Nvidia to face class-action lawsuit over GeForce GTX 970 memory

[MustardCutter]

I did pick the numbers for the 1.795l 1.8l engine example from where the sun doesn't shine, guys...

Anywho, as far as I'm aware it does have 4gb as sold and it scales only slightly worse than the 980's 4GB performance-wise when greater than 3.5gb is used. It just strikes me as a lot of fuss about something minor. It's questionable business practice but I'd hardly call it the lie of the century. I imagine most companies in this world have committed far more dubious acts that have gone un-noticed.

[TRUTHnTRINITY]

https://www.change.org/p/nvidia-refund-for-gtx-970 so this petition is going on for the GTX 970. If you're a consumer that has purchased a GTX 970 and feel the were mislead by NVIDIA please sign.

[spl]

I have a GTX970 but I won't be signing that petition. The level of performance delivered is as the review benchmarks promised so I'm happy. I'm more bothered about Samsung's response to the 840 Evo issues, offering a temporary fix and calling it permanent. Over nine months now since that guy first discovered the problem and there still isn't a proper fix. Seems that deserves a lot more attention than this GTX970 business.

[ZaO]

I did pick the numbers for the 1.795l 1.8l engine example from where the sun doesn't shine, guys...

Anywho, as far as I'm aware it does have 4gb as sold and it scales only slightly worse than the 980's 4GB performance-wise when greater than 3.5gb is used. It just strikes me as a lot of fuss about something minor. It's questionable business practice but I'd hardly call it the lie of the century. I imagine most companies in this world have committed far more dubious acts that have gone un-noticed.

Well I'll agree to disagree with you that it's not a big deal, and leave it at that At least it's looking like there's some sollutions appearing for those who're troubled by this issue..

[GuidoLS]

To be fair, which isn't always the case... Nvidia DID say they were going to allow people that wanted portable bbq grills to start overclocking their toteables again before any lawsuit was ever announced, so the title is a bit misleading, although I doubt it was intentionally so.

[CoolStoryBro]

redudnt to my next post

[CoolStoryBro]

I have a GTX970 but I won't be signing that petition. The level of performance delivered is as the review benchmarks promised so I'm happy. I'm more bothered about Samsung's response to the 840 Evo issues, offering a temporary fix and calling it permanent. Over nine months now since that guy first discovered the problem and there still isn't a proper fix. Seems that deserves a lot more attention than this GTX970 business.

Samsungs 840 EVO issues fix is in the works as a firmware update. NVidia's issue is false advertised hardware that isn't on the card that can't be fixed with anything. I owned two.. well still own officially as mine are in shipment back to the vendor i purchased them in October '14 for full refund RMA because of the drastic FPS drops (20 fps sometimes as much as a 30fps drop) when the ram usage exceeds 3.3-3.5gb of ram. I'm glad someone with the expertise was able to figure out what the issue was with these cards.

[Tunnah]

Has anyone experienced slowdown in real world usage yet ?

[spl]

Has anyone experienced slowdown in real world usage yet ?

I haven't but apparently CoolStory has! If you check out Techspot's Evolve review, you'll see that the GTX970 trails the GTX980 by about what you'd expect (or even a little less gap than you'd expect) at 1080p but at high resolution the proportional gap increases significantly which suggests this will have far more impact on those who game at higher resolutions than 1080p. I'm sticking with 1080p until AMOLED is affordable but it's more of a big deal to those gaming at 1440p, 1600p or 4K.

[spl]

Samsungs 840 EVO issues fix is in the works as a firmware update. NVidia's issue is false advertised hardware that isn't on the card that can't be fixed with anything. I owned two.. well still own officially as mine are in shipment back to the vendor i purchased them in October '14 for full refund RMA because of the drastic FPS drops (20 fps sometimes as much as a 30fps drop) when the ram usage exceeds 3.3-3.5gb of ram. I'm glad someone with the expertise was able to figure out what the issue was with these cards.

Actually we got a firmware update to address this last year and it hasn't worked. Many people (myself included) trusted Samsung and took them at their word that last year's update would do the trick. At least I've been lucky it seems to have one of the drives not affected.

The nVIDIA issue isn't about anything not being on the card. Every last byte of that 4GB is present. Just 512MB of it is being wasted because the memory controller isn't accessing it at full speed. Personally I'd have preferred that the card was just 3.5GB if drivers / DirectX / game engines / whatever can't know better than to leave that 512MB alone unless absolutely necessary. A lot of the problem is that games (such as Advanced Warfare) will use it even when they don't need it - they'll just use all the memory they can get, resulting in the slow-accessed memory being used. Other problem of course is those gaming at very high resolutions who actually do need it.

[CoolStoryBro]

Actually we got a firmware update to address this last year and it hasn't worked. Many people (myself included) trusted Samsung and took them at their word that last year's update would do the trick. At least I've been lucky it seems to have one of the drives not affected.

The nVIDIA issue isn't about anything not being on the card. Every last byte of that 4GB is present. Just 512MB of it is being wasted because the memory controller isn't accessing it at full speed. Personally I'd have preferred that the card was just 3.5GB if drivers / DirectX / game engines / whatever can't know better than to leave that 512MB alone unless absolutely necessary. A lot of the problem is that games (such as Advanced Warfare) will use it even when they don't need it - they'll just use all the memory they can get, resulting in the slow-accessed memory being used. Other problem of course is those gaming at very high resolutions who actually do need it.

This an old post and this probably won't be read, but i'd like to post it anyways for anyone who would pass by questioning a purchase of a gtx 970.

The GTX 970 does have 4GB of VRAM. That wasn't the hardware i was talking about that is missing tho.. They advertised the memory and the controllers of the memory to be the same as the 980 which is false.. The issue is

A quote

"The GTX 970 has 56 ROPs and 1792 KB of L2 cache compared to 64 ROPs and 2048 KB of L2 cache for the GTX 980. Before people complain about the ROP count difference as a performance bottleneck, keep in mind that the 13 SMMs in the GTX 970 can only output 52 pixels/clock and the seven segments of 8 ROPs each (56 total) can handle 56 pixels/clock. The SMMs are the bottleneck, not the ROPs.”

"If you don’t speak graphics card, PCPer helps break down the architecture of the GTX 970. Because the GTX 970 only has seven ports connecting memory controllers and cache, one of those ports would always be burdened with twice as many requests."

This is where after 3.5gb of VRAM is used the card has to basically piggy backs the bandwith on the 7th ROP because there isn't an 8th ROP on each SMM on the GTX 970 and it slows all the RAM down to the slowest speed that it's using to push the bandwith on the last 500mb. I personally experienced this gaming at 1440p and 4k resolutions in games. For example a game i tested personally: Shadows of Morder went from 60 solid fps down to 38-42fps whenever the VRAM usage went over 3.5gb.


More from an article.

PCPer explains “if the 7th port is fully busy, and is getting twice as many requests as the other port, then the other six must be only half busy, to match with the 2:1 ratio. So the overall bandwidth would be roughly half of peak. This would cause dramatic underutilization and would prevent optimal performance and efficiency for the GPU.”
Nvidia avoided that problem by dividing the memory into a 3.5GB pool and a 0.5GB pool. Few games (currently) require more than 3.5GB of VRAM, so the primary pool can be accessed at maximum bandwidth.
PCPer writes: “Let's be blunt here: access to the 0.5GB of memory, on its own and in a vacuum, would occur at 1/7th of the speed of the 3.5GB pool of memory."


Entire Article can be found on PCGAMER with quotes from NVIDIA (These forums won't let me link urls)
t