what they intend to do about it is what I'm eagerly waiting for. But I'm fed up with people making comparisons to phones, SSD's etc, it's just ridiculous. But hey, if Nvidia DO turn around and tell us all to just shut up, I'll gladly eat humble pie... but I simply don't see that happening based on the evidence I've seen of a serious issues with these cards.
 
Last edited:
Nope, hence all the scare mongers having a field day :)

This.

There needs to be a defining standard of testing and that will be what nVidia are doing. They will have a better way of testing and if they find something, they will correct it. No need for such hyperbole and having a pop at each other.

I remember the 320.18 card killing drivers and the most vocal were those that used AMD cards. Many voices make loud noises :D
 
Status
Not open for further replies.
Back
Top Bottom