Doesn't anyone have anything to add to this news? I imagine it effects SETI users in the fact that you don't ONLY crunch SETI with your GPU's? Right?

The brief story is only 3.5GB of the cards RAM has full speed. The other 512MB is pooled into half speed or less.

Wonder if this should affect users considering a 970 vs a 980? All I know is AMD is having a field day with this!

All the best,
Aaron Lephart

TechVelocity.com

---

Link the article so the rest of us can look at it, then you will probably get some responses.

A quick Google search brought me to this article from my favorite hardware review site, AnandTech.

"Ryan Smith @ AnandTech" wrote:

In any case, the one bit of good news here is that for gaming running out of VRAM is generally rather obvious. Running out of VRAM, be it under normal circumstances or going over the GTX 970’s 3.5GB segment, results in some very obvious stuttering and very poor minimum framerates. So if it does happen then it will be easy to spot. Running out of (fast) VRAM isn’t something that can easily be hidden if the VRAM is truly needed.

To that end in the short amount of time we’ve had to work on this article we have also been working on cooking up potential corner cases for the GTX 970 and have so far come up empty, though we’re by no means done. Coming up with real (non-synthetic) gaming workloads that can utilize between 3.5GB and 4GB of VRAM while not running into a rendering performance wall is already a challenge, and all the more so when trying to find such workloads that actually demonstrate performance problems. This at first glance does seem to validate NVIDIA’s overall claims that performance is not significantly impacted by the memory segmentation, but we’re going to continue looking to see if that holds up. In the meantime NVIDIA seems very eager to find such corner cases as well, and if there are any they’d like to be able to identify what’s going on and tweak their heuristics to resolve them.

Obviously Ryan's testing is focused on the gaming aspects of the card and not the distributed computing aspect, but it seems that this may not be a big deal unless someone modifies their cc_config.xml file to run as many workunits as needed to fill all available RAM on the video card, which doesn't sound to smart to me if you're overtaxing the GPU itself for the sake of maxing out all RAM.

Meh. While it is noteworthy and interesting, I don't think it is going to have as massive of an impact as some of the biggest detractors are going to make of it.

There has been some talk on The Guru of 3D site http://www.guru3d.com/news-story/does-the-geforce-gtx-970-have-a-memory-allocation-bug.html. It seems that there is a big discussion, acknowledged by NVidia, concerning the last 400 - 600 MB on 4GB devices.

---

I don't buy computers, I build them!!

The consensus I've seen from reading about this is that it's definitely a big deal for NV to have made a mistake on the official specifications, but in terms of real-world performance - graphical or otherwise - there's little evidence towards a noticeable detrimental impact.

---

Soli Deo Gloria

Okay here's my take. A 780 Ti most commonly comes with 3GB, which I think is fine. A 780 Ti is also slightly faster than a 970. A 970 has 4GB, but maybe not 4GB of the good stuff. As long as it has 3GB of the good stuff I'm fine with it.

Okay here's my take. A 780 Ti most commonly comes with 3GB, which I think is fine. A 780 Ti is also slightly faster than a 970. A 970 has 4GB, but maybe not 4GB of the good stuff. As long as it has 3GB of the good stuff I'm fine with it.

Same here.

---

I run three projects at the same time. Seti, Milkyway and Einstein. Most efficient use of my 970's are three tasks per card. The most ram I have seen utilized so far has been 2.5GB. So still 1 GB of unused ram before I would ever run into this problem with distributed computing. As the other poster said ..... it would make no sense to overtask the cards just to use up all available memory of the 4GB on board. This may be a problem with gamers who run 4K resolution on multiple displays at the highest gaming settings but is inconsequential to distributed computer users.

Cheers, Keith

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

The consensus I've seen from reading about this is that it's definitely a big deal for NV to have made a mistake on the official specifications, but in terms of real-world performance - graphical or otherwise - there's little evidence towards a noticeable detrimental impact.

I can see a difference when using large displays and have heard of some issues with people playing games while trying to use gpgpu but other than that under normal circumstances there shouldn't a huge issue.
Just to point to the fact look at how long it took people to publicize it - not many people are affected.

It seems that a class action suit has been brought against NVidia and Gigabyte.
http://www.guru3d.com/news-story/nvidia-faces-class-action-lawsuit-over-gtx-970-memory-debacle.html

---

I don't buy computers, I build them!!

It seems that a class action suit has been brought against NVidia and Gigabyte.
http://www.guru3d.com/news-story/nvidia-faces-class-action-lawsuit-over-gtx-970-memory-debacle.html

Some people don't have enough washing or ironing to keep them occupied.

---

Grant
Darwin NT

Doesn't anyone have anything to add to this news? I imagine it effects SETI users in the fact that you don't ONLY crunch SETI with your GPU's? Right?

Yes, only use the GPU's for distributed processing. See my earlier post in the thread. Of absolutely no concern for most Seti crunchers.

Cheers, Keith

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

Hello Keith,
I have two 970s and I'm trying to increase the number of tasks per card. Unfortunately, I'm not exactly a computer guru so I don't really manage to write a good app_info.xml. Up to now I managed to successfully delete the already existing tasks. Would you mind telling me how you've configured your boinc to increase the number of tasks? An example of app_info.xml would be great (I assume that's what you're using?).
I'm running boinc (seti@home) on a windows 7 ultimate (64-bit), with 2 970s.
Any help is greatly appreciated.
Regards,
Radu

Nevermind, Keith. I've discovered Lunatics in the meantime. :)

Sorry about not replying earlier but I stopped following the thread. Glad you found the Lunatics installer. Makes it a whole bunch easier to optimize thruput without having to hand edit a app_info. I did that for a long while until the Lunatics installer made it so easy. Let me tell you, hand editing an app_info for multiple applications wasn't fun and very easy to screw things up. Many times I lost tasks because of a simple syntax error. Now that resends are turned off, that would be even more disastrous. The other app I found very handy is the SetiPerformance app by efMer. That is what helped me determine that the most optimized thruput for my twin 970's was three tasks per card. I also reduce my total concurrent CPU tasks to make sure there are free CPU's to feed the graphic cards. You also might look into optimizing the app_config files for each project. Also you should be running an optimized MBCuda.cfg file and also an optimized ap_cmdline_win_x86_SSE2_OpenCL_NV file for whenever we get the AP database back. Each little tweak helps the daily thruput.

Cheers, Keith

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)