GeForce GTX 970s seem to have an issue using all 4GB of VRAM, Nvidia looking into it

Status
Not open for further replies.
I google translated the creators post and he says the last 250mb on his Titan is degraded.

I think we need to wait as that doesn't seen right.
 
Not sure if this is relevant to anything, but after restarting my machine, I find that Windows 8.1 is using about 400MB of VRAM all by itself according to GPU-Z.

Is this being taken into account with this test?
 
Not sure if this is relevant to anything, but after restarting my machine, I find that Windows 8.1 is using about 400MB of VRAM all by itself according to GPU-Z.

Is this being taken into account with this test?

See headless mode, and most people are missing this hence the myriad of incorrect results now popping up.

Even when people run basic aero mode there is VRAM allocated so the only way to run this is by using the IGP else it's inaccurate.
 
See headless mode, and most people are missing this hence the myriad of incorrect results now popping up.

Even when people run basic aero mode there is VRAM allocated so the only way to run this is by using the IGP else it's inaccurate.

General question, does Windows give up its VRAM allocation if a game is rendering in exclusive full-screen mode? If so, could the benchmark be recoded that way to keep from having to do the hassle of IGP?

I have an MSI 980, so I'm not terribly concerned at this point, I'm just curious since I don't do any Windows or gfx programming.
 
I just ran the test while using iGPU with my Gigabyte G1 (Hynix) idle. I'll have the EVGA FTW+ in a couple hours that I can run tests on also.

Update: Added G1 as primary

iGPU Primary Out



Gigabyte G1 (Hynix) Primary Out

 
Everyone should at the very least
  • disable Shadowplay and
  • make note of their prior windows/application VRAM usage levels
before running this and posting their results. But it's still not particularly meaningful if you are using the same GPU for actual desktop graphics at the same time.

I assume this is what you mean by VRAM usage:

sVqfhSn.png


So, that's before. During it seemed to go up to 3004 and stay there, but like others have said, the test crashed my video driver about every time.

 
I assume this is what you mean by VRAM usage:

sVqfhSn.png


So, that's before. During it seemed to go up to 3004 and stay there, but like others have said, the test crashed my video driver about every time.

In windows 7 you can run it with basic theme or by killing dwm.exe, maybe same for win 8.
 
General question, does Windows give up its VRAM allocation if a game is rendering in exclusive full-screen mode? If so, could the benchmark be recoded that way to keep from having to do the hassle of IGP?

I have an MSI 980, so I'm not terribly concerned at this point, I'm just curious since I don't do any Windows or gfx programming.

Not totally sure, I run basic theme on my win 7 steam box to not have to worry about it whilst gaming.

User on nvidia forum has shown GTX 980 on aero mode to have degraded results, but normal results with basic/classic windows 7 theme.
 
All I know is this:

I upgraded from an EVGA GTX560Ti to my EVGA GTX970 SC and I've had zero problems playing anything I own at max graphics levels since. If it's not using all of the memory then I can't tell by playing games, so I'm not terribly worried about it.

Still, subbed the thread as I'd like to keep informed on the matter. Hopefully this can get sorted out for the people who are affected by this.
 
Not totally sure, I run basic theme on my win 7 steam box to not have to worry about it whilst gaming.

User on nvidia forum has shown GTX 980 on aero mode to have degraded results, but normal results with basic/classic windows 7 theme.

This post on the nvidia forums shows the opposite, the guy got better results with aero and worse results with basic. Weird.

https://forums.geforce.com/default/topic/803518/geforce-900-series/gtx-970-3-5gb-vram-issue/post/4430284/#4430284
 
kind of glad i stayed team red. my new 290 was only 215 new and has been working like a charm.

Hope this issue still gets fixed for you guys though.
 
Got a good deal on 2 launch G1 970's. Haven't even had a chance to put them through their paces. Not sure how to rest this as my x99 build is still not finished.
 
I was super close to grabbing a GTX 970 in the next month or two, but I'm gonna keep an eye on this and see how it pans out before I drop ~£280 on one.
 
Since installing my GTX 970 I've had problems with periodic stuttering/pauses when playing modern games, and, while it can't be 100% related to this (for instance I'm guessing CS:GO at 1920x1080 isn't taking nearly 4GB of RAM), hearing about this has definitely made me curious if it might be related, especially as I have the memory guzzling stuff maxed on most games and use DSR on some.
 
Since installing my GTX 970 I've had problems with periodic stuttering/pauses when playing modern games, and, while it can't be 100% related to this (for instance I'm guessing CS:GO at 1920x1080 isn't taking nearly 4GB of RAM), hearing about this has definitely made me curious if it might be related, especially as I have the memory guzzling stuff maxed on most games and use DSR on some.

If CSGo is stuttering on you with a 970 @ 1080p, the 970 isn't your problem. I run that game at 4K on 16:10 and have no hitches.
 
Everyone should at the very least
  • disable Shadowplay and
  • make note of their prior windows/application VRAM usage levels
before running this and posting their results. But it's still not particularly meaningful if you are using the same GPU for actual desktop graphics at the same time.

Also, I second the request for source code of the benchmark, I'd like to see how it decides where an allocated chunk is in physical memory, because I don't know a reliable way to do that.

This should be it:

For anyone that is interested, here is the link to the source code of the benchmark (post 20 by Nai)
.
 
I was super close to grabbing a GTX 970 in the next month or two, but I'm gonna keep an eye on this and see how it pans out before I drop ~£280 on one.

Same here. January I'm trying to get through my Christmas games. February I'm getting the New 3DS XL with Majora's Mask. A 970 was going to be my toy I'd buy in March.

I'll be watching, but will be really disappointed if they don't fix this since the 970 really fits what I'm willing to spend on a graphics card and is very fast even when compared to the 980. The 980 is great, but $500+ for a graphics card is a little much since I'm not primarily a PC gamer.
 
Dammit. I was originally going to replace my 560 Ti with the 960. Then there was the reaction to the 2GB so I thought maybe I should look into the 970 instead, maybe spend an extra $130. Now this.
 
Just a thought (speculation). Could the drivers be "soft-capping" 970 vram use to about 3.5gb *because* Nvidia already knows about this vram performance issue?

Because allegedly the 980 goes right to 4gb of use in game scenes and settings where the 970 gets allocated much less, and you have to crank it much further to get the driver to feed the vram past that point on 970.

It fits with the "hardware design cause" scenario and implies horrible things about Nvidia, so this should be proven before believed.
 
Just a thought (speculation). Could the drivers be "soft-capping" 970 vram use to about 3.5gb *because* Nvidia already knows about this vram performance issue?

Because allegedly the 980 goes right to 4gb of use in game scenes and settings where the 970 gets allocated much less, and you have to crank it much further to get the driver to feed the vram past that point on 970.

It fits with the "hardware design cause" scenario and implies horrible things about Nvidia, so this should be proven before believed.

It's very possible.
 
Just a thought (speculation). Could the drivers be "soft-capping" 970 vram use to about 3.5gb *because* Nvidia already knows about this vram performance issue?

Because allegedly the 980 goes right to 4gb of use in game scenes and settings where the 970 gets allocated much less, and you have to crank it much further to get the driver to feed the vram past that point on 970.

It fits with the "hardware design cause" scenario and implies horrible things about Nvidia, so this should be proven before believed.

Honestly given Nvidia's past history of nonsense (bumpgate, denver, etc) this would not surprise me at all. The seeming linear relationship with disabled MXs and the RAM slowdown is also quite disturbing. I'm hoping this isn't the case as I too was looking at a 970 and thinking 'hmmm' pretty hard.
 
I literally just bought this card yesterday for my new build -.- Luckily I have two weeks to send it back if I don't open it, but I still hope they manage to fix it somehow.
 
I literally just bought this card yesterday for my new build -.- Luckily I have two weeks to send it back if I don't open it, but I still hope they manage to fix it somehow.
If the hardware angle is true, the only fix will be a loyalty uprade to a GTX980.
 
Any chance I could talk Gigabyte into giving me a refund on my two 970s either now or if/when this issue becomes more highly documented? My Newegg return period is long gone and I'm sitting on nearly $800 worth of unsatisfactory performance that I don't believe Nvidia will do anything about. I need more VRAM, that 3.5 GB wall is affecting a lot of the more demanding games I play and I really don't want to take a $60+ loss on each of these things.
 
I wouldn't expect a 980, but I do except a replacement 970 that utilizes all 4GB of RAM properly. That is if this can't be fixed through software.
 
I wouldn't expect a 980, but I do except a replacement 970 that utilizes all 4GB of RAM properly. That is if this can't be fixed through software.

I doubt that Nvidia would replace a card like this. At most, maybe a free game or some such as compensation, that is if it turns out to be a hardware issue.
 
I wouldn't expect a 980, but I do except a replacement 970 that utilizes all 4GB of RAM properly. That is if this can't be fixed through software.

THe only way that could possibly happen is if this fuck up your gaming experience with anomalies. If the only way to know about it is to read about it on an internet gaming forum, then there's not much chance. THey can just say it was never meant to be a 980 in the first place.
 
So if this does turn out to be a hardware issue and not software surely Nvidia would have to give everyone either a new revision 970 (if one is made) or 980?????
 
My off day today was supposed to be spent installing my new EVGA 970 into my X51....Now I wonder if I should send mine back and wait :(
 
I doubt that Nvidia would replace a card like this. At most, maybe a free game or some such as compensation, that is if it turns out to be a hardware issue.

It's false advertizing. They misled customers with erroneous specs, that's potential class action lawsuit territory.

Microsoft had to bite the bullet with RROD and, if this ends up being a hardware issue, we need to make it very clear to Nvidia that we expect them to do the same. At the very least, they have an obligation to replace the defective cards and offer free installations at participating retailers.
 
THe only way that could possibly happen is if this fuck up your gaming experience with anomalies. If the only way to know about it is to read about it on an internet gaming forum, then there's not much chance. THey can just say it was never meant to be a 980 in the first place.

Stuttering/hitching/freezing due to premature lack of VRAM; that's exactly the issue I'm having with SLI 970s when I push the Vram usage up into and past the 3.5 GBs area (970s try to avoid going above) despite retaining under 99% GPU load. Space Engine and Skyrim both clearly exhibit the issue.
 
It's false advertizing. They misled customers with erroneous specs, that's potential class action lawsuit territory.

Yep. It's the equivalent of plugging in a new monitor and having the refresh rate or resolution be different than what's on the box, or getting a new cpu set up and having the its GHz be smaller. VRAM is a huge, huge selling point for graphics cards nowadays and it's one of the primary specs you cannot and should not lie or mislead customers regarding. I am going to be pissed as shit if this ends up being a hardware issue and Nvidia knew about it the entire time (how could they not?).
 
We don't know shit yet, especially since we have people saying that that test we're doing might not matter at all. So, anyone dreaming of a free upgrade to 980 at this point is living in fantasy land.
 
I doubt that Nvidia would replace a card like this. At most, maybe a free game or some such as compensation, that is if it turns out to be a hardware issue.

A free game wouldn't make up for it if this was true. This isn't a loss of service for a small amount of time. This is a hardware issue.
 
Yep. It's the equivalent of plugging in a new monitor and having the refresh rate or resolution be different than what's on the box, or getting a new cpu set up and having the its GHz be smaller. VRAM is a huge, huge selling point for graphics cards nowadays and it's one of the primary specs you cannot and should not lie or mislead customers regarding. I am going to be pissed as shit if this ends up being a hardware issue and Nvidia knew about it the entire time (how could they not?).

If it's not 4GB, this is lawsuit material. The question is, who is going to sue?

The bigger news sites need to report this.

If this ends up being an unfixable hardware design issue, then #1) I don't for one second believe Nvidia didn't already know about this and #2) that may be partially why it was priced so much lower than its slightly bigger brother. Scummy if true.

And very interesting that none of the big review sites reported this. I wonder why.

If there's enough backlash, then hopefully they do something about it. Drop the 980 price plz.

I just google'd for the issue and found:

http://www.craveonline.com/gaming/a...tx-970-cant-use-4gb-vram-nvidia-investigating
http://www.lazygamer.net/general-news/nvidias-gtx970-has-a-rather-serious-memory-allocation-bug/
http://news.softpedia.com/news/NVIDIA-GeForce-GTX-970-Can-t-Use-All-4-GB-of-Memory-470953.shtml

It's being picked up, but definitely needs to hit the bigger sites!
 
Status
Not open for further replies.
Top Bottom