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

[james.miller]

I thought you had posted that video because you had noticed the stuttering.

I don't think you understand the issue & I actually see it, albeit minor, in your video... does that microstutter every couple of seconds not bother you? THAT is the issue... it is more visible on the previously posted FC4 video (ignore the colour splashes).

There is a slight regular hitching in there but that's down to AC:U and using shadowplay when the gpu is working flat out - its going to hitch a little but,
it's a million miles away from what's happening in that FC3 video.

Using the extra memory / paging 10% of the textures will not cause the FPS to suddenly drop by 70%.

when you run out of memory performance drops off a cliff with any hardware in my experience.

The stutter is clear in the video you posted of AC:U.

I certainly don't think it is.

 

[Kaapstad]

Seen this posted on the geforce forums by someone,sounds quite interesting and a possible reason maybe ?




"OK, this makes more sense:

There seems to be a lot of confusion about this, so I'll just post how I interpretted, the statement:
What they're saying is that there are two memory modules, 3.5GB + 0.5GB, they both perform the same, but the 3.5GB is the primary memory pool, so it will always access that first, and then it will access the 0.5GB if it needs to.

The problem is that memory testing programs, and monitoring programs, cannot access that 0.5GB module, so when a monitoring program like MSI Afterburner is monitoring memory use, it only sees the 3.5GB memory module, if the card is using >3.5GB, monitoring programs will only show 3.5GB, as they cannot monitor the 0.5GB.

Similarly, the memory tester that was hot on this issue, cannot access the 0.5GB, but knows that is a 4GB card, so when it tries to get the final 0.5GB, it gets "redirected" to system RAM, which has much less bandwidth than video memory.

When people see >3.5GB reported by monitoring programs, I'm going to guess what's happening there is they are beginning to use >4GB, causing system memory to be allocated, which is then added to the monitoring program's measurement, making it look like >3.5GB is causing massive slowdowns, when in reality it's >4GB that's causing massive slowdowns.

Anyway, that's my interpretation, I could be completely wrong, so... Grain of salt and all that.

Any comments on this? "


Does sound like a possible reason to me but who can say at the moment.One things for certain ,Nvidia have to make some proper response and fast if they dont want to hurt sales more than they possibly have been already.

That is the conclusion I came to several pages back

You only have to look at the VRAM usage reported in GPUZ when myself and String were comparing cards on the Shadow of Mordor bench 980 v 970 using the same settings and both cards getting good bench results. It looked like my 980 was using about 0.7gb more VRAM where in fact it was using about the same.

 

[eyetrip]

Well, the stuttering in the fc4 video posted is very much like how dragon age runs for me once I get over 3.5.

You know I have not been ecstatic with the 970's since I got them, but assumed nvidia was just in a bad way with drivers - and please don't think I am saying this now that this issue has come up, as I felt like that long before.

Performance been relatively bad on 4k, with regard getting things smooth.

So nvidia have confirmed there is a 'design' implementation that explains the behaviour that led people to start posting about this?

So our bandwidth is also gimped?

 

[Silent_Scone]

check the usage in my ac:u video, if it was swapping ram i dont think i'd manage 10fps, let alone 30 lol

That game...

I have buffer overflow on my 980s when at 1440p and Ultra settings. Just about copes with FXAA.

Stuttering and hitching ensues if I use MSAA.


I don't see the point in going round in circles with this issue until Nvidia clear it up, seems people are getting their wires crossed

 

[crinkleshoes]

There is a slight regular hitching in there but that's down to AC:U and using shadowplay when the gpu is working flat out - its going to hitch a little but,
it's a million miles away from what's happening in that FC3 video.






I certainly don't think it is.

Well... it isn't doing it in the FC4 video when under 3.5GB usage...

when you run out of memory performance drops off a cliff with any hardware in my experience.

That's not quite what's occurring though, is it?

 

[J.D]

Well, the stuttering in the fc4 video posted is very much like how dragon age runs for me once I get over 3.5.

You know I have not been ecstatic with the 970's since I got them, but assumed nvidia was just in a bad way with drivers - and please don't think I am saying this now that this issue has come up, as I felt like that long before.

Performance been relatively bad on 4k, with regard getting things smooth.

So nvidia have confirmed there is a 'design' implementation that explains the behaviour that led people to start posting about this?

So our bandwidth is also gimped?

I'm not so sure this is the case. As I've said previously, it's a bit insane that Nvidia would risk this and the explanation above provided by Dash RipRock (through quotation) above makes sense. I'm not saying there is no problem, I'm still waiting to see but it does defy logic that Nvidia would risk this when there are teams dedicated to benchmarking professionally, never mind the amount of users who test their cards in a series of benchmarks for tables in forums across the world.

 

[Silent_Scone]

Well... it isn't doing it in the FC4 video when under 3.5GB usage...



That's not quite what's occurring though, is it?

Unity can use over 4GB VRAM easily, no questions there.

 

[Buckster]

well the 660Ti was advertised as a 2gb 192bit card

 

[StrongSad]

This is quite serious really....

Thought I would have a look at AC-U on my PC, (not my game BTW) to have a look and see if I can detect any issues.....

7 GB patch.......

What I do find interesting is that none of the Tech reviews I have seen, have picked up on this.
I only found out from here. Yet there is review after review saying that this card has golden undercarriage, yet not one has said "hey, by the way, it spins down at anything over 3.5GB memory usage... go buy one!"

 

[Dash RipRock]

I'm not so sure this is the case. As I've said previously, it's a bit insane that Nvidia would risk this and the explanation above provided by Dash RipRock (through quotation) above makes sense. I'm not saying there is no problem, I'm still waiting to see but it does defy logic that Nvidia would risk this when there are teams dedicated to benchmarking professionally, never mind the amount of users who test their cards in a series of benchmarks for tables in forums across the world.

It does seem rather far fetched that nvidia would put 4 gig on a card and make the last 0.5 gig so bandwidth starved and then hope that no one would notice, it has to be said

 

[Orangey]

Is this two-level memory then? Like AMD is rumoured to be using?

 

[fs123]

Seen this posted on the geforce forums by someone,sounds quite interesting and a possible reason maybe ?




"OK, this makes more sense:

There seems to be a lot of confusion about this, so I'll just post how I interpretted, the statement:
What they're saying is that there are two memory modules, 3.5GB + 0.5GB, they both perform the same, but the 3.5GB is the primary memory pool, so it will always access that first, and then it will access the 0.5GB if it needs to.

The problem is that memory testing programs, and monitoring programs, cannot access that 0.5GB module, so when a monitoring program like MSI Afterburner is monitoring memory use, it only sees the 3.5GB memory module, if the card is using >3.5GB, monitoring programs will only show 3.5GB, as they cannot monitor the 0.5GB.

Similarly, the memory tester that was hot on this issue, cannot access the 0.5GB, but knows that is a 4GB card, so when it tries to get the final 0.5GB, it gets "redirected" to system RAM, which has much less bandwidth than video memory.

When people see >3.5GB reported by monitoring programs, I'm going to guess what's happening there is they are beginning to use >4GB, causing system memory to be allocated, which is then added to the monitoring program's measurement, making it look like >3.5GB is causing massive slowdowns, when in reality it's >4GB that's causing massive slowdowns.

Anyway, that's my interpretation, I could be completely wrong, so... Grain of salt and all that.

Any comments on this? "


Does sound like a possible reason to me but who can say at the moment.One things for certain ,Nvidia have to make some proper response and fast if they dont want to hurt sales more than they possibly have been already.

What if Nvidia are not telling the entire truth and they have somehow made the drivers or gpu allocate some system memory to make it up to 4GB, much like how windows uses a swap file.
This would explain why the MSI Kombuster test also causes a large increase in system memory usage.

 

[Kaapstad]

What if Nvidia are not telling the entire truth and they have somehow made the drivers or gpu allocate some system memory to make it up to 4GB, much like how windows uses a swap file.
This would explain why the MSI Kombuster test also causes a large increase in system memory usage.

To deliberately mislead people would be very unwise and something NVidia or most companies would not do.

 

[fs123]

To deliberately mislead people would be very unwise and something NVidia or most companies would not do.

I'm just thinking out of the box since I know many old onboard gfx chips actually did reserve system ram as gpu memory. Surely Nvidia haven't done that.

 

[J.D]

It does seem rather far fetched that nvidia would put 4 gig on a card and make the last 0.5 gig so bandwidth starved and then hope that no one would notice, it has to be said

lol yeah a bit like....



I hope I've got this right about gifs this time .

I can't see it but we'll wait and see.

 

[Dash RipRock]

What if Nvidia are not telling the entire truth and they have somehow made the drivers or gpu allocate some system memory to make it up to 4GB, much like how windows uses a swap file.
This would explain why the MSI Kombuster test also causes a large increase in system memory usage.

Again,something like that would be EXTREMELY easy for the tech savvy people to spot and wouldnt be worth the loss of image that would ensue.Such an action would defy logic if you want return custom for your products.

 

[Dash RipRock]

lol yeah a bit like....



I hope I've got this right about gifs this time .

I can't see it but we'll wait and see.

 

[Amatsubu]

Just gave Unity @1080p a whirl. GPU-Z reports vram maxing out at 3531mb but the game doesn't really stutter. There are drops to 45fps in more demanding places but it's usually 55-60fps.

This card won't last long if the problem is real and there are titles that need as much vram at 1080p.

 

[Andy__C]

It does seem rather far fetched that nvidia would put 4 gig on a card and make the last 0.5 gig so bandwidth starved and then hope that no one would notice, it has to be said

I couldn't help but think this..... Especially with how much the card was celebrated. That and every reviewer and board partner not noticing either.

 

[J.D]

That video haha, nutter. You're setting yourself up for some egg on your face if this turns out to be the inability of these benchmarks to address this additional Vram. Games from some tests seem fine for some, wonky for others so I'm not convinced this is a hardware problem yet. Time will tell.

I hope it's not true.