-
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
- Thread starter kitch9
- Start date
- Status
Except the 970 will only use the extra 500MB if it needs to.
If the 970 and 980 perform similarly taking into account the difference in SMX units they will use the same amount of VRAM. Both the 970 and 980 scores scaled correctly in relation to the number of SMX units available. This included the minimums which were fine on the 970 (the mins dipping is usually the first sign of the VRAM running out).
It is the monitoring software that can not show this.
This would have resulted in the 970 result taking a performance hit, something that did not happen.
From what I can make out NO it doesn't, but this thread seems to be more about people 'discussing' how they think it uses the 3.5 + .5 memory makeup.
So unless anyone can point out otherwise, regardless of what the results of it are, the card doesn't seem to be 'as specified'
Soldato
- Joined
- 24 Aug 2006
- Posts
- 10,113
- Location
- Gibraltar
Issues with the new compression method nv is using?
This would have resulted in the 970 result taking a performance hit, something that did not happen.
Not if, as Stoeeh was saying, the benchmark at those settings requires <3.5GB
They don't though as my GTX 980 proves.
@Kaap
I'm not convinced, I have already posted earlier in the thread shots with 4GB of VRAM in use on my 970 so clearly it is reporting it there.
I am also of the opinion that just because monitoring software shows X amount of VRAM in use does not necessarily mean that that figure is actually required. By that I mean that excess VRAM is being used up for caching or whatever.
I'm sure this has been argued to death on the forums over the years over which stance is right
but in this case I think the 970 won't 'cache' over 3.5GB
I'm not convinced, I have already posted earlier in the thread shots with 4GB of VRAM in use on my 970 so clearly it is reporting it there.
I am also of the opinion that just because monitoring software shows X amount of VRAM in use does not necessarily mean that that figure is actually required. By that I mean that excess VRAM is being used up for caching or whatever.
I'm sure this has been argued to death on the forums over the years over which stance is right
but in this case I think the 970 won't 'cache' over 3.5GBAgain dsr is used but this time in Mordor on a 970 with ultra settings. Gpu monitoring program cleary showing more than 3.5GB in use:
https://www.youtube.com/watch?v=BQZW8sgeZyg
Whoever the owner of the video is seems to think the stutter went with a driver update, just to add more speculation to the issue. It's all very confusing haha.
Associate
- Joined
- 12 Jun 2010
- Posts
- 191
To quote myself here,could this then be a possible case of this:-
"Just to throw more speculation into the mix .....how about this for an explanation.Taking shadow of mordor as a test case when Kaapstad said his 980 was using 4gb and at the same settings the 970 was reporting 3.5 gb,maybe (if we assume monitoring software is not reading the last 0.5gb correctly) the 970 WAS actually using 4gb,but because of this sectioned off 0.5 gb it`s just not showing.
If we then push beyond that by adding settings that would increase the 980 usage beyond 4GB,lets say to 4.3gb would it be possible that the monitoring software could potentially be reading this as only 3.8 gb on the 970,thus giving appparant stutter at less than 4gb.
This does sort of hinge on the 980 stuttering at 4.3 GB usage of course,but while everyone else is throwing all sorts speculation out there ....why not eh"
Although the bit about the driver correcting it is a strange one
Which goes back to the caching/obsolete data possibility
I don't think it can be used as evidence either way is all.
Anyhow, I am going to avoid too much speculating, I'm just gonna wait out and see how it unfolds. I don't see it as a big issue unless it turns out to be more than a few % impact but understand it could be annoying for some thinking they are buying one thing and getting another.
I wouldn't be surprised if (actually I expect) these quirks to be present in many other GPU's and complex ICs past and future.
Last edited:
Which goes back to the caching/obsolete data possibility
I don't think it can be used as evidence either way is all.
Anyhow, I am going to avoid too much speculating, I'm just gonna wait out and see how it unfolds. I don't see it as a big issue unless it turns out to be more than a few % impact but understand it could be annoying for some thinking they are buying one thing and getting another.
What can be proved is the 970 ran fine with the same setting as my GTX 980 which used 4gb in the process.
Also if there was any benefit in not caching don't you think this would be applied to the 980 as well as the 970.
No I think the monitoring software is the problem.
Permabanned
- Joined
- 15 Dec 2008
- Posts
- 3,417
Thing is Greg we shouldn't have to that.
Must admit I've been reading this with interest and to try and test it I started playing Assassins Creed Unity on the spec in my sig. I tend to keep afterburner on my second screen and this seems to monitor the full 4GB as I hit 4050mb memory used but had no issues.
I run the game at 1440p with everything maxed except for AA which is on FXAA. I've seen memory usage range from 3500-4000 with no obvious hitches or stutters. I'm making between 35-50fps depending on the scene but with gsync everything looks and feels buttery smooth.
Has anyone else tried the monitoring with afterburner to see what that shows?
I run the game at 1440p with everything maxed except for AA which is on FXAA. I've seen memory usage range from 3500-4000 with no obvious hitches or stutters. I'm making between 35-50fps depending on the scene but with gsync everything looks and feels buttery smooth.
Has anyone else tried the monitoring with afterburner to see what that shows?
Sure if it is caching with the available ram there could be a tangible benefit and it may or may not account for a portion of the few % difference that Nvidia reported.
OK, must resist topic/forums and stop procrastinating my day!
I am downloading Middle Earth onto a PC with cards that don't need to worry about caching to see how much VRAM can be used @1080p.
I think you're right, monitoring software (some? all?) doesn't seem to see the 3.5-4GB memory.
What should you do then? I had 2GB cards and when playing FC3 with everything maxed at 5860x1080, it would go over the 2GB buffer, so I lowered AA till it was under the 2GB buffer. If people are using more than the available 4GB (even though it is showing 3.5GB because of a faulty reading), they should do the same.
It isn't rocket scienece.
4GB and over will stutter on a 4GB GPU and cause problems.
Soldato
- Joined
- 6 Jan 2013
- Posts
- 22,179
- Location
- Rollergirl
I've done some videos with GPU-z screen caps to maybe help with this debate...
SOM @ 1080p - Full Eye Candy
SOM @ 3840*2160 (200%) - Full Eye Candy
SOM @ 3840*2160 (200%) - Without AA or AO
And here's 2 970's in SLI @ 1080p with full eye candy...
... And 2 970's in SLI @ 1080p with AA and AO disabled...
SOM @ 1080p - Full Eye Candy
SOM @ 3840*2160 (200%) - Full Eye Candy
SOM @ 3840*2160 (200%) - Without AA or AO
And here's 2 970's in SLI @ 1080p with full eye candy...
... And 2 970's in SLI @ 1080p with AA and AO disabled...
Last edited:
I've done some videos with GPU-z screen caps to maybe help with this debate...
SOM @ 1080p - Full Eye Candy
SOM @ 3840*2160 (200%) - Full Eye Candy
SOM @ 3840*2160 (200%) - Without AA or AO
And here's 2 970's in SLI @ 1080p with full eye candy...
Fuark those dips are painful
- Status