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

[ubersonic]

If anyone here is having this issue can you please try enabling classic (not basic) theme and seeing what that does (fixes the issue for 780ti users).

To enable classic in Windows 7/8 you right click computer, properties, advanced system settings, performance settings, and check the box for performance settings and it will switch to classic 9x style shell.

 

[Divine Pain]

Wow, this is a pretty big deal, surprised this has only come to light now

 

[~>Dg<~]

gibbo sweating in corner after 100,000 970s sold

 

[ReCoN1989]

Technically speaking misrepresentation is a criminal offence. Advertising a card with 4gb of ram when it is not usable would be an offence as it undoubtedly persuaded many people to buy the card.

 

[chef-borjan]

All I want to see is this benchmarking tool run with a 970, then swapping in a 980 and show the difference.

 

[Fanatic]

The plot thickens, seems somebody has managed to replicate the issue, with a GTX780ti :O

With the 780ti the same issue would occur as it approached it's 3GB limit, though oddly only when Windows Aero was enabled.

Isn't that just because Windows Aero uses about 150MB of VRAM? So with it on you are going to hit your VRAM limit sooner.

 

[Rroff]

In previous cases of hardware having performance issues in later memory blocks its been solvable with a firmware update but who knows.

Isn't that just because Windows Aero uses about 150MB of VRAM? So with it on you are going to hit your VRAM limit sooner.

Quite possible - wondered the same thing when I saw the post. Windows 7/8 does some funky virtualisation stuff to with DX that can hit VRAM performance hard using some older techniques for direct memory read/writes especially with aero in the mix

 

[Legend]

This is POTENTIALLY A HUGE DEAL if it's a hardware fault that cannot be fixed. Could be VERY expensive indeed for Nvidia, as there will be tens of thousands of us who've bought these cards up in arms about it, and I honestly cannot see they will have a legal leg to stand on.

 

[ReCoN1989]

I expect so, this problem seems to be linked to the cut down Maxwell architecture. The way to test it is by using a IGPU so usage on the dGPU will be 0%.

If it is not possible to enable some disabled parts with a bios update because they have been removed then I don't see how this will be fixed.

 

[Phixsator]

Well before people get all aggro perhaps we should let nVidia have a chance at fixing it.. I dont mind seeing a bit of burn on nVidia's back when they are not being fair to consumers but one has to be fair and offer a chance for them to remedy the situation I believe.

 

[J.D]

At first I thought it was just in situations regarding this software this happened but it seems as some users have replicated this in their gaming.

This is crazy but can you expect transparency if true?. Interesting to see what happens here. Stupid move from Nvidia if this can't be fixed with a firmware update as to what Rroff stated above.

Something is bothering me though as I've seen nothing but great reports about the 970 from users. I take it they never reached these Vram levels but if so, will they return the cards regardless?. Strange times indeed.

 

[chef-borjan]

Well before people get all aggro perhaps we should let nVidia have a chance at fixing it.. I dont mind seeing a bit of burn on nVidia's back when they are not being fair to consumers but one has to be fair and offer a chance for them to remedy the situation I believe.

Agreed.

Need to wait for official confirmation first as well, or at least a more comprehensive in-depth test than a few screen shots on forums.

 

[Rroff]

At first I thought it was just in situations regarding this software this happened but it seems as some users have replicated this in their gaming.

This is crazy but can you expect transparency if true?. Interesting to see what happens here. Stupid move from Nvidia if this can't be fixed with a firmware update as to what Rroff stated above.

Something is bothering me though as I've seen nothing but great reports about the 970 from users. I take it they never reached these Vram levels but if so, will they return the cards regardless?. Strange times indeed.

Something about the 9xx cards has never inspired me, glad I decided to hold onto the 780 until something better comes along.

 

[Scramz]

This issue was raised a few weeks back and Nvidia have said they are looking into it but nothing else since from them. Not a good sign.

 

[ReCoN1989]

Taken from Overclock.net

=.= From the Nai's Benchmark, assuming if the allocation is caused by disabled of SMM units, and different bandwidth for each different gpus once Nai's Benchmark memory allocation reaches 2816MiBytes to 3500MiBytes range, I can only assume this is caused by the way SMM units being disabled.

Allow me to elaborate my assumption. As we know, there are four raster engines for GTX 970 and GTX 980.
Each raster engine has four SMM units. GTX 980 has full SMM units for each raster engine, so there are 16 SMM units.

GTX970 is made by disabling 3 of SMM units. What nvidia refused to told us is which one of the raster engine has its SMM unit being disabled.
I found most reviewers simply modified the high level architecture overview of GTX 980 diagram by removing one SMM unit for each three raster engine with one raster engine has four SMM unit intact.

First scenario
What if the first (or the second, third, fourth) raster engine has its 3 SMM units disabled instead of evenly spread across four raster engine?

Second scenario
Or, first raster engine has two SMM units disabled and second raster engine has one SMM unit disabled?

Oh, please do notice the memory controller diagram for each of the raster engine too. >.< If we follow the first scenario, definitely, the raster engine will not be able to make fully use of the memory controller bandwidth.

I agree that this is the most likely explanation, which would mean the issue is hardware-related and cannot be fixed. Here's an illustration of GM204:
** No hotlinking **
http://images.bit-tech.net/content_images/2014/09/nvidia-geforce-gtx-980-review/gtx980-17b.jpg

It does seem that each of the four 64-bit memory controllers corresponds with each of the four raster engines and in the same way that the 970's effective pixel fillrate has been demonstrated to be considerably lower than the 980's even though SMM cutting leaves the ROPs fully intact (http://techreport.com/blog/27143/here-another-reason-the-geforce-gtx-970-is-slower-than-the-gtx-980), the same situation may apply to bandwidth with Maxwell. However, the issue may be completely independent of which SMMs are cut and may simply relate to how many.

GM206's block diagram demonstrates the same raster engine to memory controller ratio/physical proximity:

http://cdn3.wccftech.com/wp-content/uploads/2015/01/GM206-Block-Diagram.jpg

I expect a cut-down GM206 part and even a GM200 part will exhibit the same issue as a result, it might be intrinsically tied to how Maxwell as an architecture operates. Cut down SMMs -> effectively mess up ROP and memory controller behavior as well as shaders and TMUs. I also don't think there's a chance in hell Nvidia were unaware of this, but I could be wrong.

Yeah, it does seem the driver prioritizes on using 3.5Gb, otherwise the performance impact of using more than that would be too apparent.
But it would depend on how GTX 970 SMM being disabled.

64bit memory controller, total 4 memory controllers = 256 bit memory controller.


Assuming if there are 3 raster engines with each three has one SMM disabled leaving 1 raster engine with 4 SMM intact.
Mathematically ;
16 SMM = 256 bit = 4096 Mb
13 SMM = 208 bit = 3328 Mb

208 bit = effective width after disabling SMM with 256 bit being actual memory controller width.

I could be wrong on this oversimplify calculation. Still worry if one raster engine has three disabled SMM >.<

 

[Bonjour]

Could this be a repeat of the great NVidia FX shader model 2 scandal of 2003...

 

[ICEx21]

At first I thought it was just in situations regarding this software this happened but it seems as some users have replicated this in their gaming.

This is crazy but can you expect transparency if true?. Interesting to see what happens here. Stupid move from Nvidia if this can't be fixed with a firmware update as to what Rroff stated above.

Something is bothering me though as I've seen nothing but great reports about the 970 from users. I take it they never reached these Vram levels but if so, will they return the cards regardless?. Strange times indeed.

This problem has been noted for a while now, unfortunately I did minimum research before buying these and really regret it now, hopefully if this doesn't get fixed soon I can get a refund and then I plan on going back to my 290.

 

[Kronos]

I am subscribing to this thread as you guys seem to have your finger on the pulse. And I don't want to get my hard earned together and throw it at an inherently faulty GPU.

I shall watch this with great interest.

 

[Scoiexx]

Wow this is unexpected, in the 1080 range, what game currently would get to 3.5+ vram usage. Really want to test this.

 

[J.D]

Something about the 9xx cards has never inspired me, glad I decided to hold onto the 780 until something better comes along.

Something about the 9xx cards has never inspired me, glad I decided to hold onto the 780 until something better comes along.

Efficiency release it seems to bring in the pennies. Good on you for sticking with the 780. Like you, the 9 series has been a let down for me. When I seen the 780s for £199 and the Ti's for £299 I wondered how the 970s would sell but still they flew off the shelf. Marketing for you .

If Nvidia has messed up here, good luck to them. 100,000 cards multiplied by £300 = beans on toast after the refund in the Nvidia lunch room. There's a lot more than 100k cards moved most likely though. It's stupidity if they knew about this and expected nobody to find out, well it might lead people who prefer Nvidia to try AMD. Saying that, I'm just guessing and presuming. Patience = wisdom here so on that note, I'll shut up .

This problem has been noted for a while now, unfortunately I did minimum research before buying these and really regret it now, hopefully if this doesn't get fixed soon I can get a refund and then I plan on going back to my 290.

I'm sorry to hear this. Even more so as I know you'd have at least considered crossfiring the 290 you already had. I've done similar things when buying, as have many others so don't be too hard on yourself.

As I said previously. I've seen nothing regarding Vram problems apart from this and most users seem extremely happy with their purchases. If this Vram limitation isn't so much a problem for you when gaming then at least you can enjoy your cards and watch how this unfolds.