Graphics Nvidia’s GTX970 has a rather serious memory allocation bug

here a link to tell you more about the gtx970 problem http://hexus.net/tech/news/graphics/79925-nvidia-explains-geforce-gtx-970s-memory-problems/

 

they also said that the 970 had 64 ROPs, they said that it has no major performance effects by posting FPS > which given how much stick AMD got with FCAT results. I'd like to see proper FCAT results for the 970 (should point out its fairly academical to me, not looking at getting 970, just general tech interest (though will admit to a slight dislike to nvidias proprietary crap))

 

Don't get me wrong, I'm not saying Nvidia is beyond reproach as they should have said that the 970 came with 52 not 64 ROP's and only 1792 KB L2 cache versus the 980's 2048 KB.

But look at it this way, would people prefer a 970 with 192-bit memory bus and a 3GB frame buffer, or a 970 with 256-bit memory bus and a 3.5GB/0.5GB memory configuration ?

I would like to see more testing done, like as you said PCPER done with AMD though.

 

a GTX 970 with 3.5GB of ram and a big cache......

 

If the numbers added up so would i, but without the time to work it out i don't think that would be possible.

 

https://www.youtube.com/watch?v=b74MYv8ldXc


^^ video from PCPER

 

BT? ge your ass into gear - run some FCAT and benchies @ super hi res with 290 and 290x using every scrap of ram they poses and FCAT the results.

 

Well, unless time-travelling quantum gremlins have caused all past results to become invalid, all previous FCAT testing of the 970 is still correct. PCPER's own testing of the 970 at launch show no dramatic spike in frame variance between 2560x1440 and 3840x2160. Techreport's testing of the 970 at 4k also show no dramatic performance deficit.

I don't think BT have any FCAT capture hardware. It's a pretty specialised rig.

 

except that users are reporting spiking in games....

 

Well if FCAT testing shows no dramatic spike in frame variance i would suggest it's the users setup, or a badly written game (bug)

 

which is how this started didn't it - till someone got a benchmark to show something weird in VRAM allocation and the internet turned into nerd rage when nv came out with all the PR BS they have about crossbars and disabled ROPs and SMM....

 

This just landed in my in box don't know if it adds anything?http://www.pcgamer.com/why-nvidias-gtx-970-slows-down-using-more-than-35gb-vram/

 

IMO anandtech hit the nail on the head with:;

using under 3.5GB at full speed , or get c**k blocked using the last 512MB exclusively which can slow the entire lot down

 

Have a read of the Anandtech article Kronos. Probably got the most info at the moment that I've seen.

 

Cars and speed limits are not a great analogy

 

yeah all those people who have issues playing SoM because the game is using all 4GB and texture caching in the slow ram.... they should just be quiet?

and your car analogy is pants.

 

Will do cheers.

 

1440 or 1600 as a minority? really??

im going 4k soon (well that or a titan lol)

 

No worries mate.

For what it's worth, I'd still get the 970 if that's what you want. Or if you want to spend a little less pick up a 290. I can't imagine you'll be unhappy with either choice.

 

I second jrs, most people have 1080p or less screens, even among enthousiasts like us here. Most people just change their screen when the previous one died, not because a new model has been released.

I'm considering changing my 17" Eizo I bought in 2004 only because most softwares nowaday aren't 5:4 friendly anymore. I'll buy a 23/24" 1080/1200p monitor as I'm the quality over quantity guy. I'd rather have a great 1080p screen than a so-so 32" 5k one.