GTX 970 VRAM (update: 900M GPUs not affected by 'ramgate')

GTX 970 owners will get to choose 1 free game from Steam

 

Where did you hear this? Just purchased a 970 last week. Thanks

 

As of today GameNab has been closed.
http://gamenab.net/

 

It was a semi joke..

 

Putting this 970 VRAM issue to bed, esp. for Cloudfire: http://www.pcper.com/reviews/Graphics-Cards/Frame-Rating-GTX-970-Memory-Issued-Tested-SLI

 

...and hello GTX 960 Ti [OEM]

 

Everybody should watch this amazing truth about Nvidia!

 

Only been linked about a half dozen times already...

 

http://linustechtips.com/main/topic/302368-the-gtx-980m-also-has-memory-issues/

So.... could it be actually true that Nvidia has lied (again) and this same issue affects the 980M? Does anyone with a 980M experience stuttering at 3.5GB of vram? (for 4gb variants)

 

980M is unaffected. It is Gamenab who is lying.

 

So does this thread prove that if you beat a dead horse long enough, it will somehow magically spring back to life? And yes I fully realize the irony of me posting in this thread yet again.

 

Bad news everyone, seems the GTX 980M used in the ASUS G751JY has some minor problems aswell...



Now we need to wait for a game which uses more than 3,5 GB VRAM to see the issue

 

Jee! Just put someone in the first post info that test is legit with Optimus ONLY already!

P.S. This may be done by moderators or members too.

 

Probably because the Nvidia mobile GPUs are based on their less powerful desktop siblings, in this case: 980M <=> 970.

 

It was exactly 3.5G though.

 

as said before, no optimus on that machine, do dont care...

 

Just about any game with DSR or Dying Light.

 

Not entirely true. BF4 for example as tested by PCperspective doesn't take more until resolution goes higher than 1.2x4K.

 

Another evidence for two-way 980M SLI (from a MSI GT80):
http://i.imgur.com/h8igLDO.png

That's about the bumpgate (2008). Let's see what will be their behavior this time.

 

looks fine

 

lol the last 500MB DRAM is running at 6-8GB/s.....

But its not. Stop using that broken benchmark

 

What's wrong with the benchmark?

 

Did you read my post?
That it report 6-8GB/s. GTX 980/980M/970M only have one memory bank

 

Read again.

 

Should probably close this thread...

 

Read what?

 

Dude, now I understand what sort of mental confusion is going on with you.
You still didn't get it. Stop spreading lies.
https://forums.geforce.com/default/...gtx-970-why-frame-time-analysis-is-important/
Read the article. Read the comments.

 

You didnt (wasnt able to) answer my question. The link you posted didnt either. I can post other tests that found nothing out of the ordinary with FCAT. But that doesnt matter. This is about GTX 980M.

The benchmark reported 6-8GB/s on 980M.
The proposed issue with GTX 970 is because it have two memory banks. GTX 980M don`t.


Now why should this be like bumpgate like the fear mongering you tried on the previous page? Why should mobile cards be affected? Explain with your own words

 

Yeah, not sure why its even on this forum. It was relevant before it was discovered that mobile chips doesnt have two memory banks anyway

 

Any last comments on this "issue"? Otherwise, as suggested, the thread will be closed.

 

Yeah. Last comment will be in courtesy of the author whose Benchmark revealed us the truth:
If benchmark needs fixed hands and head to use then it is not the program which is broken.

It was told tenth of times twice per page that you need Optimus to get legit results! If you don't, speed is skewed.
How this easy rule could become such a big wall for somebody? Guess they were too busy praying fort memory controllers.

 

Yes, edit the first post so people know this only affects DESKTOP 970 cards, and none of the mobile cards suffer from this issue so they can all take a deep breath and relax.

Also this:

 

Done

Sent from my Nexus 5

 

LAPTOP GPU`S RULES 4EV3R OMG!!1!

 

Can we stop with the -gates already? That's so 1972. I get enough of this crap in the daily news.

 

Gategate confirmed!

In all seriousness, I wonder whether all this furore over the 970 has scared NVIDIA off from ever releasing another GPU with a partially disabled ROP/L2 cache cluster ever again.

 

Pffft, no way. Most people have a short memory and I'd say 99% of users who buy the 970 don't care, don't know, or aren't affected by this at all.

 

What was the benefit to Nvidia, did they save on materials cost with this approach?

 

They sold more 970 cards than they would have if they'd advertised it as a 3.5GB 224-bit card.

 

Well, how much more expensive would it have been to include normal 4GB video memory?

 

A lot more expensive in the long run. They would have to use non-defective GM204 chips with the full amount of ROPs and L2 cache, for starters. There's a reason 970 is so much cheaper than 980.

 

I see, so they found a crafty way to save on materials cost using these defective chips - now I wonder why Nvidia and AMD don't do this a lot more often.

 

Haha, nice. Your sarcasm was pretty crafty, too.

But seriously, everything would be at peace if Nvidia had actually released the correct specs for the 970, like they did for every other cut-down GPU ever.

 

They always did. Except this time the architecture of Maxwell's chip projecting brought such consequences.

 

Well, we had a rollercoaster of a conversation over dekstop GPUs and all. Thread is closed since this does not have bearing with laptop GPUs.